Bank bailout under TARP in the US

Transcription

1 BSG Workng Paper Seres Provdng access to the latest polcy-relevant research Bank balout under TARP n the US BSG-WP-2016/010 January 2016 Mthul Ncube, Blavatnk School of Government, Unversy of Oxford Copyrght for all BSG Workng Papers remans wh the authors.

2 Bank Balout under TARP n the US + Mthul Ncube* Unversy of Oxford January 2016 Abstract In response to the recent global fnancal crss, the US government launched the Troubled Assets Relef Program (TARP), the largest government balout n US hstory. Ths paper examnes the market responses to the TARP- related events as reflected n stock returns and tal rsk. Our emprcal strategy perms a counterfactual nterpretaton of the data and provdes credble emprcal evdence to answer the queston what would have happened to those banks that dd n fact receve balout funds f they had not receved the balout. It s found that the market responded favorably to the announcement of TARP, whch suggests that the launch of the balout program ndeed helped restore nvestors confdence n the fnancal system. However, the market tended to react negatvely to the recept of TARP balout funds. Banks that receved larger balouts experenced greater stock prce declnes. Ths ndcates that, nstead of creatng a certfcaton effect, the recept of balout funds conveyed an adverse sgnal to the market. Besdes, our emprcal evdence suggests that TARP faled to reduce tal rsk. JEL Classfcaton Codes: G18, G21, G28 Keywords: TARP balout; abnormal returns; tal rsk; counterfactual. *Professor of Publc Polcy, Blavatnk School of Government, & Fellow of St Antony s College, Unversy of Oxford, Oxford OX1 4JJ, UK; emal: mthul.ncube@bsg.ox.ac.uk; +The author thanks the Afrcan Development Bank who supported the work fnancally, and Meng Qng and Patrck Asea for ther nput nto the broader project. Comments from an anonymous referee and Peter Tufano are acknowledged. All errors are mne.

3

4

5 v

6 1. Introducton It s very hard to know the counterfactual. Mervyn Kng, the Governor of the Bank of England, before the Treasury Select Commtee on the Fnancal Crss, March Banks play a val role n an economy; they fnance the assets that lead to economc expanson and job creaton. Yet almost everyone loves to hate the Troubled Asset Relef Program (TARP) whch authorzed the US Department of the Treasury to nject capal nto banks n the mdst of the worst global fnancal shock snce the Great Depresson. TARP, the largest government balout n US hstory, flooded banks (n some cases whether they needed the funds or not) wh money to prevent bank runs, prop up lendng, and convnce consumers and nvestors that the entre bankng sector was safe. Yet, whether TARP restored nvestor confdence by stablzng fnancal markets and helped banks survve the global fnancal meltdown remans an open queston. Whether TARP hndered or even worsened the fnancal crss s the object of serous contenton among polcymakers, academcs and the general publc. 1 The heated debate over TARP has underscored the fact that economc theores can rarely explan wh certanty whether one set of polces are superor to another or are certan to succeed n a gven crcumstance. Indeed, for every example of success wh TARP, opponents are quck to show falures wh them. Crcs argue that condons would have been better off whout TARP mplyng that fnancal markets would have recovered faster and stronger f there were no balouts. They also pont to apparent successes wh an alternatve set of polces. Taylor (2009) concludes that government actons and nterventons caused, prolonged and worsened the fnancal crss. Whle, Blnder and Zand (2010) argue, 1 An mpressve lst of academc economsts wrote to the US Congress protestng TARP. It rapdly became a favore punchng bag of lberals, conservatves, Republcans, Democrats and the publc. A Bloomberg poll n October 2009 asked how TARP had affected the economy. Forty-three percent of respondents sad weakened the economy; 21 percent sad made dfference; only 24 percent sad helped, wh 12 percent unsure one way or another. Commentators n newspapers from the Wall Street Journal to the New York Tmes dsparaged TARP. On October 3, 2008, when TARP became law, one member of Congress even went so far as to say, I don t thnk s too much of a stretch to say ths may be the day Amerca ded. 2 Ths problem s referred to as the fundamental problem of causal nference (Holland, 1986). 3 Recognzng the dffculty of pnpontng causal effects n emprcal socal scence research, a large and nfluental body of work has developed methods for credble causal nference of the effects of a polcy, program or treatment, ncludng Abade et al. (2002); Angrst et al. (1996); Angrst (2004); Heckman (1990); and Heckman (1979). 4 See Bloomberg, October 3, 2008, quotng Representatve John Yarmuth n hs decson to reverse hs vote n favor of the bll the stock market drop on Monday served as a wake-up call to a lot of people. 5 In addon to the nne nstutons dentfed by the US Treasury lst, Wachova, that had sgned a defnve merger agreement wh Wells Fargo, also receved a capal njecton. Of the ten nstutons that receved TARP capal on October 14, 2008 three were nvestment banks at the tme and were not requred to report as bank holdng companes. Hence, comparable fnancal statement data and capal ratos for these three nstutons are unavalable and we exclude 1 them n our analyses requrng fnancal characterstcs. In all tests, we also exclude Wachova due to s merger agreement wh Wells Fargo.

7 If polcymakers had not reacted as aggressvely or as quckly as they dd, the fnancal system mght stll be unsettled, the economy mght stll be shrnkng, and the costs to U.S. taxpayers would have been vastly greater. Evaluaton of publc polces s a central task of economcs yet as one observer noted wryly Judgng the mers of economc polces s a taxng task. It remans mpossble to assess the consequences of a path not taken. TARP passed; we know what occurred. We cannot say wh certanty what would have occurred f TARP had not passed or f the government had pursued another opton. Ths s precsely the objectve of ths paper. We ask the counterfactual queston: what would have happened to those banks that dd n fact receve balout funds f they had not receved the balout? Clearly, ths exact counterfactual s not observable; a sngle bank cannot smultaneously receve and not receve a balout. 2 To resolve ths problem we use propensy score stratfcaton matchng to select a control group of non- balout banks that s closely matched to the group of baled- out banks to artfcally create such twns. We then use the matched groups to estmate the markets response to banks balout decson n terms of stock market returns and systemc tal rsk. A causal effect s defned as the dfference n outcome between a world n whch the bank receves the treatment and a counterfactual world n whch the same bank does not. 3 The treatment s acceptance of TARP balouts. Estmates of effects n ths framework are the effect of treatment on the treated (ETT) rather than the effect of treatment for the entre populaton. In other words, n estmatng the effect of TARP balouts, we are estmatng the effect on those banks n the data who actually accepted balout funds, not the hypothetcal effect of balouts on any bank whch could concevably have receved balout funds. In addon, the estmates from the model are estmates of the average treatment effect, rather than the effect on each ndvdual bank. To the best of our knowledge ths s the frst paper to examne the markets response to TARP- related events as reflected n stock returns and tal rsk. Most promnently, we allow for non- random selecton nto the TARP balout program by usng propensy score matchng methods. Ths strategy perms a counterfactual nterpretaton of the data and provdes the frst credble and robust emprcal evdence that n the absence of balouts there would have been greater tal rsk and more negatve abnormal returns than was the case wh balouts. 2 Ths problem s referred to as the fundamental problem of causal nference (Holland, 1986). 3 Recognzng the dffculty of pnpontng causal effects n emprcal socal scence research, a large and nfluental body of work has developed methods for credble causal nference of the effects of a polcy, program or treatment, ncludng Abade et al. (2002); Angrst et al. (1996); Angrst (2004); Heckman (1990); and Heckman (1979). 2

8 Admtedly, ours s a very narrow queston. We do not attempt to evaluate the costs and benefs of TARP, neher do we evaluate the mpact of TARP on the real economy or the mpact of another set of balout polces. Other papers have attempted to address these questons. For nstance, Verones and Zngales (2010) estmate the costs and benefs of TARP (whch they refer to as Paulson s gft) and show that ths government nterventon ncreased the value of banks fnancal clams by $130 bllon at a taxpayers cost of $21 44 bllon wh a net benef between $ bllon. Talaferro (2009) studes the way banks used new capal under TARP. He fnds that partcpatng banks used roughly thrteen cents of every program dollar, to support new lendng, whle they retaned a consderable porton, about sxty cents of every dollar, to shore up ther capal ratos. Bayazova and Shvdasan (2012) study selecton nto TARP and subsequent stock prce reactons. Ivashna and Scharfsten (2010) demonstrate a relatonshp between cred lne commments and loan growth durng the 2008 crss. A- Sahala et al. (2009) do not fnd strong evdence that eher macroeconomc or fnancal polces had an advantage n calmng nterbank markets durng the global fnancal crss. Duchn and Sosyura (2009, 2014) study the polcal n nfluences on TARP fund dstrbutons. Harvey (2008), Bebchuk (2009), and Coates and Scharfsten (2009) crque the desgn of TARP and dscuss varous neffcences that could be created by the program. Berger and Roman(2014) and L(2013) analyze the dfferences between TARP and non- TARP banks n terms of markups and loan supply. Kotter and Noth(2015) nvestgate the TARP dstorted prce competon among US banks. The emprcal methods used n prevous studes do provde credble emprcal evdence of causaly between TARP balouts and outcomes of polcy nterest. In other words, the methods cannot estmate the average treatment effect on the treated. Neher do any of the prevous methods account for unobservable heterogeney. The decson to receve balouts s not exogenous to banks. Each bank self- selects nto eher the balout or no- balout regme, therefore estmates that do not account for self- selecton may be based. In order to correct for such bas, Heckman selecton or nstrumental varable approaches could be used. Yet, these approaches stll assume that the outcome equatons would dffer only by a constant term between balout and non- balout banks. In realy, dfferences between the two groups may be more systematc, that s, there may be nteractons between balout choce and the other determnants of bank outcomes. Wh the excepton of Duchn and Sosyura (2010, 2014), Berger and Roman(2014), and L(2013), prevous studes do not construct an approprate counterfactual group of banks that do not accept balout funds. Constructng an approprate counterfactual group of banks s essental for studyng the mpact of TARP. For example, suppose frm value s seen to declne after TARP, whout a counterfactual you would not be able to determne whether would have declned even more f frms that actually dd accept balout funds had not accepted balout funds. Even though ths cannot be observed, ther hypothetcal behavor can be proxed by the behavor of a sample of other banks that dd not accept balout funds. 3

9 Attempts to use propensy score matchng alone- whout some form of structural models futle because cannot capture the polcy mpact of most nterest the average effect of the treatment on the treated (ATT). Propensy score matchng can deal wh structural dfferences between balout and non- balout banks, but only to the extent that these dfferences are based on observables. When there are unobserved factors that smultaneously nfluence banks balout decson and say the fnancal health of the banks, such as manageral sklls, ably, or motvaton, then propensy score matchng may stll result n based estmates. In ths paper use make use of varous methodologes dependng on the queston. We use a smple event- study methodology based on a three- factor Fama- French model, to analyse the mpact of TARP on stock returns, and construct a counterfactual of untreated banks whch are then compared wh treated banks. The untreated banks that constute the counterfactual, are constructed usng a propensy score matchng method. We also make use of the buy- and- hold nvestment approach to analyzng returns and how they are mpacted by TARP for both treated and untreated banks. The mpact of TARP on tal rsk s analysed usng value at rsk(var) and changes n VAR for treated and untreated banks. The rest of the paper proceeds as follows. In Secton 2, we dscuss the background and events leadng up to TARP. Secton 3 descrbes the dataset and the characterstcs of banks n our sample. Secton 4 presents emprcal evdence on the mpact of TARP balout events on stock returns. Secton 5 nvestgates the valuaton effect of balout sze usng the event- study methodology. In Secton 6, we look at the buy- and- hold returns of balout and non- balout banks over the TARP capal njecton perod. Secton 7 examnes the mpact of TARP balout on systemc tal rsk as measured by value- at- rsk. Secton 8 concludes. 4

10 2. Background to TARP As part of the government s measures n response to the global fnancal crss, the Troubled Assets Relef Program (TARP) was the largest government balout n US hstory (a bref hstory of US government balouts s summarzed n Appendx 1). The geness of TARP les n the days followng the collapse of Lehman Brothers and the rescue of AIG n md- September In the aftermath of these events, fundng costs for fnancal nstutons escalated sharply due to wdespread fear of a domno effect of collapse among fnancal nstutons that were unable to fund oblgatons and concerns about counterparty rsk. On September 20, 2008, Treasury Secretary Henry Paulson and Federal Reserve Charman Ben Bernanke sent a fnancal rescue plan to Congress requestng approval to stablze the fnancal system by purchasng troubled assets, prmarly those related to mortgage- backed secures (MBS), from banks and other fnancal nstutons. Though ths nal plan was rejected by Congress, a modfed verson was approved on October 3, Presdent George W. Bush sgned nto law the Emergency Economc Stablzaton Act of 2008 (EESA) whch authorzed spendng of up to $700 bllon to purchase or nsure troubled assets, n an attempt to unlock cred markets and restore confdence n the bankng system. 4 Accordng to EESA, the term troubled assets was defned as: () Resdental or commercal mortgages and any secures, oblgatons, or other nstruments that are based on or related to such mortgages, that n each case was orgnated or ssued on or before March 14, 2008, the purchase of whch the Secretary determnes promotes fnancal market stably; and () Any other fnancal nstrument that the Secretary, after consultaton wh the Charman of the Board of Governors of the Federal Reserve System, determnes the purchase of whch s necessary to promote fnancal market stably, but only upon transmtal of such determnaton, n wrng, to the approprate commtees of Congress. On October 13, 2008 the Treasury announced that would nvest drectly n the equy of a broad range of fnancal nstutons and that these equy njectons would be targeted at healthy frms. On October 14, 2008 the US Treasury unveled the detals of s Capal Purchase Program (CPP) whch allocated $250 bllon towards purchases of preferred stock and equy warrant of US fnancal nstutons. The nne largest fnancal nstutons, ncludng Bank of Amerca, Bank of New York Mellon, Cgroup, Goldman Sachs, JP Morgan, Merrll Lynch, Morgan Stanley, State Street, and Wells Fargo, were dentfed as the nal recpents of an aggregate nfuson of $125 bllon. 5 Other banks were also allowed to apply for the preferred stock nvestment by the Treasury untl November 14, Capal njecton through the purchase of preferred stock would qualfy as Ter 1 capal but not dlute the votng power of the exstng common shareholders, and thus was expected to be attractve to banks. On the same day, a program to offer government guarantees on new bank debt ssues was unveled, and the celng on the Federal 4 See Bloomberg, October 3, 2008, quotng Representatve John Yarmuth n hs decson to reverse hs vote n favor of the bll the stock market drop on Monday served as a wake-up call to a lot of people. 5 In addon to the nne nstutons dentfed by the US Treasury lst, Wachova, that had sgned a defnve merger agreement wh Wells Fargo, also receved a capal njecton. Of the ten nstutons that receved TARP capal on October 14, 2008 three were nvestment banks at the tme and were not requred to report as bank holdng companes. Hence, comparable fnancal statement data and capal ratos for these three nstutons are unavalable and we exclude them n our analyses requrng fnancal characterstcs. In all tests, we also exclude Wachova due to s merger agreement wh Wells Fargo. 5

11 Depos Insurance Corporaton (FDIC) guarantee of non- nterest bearng transacton accounts at banks was also ncreased at ths tme. The new bank debt guarantee natve was fnalzed on November 21, 2008 as the Temporary Lqudy Guarantee Program (TLGP) whch guaranteed senor unsecured bank debt, whn prescrbed lms, ssued between October 14, 2008 and June 30, Under CPP, the US Treasury would purchase non- votng senor preferred stock of qualfyng fnancal nstutons (QFIs), and banks could apply for ths njecton n amounts rangng from 1 percent to 3 percent of ther rsk weghted assets (RWA). In addon to senor preferred stock, the US Treasury would receve warrants wh a ten year lfe to purchase common stock of qualfyng banks for an amount equal to 15 percent of the preferred equy nfuson. The dvdend on the preferred stock was set at 5 percent, but would rse to 9 percent after three years. The fnancal terms of CPP capal were vewed to be very attractve for banks and substantally below the fundng costs obtanable n publc capal markets for most banks. However, CPP nfusons forbade dvdend ncreases on the common shares untl the preferred shares were repad fully and also set lms on executve compensaton whereby senor executve benef plans, severance, and golden parachute agreements had to be termnated or modfed. Followng CPP and TLGP, TARP evolved to nclude several other components ncludng the Publc- Prvate Investment Program (PPIP) to acqure troubled loans and toxc assets from fnancal nstutons and the Term Asset- Backed Secures Lendng Facly (TALF) to support the ssuance of asset- backed secures (ABS). Our analyss focuses on the CPP program because remans the cornerstone of TARP and because targets specfc fnancal nstutons, allowng us to study the characterstcs of the banks supported by the capal njectons. Henceforth, we refer to capal njectons under the CPP program as TARP nfusons. Snce the nal preferred stock nvestment of $125 bllon nto the nne fnancal nstutons on October 14, 2008, TARP capal nfusons have been made nto a large number of other fnancal nstutons. To partcpate n the program, elgble fnancal nstutons had to subm a short applcaton to ther prmary federal bankng regulator, namely the Federal Reserve, the Federal Depos Insurance Corporaton (FDIC), the Offce of the Comptroller of the Currency (OCC), or the Offce of Thrft Supervson (OTS). After recevng the applcaton, the regulators assessed the fnancal condon of the applcant based on the CAMELS ratng system. If the nal revew by the bankng regulator was successful, the applcaton was forwarded to the Treasury s nvestment commtee and then the assstant secretary for fnancal stably who made the fnal decson about the nvestment. As of July 30, 2009, over 2,700 applcatons had been submted, of whch 1,300 were sent to the Treasury, and 660 were approved for balout funds. Wh the passage of the TARP legslaton, banks across the country faced a dffcult decson: should they accept government ad that could help keep them solvent but also open them to crcsm of beng baled out? The banks choce to apply for TARP funds thus was also a functon of ther own nternal delberatons as to expected costs and benefs, manageral tastes, preferences and prvate nformaton. Eventually, 758 banks took the deal and accepted funds through TARP. But others eager to protect ther mages or unwllng to accept the program s burdens opted aganst takng the assstance. 6

12 3. Sample and Data 3.1 Sample Characterstcs To construct our man (unversal) sample, we start wh data avalable at bank holdng company level from the Bank Holdng Company Database provded by Federal Reserve Bank of Chcago. The dataset ncludes quarterly fnancal data on a consoldated bass for all domestc bank holdng companes (BHCs) wh total assets of $500 mllon or more. The consoldated bank holdng company fnancal data are desrable because the Troubled Assets Relef Program (TARP) s made at the level of holdng companes. From the unverse we obtan two sub- samples. The frst sub- sample s BHCs that accepted TARP balout funds (balout banks). The lst of balout banks sobtaned va ProPublca s TARP database. The sub- sample of balout banks s used to conduct our basc event study. The second sub- sample s matched banks that dd not accept TARP balout funds but are smlar to the balout recpents accordng to propensy score matchng methods (counterfactuals). More specfcally, for the balout sub- sample, we obtan data on TARP partcpant BHC from ProPublca s TARP database, whch can be found at The database tracks where taxpayer money has gone n the ongong balout of the fnancal system. By December 30, 2011, 926 nstutons had receved balout funds of $700 bllon (there s a separate balout of Fanne Mae and Fredde Mac). We restrct our analyss to bank holdng companes because TARP balouts are made at the level of holdng companes. Snce we retreve fnancal reportng data from Consoldated Fnancal Statements for Bank Holdng Companes- FR Y- C (Call Report), we lm our sample to bank holdng companes wh total consoldated assets above $500 mllon.in addon, we analyze publcly traded banks because our event study employs stock market data. We lm our balout sub- sample for the event study to banks that partcpated n TARP and had ordnary shares lsted on NYSE, AMEX, or NASDAQ. Panel A of Table 1 shows that CPP capal of $640 bllon was provded to 926 frms, ncludng 758 bank holdng companes who receved $236 balout funds. Of the bank holdng companes, 247 are publcly traded on the New York Stock Exchange (NYSE), Amercan Stock Exchange (AMEX), or NASDAQ Stock Market (NASDAQ). For the non- balout sub- sample, we start wh 977 bank holdng companes wh consoldated assets of $500 mllon or more as of September 30, 2008, and therefore have consoldated fnancal nformaton avalable from Bank Holdng Company Data before the announcement of TARP. After removng the bank holdng companes that announced ther partcpaton n TARP, we end up wh our non- balout sub- sample. Table 1 presents the selecton process for our balout and non- balout sub- samples. Table 1: Sample Selecton Selecton Crera Balout Amount Number Panel A: Balout Banks Frms that receved balout funds under TARP $640 bllon 926 Retan bank holdng companes only $236 bllon 758 7

13 Retan bank holdng companes that have ordnary shares lsted on NYSE, AMEX, or NASDAQ Retan bank holdng companes wh total consoldated assets of $500 mllon or more as of September 30, 2008 $227 bllon 247 $216 bllon 187 Panel B: Non- Balout Banks Number of bank holdng companes wh total consoldated assets of $500 mllon or more as of September 30, 2008 Retan bank holdng companes that have ordnary shares lsted on NYSE, AMEX, or NASDAQ Retan bank holdng companes that dd not recevetarp balout funds by September 30, 2011 N.A. 976 N.A. 318 N.A. 131 Notes: Reported are the sample selecton processes for the study. Panel A descrbes the constructon of the sub- sample of bank holdng companes that receved TARP balout funds by September 30, 2011 (.e. balout banks or treated group). Panel B descrbes the constructon of the sub- sample of bank holdng companes that dd not receve TARP balout funds by September 30, 2011 (.e. non- balout banks or control group). 3.2 Prelmnary Analyss We classfy the banks followng FDIC and Federal Reserve Gudelnes nto one of four sze groups based on perod- end book value of assets: 1. Greater than $10 bllon 2. Between $3 bllon and $10 bllon 3. Between $1 bllon and $3 bllon 4. Less than $1 bllon Asset szes of the BHCs as well as all accountng data are avalable from Bank Holdng Company Data from Federal Reserve Bank of Chcago. All domestc bank holdng companes wh total assets of $500 mllon or more are requred to fle FRY- 9C on a consoldated bass. For the bank holdng companes wh data avalable we constructed a number of demographcs, such as bank sze and age, as well as fnancal varables, such as CAMELS. The man varables used n our analyss are lsted n Table 2 along wh ther detaled defnon and data sources. Table 2: Defnon of Man Varables and Source of Data Varable Defnon Source Balout amount (BA) Balout rato (BR) Amount of TARP funds receved by a balout bank ($bllons) Rato of the amount of TARP funds receved by a balout bank to the bank s Ter 1 capal (%) Eye on the Balout Eye on the Balout; BHC Data (BHCK 8

14 Capal adequacy (CA) Asset qualy (AQ) Management qualy (MQ) Earnngs (EAR) Lqudy (LIQ) Sensvy (SEN) Bank sze (SZ) Bank age (AGE) Stock return (R) Index return (MKT) Rato of Ter 1 capal to total rsk- weghted assets (%) Rato of noncurrent loans and leases (90 days or more past due or n nonaccrual status) to total loans and leases (%) Rato of annualzed total non- nterest expense to annualzed net operatng ncome (%, net operatng ncome s measured as the sum of net nterest ncome and non- nterest ncome) Rato of annualzed net ncome to average total assets (%) Rato of cash and balances due from deposory nstutons to deposs (%) Rato of the absolute dfference between earnng assets that are reprcable whn one year and nterest- bearng depos lables that are reprcable whn one year to total assets (% as a measure of sensvy to nterest rate rsk) Natural log of the book value of BHC's total assets (n thousands of US dollar) at quarter- end Number of years snce the enty s general ledger was opened for the frst tme and/or the date on whch the enty became actve (years) Daly percentage change n stock prce (%) Daly return of the CRSP value- weghted ndex of all NYSE, AMEX, and NASDAQ frms (%) 8274) BHC Data (BHCK 8274 A223) BHC Data (BHCK B529) BHC Data (BHCK ) BHC Data (BHCK ) BHC Data (BHCK BHDM BHFN ) BHC Data (BHCK ) BHC Data (BHCK 2170) BHC Data (RSSD 9950) CRSP US Stock CRSP US Stock Notes: Reported are the man varables used n the study along wh ther defnons and the sources of data. The balout data s obtaned from Eye on the Balout database provded by ProPublca ( Accountng nformaton at bank holdng company level s collected from Bank Holdng Company Database provded by Federal Reserve Bank of Chcago ( Income and expense attrbuted to each quarter s annualzed and compared to average asset or lably balances for the correspondng quarter. Stock return data s retreved from CRSP US Stock Database. Table 3 below reports the summary statstcs of the man varables used n the study. Reported are the mean, 25th percentle, medan, 75th percentle, and standard devaton of each varable. The statstcs for the fnancal varables reported n Table 3 are computed based on the Bank Holdng Company Data 9

15 released at the end of September 2008, the latest fnancal nformaton avalable before the announcement of TARP on October 14, The summary statstcs for the four sze groups are also reported n Appendx 3. Table 3: Summary Statstcs of the Man Varables for Balout Banks Varable Mean 25 th Percentle Medan 75 th Percentle Standard Devaton No. of Obs. BA BR 29.34% 24.54% % CA % 10.73% AQ % 1.63% 2.29% MQ % 65.45% 73.33% EAR 0.02% 0.06% 0.45% 0.76% LIQ 4.20% 2.42% % SEN 15.42% 6.13% 13.56% 23.42% BZ AGE Notes: The table reports the summary statstcs of the man varables used n the study. Reported are the mean, 25th percentle, medan, 75th percentle, and standard devaton of each varable lsted n Table II. The statstcs for the fnancal varables are computed based on the Bank Holdng Company Data released at the end of September 2008, the latest fnancal nformaton avalable before the announcement of TARP on October 14, BA represents balout amount (n bllons $), BR balout rato, CA capal adequacy, AQ asset qualy, MQ management qualy, EAR earnngs, LIQ lqudy, SEN sensvy, SZ bank sze (natural log of total assets n thousands $), and AGE bank age (number of years). The detaled defnon and data source are avalable n Table 2. Table 4 presents the par- wse correlaton among the man varables. Agan, the statstcs for the fnancal varables are computed based on the latest fnancal nformaton avalable before the announcement of TARP. Table 4: Correlaton Coeffcent Matrx of Man Varables for Balout Banks Varable BA BR CA AQ MQ EAR LIQ SEN BZ BR CA 0.29*** 0.25*** 1.00 AQ 0.13* MQ ** 0.23***

16 EAR *** 0.48*** 1.00 LIQ 0.21*** SEN 0.19** *** BZ 0.74*** *** *** 0.32*** 1.00 AGE 0.57*** *** *** 0.13* 0.58*** Notes: The matrx reports the correlaton coeffcents between each par of the man varables used n the study. The fnancal varables used to estmate the par- wse correlaton coeffcents are computed based on the Bank Holdng Company Data released at the end of September 2008, the latest fnancal nformaton avalable before the announcement of TARP on October 14, BA represents balout amount (n bllons $), BR balout rato, CA capal adequacy, AQ asset qualy, MQ management qualy, EAR earnngs, LIQ lqudy, SEN sensvy, SZ bank sze (natural log of total assets n thousands $), and AGE bank age (number of years). The detaled defnon and data source are avalable n Table 2. *, **, and *** represent statstcal sgnfcance at the 10%, 5%, and level, respectvely. Apart from the bank- level varables, we also collected tme seres of the TED spread, the LIBOR- OIS spread, the VIX ndex, and the Nose measure. The four tme seres are plotted n the four panels n Fgure 1, wh vertcal reference lnes ndcatng the date that Lehman fled for bankruptcy (15 September 2008) and the date that TARP was announced (14 October 2008), respectvely. The tme seres data s obtaned from Bloomberg. 6 Fgure 1: TED spread, LIBOR- OIS spread, VIX ndex and Nose Measure Panel A: TED Spread Panel B: LIBOR- OIS Spread 6 The TED spread s the dfference between the nterest rates for the three-month Eurodollars contract, as represented by the London Interbank Offered Rate (LIBOR), and the three-month US government debt, as represented by the three-month Treasury bll nterest rate. The sze of the spread s denomnated n bass ponts. The TED spread s an ndcator of perceved cred rsk n the general economy. Ths s because the Treasury blls are consdered rsk-free whle LIBOR reflects the cred rsk of lendng to commercal banks. When the TED spread ncreases that s a sgn that lenders beleve the rsk of default on nterbank loans,.e. counter-party rsk, s ncreasng. LIBOR-OIS spread s the dfference between LIBOR and the overnght ndexed swap rate. The spread between the two rates s consdered as a measure of health of the bankng system. Three-month LIBOR s the generally floatng rate of fnancng, whch fluctuates dependng on how rsky a lendng bank feels a borrowng bank s. The OIS s a swap derved from the overnght rate, whch s generally fxed rate of nterest over the same perod. In the US, the spread s based on the LIBOR Eurodollar rate and the Federal Reserve s Fed Funds rate. LIBOR s rsky n the sense that the lendng bank loans cash to the borrowng bank, and the OIS s consdered stable as both counter-partes only swap the floatng rate of nterest for the fxed rate of nterest. The spread between the two s therefore a measure of how lkely s that borrowng banks wll default. Ths reflects rsk premums n contrast to lqudy premums. The VIX s the tcker symbol for the Chcago Board Optons Exchange Market Volatly Index, a popular measure of the mpled volatly of S&P 500 ndex optons. Often referred to as the fear ndex, represents one measure of the market s expectaton of stock market volatly over the next 30 day perod. 11

17 Panel C: VIX Index Panel D: Nose Measure 2007/01/ /07/ /01/ /07/01 Lehman TARP 2009/01/ /07/ /01/ /01/ /07/ /01/ /07/01 Lehman TARP 2009/01/ /07/ /01/ /01/ /07/ /01/ /07/01 Lehman TARP 2009/01/ /07/ /01/ /01/ /07/ /01/ /07/01 Lehman TARP 2009/01/ /07/ /01/ Notes: The fgure plots the tme seres of TED spread (dfference between 3- month US LIBOR and US Treasury Bll), LIBOR- OIS spread (dfference between the 3- month US LIBOR and the overnght SWAP rate), the VIX ndex from the Chcago Board of Opton Exchange, and nose measure over the perod of 2007 to The vertcal reference lnes ndcate the events of Lehman's bankruptcy and the announcement of TARP respectvely. 12

18 4. The Impact of TARP Balouts on Stock Returns 4.1 Emprcal Strategy We conduct a standard event study to gauge the mpact of TARP on stock returns. We are nterested n two event dates. The frst date s October 14, 2008 (the day of the announcement of TARP), and ths date s the same for all banks n our sample. The second set of dates s the date that each bank n our sample actually receved the TARP funds (the day of recept), and each bank has a unque date. We estmate bank returns usng the followng two models. The frst model s Markowz market model whch s specfed as follows R = α + β MKT + ε t [ t 0, t ] (1) M t, 1 where t0 and t1 denote the begnnng and end of the tme wndow where parameters are estmated (.e. the estmaton wndow), R s the daly stock market return of bank between tradng dates t 1 and t and MKT s defned as the daly return of the CRSP value- weghted ndex of all NYSE, AMEX and NASDAQ frms. For the second model nclude the followng Fama- French three factors model R RF = α + β MKT RF ) + β SMB + β HML + ε t M ( (2) t t SMB t HML t where SMB s a sze factor (small mnus bg) and HML s a value factor (hgh mnus low). We estmate the parameters of equaton (1) and (1) wh OLS usng a wndow startng from September 17, 2007 to September 17, 2008 (.e. the normal perod), and use the estmated parameters to predct returns n wndows of 2T+1 days around the event,.e. 21 days, 11 days, 7 days, 3 days and 1 day before and after each event, or n other words [ 10, + 10], [ 5, + 5], [ 3, + 3], [ 1, + 1], and [ 1, 0], where 0 s the day of the event. Usng the estmated parameters for the Markowz market model (1), we defne Market- adjusted return as follows aˆ = ˆ α + ε, t [ t T, t T ] (3) + Smlarly, usng the estmated parameters for the Fama- French model (2), we defne Fama- French adjusted return as follows ˆ (4) a = +α ˆ + ε RF t We compute the abnormal returns of bank as the devaton of the actual returns from those predcted by the Markowz market model (1) and the Fama- French three factors model (2). The Fama- French benchmark factors are obtaned from Kenneth R. French Data Lbrary. Market capalzaton and daly stock returns are retreved from CRSP database. For the Markowz market model, the abnormal returns are computed from the followng equaton ε = R ˆ α ˆ β MKT, t [ t 0, t ] (5) ˆ M t 1 13

19 Smlarly for the Fama- French model, we defne the abnormal returns as follows ˆ ε = ( R RF ) ˆ α ˆ β ( MKT RF ) ˆ β SMB ˆ β HML (6) t M t t SMB The ndvdual banks abnormal returns are aggregated usng t HML t εˆ from eher equaton (5) or equaton (6) for each tradng day (t) whn estmaton wndow [ t T, t + T]. The aggregated abnormal return for tradng day t s AR t = 1 N N = 1 ˆ ε (7) Average cumulatve abnormal returns ( CAR ) are derved by summng the abnormal returns over varous ntervals τ = t T t CAR τ = AR t (8) 4.2 The Announcement of TARP Table 5 presents the mean, medan and standard devaton of the followng varables around the day of the announcement of TARP: (a) raw stock returns (b) market adjusted stock returns (c) Fama- French adjusted stock returns (d) market abnormal returns (e) Fama- French abnormal returns (f) cumulatve abnormal returns for the market model (g) cumulatve abnormal returns for the Fama- French model. The statstcal sgnfcance of all the above varables are tested and ndcated at the, 5% and 10% sgnfcance levels, respectvely. Standard errors are adjusted for heteroskedastcy and autocorrelaton. As shown n the table, even though the average row stock returns of balout banks n the sample are negatve over the event wndows of [ 10, + 10] and [ 5, + 5] around the announcement of TARP, the adjusted returns, abnormal returns, cumulatve abnormal returns are unformly posve regardless of the model specfcaton and the event wndow chosen. Balout banks stocks responded to the announcement of TARP favorably, mplyng that the launch of TARP ndeed restored nvestors confdence n the fnancal system. Table 5: Returns around the Announcement of TARP Event Wndow Varable Mean Medan Standard Devaton [ 10, +10] Raw stock returns Market- adjusted Fama- French adjusted Market abnormal Fama- French abnormal Market CARs

20 Fama- French CARs [ 5, +5] Raw stock returns Market- adjusted Fama- French adjusted Market abnormal Fama- French abnormal Market CARs Fama- French CARs [ 3, +3] Raw stock returns Market- adjusted Fama- French adjusted Market abnormal Fama- French abnormal Market CARs Fama- French CARs [ 1, +1] Raw stock returns Market- adjusted Fama- French adjusted Market abnormal Fama- French abnormal Market CARs Fama- French CARs [0] Raw stock returns Market- adjusted Fama- French adjusted Market abnormal Fama- French abnormal Market CARs Fama- French CARs Notes: Summary statstcs are presented for the returns of the balout banks around October 14, 2008 (the day of the announcement of TARP). The sample of banks that accepted TARP balout funds durng the October 2008 to December 2009 perod s obtaned from ProPublcas TARP database. Stock return data s retreved from CRSP US Stock database. Reported are mean, medan, and standard devatons of row 15

21 stock returns, market- adjusted stock returns, Fama- French adjusted returns, market abnormal returns, Fama- French abnormal returns, market CARs, and Fama- French CARs n event wndows of 2 T + 1 tradng days around the announcement of TARP,.e. 21 days, 11 days, 7 days, 3 days and 1 day around October 14, The return varables are defned n the text. Table 6 shows the pont and cumulatve estmates of the average abnormal returns around the day of the announcement of TARP (.e. October 14, 2008) estmated usng one- factor market model. Fgure 2 provdes a graphcal overvew of the average CARs by plottng the average CARs aganst tradng days relatve to the day of the announcement of TARP along ther 90 percent confdence bands. The pont (daly average) and cumulatve (relatve to 10 days before the event) abnormal returns estmated usng one- factor market model confrm the observaton from Table 5. The average abnormal returns are sgnfcantly posve on the day of the announcement of TARP as well as the day after, both are greater than 4%, suggestng the event had an mmedate effect on banks' stock performance. Even f we control for the pre- event trend (average daly abnormal return of 0.56% pre- event), the balout banks cumulatve abnormal returns after the announcement of TARP are stll sgnfcantly posve. Table 6: Pont and Cumulatve Market Abnormal Returns around the Announcement of TARP Event Day Pont Estmaton CAR Estmaton Mean Std. Dev. Mean Std. Dev ** ** *** *** *** *** *** ** *** *** ** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ** *** *** ***

22 *** *** *** *** *** *** *** *** Notes: The table shows the pont and cumulatve abnormal returns estmated usng Markowz market model n a wndow of ten days before and ten days after October 14, 2008 (the day of the announcement of TARP). The pont and cumulatve estmate of the average returns for the event are reported along ther standard error. Standard errors are adjusted for heteroskedastcy and autocorrelaton. The return varables are defned n the text. *, **, and *** represent statstcal sgnfcance at the 10%, 5%, and level, respectvely. Fgure 2: The Evolvement of Market CARs around the Announcement of TARP Tradng Days Relatve to the Announcement of TARP (0 = Event Day) Notes: The fgure shows the average cumulatve returns of the balout banks n the sample n a wndow of ten days before and after October 14, 2008 (the day of the announcement of TARP), along ther 90% confdence bands. CARs plotted n ths fgure are estmated usng Markowz market model. Table 7 shows the pont and cumulatve estmates of the average abnormal returns around the day of the announcement of TARP estmated usng Fama- French model. Fgure 3 provdes a graphcal overvew of the average CARs by plottng the average CARs aganst tradng days relatve to the day of the announcement of TARP (.e. October 14, 2008) along ther 90 percent confdence bands. The pont and cumulatve abnormal returns estmated usng three- factor Fama- French model are more posve and more sgnfcant than ther one- factor market model counterparts around the event wndow of 10 days 17

23 before and after the announcement of TARP, confrmng that the TARP to a great extent restored nvestors confdence n fnancal system. The cumulatve Fama- French abnormal return over the entre event wndow s as hgh as 21.95%. The dfference between Fgure 2 and Fgure 3 may be explaned by the sze effect that large bank responded to the announcement of TARP more posvely thank the small banks. To provde further nsghts, we spl the balout banks n our sample nto 5 sub- samples based on ther book value of assets as of the quarter- end of the announcement of TARP,.e. 31 December, The cumulatve abnormal return over the event wndow of 10 days before and after the even are reported for each of the 5 sub- samples, see Appendx 4. It clearly shows that the large banks were performng sgnfcantly better than the small banks when the TARP was announced. Ths dfference may because the large banks are more lkely to be baled out f s needed n the future. Table 7: Pont and Cumulatve Fama- French Abnormal Returns around the Announcement of TARP Event Day Pont Estmaton CAR Estmaton Mean Std. Dev. Mean Std. Dev *** ** *** *** *** *** *** *** *** *** *** *** ** *** *** *** *** *** *** ** *** *** *** *** *** ** *** *** *** *** *** *** ***

24 *** *** Notes: The table shows the pont and cumulatve abnormal returns estmated usng Fama- French three- factor model n a wndow of ten days before and ten days after October 14, 2008 (the day of the announcement of TARP). The pont and cumulatve estmate of the average returns for the event are reported along ther standard error. Standard errors are adjusted for heteroskedastcy and autocorrelaton. The return varables are defned n the text. *, **, and *** represent statstcal sgnfcance at the 10%, 5%, and level, respectvely. Fgure 3: The Evolvement of Fama- French CARs around the Announcement of TARP Tradng Days Relatve to the Announcement of TARP (0 = Event Day) Notes: The fgure shows the average cumulatve returns of the balout banks n the sample n a wndow of ten days before and after October 14, 2008 (the day of the announcement of TARP), along ther 90% confdence bands. CARs plotted n ths fgure are estmated usng Fama- French three- factor model. 4.3 The Recept of TARP Funds Table 8 presents the mean, medan and standard devaton of the same set of varables as defned n Table 5, but the event date s set to be the day that each balout bank n our sample actually receved the TARP funds,.e. the day of recept. In contrast to the results presented n Table 5, the balout banks stock returns around the day of the recept of TARP funds are negatve accordng to most of the measures, especally n the event wndow of 1 day before and after the event. However, the one- factor market model and three- factor Fama- French model gve us conflctng results f alternatve event wndows are consdered. Returns estmated usng one- factor market model show a negatve market reacton to the recept of the balout funds, whle returns estmated usng Fama- French three- factor model are all 19

25 posve even though ther magnudes are farly small (close to zero). The medans of the returns are consstently negatve regardless of the model specfcaton and event wndow consdered. Our emprcal results are consstent wh the fndngs of Bayazova and Shvdasan (2012) that the recept of TARP funds dd not have meanngful certfcaton effect. Table 8: Returns around the Recept of TARP Funds Event Wndow Varable Mean Medan Standard Devaton [ 10, +10] Raw stock returns Market- adjusted Fama- French adjusted Market abnormal Fama- French abnormal Market CARs Fama- French CARs [ 5, +5] Raw stock returns Market- adjusted Fama- French adjusted Market abnormal Fama- French abnormal Market CARs Fama- French CARs [ 3, +3] Raw stock returns Market- adjusted Fama- French adjusted Market abnormal Fama- French abnormal Market CARs Fama- French CARs [ 1, +1] Raw stock returns Market- adjusted Fama- French adjusted Market abnormal Fama- French abnormal

26 Market CARs Fama- French CARs [0] Raw stock returns Market- adjusted Fama- French adjusted Market abnormal Fama- French abnormal Market CARs Fama- French CARs Notes: Summary statstcs are presented for the returns of the balout banks around the day that each bank n the sample actually receved the TARP funds. Ths event date s specfc to each balout bank, rangng from October 2008 to December The sample of banks that accepted TARP balout funds durng ths perod s obtaned from ProPublca s TARP database. Stock return data s retreved from CRSP US Stock database. Reported are mean, medan, and standard devatons of row stock returns, market- adjusted stock returns, Fama- French adjusted returns, market abnormal returns, Fama- French abnormal returns, market CARs, and Fama- French CARs n event wndows of 2 T + 1 tradng days around the date that each bank receved the TARP funds,.e. 21 days, 11 days, 7 days, 3 days and 1 day around the day of recept. The return varables are defned n the text. Table 9 reports the pont and cumulatve estmates of the average abnormal returns around the day of the recept of TARP funds usng market model. Fgure 4 provdes a graphcal overvew of the average CARs by plottng the dynamcs of the average CARs aganst tradng date relatve to the day of the recept of TARP funds along ther 90 percent confdence bands. In lne wh the results reported n Table 8, the cumulatve abnormal returns estmated usng one- factor market model reman negatve throughout the entre event wndow of 10 days before and after the banks actually receved the balout funds. The balout bank experenced sgnfcantly negatve abnormal returns mmedately after the recept of TARP funds. Although the balout banks underperformed the market before they receved the balout funds, they performed even worse after the event. The negatve cumulatve abnormal returns are stll sgnfcant even f we control for the pre- event downward trend. Ths may suggest that the recept of TARP funds conveyed a sgnal that the bank s n trouble to the market, therefore the event was nterpreted as a bad news by the outsde nvestors. 21