Carbon Taxes and Deficit Reduction. Jared C. Carbone University of Calgary and Resources for the Future

Transcription

1 Carbon Taxes and Defic Reducion Jared C. Carbone Universy of Calgary and Resources for he Fuure Richard D. Morgensern Resources for he Fuure Roberon C. Williams III Universy of Maryland, Resources for he Fuure, and NBER Preliminary Draf, May 2012 Absrac This paper looks a boh he efficiency and disribuional implicaions of a carbon ax and/or oher energy axes as par of a package of measures o address he budge defic. We build a dynamic general-equilibrium overlapping-generaions model (OLG) of he U.S. economy, including deail on governmen axes and expendures and a disaggregaed producion srucure including several energy indusries. We find ha boh he overall cos of including a carbon ax in a package of defic reducion measures and he disribuion of ha cos across generaions vary significanly based on wha oher ax and spending measures are included in ha package, and vary que subsanially based on which of hose measures he carbon ax revenue is used o offse. Moreover, he effecs of a carbon ax whin a defic-reducion package can differ significanly from he effecs of a revenue-neural ax swap (as modeled in he prior leraure). For heir helpful commens and suggesions, we hank Lin Barrage, Dallas Burraw, Alan Krupnick, and seminar paricipans a he Congressional Budge Office and he Norheas Workshop on Energy Policy and Environmenal Economics. We also hank he Biparisan Policy Commission, he Smh Richardson Foundaion, and Resources for he Fuure for financial suppor, and Daniel Velez-Lopez and Samuel Grausz for excellen research assisance.

2 1. Inroducion There is a growing consensus ha U.S. fiscal policy needs major reform in order o bring he budge defic and corresponding growh in he naional deb down o levels ha are susainable in he long run. Many oher naions face similar budge issues. A he same ime, he world faces he challenge of reducing emissions of greenhouse gases in order o lim he effecs of global climae change. Including a carbon ax as one elemen of fiscal reform may help o address boh of hese major long-run issues. This paper looks a boh he efficiency and disribuional implicaions of a carbon ax and/or oher energy axes as par of a package of measures o address he budge defic. The leraure on ax ineracions and he double dividend has considered he general issue of ineracions beween environmenal axes and he broader fiscal sysem, in he conex of budgeneural changes ha use revenues from environmenal axes o fund cus in oher axes or increases in spending. However, while policy ineres in he use of environmenal axes for defic reducion can be raced a leas o he Clinon Adminisraion s 1993 BTU ax proposal, o our knowledge, no paper in ha exensive prior leraure models he use of environmenal axes as par of a defic reducion package. Addressing his problem requires differen concepual and mehodological ools. Defic reducion is an inherenly dynamic issue whereas mos environmenal ax-ineracion models are saic. Moreover, he dynamic models ha do exis (e.g., Bovenberg and Goulder, 1996) assume infinely-lived agens which, due o Ricardian equivalence wherein governmen borrowing is fully offse by privae secor saving prevens hem from realisically modeling he effecs of changes in he iming of axes or spending. To address hese issues, we build a dynamic general-equilibrium overlapping-generaions model (OLG) of he U.S. economy, including deail on governmen axes and expendures and sufficien energy-secor deail o model he effecs of inroducing a carbon ax, oher energy axes, or modificaions o he res of he ax sysem such as changes in labor and capal income ax raes or he inroducion of a value-added ax. The model hus combines elemens of Auerbach and Kolikoff s (1987) dynamic model of fiscal policy wh elemens of environmenal ax-ineracion models such as Bovenberg and Goulder (1996). "

3 This overlapping generaions srucure has several key advanages. Firs, he model does no exhib full Ricardian equivalence, and hus can more realisically model he effecs of changes in he ime pah of governmen budge defics. Second, provides a more realisic resul for he effecs of ax policy on capal accumulaion, because (unlike a model wh infinely lived agens), does no imply infinely elasic capal supply. This is obviously crucial when considering capal axaion, bu is also imporan for analyzing oher axes ha may indirecly affec capal, such as a carbon ax. Third, he overlapping generaions srucure perms us o examine how agens in differen generaion are affeced by policy reforms. This is likely o be imporan from he perspecive of welfare analysis as well as he analysis of he behavioral response o he policies. We use he new model o examine he effecs of including a carbon ax or oher energy ax changes in a package of defic-reducion measures in he conex of differen ime pahs of defics while achieving a susainable long run deb/gdp raio. We explicly consider how hose effecs vary based on wha oher ax or spending changes are included in he package, while focusing on boh efficiency effecs and he disribuion of burdens across generaions. We find ha boh he overall cos of including a carbon ax in a package of defic reducion measures and he disribuion of ha cos across generaions vary significanly based on wha oher ax and spending measures are included in ha package, and vary que subsanially based on which of hose measures he carbon ax revenue is used o offse. Carbon axes are generally viewed as imposing a burden on presen generaions in order o benef fuure generaions. However, our resuls sugges ha his is no always he case: including a carbon ax as par of a defic reducion plan can make presen generaions beer off. The plan of he paper is sraighforward. The nex secion develops a deailed descripion of he model. Secion 3 explains he policy simulaions and presens resuls. Secion 4 concludes and discusses poenial direcions for fuure work. #

4 2. Model Logic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uses o finance he producion of governmen services and make ransfer paymens o privae households. The governmen may run a defic or surplus in a given ime period, and he modeler may exogenously impose new governmen policies ha change expendures or ax revenues o affec he size of he defic and he accumulaion of public deb over ime. Governmen policy changes are announced in = 0 wh perfec cerainy and a household is assumed o have perfec foresigh over s lifeime in responding o he effecs of policy changes. Concepually, he model has wo main componens a descripion of he dynamics of economic growh and decision-making over ime, and a muli-secoral descripion of he radeoffs beween economic acivies a a poin in ime. The logic and calibraion sraegy for he dynamic componen of he model builds direcly on he work of Rasmussen and Ruherford (1994). The muli-secoral componen builds on he work of Böhringer, Carbone and Ruherford (2011). We begin by describing he aggregae economy and how evolves over ime. We hen move on o describe he household-level choice model and how connecs o he aggregae economy. This is followed by a discussion of he muli-secoral producion side of he economy represened a each ime sep in he model, a descripion of he inial seady-sae and baseline calibraion procedures.

5 In he discussion ha follows, we use he accouns for period = 0 o illusrae he relaionships beween he variables in he model unless oherwise indicaed. Uns are chosen such ha period-0 benchmark prices are normalized o one, implying ha physical quanies and he value of hose quanies coincide. We also follow he convenion of represening prices in he model in period-0 presen-value erms excep where indicaed. Thus, in he inial equilibrium, a price a ime in he model, p = (1 + r ) where r is he benchmark real ineres rae. 2.1 Aggregae Demand Naional accouns describe he relaionship beween aggregae income, savings and consumpion. Naional income is derived from he reurn on capal ( R ), labor ( L ) and naural resources ( N ). Households also receive governmen ransfers (T ). Thus income balance requires ha, in each period, naional income is equal o naional consumpion (C ) plus saving ( S ). R+ L+ N + T = C+ S (1) Invesmen ( I ) equals privae saving reduced by public dissaving in he form of a budge defic ( D ) and raised by he amoun ha invesmen is financed from abroad via a rade defic ( B ). S D+ B= I (2) The size of he budge defic is deermined by he governmen's budge balance condion, requiring ha oal ax revenues ( ) mus cover governmen spending (G ) and ransfers minus he defic. = G+ T " D (3) Tax revenue is derived from a labor income ax ( L ), a ax on capal income ( K ), a consumpion ax ( C ), a ax on carbon emissions ( E ) applied o oal emissions ( E ), and a fuel ax ( F ) applied o he oal value of refined-oil producs consumed domesically ( F ). These ax %

6 raes, along wh he level of governmen spending, are he policy insrumens ha are he subjec of our deb-reform counerfacual scenarios. In addion, here are benchmark axes presen in he GTAP daase ha forms he basis for he calibraion of he model on he use of produced goods in inermediae and final consumpion as well as impor ariffs and expor subsidies. These ax raes are held consan in our policy scenarios bu he value of he revenues colleced from hese sources may change as he economy responds o he various policy inervenions we inroduce. The ne value of hese ax revenues is summarized by 1. " = L+ R+ C+ E+ F+# (4) L K C E F The aggregae capal sock ( K ) evolves based on a consan rae of depreciaion ( ) and he level of invesmen. K = 1 (1 " ) K + + I (5) The model is calibraed o an inial seady sae in which all of he componens of he aggregae economy grow proporionaely a a growh rae equal o he consan, exogenous rae of populaion growh ( ) plus a consan, exogenous rae of labor-augmening echnological change (). This equilibrium implies ha he benchmark reurn o capal jus covers he cos of ineres plus depreciaion R= ( r + ) K (6) and new invesmen covers depreciaion plus he growh of he economy I = ( + + " )K (7) This represens a complee descripion of aggregae demand and how evolves over ime in he inial seady sae. We now show how he overlapping generaions srucure in he model links represenaive household choices of each generaion o he aggregae oucomes described above. 1 A more deailed descripion of he axes represened in he GTAP daabase and how we make use of hem in our model is described in Appendix &

7 2.2 Household Opimizaion We refer o households enering he model in ime period g = as household g. Each household g opimizes by choosing s levels of consumpion ( c g ) and leisure demand ( l g ) o maximize he discouned sum of insananeous uily over s lifespan from g g+ N. Insananeous uily depends on he amoun of leisure and consumpion a household consumes, where hese quanies ener a consan elasicy of subsuion (CES) funcion o creae a fullconsumpion bundle ( z g ). max c, l g g g g+ N 11/ ' ) 1 * zg = 1 +, = g- 1+ ". 1' 1/ s.. z = (# c + (1 '# ) l ) g+ N g+ N " C " C 1/ " C g C g C g l N FX p c / p (% ' l ) + p n + p p &, l 0 0 ' g C g g g g g g = g = g i( FF l c g g / % g g U (8) where U g is lifeime uily, is he rae of ime preference, is he ineremporal elasiciy of subsuion, 1/(1 # ) = " is he elasicy of subsuion beween consumpion and leisure. C C Households spend heir wealh on consumpion goods, where p C is he price index of he final consumpion bundle. They finance his consumpion based on income derived from a number of differen sources. Labor income, depends on he wage rae, l p, g, he index of labor producivy over he life cycle, which reflecs he exogenous accumulaion of human capal and depreciaion of skills workers experience over he lifecycle and g, he household's period- effecive ime endowmen. I is an effecive endowmen in he sense ha reflecs he effec of labor-augmening echnological change over ime such ha, 1 = (1 + ). I grows across g+ g '

8 generaions o reflec populaion growh as well 1, 1 = (1 + " + ). Households also hold g+ + g endowmens of naural resources used in fossil energy producion, n g ( ng = N ), where g N p is he naural resource ren associaed secor k and FF is he se of fossil fuel secors. Resource endowmens grow such ha n = gk, 1 n and + gk, ng+ 1, k, + 1 = n g (1 + + ). They also receive governmen ransfers, g (" g = T ), which follow a calibraed ime pah across he g lifecycle based on social secury and oher major benef-program disribuions. They grow across generaions according o he overall growh rae of he economy. Governmen ransfers are denominaed in erms of he foreign exchange good in he model, where FX p is he price of foreign exchange and p = (1 + r ) is he presen value world price index. In addion, generaions alive a = 0 and = T hold asses (suppressed in (8)) ha accoun for he inial capal sock, he benchmark imbalance in inernaional rade and he asses required o finance pos-erminal consumpion. These aspecs of he model are discussed in more deail in secion 1.4 and 1.5 respecively. The composion of he final consumpion good ( c ), described in more deail in secion 1.3, is assumed o be common o all households. 2.3 Producion The producion side of he model is based on a muli-secoral descripion of he economy wh compeive, consan-reurns-o-scale indusries and a connecion o inernaional markes based on he small open economy assumpion. The producion funcions ha describe he echnologies used in each secor of he economy are based on nesed CES funcions. Firms combine basic facors of producion (physical capal, labor and naural resources) wh inermediae inpus. We explicly represen sixeen secors in he model, lised in Table 1. Fifeen of hese secors are aken direcly from he disaggregaed GTAP 7.1 daase (each secor name in he able is accompanied in parenheses by s hree-leer GTAP code) wh he focus on represening energy and energy-inensive secors of he economy, hose mos direcly affeced by a carbon ax. The final secor is a compose of all oher privae-secor acivies in he economy. (

9 Final consumpion (C ), governmen services (G ) and invesmen ( I ) are all produced using an idenical bundle of inpus based on hese 16 secors. SECTORS Energy Coal (COA), Crude oil (CRU), Naural gas (GAS), Refined peroleum and coal (OIL), Elecricy (ELE) Energy-and rade-inensive Paper, pulp, prin (PPP), Chemical, rubber, plasic producs (CRP), Iron and seel (I_S), Non-ferous meal (NFM), Nonmeallic mineral (NMM), Machinery and equipmen (OME), Plan-based fibers (PFB, Waer ranspor (WTP), Air ranspor (ATP), Oher ranspor (OTP) Res of indusry and services (Y) Table 1: Model Secors The srucure for he producion of all non-fossil-fuel goods (including C, G and I ) is described diagramaically in Figure 1 and algebraically (via he definion of he un-cos )

10 funcion) in equaion (17). The producion srucure for fossil fuels in he model (coal, naural gas and crude oil) are described diagramaically in Figure 2 and algebraically in equaion (22). We model inernaional rade based on he Armingon assumpion ha domesic and foreign varieies of a good are reaed as imperfec subsues by consumers. For expors o foreign markes, we accomplish his hrough he use of a consan elasicy of ransformaion, A = 1/(1 #" ), beween oupu produced by domesic firms desined for eher he home or he i expor marke. i " [# HS + (1 # ) X + ] = Y (9) HX i HX i 1/ i Xi i i where he and erms are chosen o mach benchmark levels of oupu described in our daase, X is supply desined for expor, oal oupu in indusry i in ime period. HS is supply desined for he home marke and Y is Similarly, impored goods are assumed o combine wh domesic varies of he same A good in compose where = 1/(1 #" ) is inerpreed as he elasicy of subsuion beween i i home and foreign varieies. Thus he compose of domesic ( HD ) and foreign ( IM ) i i varieies of good i demanded in producion acivy k, A i, is given as A = " [# HD + (1 # ) IM ] (10) A i A i 1/ i i Aik ik i ik i We assume he exisence of a perfec inernaional bond marke of which he domesic economy represens a small par. As a resul, he world ineres rae is fixed a r. Domesic households save via invesmen in he domesic capal sock. The domesic ineres rae may deviae from he world ineres rae in ransion because building up he capal sock requires purchase of he invesmen good in he model, which is produced using imperfecly subsuable *

11 domesic and impored goods as inpus. However, inernaional rade in goods guaranees equalizaion of domesic and inernaional raes in he seady sae 2. There is a single, homogenous labor marke across secors. We examine insances of he model where physical capal is reaed eher as inersecorally mobile or as secor specific. We assume ha neher labor or capal is inernaionally mobile. 2.4 Treamen of Benchmark Asses and Defics The model is calibraed o an inial seady sae in which he economy runs permanen budge and rade defics, boh of which grow a he benchmark growh rae of he economy ( + ). For ineremporal budges o balance in a seady sae wh permanen defics, agens mus hold asses equal in presen value o he infine-horizon defics o ensure ha heir ineremporal budges are balanced. To accoun for hese debs, households alive a = 0 in our model are assumed hold aggregae inial asses, A, equal o 1+ r A= (1 + r) K + ( B" D) r " " (11) The firs erm of he righ-hand side of he equaion capures he inial value of he capal sock consisen wh he seady sae and he second erm is he soluion o he infine series ha describes he value of fuure rade defics less budge defics. In oher words, privae households are assumed o hold an endowmen of foreign asses sufficien o cover he value of fuure rade defics bu owe a deb equal o he value of fuure budge defics o he governmen agen in he model. These asses are assumed o be held in proporion o he populaion size of each generaion. The governmen agen holds endowmens in s period-by-period budge sufficien o cover he value of he defics runs. Thus in period, " + pp D= p G + pp T (12) FX FX G 2 For example, if inernaional bonds consisenly earn a higher reurn han domesic asses, his will cause he exchange rae o depreciae, simulaing he demand of domesic goods abroad and raising he marginal produc of capal and he domesic ineres rae. "+

12 where he lef hand side of he equaion includes he period- value of governmen revenues from axes ( ) and he value of he foreign-exchange good ha covers he value of he period- FX defic ( pp D). p G is he price index for he compose good ha produces governmen services and ransfers ( T ) are denominaed in erms of he foreign exchange good. In our policy scenarios, we esablish a series of baselines in which he primary revenueraising axes in he model ( K, L, C ) and governmen spending cus ( ) are used o draw down he he budge defic o mee a arge of level of cumulaive reducion in he level of public deb by period T,. We hen perform policy experimens off of hese baselines in which we swap carbon and fuel axes for heavier reliance on he oher revenue-raising insrumens o h he same deb arge. When we affec he changes in he ax raes and spending levels o achieve he deb-reducion baselines, he levels of and (from equaion (12)) o reflec he effecs of hese policies and he size of he foreign-exchange endowmen covering he defic ( D ) changes accordingly. The level of governmen ransfers is held consan hroughou our experimens. 2.5 Treamen of Pos-Terminal Behavior Numerical models can only cover a fine number of ime periods. Neverheless, he logic of he equilibrium concep requires ha we describe he behavior of model agens over an infine horizon. Firs, we mus specific how invesmen in he capal sock proceeds in period T. We assume (and verify in our simulaions) ha he model equilibrium is in a seady sae a ime T. Accordingly, we se erminal period invesmen a a level consisen wh seady-sae growh. Thus he pos-erminal capal socks, KT, are chosen o saisfy i I it / I i,t 1 = 1+ " + (13) Second, we mus describe he behavior of households alive in period T in he poserminal par of heir lives. Tha is, because households are forward-looking in our model, he consumpion and saving choices hey make before period T depend on wha choices hey are assumed o have made afer period T. Moreover, wha hese pos-erminal choices are assumed ""

13 o be depends on he differen seady saes generaed by our policy experimens. To solve his problem, we specify rules describing adjusmens o pos-erminal consumpion and asse posions ha are consisen wh opimal behavior in a seady sae bu which are a funcion of endogenous model variables ha indicae he paricular seady sae arrived a in he model in he differen policy experimens. Specifically, we assume ha erminal households g are endowed wh pos-erminal asses (denominaed in erms of he foreign exchange good in he model) equal o agt AT g, which are chosen such ha welfare level of erminal households is equal for all erminal households in he model. Ug Ug 1 = 0, g+ N > T (14) We choose he pah of he pos-erminal full-consumpion for erminal households such ha ha shadow value of his consumpion ( ZT p which, again, is consisen wh seady-sae behavior. ) declines a he seady-sae ineres rae, r T, Z ZT a 1 pg a 1, T= pga(1 + rt), g+ N > T, g" a" N (15) 2.6 Calibraion The muli-secoral producion model is calibraed based on he social accouning marices described in he GTAP 7.1 daabase, aggregaed o he secors described in Table 1 using he sandard calibraion procedure (Shoven and Whalley 1992). The base year in he GTAP daa is The assumpions for he values of he subsuion elasicies in he producion and uily funcions in he model are described in Table 7. I is worh noing ha we use he op-level subsuion elasicies beween secor-specific naural resources ( N ) and oher inpus o producion in he fossil-energy secors in he model (coal, crude oil and naural gas) o imply supply elasicies ha are consisen wh empirical evidence on he price responsiveness of hese energy sources. We assume ha he benchmark reurn o capal ( R ) in he GTAP daase represens he seady-sae reurn in our inial equilibrium and infer he level of he inial capal sock, K, 0 "#

14 consisen wh ha reurn (based on equaion (6)). We hen use our calculaion for esimae of K o infer he level of seady-sae invesmen required. Our esimae of B, he rade defic, 0 comes from he GTAP daa. Our esimae for he benchmark budge defic is based on Congressional Budge Office (CBO) long-run projecion of ~3% of GDP ("An Analysis Of The Presiden's Budgeary Proposals For Fiscal Year 2012," CBO, hp:// The inial level of publicly-held deb is also based on he CBO esimae ha exising deb represens approximaely 70% of GDP in Benchmark expendures on governmen services (G ) is aken from he GTAP daa. The level of governmen ransfers is hen inferred based on he benchmark level of ax revenues in he GTAP daa ( ) and he accouning ideny for he governmen's budge (3). We assume an annual world equilibrium (and domesic benchmark) ineres rae of 5%. The annual rae of populaion growh is 1% and he annual rae of labor-augmening echnological change is 2%. The annual rae of depreciaion of he capal sock is 7%. For he mos par, we use he GTAP benchmark axes o calibrae he model. The excepions are ha we replace facor axes for labor and capal wh exernal esimaes. We calibrae he benchmark marginal labor ax rae o a value of 35.8% (Barro and Redlick 2011) and he marginal capal ax rae o a value of 39.9% (Babiker, Mecalf and Reilly 2003). The producivy index ha describes he rajecory of lifecycle wages comes from Auerbach and Kolikoff (1987) and follows 2 ( * a" * a ) 4.47 = / (16) a e e where a represens he age of he individual in years. Figure 3 shows his producivy index over he lifecycle, which rises unil middle age and declines hereafer. This figure also shows he allocaion of ime (measured in efficiency uns) beween work and leisure in he benchmark equilibrium, a each age. The calibraion of he lifecycle pah of income received by households in he form of benefs from governmen ransfer paymens is based on he repor of income from governmen ransfers by age from he 2010 Consumer Expendure Survey (hp:// In addion, we represen he progressivy in he income ax sysem (i.e., ha marginal ax raes "

15 exceed average ax raes) by modeling a proporional ax sysem plus a ransfer (where he ransfer equals he difference beween marginal and average raes, imes he amoun of income axed). Adding ha erm o he observed ransfer paymens in he daa gives he ransfer paymens in he model ( ). g Figure 4 shows he lifecycle pah of consumpion and he lifecycle pahs of income from labor, capal, and governmen ransfers, for he benchmark equilibrium. Labor income increases unil middle age and declines hereafer, reflecing boh he producivy index and he amoun of ime spen working, each of which also peaks in middle age. Transfer paymens are much higher in old age han earlier in he lifecycle, reflecing Social Secury and oher programs ha arge he elderly. And capal income peaks lae in life, and hen declines as households spend down heir asses. Households save o smooh consumpion over he lifecycle, so he pah of consumpion rises smoohly hroughou he lifecycle, even as he individual componens of income vary. A challenge presened by he OLG framework is ha he individual consumpion and saving choices of households in he benchmark mus be consisen wh he associaed aggregaed quanies in he economy a each ime sep. We follow he procedure described in Rasmussen and Ruherford (1994) wh some minor modificaions o accoun for he specific deails of our applicaion. We summarize he procedure and our modificaions below. The calibraion explos he srucure of he inial seady-sae equilibrium. All households solve he same generic maximizaion problem and are assumed o have idenical preferences. In he seady sae, equilibrium quanies grow proporionaely. As a resul, he seady-sae lifecycle choices of fuure generaions are idenical o he choices of curren generaions down o a scaling facor ha capures he seady-sae growh of he economy. Thus, in our calibraion, he generaion enering he model a period 0, g = 0, is used o generae reference pahs for benchmark consumpion, leisure demand and asses. These reference pahs are hen used o populae he benchmark rajecories for all generaions of households. A candidae se of values for he household-level rajecories yields a predicion on he aggregae levels of consumpion, labor supply and asses in = 0. We calibrae some of he parameers enering he household choice problem ( and a proporional scaling facor for g ) "%

16 o joinly ensure ha he individual oucomes are consisen wh uily maximizaion while he aggregae oucomes are consisen wh he economic daa used o calibrae he model. Noe ha he level of he governmen budge defic in he benchmark (approximaely 3% of GDP, as described above) implies ha he governmen deb/gdp raio is rising. Thus, by using his as he benchmark seady sae for he calibraion, we are implicly calibraing o an economy in which he governmen deb/gdp raio is rising forever, a an increasing rae. Moreover, because much of he governmen budge defic is currenly being funded by capal flows from overseas, he benchmark implicly assumes ha such capal flows will coninue forever. Figure 5 shows he implied pah of he deb/gdp raio over ime for he benchmark. Governmen deb rises o nearly 700% of GDP in 100 years. Such high levels of governmen deb are obviously unsusainable, and hus in his respec, he benchmark equilibrium is highly unrealisic. However, his is he pah ha he U.S. is currenly on, and hus calibraing o a benchmark equilibrium wh a susainable long-run deb/gdp raio would require some assumpion abou wha fuure policy changes will pu he U.S. on a susainable budge pah. Raher han make an arbrary assumpion abou hose fuure changes, we ve insead chosen o calibrae o he curren pah, even hough is unsusainable. 3. Policy Simulaions We use he model o perform wo ses of policy simulaions. In he firs se of simulaions, we simply model revenue-neural ax swaps ha impose a carbon ax and lower one of he oher axes in he model, keeping he presen discouned value of ax revenue he same as in he benchmark equilibrium. In he second se of simulaions, we model defic-reducion packages, and look a he effecs of including a carbon ax in he package of defic-reducing policies. 3.1 Revenue-Neural Tax Swaps "&

17 The firs se of simulaions model revenue-neural ax swaps. In each case, we impose a 20/on carbon ax, and lower he ax rae on labor, on capal, or on consumpion, so as o hold he presen discouned value of governmen revenue consan. Thus, he ax changes in his firs se of simulaions are very similar in naure o hose from prior papers in he ax-ineracion leraure (e.g., Bovenberg and Goulder, 1996). This provides a good comparison o he resuls from prior work, as well as illusraing he basic effecs a work in he model. In each case, he ax swap is modeled as an immediae and unanicipaed change in ax raes a ime 0, which is hen permanen hereafer. Thus, he carbon ax rae goes immediaely o 20/on in period 0, and he oher ax rae (capal, labor, or consumpion) immediaely drops by enough o offse he carbon ax revenue, keeping he presen discouned value (over all fuure periods) of governmen revenue consan. Then boh axes say consan over ime a he new raes hereafer. Noe ha his does no necessarily mean ha revenue is held equal o he benchmark in each ime period doing ha would require adjusing he offseing ax rae in each period and hus he shor-erm ime pah of ax revenue may change slighly as a resul of he ax swap. These simulaions also keep governmen spending he same as in he benchmark, so he long run pah of governmen deb will be he same as in he benchmark (wh he governmen deb/gdp raio increasing o unsusainable levels), bu he shor-erm ime pah may differ slighly. 3 By holding he carbon ax rae consan over ime, hese simulaions differ from mos policy proposals for a carbon ax, which ypically involve a ax rae ha rises over ime. We do his for simplicy: is easier o undersand and o explain he effecs of a consan carbon ax rae han he effecs of a carbon ax rae ha changes over ime. Similarly, many proposals would announce he carbon ax prior o when is imposed, so ha he economy has some ime o adjus before he policy akes effec. 4 Again, for simplicy, we assume ha he ax swap is implemened immediaely, whou any pre-announcemen. 3 An alernaive approach would be o hold revenue consan in each ime period, which would imply ha he offseing ax rae cu would need o vary slighly from period o period. This would make he resuls somewha more difficul o inerpre, because agens in he model would reac o he period-o-period ax changes. Looking a a ax swap ha keeps raes consan over ime (afer he inial change) is simpler. 4 However, Williams (2010) suggess ha is more efficien no o pre-announce environmenal policy. "'

18 Figure 6 shows he revenue raised by he carbon ax for he firs 25 years, in each of he hree cases considered (swapping a carbon ax for a lower ax rae on capal, labor, or consumpion). In each case, he 20/on ax inially raises roughly 105 billion/year (in 2004 dollars), wh he revenue rising over ime a roughly 3% per year as he economy grows. Noe ha he carbon ax revenue is no exacly he same across he hree cases, bu is very similar. Wheher he carbon ax subsues for a ax on capal, labor, or consumpion maers for he size and growh rae of he economy, as well as for he composion of he economy, and hose in urn maers for he level of carbon ax revenue. Bu he resuling differences are que small. Figure 7 shows he offseing ax cus made possible by he revenue from he carbon ax. The carbon ax can fund roughly a 4.4 percenage-poin cu in he ax on capal, a 1.46 percenage-poin cu in he ax on labor, or a 0.86 percenage-poin cu in he ax on consumpion. As noed earlier, each of hese offseing ax cus is a one-ime, permanen cu ha exacly offses he revenue from he carbon ax, such ha he presen discouned value of ax revenues remains he same as in he benchmark. There are wo reasons why he percenage-poin cu is bigges for he capal ax and smalles for he consumpion ax. Firs, he ax base is bigges for he consumpion ax (consumpion is more han eher labor or capal income by self) and smalles for he capal ax (capal income is a smaller share of he economy han labor income). In he absence of any behavioral responses o he ax change, a given amoun of revenue can fund a larger percenagepoin cu for a ax wh a smaller base. Second, o he exen ha here is a behavioral response (and hus an efficiency gain) from he ax cu, ha will imply a smaller revenue loss from any given rae cu or a larger cu for any given revenue loss. The capal ax is he mos disorionary of hese hree axes, while he consumpion ax is he leas disorionary. Thus, boh of hese reasons cause he offseing percenage-poin cu o be bigges for he capal ax and smalles for he consumpion ax. The ne effec of he ax swap on he welfare of each generaion is shown in Figure 8. The horizonal axis in he figure covers he differen generaions in he model ha are affeced by he ax swap, saring from he generaion born 50 years before he ax swap (generaion -50) and running hrough he generaion born 100 years afer he ax swap (generaion 100). The verical axis shows he equivalen variaion for he welfare effec of he ax swap, expressed as a "(

19 percenage of discouned remaining lifeime consumpion (where a posive equivalen variaion indicaes ha a given generaion is made beer off). Noe ha his measure does no include any environmenal benefs ha resul from lower carbon emissions. 5 The firs hing o noe in his figure is he difference in overall efficiency across he hree cases. Mos generaions are beer off when carbon ax revenues are used o cu he capal ax han hey are if hey are used o cu he labor ax, and beer off when revenues are used o cu he labor ax han if hey are used o cu he consumpion ax. And he overall efficiency ranking of he policies reflecs ha: using carbon ax revenues o cu he capal ax is overall he mos efficien of he hree cases, and using hem o cu he consumpion ax is he leas efficien. As noed earlier, he capal ax is he mos disorionary of he hree axes, and he consumpion ax he leas disorionary, so he revenue-recycling efficiency gain is larges when cuing he capal ax and smalles when cuing he consumpion ax. These efficiency resuls are very similar o resuls from he prior leraure on environmenal ax swaps in models ha include boh capal and labor axes (e.g., Bovenberg and Goulder, 1996). However, because hose models assumed an infinely-lived represenaive agen, hey couldn look a he disribuional effecs across generaions. Those disribuional effecs are ineresing and imporan. While using he carbon ax revenue o reduce he consumpion ax is he leas efficien of he hree cases, provides large ne benefs for generaions ha are relaively old a he ime of he ax swap. Generaions ha are a leas 35 years old a he ime of he ax swap are made beer off by he ax swap, even ignoring any environmenal benefs. Agens near he end of heir lifespans are consuming a lo, so hey benef grealy from he lower consumpion ax. The carbon ax also pushes up he price of consumpion, bu because some of he burden of he carbon ax is passed back o facors of producion (capal and labor), pus less burden on consumpion han he consumpion ax does. Thus, he ne effec is o make he oldes generaions a he ime of he swap beer off. Bu his comes a a subsanial cos for younger generaions. & We om environmenal benefs no because such benefs are unimporan indeed, hey are probably he mos imporan reason o pursue an environmenal ax swap bu because our focus here is on he non-environmenal effecs, and because of he uncerainy in esimaes of he environmenal benefs. ")

20 Using he carbon ax revenue o cu labor axes has an oppose effec on he oldes generaions. Agens near he end of heir lifespans are working very lle, and are deriving mos of heir income from capal and from governmen ransfers. Thus, hey ge very lle benef from he labor ax cu, bu sill bear a significan burden from he carbon ax. Younger generaions do beer under his policy, hough no generaion ges a ne benef from he swap (again, ignoring any environmenal gains). Ineresingly, he welfare effec across generaions in his case is non-monoonic in age a he ime of he swap: generaions ha are young a he ime of he swap bear a smaller ne cos from he swap han eher hose born earlier (for he reasons noed in he previous paragraph) or hose born laer. One effec in his labor-ax-swap case is o lower he long-run level of he capal sock (imposing a carbon ax and lowering he labor ax raises he ne ax burden on capal, hus discouraging saving and lowering he capal sock). This lower long-run capal sock lowers reurns o labor for fuure generaions, making hem worse off. Bu akes some ime o ransion o ha lower capal sock, so hose who alive a he ime of he swap are hur less by ha lower long-run capal sock han hose who are born laer. The disribuion of welfare changes across generaions in he capal-ax swap case is perhaps he mos ineresing. The oldes generaion alive a he ime of he swap (-50) is made worse off by he swap, hose who are somewha younger (generaions -45 o -25) are made beer off, hose who are younger sill (generaions -20 o 5) made worse off, and hen generaions born more han five years afer he swap are made beer off again. For he oldes generaion, he reason is fairly simple: ha generaion is geing relaively lle income from capal, having already sared spending down s savings. Thus, ha generaion ges relaively lle benef from he lower capal ax rae compared o he higher cos bears as a resul of he carbon ax. Somewha younger generaions (-45 o -25) have more asses a he ime of he swap, and hus benef more from he lower capal ax rae and are made beer off overall. And fuure generaions (10 and laer) benef because he long-run capal sock is higher (he capal ax rae cu encourages saving and increases he capal sock), hus pushing up heir reurns o labor. However, because he capal sock akes ime o adjus, generaions ha are young a he ime of he swap or born jus afer (generaions -20 o 5) don benef as much from he higher capal sock as do laer generaions. And because hey re "*