Banks, Taxes, and Nonbank Competition *

Transcription

1 Commets Welcome Baks, Taxes, ad Nobak Competitio * by George Peacchi Departmet of Fiace Uiversity of Illiois 4041 BIF, Box E. Gregory Drive Champaig, IL Phoe: (217) gpeacc@illiois.edu First Draft: November 2014 This Draft: December 2015 Abstract This paper models baks choice of capital structure ad iterest rates o loas ad deposits whe fiacial services markets are characterized by ecoomies of scope, corporate taxes, ad competitio from obaks (shadow baks). I markets with rich retail ledig opportuities but limited retail savigs, baks may choose high equity capital (low leverage) whe they are ot subject to corporate icome taxes. Whe baks are taxed, capital declies ad retail borrowers bear the tax burde. I the opposite case of markets with few ledig opportuities but pletiful retail savigs, the tax burde falls o depositors ad baks miimize capital. Whe baks face greater obak competitio for retail savigs, equilibrium loa rates icrease, ecouragig etry from obak leders. The model s predictios are cosistet with U.S. baks over the last two ceturies. Recet empirical research o how taxes affect bak behavior also supports the model. * I am grateful for valuable commets provided by participats of the 2015 FDIC/JFSR Bakig Coferece ad participats of semiars at Louisiaa State Uiversity, Uiversità Cattolica, the Uiversity of Oxford, the Bak of Filad, ad the Uiversity of Amsterdam.

2 Baks, Taxes, ad Nobak Competitio I. Itroductio I may coutries, a variety of itermediaries compete to provide similar fiacial services. This paper takes a idustrial orgaizatio approach to aalyze competitio betwee bakig ad obakig istitutios for retail ledig ad savigs/trasactios services. It focuses o how differeces i govermet support, regulatio, ecoomies of scope, ad corporate icome taxes determie the market shares of these differet istitutios. Baks, defied as itermediaries that primarily make loas ad fud them maily with deposits, have experieced chages i regulatio, taxatio, ad obak competitio over the last two ceturies. A cost of fudig-based model is preseted that explais how baks ad obaks evolved durig this period. The aalysis highlights the role that taxes play i determiig baks equilibrium iterest rates o retail loas ad deposits, their capital structures, ad the icetives for obaks to eter the market for fiacial services. Depedig o regulatio ad the market structure where a bak operates, taxes ca affect either the equilibrium retail loa rates charged by the bak or the equilibrium retail deposit rates paid by the bak. Thus, the model derives coditios uder which retail borrowers or retail depositors bear the burde of a bak s corporate taxes. The model also predicts whe, i the absece of taxes, baks will choose high equity capital versus low equity capital. The paper is related to DeAgelo ad Stulz (2015) who examie baks optimal capital structure whe loas ad deposits ca be priced differetly from competitively-priced debt. They argue that whe baks are able to provide liquidity (savigs/trasactio) services to idividuals who lack access to capital markets, high leverage is optimal for baks. Like them, the curret paper assumes baks serve retail loa ad deposit customers. However, while their aalysis assumes retail loa ad deposit rates have exogeously-fixed spreads relative to similar

3 competitively-priced securities, the curret paper s model derives the equilibrium level of these spreads from the structure of loa ad deposit markets. Whe spreads are edogeous, the model shows that, cotrary to DeAgelo ad Stulz (2015), baks may choose low leverage (high equity capital) if they are ot taxed ad retail loa demad exceeds retail savigs. If baks are subject to corporate taxes, borrowers bear some of the burde of corporate taxes ad, at the margi, loas are fuded with competitively priced debt ad equity. I cotrast, if retail savigs is larger tha loa demad, retail deposits fud a bak s security purchases at the margi. The effect of higher corporate taxes is to reduce security purchases with o effect o equilibrium retail loa rates. Rather, retail depositors bear the brut of corporate taxes. The paper the itroduces the possibility of competitio from obaks, also kow as shadow baks. I this respect, the paper relates to Haso et al. (2015) who preset a model where deposit isurace differetiates baks from obaks. While the curret paper also cosiders how the itroductio of a govermet leder of last resort ad deposit isurace chaged bak behavior, it focuses o two differet features that distiguish baks from obaks. First, ulike baks, obaks do ot simultaeously make loas ad issue retail deposits. Rather loas are fuded with competitively-priced debt ad equity by oe set of obak istitutios that ca be iterpreted as securitizatio vehicles that pool loas ad issue asset-backed securities or, alteratively, mutual fuds that ivest i corporate loas. Aother set of obak itermediaries issue savigs/trasactios accouts ad ivest the proceeds i marketable, competitively-priced debt securities rather tha retail loas. The mai example of obak savigs/trasactios providers are moey market mutual fuds. Cosequetly, obaks do ot beefit from a iformatio-related ecoomy of scope i the joit provisio of loas ad deposits. The secod mai differece betwee these obaks ad baks is that the obaks, such as loa securitizatio vehicles, loa mutual fuds, ad moey market fuds, are almost always structured to be exempt from corporate icome taxes. This tax exemptio ca give obaks a 2

4 cost of fudig advatage that, i some circumstaces, may overcome their ecoomy of scope disadvatage. Extedig the model to cosider obak competitio shows that i markets where retail savigs domiate retail ledig, depositors bear the burde of corporate icome taxes which creates a icetive for moey market fuds to eter. But as moey fuds reduce baks retail deposits relative to their retail loas, at the margi bak loas are fuded with equity ad competitively-priced debt ad the corporate tax burde is shifted to retail borrowers. Cosequetly, a icetive for etry by obak loa ivestors is created, ad subsequetly securitizatio reduces baks share of retail loas. Ha, Park, ad Peacchi (2015) derive related results, but their model is of a sigle bak. They take the bak s demad for retail loas ad supply of retail deposits as exogeous. The curret paper derives these demads ad supplies based o a market equilibrium that explicitly models competitio with other baks ad obaks. The pla of the paper is as follows. Sectio II outlies the basic model assumptios ad solves for equilibrium loa ad deposit rates whe oly baks populate a market. The model is used to uderstad the evolutio of U.S. bakig from the early 1800 s to aroud Sectio III itroduces obak competitio ad shows how, i equilibrium, loa rates ad deposit rates differ betwee obaks ad baks, as well as across baks. This sectio also outlies the coditios uder which obaks would fid it profitable to eter ad explais the rise of obak competitio from the 1970s to the preset. Sectio IV surveys recet empirical research related to the model s predictios. Sectio V cocludes. II. A Model with Oly Baks The foudatio of the paper s retail fiacial services market is the Salop (1979) circular city model. Chiappori, Perez-Castrillo, ad Verdier (1995) ad, i particular, Park ad Peacchi (2009) are otable applicatios of this type of model. The curret paper begis by solvig for equilibrium iterest rates ad market shares whe oly baks operate. What defies the relevat 3

5 market for retail fiacial services ca vary across coutries. For example, i the Uited States differet metropolita statistical areas (MSAs) or rural couties ted to costitute idividual retail markets. I other coutries, the etire atio might correspod to a sigle retail market. II.A Assumptios The basic model assumes a sigle period durig which fuds are itermediated. Baks operate i a market that has two types of retail customers: savers ad borrowers. There is a cotiuum of these customers uiformly located aroud a circle of uit circumferece. Let D equal the total volume of retail savers fuds to be ivested, which is the product of the market s desity of savers ad each saver s ivestible fuds, which is assumed fixed. Also let L be the total volume of potetial retail loas, which is equal to the product of the market s desity of borrowers ad each borrower s fixed loa amout. As a startig poit, let there be differet baks located equidistatly aroud the market of uit circumferece, so that the distace betwee each bak is 1/. Baks are assumed to have idetical productio fuctios for fiacial services at margial operatig costs of c D per uit of savigs accout balace ad c L per uit of loas, where c D icludes accout admiistrative costs ad c L combies the costs of screeig a borrower s credit ad moitorig the borrower alog with expected default losses. It is assumed that there is a ecoomy of scope i that issuig retail deposits reduces a bak s cost of screeig ad moitorig a retail borrower. Specifically, if a particular bak, say bak i, issues retail deposits i amout D i, the its per uit loa cost satisfies c L (D i )/ D i 0. This assumptio is based o Black (1975) ad Fama (1985) who argue that iformatio obtaied from deposit trasactios reduces the cost of moitorig borrowers. Empirical evidece i Mester et al. (2007) supports this view that baks are special because issuig deposits reduces their costs of makig loas by reducig default losses. The source of a idividual bak s market power is modeled by retail customers icurrig liear costs of travelig to a particular bak, equal to t D for savers ad t L for borrowers. Retail customers are assumed to obtai sufficiet gross surplus from cosumig fiacial services 4

6 such that they are always willig to icur these trasportatio costs. 1 These costs give a comparative advatage to the bak located closest to a give customer, ad therefore each bak directly competes with the two eighborig baks that are closest to it. What determies a customer s closeess to a give bak eed ot be limited to physical or geographic distace, but could iclude the distace betwee the particular attributes of a idividual bak s fiacial services ad a idividual s product prefereces. Deote the retail loa rate offered by bak i as r L,i, so that r L,i-1 ad r L,i+1 are the loa rates offered by its two eighborig baks. If a borrower is a distace x - [0, 1/] from bak i ad a distace (1/ - x - ) from bak i-1, this borrow will be idifferet betwee obtaiig the loa from these baks if 1 r + tx = r + t x Li, L Li, 1 L (1) Aother borrower located betwee bak i ad bak i+1 ad who is a distace x + [0, 1/] from bak i is idifferet betwee obtaiig the loa from bak i ad bak i+1 if 1 r + tx = r + t x Li, L + Li, + 1 L + (2) For the distaces satisfyig (1) ad (2), bak i s total loa demad is (x - + x + )L L i or r + r L L = ( + ) = + 2 tl Li, 1 Li, + 1 Li x x+ L rli, (3) Recall that c L combies the per uit of loa costs of screeig a borrower s credit, of moitorig the borrower, ad expected default losses. While screeig ad moitorig the borrower is efficiet i reducig expected default losses, for simplicity it is assumed that these activities do ot affect the lowest (miimum) ed-of-period rate of retur o ay bak s portfolio 1 Thus, i equilibrium all customers are served: the total volume of loas made by fiacial services firms equals L ad the total volume of savigs accouts issued equals D. 5

7 of loas, equal to a proportioal loss of r low per uit of loa. 2 Besides makig loas to retail borrowers, baks may ivest i default-free moey market securities that ear the iterest rate r M. Assume, for ow, that retail savers face o default risk whe ivestig their savigs i the baks deposit accouts. Later, the coditios for which deposits are riskless will be give. Deote the iterest rate paid by bak i o a deposit accout as r D,i. The if y - ad y + [0, 1/] are distaces where savers are just idifferet to supplyig fuds to bak i, this bak faces the savigs supply curve, defied as D i, of the form: D ( y y ) D r r + r D D Di, 1 Di, + 1 i = + + = Di, + 2 td (4) Besides retail deposits, baks ca also obtai fuds i a wholesale deposit market which is perfectly competitive ad where the bak is a price taker. Deote by W i the amout of wholesale fuds borrowed by bak i. If there is o risk of default o wholesale deposits, the bak must pay iterest equal to the default-free moey market iterest rate, r M. If W i < 0, the this deotes a situatio where the bak chooses to ivest (led), rather tha borrow, fuds i the wholesale market at the moey market iterest rate. I additio to retail ad wholesale deposits, bak i ca fud its assets by issuig shareholders equity i the amout E i. Besides providig a source of fudig, bak equity ca prevet rus by retail ad wholesale depositors ad may also make these deposits default-free. Bak rus are assumed to be costly because they force the bak to liquidate borrower s projects/loas prior to the ed of the period at a proportioal loss of r ru > r low. Thus, as i Diamod ad Dybvig (1983), rus are assumed to be especially disruptive i that loas liquidated prior to maturity have a value less tha their miimum ed-of-period value. Note that at the start of the period, bak i s balace sheet equatio is 2 For example, for each uit loa made to a retail borrower, 1-r low is the ed-of-period value of the borrower s project if it fails. Credit screeig ad moitorig decrease, but do ot completely elimiate, the probability of project failure, thereby reducig expected default losses but ot the miimum project retur. 6

8 Li = Di + Wi + Ei (5) Iitially, assume that deposits are uisured ad, if savers withdraw their deposit fuds prior to the ed of the period, they are serviced sequetially ad etitled to receive their iitial cotributio but o iterest. I this case, a bak ru equilibrium exists wheever 3 ( 1 r ) L < D + W (6) i ru i i Cosequetly, comparig (5) ad (6), the bak ru equilibrium ca be ruled out wheever E Lr (7) i i ru which requires sufficiet equity to cover the losses o loas due to early liquidatio from rus. 4 For simplicity, it is assumed that this cost of rus is sufficietly high that bak owers choose adequate equity to elimiate the icetive to ru so that iequality (7) holds for bak i. I additio, sice r ru > r low, it is also assumed that ( 1 r ) ( 1 ) ( 1 ) E + L D + r + W + r (8) i i low i D, i i M I other words, sufficiet equity to rule out a bak ru is eough to esure there is o default o the ed-of-period promised paymets to depositors. If coditio (7) holds, the coditio (8) holds whe ( r r ), L > Dr + Wr (9) ru low i i D i i M The iterpretatio of coditio (9) is that the relatively greater loss o loas from rus exceeds the iterest paid o retail ad wholesale deposits. If W i < 0, the the greater loss o loas from rus plus security iterest icome exceeds the iterest paid o retail deposits. Of course L i, D i, ad the retail deposit iterest rate, r D,i, are determied i equilibrium based o other parameters of the model. Oce this is doe, parameter restrictios for which coditio (9) holds ca be stated 3 If W i < 0, so that the bak ivests i moey market securities, it is assumed that these securities ca be liquidated at their iitial values. I other words, moey market securities are ot costly to liquidate. 4 Note that iequality (7) accouts for the possibility that the bak may hold liquid securities (W i < 0) to reduce liquidatio costs. 7

9 precisely. I the iterim, this restricted parameter space is assumed, so that bak i s equity is sufficiet to both rule out rus ad make its deposits default-free. Bak equity is risky due to possible loa defaults, but let r E be the certaity-equivalet rate of retur required by equity ivestors. Like the rate of retur o moey market istrumets (wholesale debt), r M, it is assumed that r E is a fixed competitive rate of retur set i atioal or global fiacial markets. Baks (ad later obaks) take these competitive rates as give. Moreover, r E ad r M are assumed to be rates of retur after persoal icome taxes paid by the margial equity ad debt ivestors. Hece, r E ad r M may be uequal if the margial equity ivestor is more or less heavily taxed at the persoal level tha the margial debt ivestor. 5 Baks are assumed to be subject to corporate icome taxes at the costat margial tax rate of τ. As i most coutries, it is assumed that a bak s iterest expese o debt is deductible prior to calculatig taxable icome. Moreover, it is assumed that r M (1-τ) < r E, a coditio which implies that the total persoal ad corporate tax burde o equity exceeds that of debt. Much empirical evidece, such as Graham (2000), supports this assumptio. Lastly, baks are assumed to choose retail loa ad deposit rates, shareholders equity, ad wholesale deposits/ivestmets to maximize the after tax, certaity-equivalet retur o equity. Ha, Park, ad Peacchi (2015) show how this objective fuctio ca be derived whe loas are default risky but markets are complete. 6 II.B Model Derivatio Give the previous assumptios, bak i s maximizatio problem is r, r, E, W Li, Di, i i ( ( )) ( ),, ( τ ) Max L r c D D r + c Wr 1 Er L i L i i i D i D i M i E (10) 5 For example, if all ivestors were idetical ad had margial persoal icome tax rates for debt ad equity equal to τ D ad τ E, respectively, the i equilibrium r M = r E (1-τ E )/(1-τ D ). 6 They show that objective fuctio (10) holds if deposits have default risk but are paid a fair risk premium (credit spread) or deposits are isured but the bak pays fairly-priced deposit isurace. 8

10 subject to the balace sheet equality (5) ad the capital costrait (7). Let λ be the Lagrage multiplier associated with the capital costrait. The Appedix shows that the first order coditios are rli, 1 + rli, + 1 re tl 2rL, i + cl ( Di ) + + λ( 1 rru ) = t ( ) r + r t c D Di, 1 Di, D L i E rdi, + + cd+ Li + λ = 0 2 Di 1 t r (11) (12) Substitutig for λ i (11) ad (12), oe fids ( ) r + r /1 τ λ = 0 (13) M E 1 r + r r t r r = c D r t 1 t ( ) ( 1 r ) Li, 1 Li, + 1 E L E L, i L i M ru ( ) r 1 r = + r t c c L D + r Di, 1 Di, + 1 D L i Di, D i M 2 2 Di (14) (15) I a symmetric Bertrad-Nash equilibrium where r L,i = r L,i-1 = r L,i+1, r D,i = r D,i-1 = r D,i+1, D i = D/, ad L i = L/, the equilibrium loa ad deposit rates are re re tl rl, i = rm ( 1 rru ) + cl ( Di ) + 1 t 1 t re tl = rm ( 1 rru ) + rru + cl ( Di ) + 1 t D rdi, = rm cd t L cl D ( D ) i i (16) (17) Equatio (16) shows that the profit maximizig loa rate reflects a weighted average of the margial cost of wholesale fudig, r M, ad the tax-adjusted cost of equity fudig, r E /(1-τ), with the weight o equity, r ru, equalig the miimum amout required to avoid rus. I additio, the equilibrium loa rate is higher the greater is the direct costs of moitorig, screeig, ad loa losses, c L, ad loa market power, t L /. Equatio (17) idicates that the bak optimally raises the 9

11 retail deposit rate util it equals the wholesale rate less operatig costs, the margial loss of market power, plus the margial beefit from lower loa costs from greater retail deposits. Sice c L / D i 0, ecoomies of scope i deposit-takig ad loa-makig result i a higher equilibrium deposit rate. Two types of equilibria ca be categorized based o the sig of W i. Sice i equilibrium E i = L i r ru = (L/) r ru, the balace sheet idetity (5) implies (L/)(1-r ru ) = D i + W i = D/ + W i. The implicatio is that W i > 0 whe D < L(1-r ru ). I other words, whe the market s equilibrium amout of deposits is less tha the proportio of loas ot fuded with equity, the bak uses wholesale deposits to fud the remaider. This type of equilibrium where W i = (L/)(1-r ru ) - D/ > 0, i = 1,, characterizes a loa rich, deposit poor retail market. I cotrast, whe W i < 0 which occurs whe D > L(1-r ru ), the retail market ca be characterized as loa poor, deposit rich. I this case, baks ivest the excess of retail deposits over loas ito moey market securities earig the wholesale retur r M. Note that eve if τ = 0 ad r E = r M, so that competitively priced equity ad debt are taxed the same, the loa poor, deposit rich equilibrium where W i < 0 ad the bak ivests the excess of deposits ito securities cotiues to hold. The equity capital costrait E i = L i r ru = (L/)r ru cotiues to bid. However, i the absece of taxes the loa rich, deposit poor equilibrium is o loger be characterized by a bidig equity capital costrait. I this situatio, a bak would be idifferet betwee issuig additioal equity or issuig wholesale deposits i order to fud its profitable loas i excess of retail deposits. This case is a couterexample to the geeral coclusio of DeAgelo ad Stulz (2015) that baks specialess i providig retail deposits implies that they will choose high leverage. The ituitio is that eve though retail depositors lack access to securities payig a competitive retur, with high retail loa demad each bak s equilibrium margial cost of issuig retail deposits is bid up to the competitive retur o wholesale deposits ad equity. At that poit 10

12 the bak stops issuig more retail deposits ad becomes idifferet betwee fudig the excess of loas over retail deposits with wholesale debt or equity sice they both have the same certaityequivalet required rate of retur. Cosequetly, each bak s equity capital costrait does ot bid. Such a loa rich, deposit poor eviromet may have described the geeral situatio of U.S. baks prior to the start of corporate icome taxatio i Figure 1 Pael A shows U.S. commercial baks proportios of total assets held as cash, securities, ad loas over the period 1834 to At least util the time of the Natioal Bakig Acts of 1863 ad 1864, cash ad securities holdigs of baks were ot a large proportio of total assets, rather, the average loa to total asset ratio was over 69 percet. Also, as show i Figure 1 Pael B, prior to these Acts deposits ad privately-issued bak otes were relatively small proportios of baks total fudig eeds. Istead, equity capital averaged over 44 percet of assets. The Natioal Bakig Acts that were passed durig the U.S. Civil War taxed privatelyissued bak otes out of existece ad required that if atioal bak issued atioal bak otes that the otes be over-collateralized by federal govermet bods. As see i Figure 1, this collateral requiremet raised the demad for govermet securities, reducig loas as a proportio of bak assets. Sice fully-collateralized atioal bakotes were ow default-free, the eed for equity capital to deter rus by oteholders was less importat compared to whe baks issued ucollateralized private otes. However, sice bakotes were o loger a source of fudig for loas, deposits came to be a more importat liability. While equity capital was still eeded to prevet depositor rus, it appears that the et effect of ote collateralizatio ad deposit expasio led to relative declie i equity capital. After baks bega payig corporate icome taxes i 1909, the tax-adjusted cost of equity capital became relatively more expesive tha wholesale ad retail deposits. The model predicts that eve with a loa rich, deposit poor market eviromet, baks equity capital costrait (7) would bid, resultig i E i = L i r ru. Shortly thereafter, the early establishmet of a govermet 11

13 leder of last resort, whose missio is to provide fuds to a solvet bak experiecig a ru, would likely decrease r ru ad further reduce baks fudig with equity. Figure 2 Pael A is cosistet with this model predictio. Whe the Federal Reserve Act of 1913 established a leder of last resort to provide a elastic currecy, average equity capital ratios of baks declied from 18.7 % i 1913 to 11.8% i However, the Federal Reserve s leder of last resort fuctio was uable to prevet widespread depositor rus at the start of the Great Depressio of the early 1930s, arguably because the Federal Reserve did ot led freely eough (Friedma ad Schwartz (1963)) ad because bak opacity preveted depositors from gaugig the solvecy of idividual baks. I respose, the Bakig Act of 1933 established the Federal Deposit Isurace Corporatio (FDIC). Federal deposit isurace was successful i stoppig rus ad, as show i Figure 2 Pael A, baks capital ratios also fell cosiderably followig its implemetatio. I additio, the figure shows that with the start of FDIC isurace, there was a flight to quality as retail deposits flowed ito baks. 7 Also as show i Figure 2 Pael B, bak ledig declied substatially durig the Great Depressio of the 1930s, both i absolute terms ad as a proportio of total bak assets. 8 The iflow of deposits ad reductio i loas is reflected i the substatial buildup of cash ad securities i bak portfolios, leadig to a loa poor, deposit rich type of equilibrium. The 1930s bakig regulatio did ot establish miimum umerical capital ratio requiremets but relied o bak supervisors subjective assessmets of whether capital was adequate. 9 This policy of regulatory discretio was icreasigly questioed durig the late 1970s. As show i Figure 2 Pael A, average equity capital to assets ratios had bee fallig sice the early 1960s ad stayed below 6% from 1977 through At the same time, Figure 2 7 Total deposits of all U.S. baks rose by 70.0% from 1933 to Total loas of all U.S. baks fell by 46% from 1929 to See Burhouse, Feid, Frech, ad Ligo (2003) for a summary of capital regulatio ad regulatory developmets durig this period. 12

14 Pael B idicates that baks asset portfolios were shiftig out of cash ad securities ito loas. Together with icreasig loa losses ad bakig idustry weakess, formal umerical capital requiremets were implemeted by federal bakig regulators begiig i Over the ext decade, (primary) equity capital to asset ratio requiremets betwee 5% ad 6%, depedig o the type of bak, were established. I 1991, the U.S. implemeted the 1988 Basel Accord which created more detailed capital requiremets. While recogizig that implicit or explicit regulatory capital requiremets varied sice FDIC isurace was established, evetually baks were subject to some form of capital costrait. Oe ca model a leverage -type equity capital requiremet as: reg E r i Di max ( Wi,0) 1 r + reg (18) where r reg is the miimum required equity capital to asset ratio, so that r reg /(1-r reg ) is the miimum equity-to-debt ratio. Let us re-derive the bak s profit maximizatio problem i (10) but with the costrait (18) replacig the previous o-ru equity capital costrait. It is straightforward to show that for a loa rich, deposit poor market where D < L(1-r reg ) so that W i 0, the equilibrium loa ad deposit rates satisfy: re tl rl, i = ( 1 rreg ) rm + rreg + cl ( Di ) + 1 t D rdi, = rm cd t L cl D ( D ) i i (19) (20) which is the same as (16) ad (17) but with r reg replacig r ru. Similarly, for a loa poor, deposit rich market where D > L(1-r reg ) so that W i < 0, the equilibrium loa ad deposit rates satisfy: ( ) t L rli, = rm + cl Di + (21) 13

15 t r r r = r c + r + L c ( D ) D reg E L i Di, M D M i 1 rreg 1 t Di (22) Here, the loa rich, deposit poor case of (19) ad (20) is very similar to the prior case where equity was costraied to equal a level that would prevet bak rus. The effect of corporate icome taxes is passed o to borrowers but ot depositors. However, with r reg < r ru, equilibrium loa rates are lower sice they reflect a weighted cost of debt ad equity fiace but with a lower weight o the relatively expesive tax-adjusted cost of equity. The equilibrium retail deposit rate is uchaged. As metioed earlier, i the absece of a corporate icome tax ad where r E = r D, the regulatory capital costrait would ot be bidig. The case of a loa poor, deposit rich market differs because previously it was assumed that equity was held to cushio possible losses from loa liquidatios due to rus. Now, a miimum required equity capital to total asset ratio meas that capital must also be issued to fud security ivestmets, ot oly loas. Sice baks hold securities i a loa poor, deposit rich market, equity fudig is based o total assets, ot just loas. The corporate tax wedge implies that depositors are affected because, at the margi, baks do ot bid as aggressively for retail deposits to fud security ivestmets. If, istead, the required capital ratio was based o riskweighted assets ad securities had a zero risk weight, the the loa ad deposit rates i (19) ad (20) would also characterize the loa poor, deposit rich market. These results are summarized i the followig propositio. Propositio I: Uder a leverage costrait where r reg is the miimum required equity capital to asset ratio, retail borrowers bear the burde of higher corporate icome taxes ad higher capital requiremets whe market loa demad is relatively high, L(1-r reg )> D. Coversely, uder this leverage costrait but whe market deposit supply is relatively high, D > L(1-r reg ), retail depositors bear the burde of higher corporate taxes ad capital requiremets. Uder a risk-based 14

16 capital requiremet where securities have a zero risk weight, retail borrowers always bear the burde of higher corporate taxes. For a give umber of baks i the market, our results show that either retail borrowers (i the loa rich, deposit poor case) or retail depositors (i the loa poor, deposit rich case) bear the burde of corporate taxes. Effectively, either retail borrowers or retail depositors bear the higher cost of fudig assets with required outside equity capital. However, bak owers (iside equity) also bear some of the tax burde. Similar to other Salop (1979)-type models, the Appedix shows that each bak ower s profit is ( D ) L t D t L c Di L + D + L i ( 1 t ) (23) where i equilibrium D i = D/. The after tax profits i (23) hold for both the loa rich, deposit poor case ad the loa poor, deposit rich case. Profits are lower i proportio to the tax rate compared to what they would be i the absece of taxes. I a more geeral loger-ru equilibrium where the umber of baks i the market,, is edogeous ad each bak must pay a fixed cost to eter, the a zero profit etry coditio where (23) equals this fixed etry cost would determie. Obviously, the higher the tax rate, τ, the lower is the equilibrium. I this log ru case, oe sees that from the equilibrium retail loa ad deposit rates (19), (20), (21), ad (22) that higher taxes, by reducig, are bore by both retail borrowers ad retail depositors. Higher corporate taxes icrease market cocetratio which raises the moopoly ret eeded to compesate baks for operatig i the market. III. Baks ad Nobak Competitio This sectio itroduces obak savigs accout providers ad obak leders operatig i the same market as baks. Nobak savigs accout providers are modeled as moey market mutual fuds (MMFs) which ivest purely i moey market securities (wholesale 15

17 debt) that pay the default-free iterest rate r M. They issue retail savig accout shares that have a margial operatig cost of c D per uit of savigs accout balace, which is assumed to be the same costat margial operatig cost as a bak deposit. However, because obak savigs providers have a mutual fud structure, they have little leeway i settig the rate of retur paid to retail savers. They must pass through the returs o their assets less operatig expeses to their shareholders. 10 Deote the retur paid by obak savigs provider i as r S,i. Sice its mutual fud structure exempts it from corporate icome taxes, it must pay the retur r S,i = r M - c D. Nobak leders are modeled as loa brokers or baks that follow a origiate-todistribute model of ledig. These itermediaries origiate loas ad immediately sell them to special purpose vehicle (SPV) securitizatios or loa-holdig mutual fuds. They differ from traditioal baks because they do ot access deposit fudig. Rather, all of their fudig is offbalace sheet at wholesale rates, havig a cost of fudig equal to r M for debt ad r E for equity. They also differ from baks because the off-balace sheet vehicles that provide fudig are exempt from payig corporate icome taxes. Examples of such corporate tax-exempt fudig vehicles are SPVs whose assets are mortgages, cosumer loas, or corporate loas ad whose liabilities are debt ad equity traches, variously referred to as mortgage-backed securities (MBS), asset-backed securities (ABS), or collateralized loa obligatios (CLOs). Aother example is mutual fuds kow as prime rate fuds that specialize i holdig corporate sydicated loas. 11 These mutual fuds are corporate tax-exempt ad typically issue oly equity shares. A fial example is a particular type of closed-ed mutual fud, called a busiess developmet compay (BDC). It ivests i small ad medium-sized eterprises (SMEs), ca issue debt up to 50% of its assets, ad yet is corporate tax exempt O this poit, see Gorto ad Peacchi (1993) who ote that MMF advisors are limited to chargig reasoable fees, ad there have bee several istaces where advisors have bee sued by MMF shareholders whe fees were alleged to be excessive. 11 These fuds ca be both ope-ed ad closed-ed mutual fuds. Stadard & Poor s (2014) reports that total assets i these prime rate fuds exceeded $170 billio i March Beltratti, Bock, Jewsikow, ad Nelso (2014) review BDCs ad their stock retur performace. 16

18 As i Ha, Park, ad Peacchi (2015), this paper models obak leders as havig a corporate tax advatage relative to baks. However, they have a disadvatage from ot issuig deposits, so that obak leder i s margial cost of per uit of loas is c L (D i ) = c L (0) c * L. For simplicity, let us take a special case of the previous sectio s assumptio ad assume that for traditioal baks which issue ay positive amout of retail deposits, they have a costat margial loa cost of c L < c * L. I other words, it is assumed that oce bak i has established a retail brach etwork, ay additio issuace of deposits does ot further reduce margial loa operatig costs. Hece, at the margi, c L / D i = 0 ad ay further margial declie i loa costs from issuig additioal retail deposits ca be igored. Suppose that there are a total of fiacial service providers i the market, ad idex these idividual fiacial service providers by i = 1,,. Iitially, let us assume that i both the retail loa market ad i the retail deposit market there are k obak fiacial service providers, so that k are baks. Later, it will be straightforward to show that the model ca be modified to allow a differet umber of obak providers i the loa market versus the umber of obak providers i the deposit market. Cosider, first, the case of k = 1, ad assume this obak is fiacial service provider i = 1, so that the baks are idexed by i = 2,,. This situatio is depicted i Figure 3 for the case of = 8 fiacial service providers, with k = 1 obak ad - k = 7 baks. A symmetric equilibrium is assumed such that baks that are equidistat from the obak set the same loa ad deposit rates. However, baks that differ i their distace from the obak will ot, i geeral, set equivalet loa ad deposit rates. To solve for the market s equilibrium loa ad savigs accout rates, cosider the profit maximizatio problems of the obak ad each bak. As was just metioed, obak providers of savigs accouts are assumed to have o discretio i settig the rates they pay customers. 17

19 They simply pass through their ivestmet returs less expeses so that r D,1 = r S, i = r M - c D is fixed. Nobak leders have more discretio. For example, a obak loa broker chooses its retail loa rate based o the rates of its eighborig baks ad the immediately sell the resultig loas to a off-balace sheet vehicle. The vehicle fuds the loas by issuig some combiatio of wholesale debt ad equity securities. This loa broker pays corporate taxes o its profit or gaio-sale from sellig the loas. But the off-balace sheet vehicle that fuds the loas is exempt from corporate taxes. Specifically, obak leder i = 1 will chose loa rate r L,1 which leads to total loas of L 1 satisfyig equatio (3). It the securitizes these loas i a off-balace sheet vehicle that issues a amout W 1 i wholesale debt ad E 1 = L 1 W 1 i equity. Thus, the obak leder s problem is to maximize the followig after-tax profit: 13, W L,1 1 * ( L L) M E ( ) Max L1 r,1 c Wr 1 Er 1 1 τ r (24) subject to the o-ru costrait (7) ad the balace sheet costrait L 1 = E 1 + W 1. The profit fuctio (24) differs i several ways from the traditioal bak s profit fuctio i (10). First, the obak does ot fud loas o-balace sheet with retail deposits. Secod, because it does ot issue retail deposits but fudig occurs i a off-balace sheet vehicle, c * L > c L due to iefficiet credit screeig ad moitorig which raises expected default losses. Third, ulike the traditioal bak, equity is ot tax-disadvataged relative to wholesale debt due to the off-balace sheet vehicle s corporate tax exemptio. The Appedix shows that whe obak leder 1 s eighborig baks set loa rates equal to r L, 2 = r L,, the obak leder 1 s optimal loa rate, r L,1, satisfies 1 * tl rl,1 = rl,2 cl re rru Mi ( re, rm )( 1 rru ) (25) 13 The term i square brackets ca be iterpreted as the ed-of-period gai-o-sale that is subject to corporate tax. It equals the differece betwee the ed-of-period value of et loa reveue mius the edof-period costs of wholesale debt ad equity fiacig. 18

20 The obak leder chooses a off-balace sheet vehicle that fiaces its loas with all equity whe r E < r M ad fiaces its loas with a proportio r ru of equity ad (1- r ru ) of wholesale debt whe r E > r M. The optimizatio problem for baks i the market is similar to our earlier solutio. Cosider the case of a loa rich, deposit poor market where retail ledig opportuities sigificatly exceed retail savigs, L >> D. The Appedix shows that the profit maximizig loa rate ad savigs rates for bak i are r t r = ( 1 δ ) ( 1 r ) r + r + c + + δ r 1 t E L L, i i, / k reg M reg L i, / k L,1 t r = r c + r c td = rm cd ( 1 δik, / ) D ( 1 δ ) δ ( ) Di, ik, / M D ik, / M D (26) (27) where for the case that is a eve umber 14 δ i k ( 2+ 3) + ( 2 3) k ( 2+ 3) + ( 2 3) i 2 2k i, / k 2 2k (28) Equatios (26) ad (27) show that baks loa ad deposit rates are a weighted average of the equilibrium rates they would have set i the absece of a obak ad the rates set by the obak. The variable δ i,/k < 1, i = 2,,, is the weight o the obak s rate ad reflects the impact of obak competitio. Baks closer to the obak are more affected by its rates: δ i,/k is a decliig fuctio of i over the rage from i = 2 to i = /2 + 1, the mid-poit of the circle, ad it satisfies the symmetry coditios: δ 2,/k =δ,/k, δ 3,/k =δ -1,/k,, δ /2,/k =δ /2+2,/k. Moreover, for a give umber of bak itervals i, i= 1,, away from the obak, a bak s rates are less 14 The Appedix discusses the case of beig a odd umber. 19

21 sesitive to the obak s rate the greater is the total umber of fiacial service providers i the market; that is, δ i,/k / (/k) < 0. I particular, sice a obak savigs accout provider is assumed to pay a competitive rate less accout operatig costs, r M c D, oe sees from (27) that bak deposit rates are always greater tha what would occur i the absece of the obak, with baks closest to the obak payig relatively greater rates. To fully solve for the obak ad baks equilibrium loa rates, first substitute (26) for the case if i = 2 ito (25) ad rearrage to obtai: re tl rl,1 = ( 1 rreg ) rm + rreg + cl + 1 t 2 L δ 2, k / (29) r L 1 rreg rm + rreg + cl rrure + 1 rru Mi re, rm + c L 1 τ is the E * where ( ) ( ) ( ) obak s et fudig ad operatig cost advatage. Thus, if L > 0 because regulatory capital requiremets ad corporate taxes are sufficietly high to offset the obak s operatig cost disadvatage, the obak loa rate will be less tha what baks would have charged i its absece. Substitutig (29) ito (26), bak i s equilibrium loa rate is re t δik, / L L rl, i = ( 1 rreg ) rm + rreg + cl + 1 t 2 δ 2, k / (30) Cosequetly, if L > 0, a bak s equilibrium loa rate is lower the closer it is to the obak. The equilibrium rates o deposits ad loas for this case of a sigle obak ad several baks exteds to the case of k > 1 obaks whe these obaks are symmetrically located aroud the deposit ad loa markets. Figure 4 shows a example of a market with k = 2 obaks with a total of = 8 total fiacial service providers ad, therefore, 6 baks. The Appedix outlies why formulas (27) ad (30) for equilibrium bak deposit ad loa rates ad formula (29) for the obaks equilibrium loa rate cotiue to hold for the case of k > 1. The 20

22 logic is that i this more geeral case, each cluster of oe obak separated by /k - 1 baks face the idetical profit maximizatio problem previously ecoutered i a market with oe obak ad -1 baks. Util ow, competitio betwee obaks ad baks have assumed a loa rich, deposit poor market where L >> D. Let us ow cosider the case of a retail loa poor, deposit rich market where L << D. Recall that whe such a market has o obaks, equatios (21) ad (22) obtai for bak s equilibrium loa ad deposit rates i the absece of obaks. Usig a similar derivatio, it is straightforward to show that for a loa poor, deposit rich market, the obak s equilibrium loa rate is: 15 * tl L rl,1 = rm + cl + (31) 2 δ 2, k / where [ ] r ( 1 r ) Mi (, ) * * L rm + cl rure + ru re rm + c L is the differece betwee the obak ad baks cost of fudig. The Appedix also shows that bak i s equilibrium loa ad deposit rates are * tl ik, / Li, = M + L+ (32) 2 δ2, k / r r c δ L t r r r = r c ( 1 δ ) + r D reg E Di, M D ik, / M 1 rreg 1 t (33) Sice if r M r E, because c * L > c, it is likely that L * < 0. The implicatio is that i this loa poor, L deposit rich market, baks will charge lower rates o retail loas compared to obaks. Hece, oe would ot expect etry by obak leders due to their higher et operatig costs. I equatio (33), baks retail deposit rates reflect their corporate tax disadvatage so that depositors bear the tax burde. However, baks retail deposit rates are higher tha i a loa poor, deposit 15 As before, the aalysis is simplified by assumig that baks have a fixed margial loa operatig cost advatage, c L < c L * whe they issue ay positive amout of retail deposits. This allows us to assume the term L i c L / D i i equatio (22) equals zero. 21

23 rich market with o obaks, equatio (22). The competitive effects of obak savigs accout providers higher rates forces baks to icrease their deposit rates. As a cosequece of the geerality of the derived loa ad deposit rates to the case of k > 1 obaks, a comparative statics exercise ca determie the effects o competitio from holdig the total umber of fiacial service providers fixed but icreasig the relative proportio of obaks. Doig so leads to the followig propositio: Propositio 2: For a market with a fixed umber of fiacial service providers, icreasig the proportio of obak providers raises baks deposit iterest rates. Whe L >> D ad r E = r D so that L > 0, a greater proportio of obak leders lowers baks retail loa rates. Whe L << D ad r E = r D so that L * < 0, a greater proportio of obak leders raises baks retail loa rates. To emphasize some implicatios our results, ote that i a loa poor, deposit rich eviromet (L << D ad L * < 0) baks retail deposits are a iexpesive source of fudig, ad obak leders competitive wholesale fudig puts them at a disadvatage. Cosequetly, obaks would lack a icetive to eter this market. Rather, whe retail deposit rates are low, obak savigs providers are most advataged ad their etry would capture high market share. The implicatios are reversed whe the eviromet is loa rich, deposit poor (L >> D ad L * > 0). Here, retail deposits are a expesive source of fudig sice baks are issuig wholesale deposits at the margi. Sice baks are also required to fud a portio of loas with tax-disadvataged equity capital, obak leders ow have a relative advatage ad would desire to eter. The above derivatio of bak ad obak competitio assumes equal proportios of obaks i both retail loa ad retail deposit markets. Yet, sice obak leders maximizatio problem is separable from that of obak savigs accout providers, the model results geeralize to differet umbers of obaks i the two markets. 22

24 Cosider the case where k = 0 i the retail loa market but k 1 i the retail deposit market. For a loa poor, deposit rich eviromet, equatio (32) cotiues to characterize baks equilibrium loa rate but with δ i,/k = 0, i = 1,,. This is the same loa poor, deposit rich equilibrium loa rate set by baks i the absece of obaks, equatio (21). I a sigificatly rich retail deposit market with k 1 of the savigs accout providers beig obaks, bak deposit rates are give by equatio (33) ad, sice 0 < δ i,/k < 1, strictly higher tha for the case of o obaks. Note that this loa poor, deposit rich equilibrium cotiues to hold as log as D i > L i (1-r reg ) for every bak so that it is optimal for each bak to ivest i securities rather tha issue wholesale deposits. However, if oly obak savigs providers eter without obak leders, L i would remai at L/ for baks but D i for baks would be strictly less tha D/(-k) as obak savigs accout providers pay higher savigs accout rates ad capture proportioally more market share. As a result, baks effective market equilibrium would ted to switch from beig loa poor, deposit rich to loa rich, deposit poor. Such a situatio would describe the U.S. startig i the 1970s ad lastig util 2001 as MMF competitio led to disitermediatio. Figure 5 shows the MMF share of total savigs/trasactio accouts, defied as the ratio of total MMF assets to the sum of total bak deposits plus MMF assets. It also shows the Retail MMF share, defied as the ratio of retail MMF assets to the sum of isured bak deposits plus retail MMF assets. 16 Furthermore, the figure graphs the ratio of baks holdigs of cash plus securities assets to total bak assets. Clearly from the late 1970s util aroud 2001, the MMF share of total savigs/trasactios accouts was risig as baks ivestmets i cash ad securities were decliig. This is cosistet with the hypothesis that greater competitio by MMFs reduced baks retail deposits, turig from what was previously a loa poor, deposit rich market equilibrium 16 Moey market fuds are classified as retail fuds or istitutioal fuds, with the latter caterig to wholesale ivestors. 23

25 to a loa rich, deposit poor equilibrium. Uder this sceario, the model predicts that the corporate tax burde shifts from baks retail depositors to their retail borrowers. If, iitially, there were o obak leders, baks equilibrium retail loa rates would rise from the level give i equatio (21) to the higher value i equatio (19). Importatly, the higher bak retail loa rates satisfyig equatio (19) ow create icetives for obak leders to eter. That is because i this situatio, equilibrium obak leder loa rates are less tha those of baks. Cosequetly, with k of leders beig obaks, obak ad bak retail loa rates satisfy equatios (29) ad (30), respectively. I particular, oe ow expects that i this loa rich, deposit poor market that baks corporate taxatio creates a ledig disadvatage that makes securitizatio via special purpose vehicles relatively more profitable. Evidece cosistet with this shift is give i Figure 6. It shows the MBS ad ABS share, defied as the ratio, i percet, of outstadig MBS ad ABS securities to the sum of outstadig MBS ad ABS securities plus outstadig bak loas. 17 Of course, there may be additioal reasos outside the scope of our model for obaks to hold loas previously held by baks. Baks may have bee discouraged from holdig particular types of loas for risk maagemet reasos. 18 I additio, ot every type of loa experieced the same degree of securitizatio. Baks ted to specialize i relatioship loas made to opaque borrowers that require moitorig of the borrower s cashflow, rather tha moitorig of a borrower s asset value (collateral). A example of such loas is a usecured lie of credit. Relative to obaks, baks have a advatage i makig such loas due to their ability to have a etwork of braches that simultaeously beefit 17 The measure of outstadig bak loas iclude real estate-related loas, agriculture loas, commercial ad idustrial loas, loas to idividuals, ad leases. Excluded are loas to depository istitutios ad other loas. 18 For example, durig the 1970s high ad volatile iterest rates caused losses o the log duratio mortgages held by may baks, precipitatig the savigs ad loa istitutio crisis. Bak regulators ecouraged mortgage securitizatio, i part via govermet sposored eterprises such as Faie Mae ad Freddie Mac. 24

26 retail deposit collectio ad retail loa moitorig. Both retail borrowers ad retail depositors ted to choose baks that are physically close. 19 Usecured loas, particularly those made to opaque small busiesses, may typically be harder to securitize. I cotrast, collateralized loas with well-established uderwritig stadards, such as mortgages ad automobile loas rated by credit scores, may be easier. I terms of our model, the differece betwee obaks ad baks operatig ad credit loss costs, c * L c, are L smaller for these loas relative to relatioship loas. IV. Survey of Prior Evidece Relatig to the Model Our model has aalyzed two mai dimesios of corporate taxes. How they affect baks choice of equity capital ratios ad how corporate taxes may ecourage obak etry, particularly i the form of tax-exempt securitizatio vehicles. This sectio examies empirical evidece o these two issues. IV.A Corporate Taxes ad Baks Choice of Equity Capital The paper s model assumes that baks must meet a miimum equity capital to asset ratio. I practice, regulators ofte require multiple miimum capital ratios, such as Tier 1 ad Tier 2 risk-weighted capital ratios ad a leverage ratio (equity capital to total uweighted asset ratio). I some istaces, baks ca meet a particular capital requiremet usig either subordiated debt or equity capital. Ashcraft (2008) focuses o how the mix of subordiated debt to shareholders equity capital i meetig a capital costrait affects risk-takig by U.S. baks. As a istrumet for baks o-risk-related icetive to use subordiated debt, he uses the corporate icome tax rate paid by commercial baks. A bak s total icome tax rate depeds o where it operates i the 19 Park ad Peacchi (2009) report that the Federal Reserve s 2004 Survey of Cosumer Fiace idicates that the media distace betwee a household ad its bak is 2 miles for checkig accouts ad 3 miles for savigs accouts ad Certificates of Deposit. Based o the 2003 Federal Reserve Survey of Small Busiess Fiaces, for bak loas made i the 1970s, 1980s, 1990s, ad , the media distaces were 2, 2, 4, ad 9 miles, respectively. Oe might coclude that distaces from their baks are becomig less importat for small busiesses but ot for depositors. 25