Party Cues in Elections under Multilevel Governance: Theory and Evidence from US States

Transcription

1 Bey Geys Ja Vermer Party Cues Electos uder Multlevel Goverace: Theory ad Evdece from US States Dscusso Paper SP II August 01 Socal Scece Research Ceter Berl WZB Research Area Markets ad Poltcs Research Professorshp & Project The Future of Fscal Federalsm

2 Wsseschaftszetrum Berl für Sozalforschug ggmbh Rechpetschufer Berl Germay.zb.eu Copyrght remas th the authors. Dscusso papers of the WZB serve to dssemate the research results of ork progress pror to publcato to ecourage the exchage of deas ad academc debate. Icluso of a paper the dscusso paper seres does ot costtute publcato ad should ot lmt publcato ay other veue. The dscusso papers publshed by the WZB represet the ves of the respectve authors ad ot of the sttute as a hole. Afflato of the authors other tha WZB: Bey Geys Norega Busess School BI, WZB ad Vrje Uverstet Brussel Vrje Uverstet Brussel, Departmet of Appled Ecoomcs, 1050 Brussels, Belgum Ja Vermer Cetre of Expertse Ecoomc Modellg & Studes of GDF Suez

3 Abstract Party Cues Electos uder Multlevel Goverace: Theory ad Evdece from US States Bey Geys ad Ja Vermer * I federal coutres, competece for polcy matters s ofte shared betee varous levels of govermet. As oly overall outcomes are observed, ths mght blur accoutablty by decreasg voters ablty to fer formato about the performace of ther leaders. I ths artcle, e aalyse ho party cues.e., poltcas party membershp actg as a cue toards ther characterstcs affect voters complete formato about poltcas a federal settg. We frst of all sho that party cues allo drect ferece regardg poltcas usg observed polcy outcomes, allevatg the accoutablty problem. Emprcal evdece from US presdetal electo results across all 50 US states over the perod provdes support for ths proposto. Yet, hle the avalablty of party cues a federal settg creases the atoal cumbets effort some cases, t may reduce effort partcularly he the regoal cumbet f of a dfferet party. Keyords: Federalsm, accoutablty, multlevel goverace, party cues JEL classfcato: D7, H30, H77 * We are grateful to James Alt, Mare-Laure Breullé, Nelly Exbrayat, Jo Fva, Achm Goerres, Lucy Goodhart, Joshua Holm, Johaes Müster, Susaa Peralta, Jör-Steffe Pschke, Stéphae Rou, Mark Schelker, Crag Volde, Jurge Wllems, three aoymous referees ad partcpats at the d aual EPSA Meetg Berl, Germay, The Ed of Fscal Federalsm Workshop Berl, Germay ad a research semar at GATE-LSE St-Etee, Frace for excellet suggestos. We also apprecated recevg parts of the data from Leo Kahae ad Jm Syder. Bey Geys gratefully ackoledges facal support from FWO-Vlaadere Odysseus grat G Opos expressed are those of the authors ad are ot attrbutable to GDF Suez. The usual caveat apples.

4 INTRODUCTION A cetral feature of may federal govermet structures s that authorty over competeces e.g., uemploymet, educato, crme, or frastructure s shared betee a atoal ad a loer level of govermet. I Stzerlad, for example, publc expedtures for socal elfare take place at both the federal 49.8% of total elfare spedg 004, catoal 8.9% ad mucpal level.3%, ad the same occurs for, amog other polcy areas, educato 8.9%, 5.0% ad 9.1% respectvely ad trasportato 59.3%, 19.9% ad 0.8% respectvely Wdmer ad Zefel, 01. I smlar fasho, publc expedtures the US o, for stace, healthcare splt to 61.6% federal-level ad 38.4% state- ad locallevel spedg 010 NIPA Table , US Bureau of Ecoomc Aalyss. Oe reaso for such complete dvso of tasks across govermet levels s that a coutry s las fal to clearly deleate the dstrbuto of poer. The 10 th Amedmet to the US Costtuto, for example, merely states that all poers ot expressly aarded by the Costtuto to the federal govermet are delegated to the states. A smlar arragemet exsts, amogst others, Belgum, Germay ad Stzerlad. The fuctoal dvso of tasks ad resposbltes across govermet levels federal systems thus freuetly resembles a marble cake rather tha a layer cake Grodzs, 1966; Volde, 005. As a drect coseuece, multlevel goverace structures are ofte argued to decrease the clarty of govermets resposblty for publc outcomes because oly overall outcomes are observed. That s, t s hard to tell hether the atoal or local cumbet s resposble for the observed level of publc performace Aderso, 006, 008, 009; Joas, 009a, b. I ths artcle, e argue that poltcas membershp of poltcal partes provdes a mechasm to allevate ths accoutablty problem uder multlevel goverace structures. We thereby explot that poltcas party membershp provdes mportat cues about ther characterstcs ad lkely behavour oce elected. Ths s supported by a substatal lterature argug that poltcal partes develop reputatos for holdg specfc polcy postos through ther electoral ad parlametary actvtes Aldrch, 1995; Müller, 000; Syder ad Tg, 00, 003, ad susta ths brad ame through party dscple Callaud ad Trole, 00; Castahera ad Crutze, 009. Eve the absece of party dscple, hoever, tra-party coheso s supported by poltcas self-selecto to partes sharg ther prefereces, ad poltcal partes preferece for farly homogeeous caddates Joes ad Hudso, Whle all partes arguably cota good ad bad poltcas from a ethcal pot of ve, the above processes mply that partes poltcas are characterzed by certa polcy prefereces assocated th ther deology. For example, left-g poltcas are more lkely to react to hgh uemploymet th demad-sde poltcs e.g., creased expedtures hle rght-g poltcas geerally prefer supply-sde polces e.g., loer taxes. Voters facg hgh uemploymet ko that oe of these ll costtute the rght recpe at a gve pot tme, but do ot a pror ko hch polcy ad party ll be more successful because the value of certa recpes mght chage over tme depedg o the cotext. The key pot s that he poltcas of the same party share such smlartes terms of deology, polcy ageda ad so o, the polcy prefereces of ay gve caddate become correlated to those of other poltcas of the same party. Ths, e argue, provdes voters th 1 Oe could argue that ths holds maly for the rak-ad-fle of the party. At hgher levels, the party may ell face a trade-off betee tra-party coheso ad the eed for charsmatc leaders able to attract e voters Padro--Muel ad Soberg 011.

5 mportat formato hch, extreme cases, may be the oly formato avalable. We refer to ths as party cues. That s, party cues are defed as the process through hch party labels of caddates crease the formato sources avalable to voters:.e., formato o the performace of oe poltca ca be used to assess aother poltca of the same party. Our cetral argumet s that such party cues ca help voters ther assessmet of caddates based o observed ecoomc outcomes electos uder multlevel sttutoal settgs. For stace, a voter mght ot ko a e guberatoral caddate, but she ca, part, assess ths caddate by lookg at the performace of a cumbet poltca of the same party a eghborg state.e., horzotally or at dfferet levels of govermet.e., vertcally. Whle the mportace of horzotal comparsos as tally hghlghted by Salmo 989 ad formalzed by Besley ad Case 995, Geys ad Vermer 008a llustrate that the formato obtaed from such comparsos s affected by the presece/absece of partsa coectos betee poltcas eghborg states.e., o hether or ot horzotal party cues ca be exploted. I ths paper, e stead focus o vertcal party cues hereby voters a multlevel goverace structure ca judge the atoal cumbet by takg to accout her partsa attachmet ad that of the regoal cumbet. Our theoretcal model frst of all shos that he the atoal ad regoal poltca are from the same party.e., poltcal poer s alged across levels of govermet, regoal publc output remas formatve to voters evaluatg the atoal cumbet eve he she has lttle or o fluece o ths output provded that tra-party correlato poltcas polcy prefereces s postve. The tuto s that, although the postve polcy outcome s attrbuted to the regoal cumbet, t rubs off o the atoal cumbet through poltcas partsa coecto. Secod, he the atoal ad regoal poltca are from dfferet partes.e., poltcal poer s ualged, regoal publc output has a eaker postve effect ad ca have a egatve effect o the atoal cumbet. Ths results from the partsa coecto betee the regoal cumbet ad the atoal opposto caddate, hch forms voters that the caddate felded the federal electo by the party of the regoal cumbet s lkely to be a better choce tha the atoal cumbet. Both predctos dcate that the avalablty of party cues creases the formato avalable a federal settg, ad suggest that regoal publc output ll affect the atoal cumbet s electo result dfferetly alged versus ualged regos. These hypotheses are cofrmed usg state-level data from te US presdetal electos betee 197 ad 008. Eve so, party cues are ot a uuely postve force, ad may carry mportat dosdes. Ideed, hle party cues make that the atoal cumbet geerally but ot alays exerts more effort he the atoal ad regoal cumbet are alged compared to the case he they are ualged, they may cause the atoal cumbet to exert zero effort he the cumbets are ualged. Hece, hle creasg the formato cotet of publc polcy outcomes, the avalablty of party cues may ell reduce poltcas effort uder certa codtos. We retur to the polcy mplcatos of these observatos belo. Ths artcle cotrbutes to the lterature o the costs ad beefts of multlevel goverace structures terms of govermet accoutablty Seabrght, 1996; Myerso, 006; Hatfeld ad Padro--Muel, 01. Seabrght 996 argues that accoutablty may be compromsed a cetralsed system because at least some regos elfare ad votes may become rrelevat to determe the re-electo of the govermet Seabrght, 1996: 61. Myerso 006 argues that poltcas ca prove ther ualfcatos at the local level a federalst structure, hch provdes formato to voters he these poltcas subseuetly compete for publc offce at the atoal level. Hatfeld ad Padro--Muel 01 sho that a 3

6 multlevel govermet structure ca help to solve a commtmet problem at the federal level. Our aalyss cocetrates o the complete formato problem dscussed by Aderso 006, 008, 009, Myerso 006 ad Joas 009a, b rather tha the redudacy effect ad commtmet problems dscussed, respectvely, by Seabrght 996 ad Hatfeld ad Padro--Muel 01. I the ext secto, e develop a smple model detalg our theoretcal argumet. The, e tur to a emprcal test of the model s ma predctos usg state-level data from US presdetal electos. Fally, e dscuss the mplcatos of our aalyss. THEORETICAL MODEL To preset the argumet most clearly ad derve testable hypotheses, e set up a smple career cocers model the sprt of Persso ad Tabell 000 ad Ashorth ad Bueo de Mesuta 006 that cludes a smple federal govermet structure th oe atoal govermet ad several regoal jursdctos. For smplcty, e lmt the umber of poltcal partes to to.e., a cumbet ad a opposto party, though these roles may dffer across jursdctos ad levels of govermet. I each jursdcto, publc output x s determed by the polcy prefereces or polcy ualty of the atoal ad regoal cumbets as ell as ther respectve efforts. The polcy ualty s represeted by the varable, hch s dra from a ubouded ormal dstrbuto th E=0 ad Var= t s crucal that s ot d, see belo. Effort, deoted by e, s costly, ad assumed to be strctly postve e 0 e retur to ths belo. The cost fucto Ce s creasg ad strctly covex th C0=0. We also assume that the atoal cumbet represeted va subscrpt ca exert a dfferet effort each rego, th her total cost of effort gve by C = Ce. Both effort e ad ualty are uobservable to voters. We ca the rte publc output each jursdcto as: x e e r r Here, ad r represet the polcy ualty of the atoal cumbet ad of the regoal cumbet rego, ad e ad e r are ther respectve efforts. 3 The eght of the atoal cumbet determg publc output a partcular jursdcto s represeted by,th 0 1 see also Solé-Ollé ad Sorrbas-Navarro, 008. Ths eght captures the effect of asymmetrc federal desgs here dfferet levels of govermet bear resposblty for a Ths mplctly assumes that the party systems at the federal ad sub-atoal level are completely aalogous. Whle full aalogy rarely occurs realty, our results go through as log as there s suffcet overlap the party systems at varous levels of govermet. Ths holds, for stace, for the US, Germay ad Belgum th both parts of the coutry, though oly to a lesser extet, say, Caada or Spa. 3 Implct such modellg s that oe set of polcy prefereces ad thus oe party could be better tha aother uder gve crcumstaces. Whle ths appears reasoable for the udmesoal polcy settg studed here, t s obvously more restrctve he thkg of mult-dmesoal polcy spaces. Note also that our smple represetato of publc output s ope to varous extesos. For example, oe could assume the presece of cost shocks, hch troduces some ose the relato betee poltcas, e ad publc output e.g., Revell, 00; Geys ad Vermer, 008a. Also, oe ca troduce a fxed tax burde related to publc output, hch allos hgher publc output to be terpreted as hgher govermet effcecy the reverse approach s take, for stace, Besley ad Case, Fally, oe mght troduce multple polcy varables. We absta from these extesos ad aalyse the most basc set-up possble to llustrate the effect of poltcas tra-party smlarty. Hoever, these varous extesos, hle complcatg the formal model, are rrelevat for our ma fdgs. 4

7 gve polcy area to varyg degrees see troducto. 4 Crucally, the varable ca, but eed ot, be correlated across poltcas. Specfcally, e model the dea that poltcas th ay gve party are to some extet terchageable Geys ad Vermer, 008a, 471 by assumg a jot probablty dstrbuto hch the s of poltcas of the same party have a postve correlato 0<<1 ad the s of poltcas from dfferet partes are depedet =0. These correlatos are commo koledge. I other ords, t s the party membershp of poltcas ad ot that of voters that acts as a cue toards poltcas characterstcs. 5,6 The tmg of the to-perod model s as follos. At the begg of perod 1, a federal govermet s establshed, as ell as a regoal govermet each jursdcto. To abstract from complcatos he allog poltcas to ga experece from multple terms offce, e assume these govermets have ot bee offce before ad o hstorcal formato s avalable about them. The, publc output comes about as a fucto of poltcas polcy prefereces ad efforts, ad output s observed by voters. 7 At the ed of perod 1, the cumbets ho are assumed to be vote-maxmzers face a electo hch they are ether re-elected or replaced by a caddate of the opposto party. Belo, e ll focus o federal-level electos ad the behavour of the atoal cumbet. I perod, poltcas aga exert ther optmal efforts ad, together th ther polcy prefereces, ths aga leads to publc output. The the orld eds. As there are o e electos perod, cumbets ll exert zero effort perod, ad voters valug publc output ll vote for the caddate th the hghest expected the frst-perod electos. They ll thereby use the frst perod s publc output to update ther belefs about the atoal cumbet ad, he possble, the atoal opposto caddate. The ex post codtoal expectato of gve the outcomes observed the frst perod.e., E x s the a eghted average of the ex ate mea of assumed to be 0 ad publc output. Based o these updated belefs, voters decde o ther vote see belo. Hece, voters are backard-lookg, usg hstorc performace to decde about ther vote because ths mght relably sgal formato about poltcas Persso ad Tabell, 000. Clearly, ths fal step reles o specfyg the voters decso-rule. Follog Revell 00 ad Solé-Ollé ad Sorrbas-Navarro 008, e assume a stochastc votg rule uder hch the probablty that a voter rego casts a ballot favour of the atoal cumbet S ca be rtte as: 4 I le th the observato that most federal systems are charactersed by symmetrc sub-atoal competeces e.g., Belgum, Germay, Spa, US, e assume that s the same for all jursdctos. Note, hoever, that our results rema vald f e allo for such asymmetres as log as >0 for all jursdctos. 5 Oe could make deped o voters partsa membershp ad assume that voters ko more about the value of th ther o party. Ths, hoever, s ot crtcal to the curret aalyss. 6 I a paper that s coceptually closest to ours, Solé-Ollé ad Sorrbas-Navarro 008 aalyse federal-level grat allocatos across alged ad ualged local-level govermets. Whle our aalyss some sese starts here thers eds as e look at ho polcy outcomes across ualged govermets affect electo results, the bggest theoretcal dfferece betee our respectve papers les the fact that Solé-Ollé ad Sorrbas-Navarro 008 treat partes as moolthc actors. Our theoretcal model refes ths assumpto by troducg the cocept of party cues. 7 We assume that voters oly observe publc output ther o jursdcto. Stll, t s possble to exted the model to the case here voters observe publc output eghborg jursdctos as ell as ther o jursdcto as log as voters caot observe output all jursdctos because the they ould vote the same ay all jursdctos. 5

8 S Pr E x E x 0 o here subscrpts o ad refer to the atoal opposto caddate ad cumbet, respectvely, ad α s a zero-mea, ormally dstrbuted radom term th varace Var, hch s ucorrelated to. 8 To evaluate S, e eed expressos for voters updated belefs about the of the atoal cumbet E x ad opposto caddate E o x. Ths mples aalysg the relato betee the polcy prefereces of the atoal cumbet ad opposto caddate ad publc output the rego. Gve the assumptos above, the polcy ualty of the atoal cumbet ad publc output.e., ad x as ell as the atoal opposto caddate s polcy ualty ad publc output.e., o ad x ll follo a multvarate ormal dstrbuto. Coseuetly, voters updated belefs cocerg the cumbet ca be rtte as see DeGroot, 1970; Thel, 1971; Meyer ad Vckers, 1997: E e f x x e f r 3 I ths expresso f e ad f er are the voters forecasts of the effort exerted perod 1 by the atoal ad regoal cumbet, respectvely, ad reflects the stregth of the partsa cue betee the atoal ad regoal cumbets. Smlarly, the updated belef regardg the opposto caddate s: E e o f o x x e f r 4 here o dcates the stregth of the partsa cue betee the atoal opposto caddate ad the regoal cumbet. Usg expressos 3 ad 4, e ca rerte euato as S f f Pr x e e 0 5 r here o 6 The coeffcet β the stochastc votg rule 5 reflects the stregth of the relato betee publc output x ad the atoal cumbet s vote share S. As dscussed, poltcas choose ther effort to maxmse ther vote share. Because the atoal vote share s a eghted average of the vote share each jursdcto, the atoal cumbet maxmzes her vote share each rego. Hece, effort perod 1 s decded by assessg the expected vote share f f euato 5. Sce the left had sde of euato 5.e., x e 1 er f f follos a ormal dstrbuto th mea e e 1 er er ad varace, the cumbet s expected vote share as a fucto of her effort level euals: 8 Oe mght also allo for a o-zero mea of α represetg, for example, a cumbecy advatage Grossma ad Helpma, 1996; Korad, 00; Mehlum ad Moee, 006. Ths does ot affect our fdgs. 6

9 S e 1 e e f e r e f r 7 Where Φ[.] represets the cumulatve dstrbuto fucto of the stadard ormal dstrbuto. The cumbet ll therefore choose effort e rego such as to maxmze S e Ce. Optmal effort s gve by the frst-order codto: e e e e C e 8 f f r r I eulbrum, the margal ga terms of vote share thus has to exactly compesate the margal cost of effort. Uder ratoal expectatos, voters forecasts of poltcas effort are correct eulbrum, such that optmal effort s characterzed by the follog expresso: 0 C e or C e 9 From euato 9, t ca be sho that effort of the atoal cumbet s hgher he the eght of the federal level regoal output creato s larger. Hoever, effort decreases th the varace of the left-had sde of the votg rule 5. Note also that as ca become egatve, a corer soluto hch effort becomes zero may arse e retur to ths belo. Euatos 5 ad 9 provde the basc gredets for aalysg the effect of party cues o ecoomc votg electos uder multlevel goverace. Hoever, before turg to ths cetral part of the aalyss, t s terestg to pot out that our model predcts more effort from the atoal cumbet a utary cotext compared to a federal cotext, because publc output has a stroger effect o her vote share such a settg ad because she has a bgger mpact o publc output. Proposto 1a The mpact of publc output o the vote share of the atoal cumbet s larger a utary cotext tha a federal cotext. Proposto 1b The atoal cumbet exerts more effort a utary cotext tha a federal cotext. Proof: Remember from euato 5 that β reflects the stregth of the relato betee publc output x ad the atoal cumbet s vote share S. To prove Proposto 1a, t the suffces to state that a utary cotext the atoal cumbet has full resposblty for publc output = 7

10 1, mplyg that β = 1, hereas a federal cotext, authorty s shared such that < 1 ad, coseuetly, β < 1 see euato 6. Usg euato 9, ths mples that the optmal effort exerted by the atoal cumbet a utary cotext e u has to satsfy: 1 C e u 0 Gve strct covexty of the cost fucto, optmal effort of the cumbet a utary settg e u ll thus be hgher tha uder a federal settg e f as log as C e u > C e f, or: 1 1 Euato 1 holds he <1 hch s, by defto, the case a federal cotext ad <1 hch s true f <1, see above. The tuto s that as a cumbet operatg a federal cotext exerts more effort, the mpact of ths o publc output s loer compared to a utary cotext because the effort s eghted by a factor <1 hereas effort couts fully a utary settg sce =1. I addto, the mpact of publc output o her vote share s loer tha a utary cotext sce <1. No, allog for the role of party cues a federal cotext, to cases must be dstgushed. I the frst case, cumbets at the atoal ad regoal level are alged such that =, hch mples, gve that there are oly to partes, that the regoal cumbet s ualged th the atoal opposto caddate o = 0. I a secod case, the atoal ad regoal cumbets are ualged = 0, ad, therefore, the regoal cumbet belogs to the party of the atoal opposto caddate o =. Substtutg ths formato to euato 6, e fd for the case of alged cumbets that: a Smlarly, the case th ualged cumbets leads to: u 3 I the absece of ay party cues = o = 0, e smply have: 4 1 8

11 Euatos ad 3 both coverge to oe he the eght of the atoal cumbet o publc output teds to oe:.e., Lm a Lm 1 ad 1 1 Lm u Lm 1. They coverge to dfferet values, hoever, he 1 1 goes to zero ad the fluece of the atoal cumbet over polcy output dsappears. I effect, euato coverges to hle euato 3 coverges to - he goes to zero:.e., Lm a Lm o o ad Lm u Lm. These o o observatos have several terestg mplcatos. Frstly, t dcates that regoal publc output ca reta a mpact o the atoal cumbet s electo result eve he she has lttle fluece o ths output provded that tra-party correlato s postve.e., > 0. Itutvely, ths result derves from the fact that regoal publc output rubs off o the atoal cumbet through her partsa coecto to the regoal cumbet ho s aarded full credt for x he euals 0 the alged case, or through the partsa coecto of the atoal opposto caddate to the regoal cumbet the ualged case. Hece, eve he there s lttle or o drect evdece upo hch to evaluate the atoal cumbet s, voters a federal system ca stll fer somethg about her va the drect formato cotaed party cues. Proposto Whe party cues exst,.e. tra-party correlato polcy ualty s postve >0, eve he the atoal cumbet has almost o fluece o regoal polcy outcomes 0, publc output ca stll affect the electo result of the atoal cumbet. Secodly, there s some cut-off value of for hch the effect of regoal publc output o the atoal cumbet becomes egatve the ualged case. It ca easly be sho that ths occurs he see belo. Ths meas that f s large eough or 1 small eough, the drect postve mpact o the atoal opposto caddate of the favourable evaluato of the regoal cumbet ca more tha offset the drect postve effect of publc output o the atoal cumbet. Coseuetly, the overall effect of publc output o the latter s electoral result becomes egatve. As a drect corollary, the atoal cumbet such a settg ll have o cetve to provde effort, leadg to a corer soluto here e = 0. 9 Proposto 3a Whe the atoal ad regoal cumbet are ualged, the effect of publc output o the atoal cumbet s electo result becomes egatve he. 1 Proposto 3b Whe the atoal ad regoal cumbet are ualged, t s optmal for the atoal cumbet to exert zero effort he. 1 Proof: u 0 he 1 0. Ths s the case he. Whe 1 0, e have a corer soluto for optmal effort see euato 9, hch becomes zero. u 9 Note that ths follos from our assumpto that effort s o-egatve. Allog for egatve effort or sabotage ; Korad, 000; Che, 003 could lead to stuatos here the atoal cumbet actvely udermes publc output regos th ualged cumbets. 9

12 Note that mples that 1 must be strctly smaller tha ½ sce 0<<1. I other ords, effort of the atoal cumbet oly becomes zero the ualged case he polcy outcomes are predomatly attrbuted to the regoal cumbet ad thereby the atoal opposto caddate ad f tra-party correlato s farly hgh. Thrdly, for all values of loer tha oe, t holds that a > u. The tuto s as follos. I the alged case, publc output that voters attrbute to the regoal cumbet ll also reflect favourably o the atoal cumbet. Ths creases the extet to hch local output traslates to vote share for the atoal cumbet. I the ualged case, hoever, as dscussed above, attrbuto of publc output to the regoal cumbet ll reflect favourably o the atoal opposto caddate through these poltcas partsa coecto. Nevertheless, the level of effort exerted by the atoal cumbet ll ot alays be hgher the alged relatve to the ualged case. The reaso s that the alged case, the traparty correlato ot oly creases the mpact of output o the vote share see above, but also creases the varace of the left-had sde of the votg rule euato 5 hch reduces effort. As a result, effort ca be hgher the ualged case uder certa codtos. Partcularly, he s relatvely large, e ca sho that the varace effect domates, ad that effort ll be smaller the alged compared to the ualged case. Whe s relatvely large, the varace effect loses mportace relatve to the effect of publc output o the vote share, ad effort ll be greater the alged compared to the ualged case. Of course, effort ll also be greater the alged case he, sce that case 1 effort s zero the ualged case. Proposto 4a The effect of regoal publc output o the vote share of the atoal cumbet s greater he the atoal ad regoal cumbet are alged, compared to the case he they are ualged, heever the atoal cumbet s ot fully accoutable for regoal output.e., <1. Proposto 4b The atoal cumbet exerts more effort he the atoal ad regoal cumbet are alged, compared to he they are ualged he or he 1. Proof: See Appedx B. As metoed above, a atoal cumbet has less cetve to exert effort a federal cotext tha a utary cotext. A fal uesto s therefore hether party cues crease the cetve for the atoal cumbet to exert effort a federal frameork. The aser s that, the alged cotext, the mpact of publc output o the vote share creases, hch creases effort exerted by the atoal cumbet. Hoever, the varace of the left-had sde of the votg rule see euato 5 also creases, hch has a egatve effect o effort. Therefore, the alged case, party cues ll crease/decrease effort depedg o the relatve sze of the varace of ad. I the ualged case, party cues decrease the mpact of publc output o the vote share as hgher output helps the atoal opposto caddate. Therefore, the ualged case, party cues alays reduce effort by the atoal cumbet. 10

13 Proposto 5a Party cues crease the mpact of publc output o the atoal cumbet s vote share he the atoal ad regoal cumbet are alged. The reverse occurs he the atoal ad regoal cumbet are ot alged. Proposto 5b Party cues crease the atoal cumbet s effort he the atoal ad regoal cumbet are alged heever. Party cues alays decrease effort he the atoal ad regoal cumbet are ot alged. Proof: See Appedx B. Overall, our model cofrms that a federal cotext, the accoutablty of the federal cumbet ca become sgfcatly mpared. Hoever, t also pots to the mportat role of party labels ad the correlato polcy prefereces or polcy ualty amog poltcas of the same party. We deed fd that the avalablty of party cues mproves accoutablty for publc polcy outcomes a federal settg, sce t allos voters to extract drect formato about poltcas from observed polcy outcomes. Ths mprovemet the formato cotet of publc polcy outcomes geerates the testable hypotheses that a publc polcy outcomes reta a mpact o the atoal cumbet s electo result eve he she has lttle drect fluece o such outcomes ad b the effect of regoal polcy outcomes o the atoal cumbet s vote share depeds o hether the atoal ad regoal cumbets are ualged. We test these predctos the emprcal secto belo. Nevertheless, party cues may at the same tme reduce poltcas effort uder certa codtos, hch e ll retur to more detal the cocludg secto of ths artcle. EMPIRICAL ANALYSIS EMPIRICAL MODEL Our emprcal aalyss explots data o US presdetal electo outcomes across all 50 US states over the perod The US federal structure, ts to-party system th the same partes operatg at both the federal ad state-level ad the dvso of poer betee both partes across states provdes a cotext close accordace th our theoretcal model. Follog the vast lterature o ecoomc votg for reves, see Naestad ad Paldam, 00; Les-Beck ad Stegmaer, 007, our cetral estmato euato takes the follog form th subscrpts ad t referrg to state ad tme respectvely: Votes,t = α + 1 StateEco,t + Cotrols,t + t +,t The depedet varable Votes,t s the share of the to-party-vote obtaed by the cumbet-party caddate state year t. For the 008-electo, t thus represets the share of votes cast favour of Joh McCa from those cast for ether McCa or Obama, as he represeted the party of the prevous cumbet.e., George W. Bush. Stll, all results reported belo rema vald he e defe the depedet varable as the cumbet-party vote total as a share of all votes cast detals upo reuest. The cetral explaatory varables are captured the vector StateEco,t. Frst, e clude state-level per capta persoal come 11

14 groth over the to years pror to the electo, measured 000 dollars, obtaed from the US Bureau of Ecoomc Aalyss. 10 Secod, e troduce total per capta state debt outstadg at the ed of the electo year, lkese measured 000 dollars, obtaed from the US Cesus Bureau. 11 The former s used to measure the effect of ecoomc codtos o electo outcomes Les-Beck ad Stegmaer, 007, hle the latter captures the fscal coservatveess of the US populato Peltzma, 199; Geys ad Vermer, 008b. It s mportat to pot out here that state-level fscal outcomes are more lkely to be predomatly drve by state-level poltcal decsos compared to state-level ecoomc groth here atoal polces may have a stroger fluece; terms of our theoretcal model, ths mples that s smaller for state-level debtedess tha for ecoomc groth. Ths dfferece s mportat for our emprcal aalyss as t mples that state-level fscal outcomes should ot greatly affect federal electo outcomes, ulke state-level ecoomc groth see proposto. As ay such effects to the extet that they do occur operate va cumbets party-poltcal coectos, the aalyss of state-level debt provdes a stroger test-case for our theoretcal predctos. Our vector of cotrol varables cossts of four varables, follog Kahae 009. Frst, e clude a dummy varable eual to 1 f the curret presdet rus for re-electo 0 otherse, measurg the sttg presdets cumbecy advatage Far, Secod, e troduce the level of voter turout, measured as the percetage of the votg age populato that cast a ballot. Thrd, e cotrol for the home-gro effect, hch argues that presdetal electo caddates have a advatage ther state of org Kjar ad Labad, 00; Mxo ad Tyro, 004; Kahae, 009, by cludg to dummy varables. Oe.e., Home IPC s set eual to 1 f a state s the home-state of the cumbet presdetal caddate a gve year 0 otherse, hle the other.e., Home RPC euals 1 for the home-state of the rval party caddate a gve year 0 otherse. Fally, e clude state α ad year t fxed effects throughout all estmatos. Especally the latter are crtcal as they capture tmespecfc effects that are varat across states e.g., the fluece of the federal-level cumbet. Hece, by cludg them the regresso model, e estmate state-level ecoomc effects cotrollg for ay fluece of federal-level ecoomcs. Crucally, e estmate the above regresso euato separately for states here the goveror s alged or ualged terms of partsa attachmet th the US presdet. 1 Ths 10 Whle the to-year perod as chose to match the tme perod betee mdterm ad presdetal electos, our results are robust to usg the groth rate state-level per capta persoal come over 1, 3 or 4 years. The same s true he employg state-level GDP groth rather persoal come groth. 11 Whle outstadg debt s admttedly a stock varable, e prefer ths over the groth of debt for to reasos. Frst, voters are more lkely to obta formato about the stock of debt rather tha ts groth rate through the meda. Secod, poltcas hertg hgh debt become assocated th ths f t s ot dealt th suffcetly uckly much lke hertg a ar; see Mueller, Note also that cludg other fscal varables such as total tax reveues, total o reveues.e., total reveues mus federal-level grats, budget defct as a share of total reveues or terest repaymets all measured 000 dollars ad per capta does ot affect our ma coclusos. As the hgh correlato betee such fscal varables geerates sgfcat multcollearty problems he troducg more tha oe of them, e costra ourselves to publc debt the fal model. The latter varable produces the strogest results terms of R² ad statstcal sgfcace, ad alays retas statstcal sgfcace he troducg ay other fscal varable. 1 I the US poltcal system, a state goveror faced th a legslature cotrolled at least part by the other party may have lmted ablty to mplemet her preferred legslatve ageda Fora, 199; Alesa ad Rosethal, 1995; Schelker, 01. As a alteratve ad more strget measure of partsa ualgmet, e therefore combed formato about both the state goveror ad the state legslature.e., alged states the have to have a goveror as ell as house ad seate majortes from the presdet s party. Although the alged sample becomes farly small ths settg N=66, our results rema ualtatvely uchaged detals upo reuest. We are grateful to a aoymous referee for ths sght. 1

15 separato allos evaluatg hether ecoomc codtos have dfferet effects across both types of states, as predcted by our theoretcal model. Idetfcato of such effects s feasble sce there s substatal varato the partsa algmet of goverors ad presdets across states as ell as th states over tme. For each presdetal electo year our sample, betee ad US states have a goveror that s ualged th the US presdet ad all US states shft ther algmet status at least oce durg the sample perod see Table A1 Appedx A. Hoever, a key detfyg assumpto uderlyg ths approach s that the selecto of states both subsamples s depedet of ay factors that may smultaeously affect presdetal electo results at the state level ad state-level ecoomc varables. From ths perspectve, t s reassurg that the alged ad ualged subsamples are ot sgfcatly dfferet alog a seres of observable dmesos e.g., state persoal come groth, state GDP groth, total debt, turout rates, fscal defct, total tax reveues, total o reveues.e., total reveues mus federal-level grats, federallevel grats, terest repaymets, home-state of the cumbet or opposto caddate, status as ol producer dummy=1 f more tha 1% of US ol producto, populato sze, age composto, term lmt legslato, ad so o detals upo reuest. Eve so, e dscuss several possble threats to our smple detfcato strategy more detal belo. Before turg to the results, e should also ote that, rather tha separate the sample, e could also employ the full sample ad add teractos betee our ecoomc varables ad dcator varables desgatg hether the state goveror s of the same or a dfferet party tha the US presdet. Whle ths methodologcal choce does ot affect our coclusos see belo, e prefer usg separate samples as e rely o a fxed-effects estmator. The resultg devatos-from-state-meas become less meagful he states shft th the sample perod from havg a goveror alged th the US presdet to havg a ualged goveror as occurs freuetly, see table A1. Ths problem does ot occur he relyg o separate samples. EMPIRICAL RESULTS Our basele fdgs are summarzed Table 1. Colums through 3 report results cludg state-level persoal come groth as the ecoomc varable, hle Colums 4 through 6 also clude state debt. I both cases, e report results for the full sample Colums ad 4, as ell as those separated for states here the goveror s of the same Colums ad 5 or a dfferet Colums 3 ad 6 party tha the US presdet. To correct for the varyg sze of the US states, e rely o heteroscedastcty-cosstet stadarderrors all regressos ad egh all regressos by the votg age populato of To start our dscusso th a bref look at the results for the cotrol varables, e see that voter turout ever adds sgfcatly to the model, hle cumbecy status has, le th expectatos, a very large ad postve effect o the cumbet-party vote share. Ths cumbecy effect s cosstetly stroger ualged compared to alged states, hch cofrms earler fdgs that the cumbecy effect s greatest dstrcts here voter partsashp s alged agast the cumbet Asolabehere et al., 000, 18; see also Erkso, 1971; Hrao ad Syder, 009. Ths may reflect the dea that cumbecy ad the esug ame recogto ad ablty to utlze the drect offce-holder beefts Hrao ad Syder, 009, 93 s especally mportat poltcally less sympathetc evromets, or 13 Whle ths eghtg scheme teds to capture the ueual mportace of states the presdetal race see also Kahae, 009, droppg these eghts leaves our fdgs ualtatvely uaffected although sgfcace levels ted to be reduced somehat. 13

16 that cumbets vulerable stuatos must ork especally hard to rema place Asolabehere et al., 000, 19; see also Erkso, Both varables capturg homegro effects also provde sgs le th theoretcal predctos. Icumbet-party caddates obta a better electoral result ther home state though ot sgfcatly so, but do sgfcatly orse the home state of ther oppoet. The latter effect s exclusvely drve by states here the cumbet s of the opposg party compared to the federal-level cumbet party caddate. Turg to the cetral ecoomc varables, Colum llustrates that ecoomc groth sgfcatly beefts the cumbet party caddate. Crucally, hoever, Colums ad 3 llustrate that the postve effect of ecoomc groth s more tha three tmes as strog he the state goveror ad US presdet belog to the same party, compared to the stuato here both cumbets belog to dfferet partes le th proposto 4a. Iterestgly, as sho the ext-to last ro Table 1, the dfferece betee both effects s also statstcally sgfcat at covetoal levels Ch²=3.03; p=0.08. The explaato les the fact that, as dscussed secto, postve ecoomc codtos ualged states ot oly beeft the federal-level cumbet party caddate that state, but also the caddate of the opposto party through hs partsa coecto th the state-level cumbet. Noetheless, ths drect effect o the federal-level opposto caddate s ot strog eough to offset the drect effect o the federal-level cumbet party caddate see proposto 3a. Addg state-level debt to the regressos Colums 4 through 6 ote that e lose oe year of observatos as e lack data o state-level fscal varables 008 does ot affect the ualtatve ature of the above fdgs, although the dfferece betee the coeffcet estmates of state persoal come groth both groups s o o loger statstcally sgfcat Ch²=1.61; p> Moreover, the coeffcet estmate of state-level debt tself s uexpectedly postve ad statstcally sgfcat Colum 4. Ths appears to go agast the commo ve of fscally coservatve voters Peltzma, 199; Geys ad Vermer, 008b. Separatg states here goverors belog to the same or a dfferet party tha the US presdet Colums 5 ad 6, hoever, llustrates that state-level debt has a statstcally sgfcat egatve effect the former ad a statstcally sgfcat postve effect the latter. Ths supports both the fscal coservatveess of the US populato as ell as our theoretcal propostos. Ideed, states here both cumbets.e., at federal ad state level are from the same party, fscally coservatve voters ll take hgh debt levels as a bad sgal cocerg the federal-level cumbet party caddate because she s lked to the fscally rresposble goveror through both poltcas partsa attachmets supportg proposto 4a. Fscally coservatve voters states th a goveror from the atoal opposto party, o the other had, terpret hgh debt levels at least partly as a bad sgal about the federal-level opposto caddate because her partsa lk to the fscally rresposble state-level goveror. I ths case, ths drect egatve effect o the electo prospects of the opposto caddate more tha offsets the drect egatve effect o the federal-level cumbet party caddate such that the latter effectvely gas from hgh levels of state debt supportg proposto 3a. Note also that the effects o state-level debt are statstcally much stroger tha those for ecoomc groth dscussed above, th the dfferece betee both types of states beg statstcally sgfcat ell beyod the 99% 14 Ths reduced statstcal sgfcace s due to the cluso of state-level debt, rather tha to 008 beg dropped from the sample. Ideed, re-estmatg the models Colums through 3 o the sample thout 008 gves very smlar results as those preseted Colums through 3. 14

17 cofdece level Ch²=17.50; p< Ths s especally terestg gve that statelevel fscal outcomes are predomatly drve by state-level poltcal decsos.e., lo terms of our theoretcal model ad should ot ormally affect federal electo outcomes except through local cumbets party-poltcal coectos to the atoal cumbet/opposto caddates. From ths perspectve, the clear-cut effects of state-level debt provde strog evdece favor of our theoretcal propostos. Table 1 about here These results rema vald uder to addtoal robustess checks besdes those reported foototes 9 ad 11 through 14. I the frst of these, e added the vote share of the curret presdetal cumbet-party caddate the prevous electo to the model. Properly specfed, ths s ot a lagged depedet varable such that e ca gore problems assocated th such varables for pael estmatos. Whle the troducto of such lagged electoral success creases the explaatory poer of the model ad mostly dsplays egatve sgs suggestg a cost of rulg ; Frey ad Scheder, 1978; Geys, 010, our cetral fdgs are robust to ths addto. Secod, although federal-level effects are cotaed our year effects, e also expermeted th the cluso of federal-level GDP groth. Ths, as expected, alays has a sgfcat postve effect o the cumbet party caddate s vote share, but, crucally, ts addto does ot affect the fdgs reported above. THREATS TO IDENTIFICATION 16 The aalyss above dsregards to potetal threats to our detfcato strategy. Frst, although there s substatal varato partsa algmet across tme ad space, chages partsa algmet may derve from both chages the US presdecy or the state goverorshp. Hoever, oly the former ca reasoably be treated as exogeous to the statelevel, hle the latter may be edogeous to the local ecoomc stuato especally he the presdet has some effect o local outcomes. We try to accout for ths to ays. I the frst, e replcate our aalyss uder to codtos: restrctg the sample to those cases here there as a o chage goverorshp, ad b a chage goverorshp. The uderlyg detfyg assumpto here s that selecto to the alged versus ualged codto s predomatly related to exogeous chages the presdecy for sample a, hle t s determed by possbly edogeous guberatoral shfts sample b. Hece, f self-selecto affects our results, ths should be largely cleared out of the results o the frst sample, ad cocetrate the latter sample. The fdgs are summarzed Table. Table about here 15 Ths dfferece statstcal sgfcace s also bore out he addg teracto terms cosstg of our ecoomc varables ad dcator varables desgatg hether the state goveror s of the same or a dfferet party tha the US presdet the full-sample regressos. Specfcally, the dfferece the effect of ecoomc groth betee both types of states s statstcally dstgushable from zero p>0.10, but t s statstcally sgfcatly dfferet from zero for total debt p<0.05 detals avalable upo reuest. 16 We are grateful to Jo Fva, Lucy Goodhart ad to aoymous referees for useful dscussos ad suggestos o ths pot. 15

18 Table dcates that our earler results are etrely drve by the uchaged-goveror subsample. Ths s reassurg as t dcates that local fscal codtos matter for presdetal electos oly he local cumbets have presded over the state log eough ad thus ca truly be kept accoutable for these codtos. More mportatly, hoever, t mples that potetal self-selecto s ot drvg our results as the potetal edogeety problem dscussed above s smallest ths sample. The secod ay e try to tackle potetal self-selecto effects from the partal edogeety of guberatoral electos s to restrct the sample to those states here goverors arroly lost/o the prevous electo as arro electos volve some degree of radomess; Lee, 008; Petterso-Ldbom, 008. Ufortuately, movg closer to the 50% electo threshold reduces the umber of observatos to the pot here credble regresso aalyss becomes uorkable. Whle the 3% ad 5% marg e report belo may ot be as close as e ould lke, sample sze costrats prevet explotg eve closer electos see also Schelker, 01. Data o guberatoral electo margs ere retreved from Lst ad Sturm 006. Table 3 about here Table 3 llustrates that, f aythg, our results ths restrcted sample become substatally stroger. Ideed, the dfferece the estmated effects of state-level ecoomc groth ad state-level debt as ell as the statstcal sgfcace of ths dfferece creases compared to our basele results Table 1. Hece, oce aga, e ca coclude that potetal selfselecto of goverors does ot appear to be drvg our results sce these results persst a sample here such self-selecto s arguably mmal due to the radomess heret close electo outcomes. Secod, as metoed, uobserved factors smultaeously affectg presdetal electo results ad state-level ecoomc varables are uproblematc for our aalyss as log as such elemets do ot have a dfferetal effect across the alged ad ualged subsamples. Such dfferetal mpact appears hghly ulkely for geeral shocks such as, for stace, ecoomc recessos or flato ad ol-prce shocks remember that both samples do ot sgfcatly dffer terms of cotag states th substatal ol producto. Hoever, US presdets themselves could have dfferet mpacts o ecoomc codtos dfferet states as they ca dfferetate ther effort across states. Of partcular cocer here s the presdet s fluece over the dstrbuto of federal grats ad the fact that such grats are ofte sgfcatly redrected to alged loer-level govermets Asolabehere et al., 00; Asolabehere ad Syder, 006; Solé-Ollé ad Sorrbas-Navarro, 008. Ths may mply that shfts federal grats could both crease presdetal popularty ad affect ecoomc codtos more strogly alged compared to ualged states. Ths s mportat for our aalyss sce, le th prevous ork, the real groth rate of per capta federal grats s sgfcatly hgher alged compared to ualged states our sample.e., p=0.066 he comparg oe-year groth rates federal grats; p=0.149 he comparg to-year groth rates. 17 To assess hether such dfferetated grat polces are drvg our results, Table 4 17 Note that f US presdets have a cetve to egage strategc grat allocatos especally he presdetal popularty decles e.g., as a vote-buyg strategy, federal grats are edogeous ad causato rus from popularty to grats rather tha from grats to popularty. Ths s of relatvely mor cocer here 16

19 reports fdgs here e drectly cotrol for the real groth rate of per capta federal grats to the state the to-year perod pror to the presdetal electos. Table 4 about here Table 4 clearly llustrates that our results our ot drve by dfferetated grat polces alged ad ualged states. Ideed, our core fdgs for state persoal come groth stroger postve effect alged states ad state debt egatve effect alged states ad postve effect ualged states persst eve he e cotrol drectly for the fact that alged states o average beeft from a stroger crease federal grats the to years pror to the presdetal electos the same holds he usg the oe-year groth federallevel grats, detals upo reuest. Moreover, lookg at the full sample results colum, the groth of federal-level grats appears to have a statstcally sgfcat egatve effect o the cumbet presdet s electo results. If creases federal grats are used as a votebuyg strategy by the cumbet presdet, ths appears to be falg. Oe possble explaato for ths couter-tutve fdg s suggested colums ad 3. Here e observe that the effect of the groth of federal grats s eakly postve alged states, but sgfcatly egatve ualged states, th the dfferece betee both effects beg statstcally sgfcat at covetoal levels Ch =7.88, p<0.01. Wth our theoretcal frameork, exactly such dfferetated effects ould arse he state goverors are able to capture most of the poltcal esteem from the crease federal-level grats e.g., by clamg credt for mproved publc provsos but suppressg that federal grats made them possble. Ideed, as such capture mples that decles, the groth federal grats ll maly have a drect effect o presdetal electo outcomes.e., through the partsa lk th the goveror. Ths drect effect ll be postve alged states ad egatve ualged states. The reaso s that the credt aarded to the goveror ll rub off o the cumbet caddate the presdetal race alged states, but ll beeft the opposto caddate ualged states. CONCLUSION Whe goverg poer s shared betee a atoal- ad loer-level of govermet, accoutablty has bee argued to decle as oly overall publc polcy outcomes are observed Aderso, 006, 008, 009; Joas, 009a, b. I ths paper, e argue that voters ca obta addtoal formato about poltcas through ther party-poltcal attachmets because poltcas party membershp provdes cues about ther characterstcs ad lkely behavour oce elected Joes ad Hudso, 1998; Callaud ad Trole, 00; Syder ad Tg, 00, 003; Geys ad Vermer, 008a. As a result, e sho that the performace of cumbets at sub-atoal levels of govermet ca help voters evaluate the atoal cumbet. Ths ould evdetly allevate the above-metoed complete formato ssues ad eake adverse selecto problems uder multlevel goverace structures. Though ot explctly addressed ths artcle, the same le of argumet lkese suggests that the performace of cumbets at the federal govermet level may assst voters judgg local-level poltcas. sce t ould bas our estmates doard :.e., t ould duce a egatve relato betee popularty ad groth ad a postve oe betee popularty ad debt alged states ad vce versa. 17

20 Evdece from presdetal electo results across all 50 US states over the perod s supportve of our theoretcal predctos. Specfcally, e fd that state-level publc performace flueces presdetal electo outcomes eve for polcy areas here atoallevel caddates ca be expected to have very lttle fluece e.g., state-level debt. Moreover, ad crucally, such effects deped o the presece/absece of a partsa afflato betee the state goveror ad the US presdet. For stace, state-level debt has a statstcally sgfcat egatve effect states here the goveror belogs to the same party as the US presdet as a result of party cues bad local performace here reflects badly o a cumbet presdet from the same party ad a statstcally sgfcat postve effect states ru by goverors assocated th the atoal-level opposto party as bad local performace o hurts the presdetal opposto caddate. Ths s mportat for our theoretcal argumet sce o such dfferece ould be expected he local codtos as such gude voters behavour; that s, a gve ecoomc outcome ould the duce the same respose voters regardless of the party resposble for t. These fdgs are ot drve by potetal self-selecto of states to ether subsample resultg from guberatoral chages, or from the substatal re-drecto of federal-level govermet grats to alged loer-level govermets. I fact, the groth of federal-level govermet grats has a smlar dfferetated effect o presdetal electo outcomes across alged ad ualged states, hch ca be credbly explaed th the cofes of our theoretcal argumet. Take together, poltcas membershp of a poltcal party provdes a mportat mechasm.e., party cues to allevate voters complete formato about atoal poltcas uder multlevel goverace. Noetheless, the avalablty of party cues a federal settg may also have a mportat draback, sce they ca lead to a reducto poltcas effort uder certa codtos. Partcularly, they may cause the atoal cumbet to exert zero effort he the atoal ad regoal cumbets are ualged. Sce a smlar predcto does ot materalze he poltcas across govermet levels are alged, ths suggests a ratoale for formg smlar coalto govermets the federal ad regoal areas Sede, 00, 80. Such alged or cogruet govermets have bee a freuet feature of Belga poltcs sce the drect electo of regoal parlamets the md-1990s, but appear uusual outsde the Belga system Sede, 00. Our aalyss dcates, hoever, that such cogruece across govermet levels ca have clear beefts terms of the cetves of poltcas. Although our aalyss usg the US poltcal system provdes substatal supportve evdece regardg the predctos of our model, ad llustrates that party cues are a mportat dmeso electos a multlevel goverace cotext, more ork s clearly reured. Further research should, for stace, verfy the exstece of smlar party cue effects poltcal cotexts th more tha to partes takg to accout potetal dffcultes posed such settgs by coalto govermets. The Germa sttutoal settg appears a frutful testg-groud for such exteso. Gve the close mappg of the Flemsh ad Walloo party systems at dfferet levels of govermet, the Belga settg provdes smlar beefts though here, of course, addtoal sttutoal complexty should be accouted for. Also, our emprcal aalyss oly establshes that state-level ecoomc outcomes ca fluece presdetal electo results eve for polcy felds here the presdet arguably has lttle to o fluece. Future research should address the reverse predcto that voters mght ell react at the state level to federal-level outcomes over hch state-level poltcas have lttle or o fluece. Prelmary, though suggestve, evdece ths drecto usg Caada data s provded Géleau ad Bélager 005. They deed sho that provcal cumbets Caada are pushed for atoal ecoomc deterorato he the cumbet 18