How do zombie banks affect the real economy : an insight into the U.S. states

(1)

1

University of Amsterdam

MSc Business Economics, Finance track

Master Thesis

“How do zombie banks affect the real economy: An insight into

the U.S. states”

Student: Karavasilis Iordanis

Student number: 10551085

Thesis supervisor: Martynova Natalya

5 July 2014

(2)

2

ABSTRACT

In this paper, I analyze the effects of zombie banks on several macroeconomic indicators such as individual income, unemployment rate and productivity. Zombie banks are financial institutions that have an economic net worth below zero. Their existence depends on governments’ bailouts. Zombie banks’ abnormal business lending makes them dangerous. I employ paneI data regressions using as entities the U.S. states. I document that the existence of zombie banks decreases individual income. Moreover, zombie banks trigger increases in unemployment rate while they don’t affect productivity. Finally, our findings suggest that the effects of zombie banks remain constant at least after a year.

Acknowledgments: I thank my supervisor, Natalya Martynova, for helpful comments and ideas and my parents for their support.

(3)

3

1. Introduction

The official definition for zombie banks is: “Financial institutions with an economic net worth below zero but continue to operate because their ability to repay their debts is shored up by government credit support”, (Kane, 1987). Zombie banks’ attraction to “horrifically unfair and inefficient funding and lending strategies” deteriorates the economy and makes them dangerous (Kane and Rice 2001). They don’t prefer to lend to prudent borrowers. On the contrary, they gamble in the resurrection of insolvent borrowers in order to comply with Basel capital standards, cover their losses and generate enough earnings which will finally restore their net worth.

Zombie banks have attracted a lot of attention in the recent years, especially after the financial crisis of 2007-20081. Wolf2 (2009) indicates that a sizeable proportion of American financial institutions is insolvent. Whitney3 (2011), also states that the top U.S. financial institutions which dominate most of the lending in U.S. are zombie banks and they have negative operating leverage. In addition, the analyst claims that financial giants such as Morgan Stanley and Bank of New York face structural economic problems and huge uncertainty. Further, Krugman (2009) classifies Citigroup and Bank of America, as zombie banks. The solvency of these major banks is critical for the whole financial system (Whitney, 2011). Finally, “Failed bank list”4, provided by Federal Deposit Insurance Corporation, gives an essence of the increase in the number of insolvent banks in U.S. after the recent financial crisis. This list records only 27 bank failures from 2000 until 2007 and around 500 bank failures from 2008 until 2014.

1_{Also known as the Global Financial Crisis and 2008 financial crisis.}

2

Martin Wolf is the associate editor and chief economics commentator at Financial Times. He is considered one of the most influential writers on economics all over the world.

3_{Meredith Ann Whitney is the co-founder and Chief Investment Officer of Kenbelle Capital LP, a}

hedge fund based in New York.

4

This list includes banks that have failed since the 1st of October, 2000. It includes useful information

(5)

5

“Zombies” are usually born through financial distress. The emergence of zombie banks is due to several reasons. Regulators’ laxity in their inspections and political unconcern are the foundations of the emergence of zombie banks. According to Kane (2000) banking insolvency and career pressure create bad incentives for bankers and regulators. Officials and supervisors are often “tempted to promote their bureaucratic and career interests” instead of forcing recapitalization (Kane et al. 2001). Further, Kane and Rice (2001) indicate that a politically directed loan to powerful parties conceals an opportunity loss which comes from the interest gap between the yield needed to compensate the bank for the default risk it takes and the loan’s expected rate of return. Thus, the larger the amount of politically preferred loans, the more likely it is for a bank to become insolvent.

Another reason for the emergence of zombie banks is depositors’ indifference to control banks’ operations due to guarantees that are provided to them (Hormar and Wijnbergen 2013). Finally, in U.S., Federal Deposit Insurance Corporation5 had liquidity and manpower constraints which reduced its ability to deal with “zombies”. Consequently, these insolvent financial institutions were left in this state for a long period of time (Zwick, 2012).

Keeping zombie banks “alive” has several pros and cons. Limiting spillovers and avoidance of financial panic and liquidity problems are some of the main benefits of the bailouts to zombie banks. On the contrary, non- optimal allocation of resources, high inefficiency in some banks and moral hazard problems constitute the most important negatives of the existence of these insolvent financial institutions. Also, “zombies” affect adversely the economy as they don’t utilize the amount of money of governments’ bailouts.

Zombie banks’ future earnings are very unpredictable because they have a huge amount of nonperforming assets on their balance sheets. As most of the loans are

5

Federal Deposit Insurance Corporation (FDIC) is indebted to cover the difference between the deposits assumed and the assets that the acquirer purchases when it closes an institution. In section 3.1 I refer more about the operations of FDIC.

(6)

6

nonperforming, zombie banks record significant losses which damage taxpayers, as they pay the costs of banks’ insolvencies. In particular, Tarr (2009) stresses that trillions of dollars are transferred from U.S. taxpayers to insolvent banks. Also, according to Calderon and Schaeck (2013) the rescue costs that taxpayers pay for governments’ interventions in the banking sector tend to be larger than the improvements for bank customers (both depositors and borrowers).

As long as creditors are confident that government will continue to provide bailouts to a zombie bank in order to satisfy its obligations and liabilities, this insolvent financial institution is able to operate and even to grow (Kane, 2000). If investors become doubtful whether government is willing to continue supporting the existence of zombie banks, bank runs by uninsured depositors are revealed. Due to banking interconnectedness nowadays these bank runs affect also healthy financial institutions which may face liquidity problems through massive deposit withdraws. Also, in case of uncertainty, silent runs6 take place (Kane and Rice 2001). Silent runs constitute another way to turn sound financial institutions to insolvent ones by reducing their profit margins.

Trying to find the best way to manage insolvent financial institutions and prevent their evolution has become a major topic across countries. Basel III, a global regulatory standard, attempts to deal with deficiencies in financial regulations. Mainly, Basel III tries to strengthen bank capital requirements by reducing bank leverage and increasing bank liquidity. However, the fact that zombie banks do not reveal their actual assets, liabilities and shareholders’ equity in their balance sheets makes their detection, management and application of Basel’s III regulations much more difficult. In addition, previous applied strategies are not effective. Having an insight in U.S. and according to Soros (2009), bailing out banks is not a sufficient strategy to make American banks solvent. Furthermore, Tarr (2009) clearly states that U.S. programs which include purchasing toxic assets, are ineffective and risk the credit of the U.S. government. The author claims that purchasing toxic assets of

6

“Silent run” is a situation in which depositors withdraw massively their funds via electronic and wire transfers.

(7)

7

zombie banks is a wrong strategy as “zombies” persist and continue their abnormal business lending. On the other hand, the cost of restructuring insolvent banks seems enormous. Krugman (2009) indicates that the funds needed to bring Citigroup and Bank of America back to life exceed their current worth. According to the best estimate of Veronesi and Zingales (2009), the cost of U.S. government to restore top ten banks to financial health is up to $1.2 trillion while, according to Whalen (2009), including Fannie Mae, Freddie Mac and AIG in the sample with the big banks ends to raise the cost to $4 trillion if default rate is high.

The key role that insolvent financial institutions played in the recent global financial crisis, their vital importance for the whole financial system, the growing number of zombie banks during the recent years and the absence of papers that investigate the direct effects of the existence of “zombies” on the real economy lead to settle the next research question: How do zombie banks affect the real economy?

There is a need to identify the way that the operations of zombie banks affect the real economy. Thus, the purpose of this thesis is to detect which banks can be classified as “zombies” and to investigate which sectors of the economy they affect. The paper contributes to the existing literature by looking whether the presence of zombie banks in the U.S. states affects adversely several economic measures such as Individual Income, Unemployment rate and Quantity index Real GDP which measures the real output of the U.S. states. To my knowledge this has not been encountered in previous studies.

I use U.S. data from 2004 until 2012. The research method is empirical analysis and theoretical review. In empirical research I use the definition of Kroszner and Strahan (1996) to detect zombie banks.

The thesis is organized as follows: Section 2 provides a summary of previous theoretical and empirical studies related to the topic of this paper. Throughout Section 3 a brief description of U.S. banking regulation nowadays is given. Moreover,

(8)

8

Section 4 presents description of data, outlines the methodology and develops the hypotheses. Section 5 presents the results and section 6 the robustness checks of the findings. Finally, Section 7 provides a summary of the paper and concludes.

2. Literature review

Pervious literature that studies the effects of zombie banks on the real economy is not so abundant. The existing papers are classified into three categories. The first category refers to the effects of zombie banks on different economic indicators. The second one focuses on the way governments’ interventions to the banking sector affects the economy. Literature on government interventions is relevant because zombies’ existence depends on governments’ bailouts and the ways that governments’ intervene influence “zombies” financial activities. As mentioned before zombie banks lend to insolvent borrowers. This zombie lending turns insolvent borrowers to zombie firms. Consequently, one of the main effects of zombie banks on the economy is through lending to zombie firms. The third stream of literature examines the connection between zombie banks and zombie firms and the consequences of this relationship.

2.1 Effects of zombie banks

The approach of this paper builds on previous work by Zwick (2012). Zwick (2012) studies the effects of zombie banks on the U.S. counties. The author focuses on the relationship between unhealed zombie banks and lending. Further, the author investigates how the losses generated from zombie banks affect the employment in manufacturing, services and mining sectors. Zwick (2012) aggregates the deposits on a county level to find out what fraction of deposits is held by zombie banks at a given point of time and how these deposits affect local economies.

(9)

9

The study of Zwick (2012) is limited to include only listed zombie banks because the author makes use of FDIC’s and CalculatedRisk’s7 web sites to receive data about bank failures in order to identify zombie banks, from 2007 until 2011. Therefore, in this research, I employ a more empirical way to identify “zombies”. Moreover, as Zwick (2012) states, his strategy concerning employment is “admittedly imperfect” and so there is some room for improvement. Thus, in this paper, I include more control variables to address concerns about omitted differential economic factors which drive the results.

The main finding of Zwick (2012) is that in areas where zombie banks operate8 even healthy banks are unable to “pick up the slack”, which leads to a slower growth. In addition, the author concludes that “zombies” are mainly small and medium sized financial institutions. Further, Zwick (2012) indicates that employment recovers more slowly in counties where “zombies” operate.

In a theoretical paper Tarr (2009) claims that abnormal lending of American zombie banks slows down the economic recovery of the country. Further, Acharya and Mora (2013) empirically report the correlation between failed and nearly failed banks and Certificates of Deposit rates9 indicating that as a bank becomes insolvent, it offers higher rates10 to attract deposits. The authors define a bank as failed or nearly failed if it experienced a decrease in stock price of more than 90%. Moreover, as depositors realize that bank deposits are riskier than other instruments which offer the same liquidity and payment services they insist in a further increase in the deposit interest rates. In U.S., the effect of the rise of deposit rates was the creation of a possible stress on deposit funding of banks, as 62% of country’s deposits were uninsured at the beginning of the crisis (Acharya and Mora 2013).

7

CalculatedRisk is an economics web log which provides an unofficial list of bank failures.

8_{Especially in areas where the deposits of zombies are above ten percent of total deposits.}

9

Certificates of Deposit (CD) rates refer to the interest rate of Certificates of Deposit. This interest rate is a fixed rate of interest payable on a set maturity date.

10_{The inclusion of articles that examine the correlation between insolvent financial institutions, the}

trend of interest rates and the safety of deposits is essential as the variable named “Individual Income” that I use includes personal interest income, which is the income that individuals receive from different sources that include interest. Consequently, interest rates affect Individual Income.

(10)

10

2.2 Effects of governments’ interventions

Calderon and Schaeck (2013) focus on the way that zombie banks affect banking competition and deposit and loan interest rates when governments intervene during banking crises. Likewise this research, Calderon and Schaeck (2013) employ Tangible Capital11 in order to classify a bank as a zombie. Further, the authors use cross-country data. They calculate the number of zombie banks in percent of the total number of banks and “zombies” loan and deposit market shares to detect their effects and provide robustness checks. In this paper, I also employ the market share of deposits of zombie banks to find how they affect the real economy. However, as my study is limited in the U.S. states I use cross-state data.

Calderon and Schaeck (2013) express the view that zombie banks cause a larger unwanted increase in the banking competition when governments intervene. Too much competition causes moral hazard problems and incentivizes banks to take excess risks. According to Cetorelli and Gambera (2001) banking competition affects access to finance, availability of credit and the economic growth of countries. Furthermore, Calderon and Schaeck (2013) conclude that when governments intervene in the banking sector of areas where “zombies” dominate, there is a reduction in average deposit rates as banks are safer after interventions and depositors require lower returns. This finding suggests that banks have access to cheaper funding costs. Simultaneously, the authors observe a larger reduction in average loan rates. This reduction of the loan nominal rates indicates moral hazard problems that the excess banking competition causes. As the decrease of loan rates is larger than the reduction of deposit rates, moral hazard problems outweigh advantages that rise from the cheaper funding cost (Calderon et al. 2013).

11

The definition of Tangible Capital is explained in details in Section 4.2 which refers to Methodology of this paper.

(11)

11

In another empirical paper, Homar and Wijnbergen (2013) claim that when governments intervene in the banking sector with measures12 which fail to face the problem of undercapitalized financial institutions, zombie banks have high incentives and they consider an attractive strategy to just hold on their bad loans, no matter if these bad loans are negative Net Present Value projects. In addition, Giannetti and Simonov (2009) claim that after governments’ bailouts, recapitalized financial institutions not only continue to lend to their clients but they also provide them larger loans.

The continuation of this abnormal business lending destroys value as it ties up resources in insolvent firms, bereaves credits for new economic activity and forms a drag on economic recovery after recessions (Homar and Wijnbergen 2013, and Tarr 2009). In addition, Giannetti and Simonov (2009) declare that after the announcement of governments’ intervention in the banking sector, both high and low quality borrowers of the banks that receive bailouts observe an increase in their values and face positive abnormal returns. However, the low quality clients experience higher abnormal returns especially when governments’ injections are not able to reestablish high levels of banks capitalization. This fact explains the reduction of the market share of more healthy and productive firms.

Further, according to Giannetti and Simonov (2009) some of the firms that are clients of financial institutions which receive bailouts, increase their investment but they neither create more jobs nor increase their growth of sales in the following two years after the recapitalization. Finally, the authors indicate that when weak banks are merged by stronger financial institutions the borrowers of the last ones face negative abnormal returns. This in another way that zombie banks which are weak financial institutions, hurt healthy firms when they are merged.

12

The measures that fail to face the problem of undercapitalization are blanket guarantees and liquidity support.

(12)

12

2.3 Effects of zombie firms

The importance of zombie banks was revealed since the banking stress that African13 countries experienced and Japan’s lost decade14. Peek and Rosengren (2005) examine the misallocation of bank credit to poor performing firms. During poor performance periods, the banks’ of less productive firms are more willing to lend to them than other financial institutions. The result of the misdirect bank lending is the conversion of insolvent borrowers to zombie firms (Ahearne and Shinada 2004, Caballero et al. 2006, and Okamura 2011). In an empirical study Okamura (2011) employs Moody’s credit ratings15 to classify a bank as a “zombie”. The author finds that undercapitalized banks are precursors of zombie firms. Zombie banks increase zombie firms’ life by 4.7 years and unlike strongly capitalized ones, these insolvent financial institutions allow firms in financial distress to become more indebted by 20.8%.

Ahearne and Shinada (2004) and Caballero, Hoshi, and Kashyap (2006) focus on zombie lending and give a bank-based explanation for Japan’s slowdown after the asset price collapse in the beginning of 1990s. The paper closest to this research is the one of Caballero, Hoshi, and Kahyap (2006). The authors study the relationship between zombie firms on the one hand and job creation, wages and productivity on the other hand. They empirically detect zombie firms16 and they use the percentage of assets of these insolvent firms to identify their effects both over time and across different sectors. In this paper, I focus on the detection of insolvent financial

13_{The banking stress that African countries experienced refers to the years 1980 to 1999. The level of}

government corruption played a significant role in the increase of the average length of time an African banking system spent in crisis (Kane and Rice, 2001).

14_{Japan’s lost decade refers to the years 1991 to 2000. It is the time after the Japanese asset price}

bubble’s collapse.

15

Banks rated by Moody’s Baa 1 and below are classified as undercapitalized banks. On the other hand, banks rated by Moody’s as A2 and above are treated as strongly capitalized.

16

Caballero et al. (2006) classify the firms as zombies based on their estimation of whether these firms receive subsidized credit or not. The authors try to detect zombie firms by comparing the minimum required interest payment of each firm each year with the actual interest payment of this firm. A firm is classified as zombie for a specific year if the actual interest payments are lower than the minimum required ones. Also, in order to avoid the misclassification of non-zombies the authors take into account the quality of the corporate bonds of firms and the low funding costs of some of them.

(13)

13

institutions. Further, Caballero et al. (2006) calculate the productivity per firm while I use Quantity index Real GDP17 per U.S. state as a measure of productivity.

Zombie firms decrease the profits and investments of healthy companies and prevent them from expanding (Caballero, Hoshi, and Kahyap 2006). According to Caballero et al. (2006), these insolvent borrowers depress market prices for their products and increase the wages of their workers whose productivity decreases. Further, they congest the markets in which they operate as their existence strangles the market share of more productive firms (Ahearne and Shinada 2004, and Caballero et al. 2006). Consequently, in industries where “zombies” dominate, there is a reduction in job creation and a lower productivity growth. Finally, in another empirical study, Hoshi and Kim (2012) examine the effects of zombie lending on the Korean economy. The authors claim that zombie firms create a productivity gap between them and healthy firms as their concentration in the industry increases. In addition, Hoshi and Kim (2012) record a decrease in employment growth, profits and investment of non-zombie firms as the proportion of zombies rises in an industry. Thus, even healthy banks cannot find good lending opportunities (Caballero et al. 2006).

Some papers (Acharya and Mora 2013, and Zwick 2012) investigate the effects of zombie banks on several economic variables. Others (Calderon and Schaeck 2013, Giannetti and Simonov 2009, and Homar and Wijnbergen 2013) examine the way that governments’ interventions affect the existence of zombie banks, recession duration, firms’ value and operations, deposit and loan rates and banking competition. Moreover, most of the papers (Ahearne and Shinada 2004, Caballero et al. 2006, Hoshi and Kim 2012, and Okamura 2011) focus on the detection and the effects of zombie firms on the economy. To my knowledge there are no papers that test empirically the effects of zombie banks on key macroeconomic indicators in the U.S. states.

17

The definition of Quantity index Real GDP is given in Section 4.1 which describes the Data used in this research.

(14)

14

3. U.S. banking regulation nowadays

As regulators’ laxity is the foundation of the emergence of zombie banks, I consider interesting to describe briefly U.S. banking regulation nowadays, emphasizing on the Federal Deposit Insurance Corporation (FDIC) resolution of failed banks.

3.1 Overview of U.S. regulatory system- FDIC

In U.S. banks are regulated both at federal and state level. Since 1989, year when the Federal Deposit Insurance Corporation Improvement Act was enacted, all commercial banks that hold deposits are obliged to obtain FDIC insurance and to have a primary federal regulator18. Any bank that carries the FDIC insurance falls under the regulatory supervision and authority of the FDIC.

Banks are classified into five categories19 according to standards for the leverage ratio, total capital ratio and Tier 1 capital ratio. Federal Deposit Insurance Act (FDIA) demands regulators to take “Prompt Corrective Action” to handle banks that are weakly capitalized and establishes a system that insures deposits held by banks and savings associations. When a financial institution is characterized as undercapitalized, it is obliged to submit a capital restoration plan. “Significantly undercapitalized” institutions are indebted to decrease their total assets, issue more stock and stop obtaining deposits from other financial institutions. Moreover, FDIA demands a receiver to be appointed in maximum 90 days if a bank is classified as “critically undercapitalized”20. Also, in order to ensure the safety of the deposits FDIA authorizes FDIC to slow up the mandatory resolution.

18

Bank’s primary federal regulator could be either the FDIC, or the Federal Reserve Board or the Office of the Comptroller of the Currency (OCC).

19_{The categories are: “well capitalized”, “adequately capitalized”, “undercapitalized”, “significantly}

undercapitalized” and “critically undercapitalized”.

20

A bank is classified as critically undercapitalized if its tangible equity is 2% or below of average quarterly tangible assets.

(15)

15

The protection of the depositors until the amount of their insured deposits21 and the management of the receivership for failed banks after their resolution are the two main actions of FDIC in every bank failure. The bank loses its authority and is taken over by FDIC when it is declared as insolvent, “critically undercapitalized“ or unable to satisfy deposits withdrawals. To keep a bank solvent FDIC has the authority to intervene in bank’s operations, force it to liquidate its assets and reduce its dividends, staff and executive salaries and bonuses. Moreover, FDIC has the ability to regulate bank’s real estate loans and provide it with regulations on its earnings.

FDIC uses three basic resolution methods for failed and failing institutions: deposit payoffs, purchase and assumption transactions and open bank assistance transactions. FDIC’s resolution activities begin formally by receiving a “failed bank letter”. Afterwards, FDIC sets a planning team which comes in contact with the CEO of the failed bank to discuss logistics and to obtain loan and deposit information. Next, FDIC decides all the possible resolution methods and provides bidders with the information package.

Bidders submit their proposals to the FDIC. Bidders’ offers consist by two amounts: the premium, for the franchise value of the failed institution’s deposits and the amount that the bidder is willing to pay for institution’s assets. FDIC is obliged to choose the least costly option to the deposit insurance fund. FDIC submits a written recommendation to the FDIC Board of Directors22 requesting approval of the proposed resolution. After the approval FDIC informs the acquirer and all successful bidders. The last step in the procedure is the closure of the failed bank and the transfer of the assets and the deposits to the acquirer.

21

The amount of insured deposits is $250,000 per depositor, per insured bank for each account ownership category.

22

(16)

16

4. Data and Methodology

In this section I describe the data used. Further, I develop the methodology and the hypotheses.

4.1 Data

The examination period extends from 2004 to 2012. I use the following data sources. WRDS (Wharton Research Data Services) provides data on bank financials23. The sample consists of 5,965 bank-year observations across states and over time for 979 financial institutions operating in 50 U.S. states24.

Moreover, U.S. Bureau of Economic Anlasyis (BEA) provides data on macroeconomic indicators such as Nominal Gross Domestic Product (GDP), Real GDP, Quantity index Real GDP, Aggregate Personal Income, Compensation of Employees, Number of Employees of the U.S states. Finally, I use Bureau of Labor Statistics (Local Area Unemployment Statistics) to obtain data for Unemployment rate and Census Bureau to obtain data for the Population of each state. The frequency of the data is yearly.

Firstly, I present a brief description of the variables25. Nominal GDP by state is the value added in production by the labor and capital located in a state. Nominal GDP for a state is the sum of the gross domestic product by state originating in all industries in a state. For each industry, GDP by state is composed of three components: Compensation of employees, Taxes on production and imports less subsidies, Gross operating surplus. Real GDP represents the real estimates of Gross domestic product by state and is measured in chained (2005) dollars. Quantity index Real GDP is an index number that measures the change in the level of quantity from

23

I obtain American banks’ deposits, assets, liabilities, common equity, intangible assets and the states that each bank operates, using Bank fundamentals annual provided by WRDS- Computstat Bank.

24

From the 50 U.S. states Wyoming is excluded as no data for its banks was available. In order to replace it we use District of Columbia.

25

(17)

17

a base year (2005), apart from any changes in relative prices. Quantity index Real GDP is an essential measure to represent the productivity of the U.S. states.

Aggregate Personal Income by state is the income that is received by all persons who live in the state from all sources. It is calculated as the sum of wages and salaries, supplements to wages and salaries, proprietors' income with inventory valuation and capital consumption adjustments, rental income of persons with capital consumption adjustment, personal dividend income, personal interest income, and personal current transfer receipts, less contributions for government social insurance. Compensation of Employees, is the sum of wages and salaries and supplements to wages and salaries. Number of Employees measures the average annual number of full-time and part-time jobs. Lastly, Population by state is the midyear (July 1) population estimates for each state.

As the Aggregate Personal Income represents the income that is received by all persons who live in the state, I create a new variable named “Individual Income”. Individual income equals the Aggregate Personal Income divided by the Population of each state per year. This new variable indicates the average amount of money that each person receives in a specific state for a certain year. Also, I consider useful the calculation of the Inflation rate per state and per year. In order to calculate the Inflation rate I follow the next procedure. I use the formula (Nominal GDP * 100) / Real GDP (Mankiw, 2008) to calculate the Deflators of Gross Domestic Product for each year. Next, I change the basis year from 2005 to 2003. I recalculate new deflators by setting the new deflator of 2003 equal to 100 and then I use the formula (Deflatort *100)/ Deflator2003 to compute the New Deflator for each year t. Finally, I compute the Inflationt for each year by using the formula {(NewDeflatort - NewDeflatort-1)* 100}/ NewDeflator t-1 (Mankiw, 2008). The last variable that I calculate is named “Ratio of Deposits”. This variable represents the percentage of deposits held by zombie banks in a specific state and during a certain year to the aggregate amount of deposits of all banks in the same state and during the same period of time. I calculate Ratio of Deposits by dividing the amount of deposits held

(18)

18

by zombie banks in a certain state and during a specific year by the whole amount of deposits of all banks in the same state and during the same year. Lastly, I multiply the result of the division by 100 to calculate the percentage. Table 1 presents an analytical description of all variables and the sources from which I obtain all the necessary data.

Table 2 provides some descriptive statistics of the data. In particular, the mean of the Nominal GDP is equal to $282,303.9 million as reported in column (1) of Table 226. Further, the average of the Real GDP is $262,613 million. It is reasonable that the latest mean is smaller than the first one, as it is measured in chained (2005) dollars and it doesn’t include the influence of Inflation. The standard deviation of Real GDP is equal to $308,369.9 million as reported in column (2). The mean of Quantity index Real GDP equals 103.1447. This fact means that in average the level of real output in the U.S. states generally increased since 2005, which is the base year. The annual Individual Income ranges from $25,265.43 to $70,143.56 as reported in columns (3) and (4). The mean of the Unemployment rate is equal to 6.289%, while it varies between 2.5% and 13.8%. The mean of Ratio of Deposits held by zombie banks is 0.74%. This fact implies that in average the deposits that zombie banks hold consist the 0.74% of the whole amount of deposits of all banks. The average Population is 6,219,060 persons and the minimum value of this variable is 567,136 persons recorded in District of Columbia in 2005. The minimum Number of Employees is recorded in Vermont in 2010 and is equal to 310,622. On the contrary, the maximum value of this variable is 16,300,000 employees. Also, column (1) indicates that the mean of Inflation rate is equal to 2.475%. Lastly, the mean of the Compensation of all Employees is $158,508.9 million while its standard deviation is $183,957.7 million.

To give to the reader a better feeling of the most crucial variables of this research I present some figures. Figure 1 shows the average Nominal Gross Domestic Product of the U.S. states through time. We observe that the amount of average GDP

26

The amounts in Table 2 are computed after I drop observations in which there is no data for deposits of banks and observations in which the percentage of Inflation rate is too high (above 7%) or too low (below -7%).

(19)

19

increases from 2004 until 2008. In particular, in 2008 average GDP is equal to $280,000 million approximately. Next, we document a decline during the recession period in U.S.27 followed by another increase which continues until 2012, year in which average GDP outreaches $300,000 million. Figure 2 plots the time series of the average Quantity index Real GDP of the U.S. states. The index of real output declines sharply in 2009 while it rises from 2004 until 2008 and from 2010 until 2012. Further, in Figure 3 I present scatter plots of the Ratio of Deposits (percent) held by zombie banks and the Individual Income of some of the states. It is clear, that in areas where the Ratio of Deposits of zombie banks is above 5% of the aggregate amount of deposits the annual Individual Income hovers between $30,000 and $40,000, while in areas where the presence of zombie banks is not so intense the annual Individual income reaches the amount of $50,000. This fact implies a negative correlation between the existence of zombie banks and the income that each person receives. Figure 4 presents the average Unemployment rate of the U.S. states through time. We see that the Unemployment rate is double in 2010 in comparison to the one in 2007. Panel A of Figure 5 shows the time series of the average Ratio of Deposits of the U.S. states while Panel B illustrates the number of U.S. states with at least one zombie bank through time. We observe the emergence of zombie banks in the mid of 2008. Moreover, we document that the graph of Ratio of Deposits held by zombie banks moves up and down reaching its maximum value of approximately 5% of all deposits in the mid of 2011. However, the number of U.S. states with at least one zombie bank reaches its maximum value in 2010, year in which the deposit percentage of “zombies” is below 2%. This fact, suggests that the most of the deposits of zombie banks comes from small and medium sized financial institutions in 2010. Finally, Figure 6 provides a map of the U.S. states with at least one insolvent financial institution during the examination period. Mainly, eastern states are more likely to be marked as “zombie” states28.

27

During years 2004 to 2012 there was one recession period in U.S. economy which started from December 2007 and ended in June 2009 according to National Bureau of Economic Research (NBER).

28

(20)

20

Finally, Table 3 documents the correlation matrix of all my variables. This table helps us to avoid possible multicollinearity problems in our equations.

4.2 Methodology and hypotheses

Firstly, I describe a way to identify zombie banks. Zombie banks are financial institutions that have an economic net worth below zero. Nevertheless, zombie banks do not reveal their actual assets and liabilities. Consequently, I can’t use the definition of Kane (2000), “Total Assets – Total Liabilities”, to detect them. The classification of American financial institutions as zombie banks is based on the definition of Kroszner and Strahan (1996). In particular, I compute Tangible Capital. Tangible Capital is equal to “common equity minus intangible assets”. If Tangible Capital is negative for a specific year, then the bank is classified as a zombie bank for this specific period of time.

In next step, I employ Panel Data regressions. The entities of the regressions are the U.S. states and the frequency of observations is yearly. The advantages of using Panel Data regressions are that more observations are included and there is more control for omitted variables. I control for omitted variables using entity and time fixed effects. Entity fixed effects control for omitted variables that change across the states but are time-invariant. On the other hand, time fixed effects control for omitted variables that are common to all entities but vary over time. The inclusion of both fixed effects allows testing for a causal effect of the independent on the dependent variables. Finally, I include clustered standard errors in order to correct for autocorrelation that leads to standard errors which are often too low. Clustered standard errors control for omitted factors that might, for a given entity, be correlated over time. For example, one of these factors might be a downturn in the local economy. The independent variable is the Ratio of Deposits. As cited, Ratio of Deposits is the percentage of deposits held by zombie banks in a specific state and during a certain year to the aggregate amount of deposits of all banks in the same state and during the same period of time. I use the Ratio of Deposits as my

(21)

21

independent variable as it implies how significant is the amount of deposits held by zombie banks for a specific state and during a certain year. I employ three panel data regressions which examine the correlation of the above mentioned independent variable to Individual Income, Unemployment rate and Quantity index Real GDP. In particular, the equations and the hypotheses that I develop are listed below.

1) Ii,t = ai + lt + b*Di,t + c*GDPindexi,t + d*Inflationi,t + ei,t

where Ii,t is Individual Income, Di,t is the Ratio of Deposits of zombie banks, GDPindexi,t represents the Quantity index Real GDP, Inflationi,t is the Inflation rate, ei,t is the error term, ai is a state-specific intercept which controls for variables that are constant over time but different across states and lt is a time-specific intercept which controls for variables that are constant across states but differ over time. Null hypothesis: The increase in Ratio of Deposits of zombie banks is expected to decrease Individual Income. This outcome would defend the view that zombie banks create zombie firms (Okamura, 2011) which in turn decrease the earnings of healthy firms (Caballero et al. 2006, and Hoshi and Kim 2012). Thus, solvent firms are forced to decrease wages and salaries, which consist the main source (approximately 2/3) of Individual Income in U.S., in order to stay profitable. Consequently, the higher the Ratio of Deposits of zombie banks the larger loans they provide to insolvent borrowers, hurting more the earnings of healthy companies.

As control variables I use GDPindex as an increase in the productivity of a country increases country’s average income (Mankiw, 2008) and Inflation rate in order to control for the change in the value of dollar through time.

2) Ui,t = ai + lt + b*Di,t + c*RealGDPi,t + d*Inflationi,t + f*Populationi,t + ei,t

where Ui,t is the Unemployment rate, RealGDPi,t represents the real estimates of Gross Domestic Product, Populationi,t represents the number of persons that live in a specific state for a specific year.

(22)

22

Null hypothesis: The increase in Ratio of Deposits of zombie banks is expected to decrease Unemployment rate. This is a very interesting hypothesis as from the one side zombie banks create zombie firms which prevent healthy firms from expanding and investing and reduce their earnings (Caballero et al. 2006, and Hoshi and Kim 2012). Thus, solvent firms are forced to dismiss employees. From the other side, zombie banks allow zombie firms to continue operating (Okamura, 2011) and not go bankrupt. The more deposits zombie banks have in their balance sheets, the more zombie firms operate. This fact increases the total number of firms operating in the economy and as a result the total number of jobs. Typically, I expect the increase in the total number of jobs to be larger than the layoffs of healthy firms.

As control variable I use Real GDP according to Okun’s law (1962). Okun’s law (1962) supports the view that changes in RealGDP affect Unemployment rate. More specifically, it implies that there is a negative correlation between Real GDP and Unemployment rate. Also, I employ Inflation rate as control variable according to Philips curve which implies a negative relationship between the Unemployment rate and Inflation rate for the short term29. Moreover, I use the Population in order to control for sudden changes in the population30 of each state, as a sudden increase (decrease) in the population tends to increase (decrease) the unemployment rate.

3) GDPindexi,t = ai + lt + b*Di,t + c*Ui,t + d*Compensationi,t + ei,t

whereCompensationi,t represents the Compensation of Employees.

Null hypothesis: The increase in Ratio of Deposits of zombie banks is expected to decrease GDPindex. This outcome would show that zombie lending not only bereaves credits from healthy firms discouraging further investment and decreasing their productivity (Caballero et al. 2006, and Hoshi and Kim 2012), but it also allows zombie firms whose workers’ productivity declines (Caballero et al. 2006) to continue operating.

29_{As I use Panel Data regression each year’s Unemployment rate is regressed on each year’s Inflation}

rate. Consequently, as the frequency of data is yearly, the short term correlation between Unemployment rate and Inflation rate is valid. Therefore, Philips curve can be applied in this case.

30

(23)

23

As control variables I use the Unemployment rate as a change in the Unemployment rate affects the level of productivity in the economy (Mankiw, 2008) and the Compensation of Employees as the lower salaries increase the amount of production that is offered (Mankiw, 2008).

Finally, I take into account the fact that the existence of zombie banks may not have an immediate effect on the above dependent variables. The losses that these insolvent financial institutions generate may need some time in order to spill into the real economy. Therefore, I replace the Ratio of Deposits with the lag value of Ratio of Deposits in my equations to observe how previous year’s (e.g. 2004) Ratio of Deposits of “zombies” affects next year’s (e.g. 2005) Individual Income, Unemployment rate and Quantity index Real GDP. The significance of all the coefficients is measured at 10%, 5% and 1% significance level.

5. Results

In this section I discuss and analyze the results of the regressions.

5.1 Discussion and analysis of results

In tables 4a to 6b each raw reports the coefficient estimates and the t-statistics (in parentheses) for the explanatory variables. In each table column (1) introduces the univariate regression. In column (2) I include control variables, while in column (3) I control for entity fixed effects. Column (4) presents the results of the regressions after the introduction of both entity and time fixed effects. Finally, column (5) uses cluster standard errors on the state level. I drop observations in which there is no data for deposits of banks and observations in which the percentage of Inflation rate is too high (above 7%) or too low (below -7%). The results are not significantly affected when these observations are included.

(24)

24

Table 4a presents the effects of Ratio of Deposits of zombie banks on Individual Income. Irrespective of the inclusion of control variables the coefficient of the Ratio of Deposits is positive but not statistically significant. The inclusion of control variables in column (2) increases the efficiency of the estimation but has little effect on the significance of the coefficient of our independent variable. Controlling for entity fixed effects doesn’t have any remarkable influence in our results. Nevertheless, the inclusion of both entity and year fixed effects increases the validity of our estimation. All the analyses therefore include both entity and time fixed effects. The sign of the coefficient of Ratio of Deposits changes in column (4) as it becomes negative but the coefficient remains statistically insignificant. In the last column of Table 4a I find strongly significantly31 negative effects of the Ratio of Deposits on Individual Income. This estimate suggests that omitted factors which, for a given entity, are correlated over time drive our results. Our findings indicate that the presence of “zombies” decreases Individual Income. Further, the volume of Ratio of Deposits matters for the changes in the dependent variable. In particular, an 1% increase in the Ratio of Deposits of zombie banks reduces the annual income that each person receives by $15.45. However, the Population estimates that I use to divide Aggregate Personal Income, include infants and children who are more likely not having any income from any source. Thus, the decrease in the income that each person receives may be even larger than $15.45 as the number of individuals that work and receive income is remarkably lower than the Population estimates. The control variables exhibit intuitive signs.

Table 4b documents the effects of previous year’s Ratio of Deposits of zombie banks on next year’s Individual Income. I replace the Ratio of Deposits with the lag value of Ratio of Deposits. The results in Table 4b are pretty much the same as the ones in previous table. In column (5) we observe that the coefficient of the lag value of Ratio of Deposits is statistically significant32 and equal to -13.94, implying that an 1% increase in previous year’s Ratio of Deposits of “zombies” reduces next year’s Individual Income by approximately $14. Consequently, the findings claim that

31

The coefficient of Ratio of Deposits of zombies is significant at 1% significance level.

32

(25)

25

zombie banks have both an immediate and a more long term effect on Individual Income, as indicated in Table 4a and Table 4b respectively. Thus, the null hypothesis that the operations of zombies decrease Individual Income is approved.

Table 5a presents an attempt at measuring the effects of zombie banks on Unemployment rate at the state level. In all columns the Ratio of Deposits is significantly positively related to Unemployment rate. In particular, column (1) indicates that the largest increase in Unemployment rate coincides with the largest zombie bank presence. The inclusion of control variables in column (2) increases the efficiency of our estimation. The coefficient of Ratio of Deposits remains statistically significant at 1% significance level in columns (2) and (3) but when I control for both entity and time fixed effects the coefficient is statistically significant at 5% significance level. In the last column the results suggest that an 1% increase in the Ratio of Deposits of zombie banks tends to increase by approximately 0.017% the Unemployment rate.

Table 5b examines the effects of previous year’s Ratio of Deposits of “zombies” on next year’s Unemployment rate. Again, I replace the Ratio of Deposits with the lag value of Ratio of Deposits. The coefficient of the independent variable is remarkably larger than the coefficient of Ratio of Deposits in Table 5a. More specifically, column (5) implies that an 1% increase in the lag value of Ratio of Deposits raises next year’s Unemployment rate by 0.032%. This fact indicates that as the time goes by, the effect of the existence of zombie banks on Unemployment rate increases. Thus, the null hypothesis that the more deposits zombie banks have in their balance sheets, the more zombie firms operate increasing the total number of jobs is rejected. A possible explanation is that zombie banks lend to certain zombie firms and do not expand their zombie lending to other insolvent borrowers. Therefore, a specific amount of zombie firms continues to operate which does not increase the total number of jobs. On the contrary, zombie firms decrease the earnings of healthy firms, discourage their entry and further investment and depress job creation (Caballero et al. 2006, and Hoshi and Kim 2012). Our results are in line not only with

(26)

26

Caballero et al. (2006) and Hoshi and Kim (2012) but they are also consistent with the findings of Zwick (2012) who supports that in areas where “zombies” operate, employment recovers more slowly.

Table 6a documents the effects of Ratio of Deposits of zombie banks on Quantity index Real GDP. As cited, Quantity index Real GDP is a measure of real output. The regression analysis suggests that zombie banks do not affect the productivity of the U.S. states as the key coefficient is not statistically significant. The Ratio of Deposits may have negative coefficient but it remains statistically insignificant in all columns of Table 6a. Consequently, the interpretation of the results is unfeasible.

The findings of Table 6b, which shows the effects of previous year’s Ratio of Deposits of zombie banks on next year’s Quantity index Real GDP, are slightly different than the ones of Table 6a. In particular, column (1) implies a weakly significantly33 negative correlation between the lag value of Ratio of Deposits held by zombie banks and the value of GDPindex. The outcome indicates that an 1% increase in previous year’s Ratio of Deposits tends to decrease next year’s value of GDPindex and as a result the real output of the states, by 0.074 units. However, the facts that this result is statistically significant only at 10% significance level and only in column (1), in which the validity of the estimation is extremely low, indicate that this finding should not be taken as definitive. In conclusion, the existence of zombie banks doesn’t have any effect on the real output after a year. Thus, the null hypothesis that zombie banks reduce Quantity index Real GDP is rejected. Possible explanations may be that the effects of zombie banks on the productivity of the states become visible in the long-run or that in case of U.S. zombie banks haven’t turned so many insolvent borrowers into zombie firms in order to affect the productivity. Hoshi and Kim (2012) support the last explanation as they indicate that the productivity gap between “zombies” and solvent firms rises as the concentration of zombie firms in the industry increases. Also, Ahearne and Shinada (2004) and Caballero et al. (2006) claim that industries with high proportion of zombie firms exhibit lower productivity.

33

The coefficient of lag value of Ratio of Deposits in column (1) of Table 6b is significant only at 10% significance level.

(27)

27

In brief, the estimates for the coefficient of Ratio of Deposits held by zombie banks tend to be negative and strongly significant for the Individual Income regressions, positive and strongly significant for the Unemployment rate regressions and negative but insignificant for the Quantity index Real GDP regressions.

6. Robustness checks

In this section I provide the results from robustness checks. I examine the validity of our results by replacing the Ratio of Deposits with the Ratio of Liabilities of zombie banks. The Ratio of Liabilities is an essential measure to examine the robustness of our findings as governments provide bailouts to zombie banks in order to satisfy their total liabilities and not only the claims of their depositors. Thus, the amount of bailouts (via which zombie banks affect the economy) is based on zombies’ total liabilities and not only on their deposits. In our sample there are several cases in which the deposits of zombie banks constitute a small amount of total liabilities. Further, there are zombie banks with zero deposits. Thus, the significance of the presence34 of zombie banks changes remarkably when I employ the Ratio of Liabilities instead of the Ratio of Deposits.

6.1 Discussion of robustness checks

Tables 7 to 9 report a set of robustness checks. The tables have the same structure as the previous ones.

Table 7 presents the effects of Ratio of Liabilities of zombie banks on Individual Income. I illustrate the effects based on the coefficients when I control for clustered standard errors. The regressions show that our results are robust. More specifically,

34

(28)

28

the coefficient of Ratio of Liabilities is significantly35 negatively related to Individual Income. The findings suggest that an 1% increase in the Ratio of Liabilities of zombie banks reduces annual Individual Income by $14.32. This amount is quite lower than the one when I use the Ratio of Deposits as independent variable. However, there is no doubt that the presence of zombie banks indeed decreases the average income that each person receives.

Table 8 documents the effects of our independent variable on Unemployment rate. The regression analysis confirms our claim that the presence zombie banks increases Unemployment rate. In all the regressions for Unemployment rate the coefficient of Ratio of Liabilities is positive and clearly36 statistically significant. In particular, the coefficient of our independent variable in column (5) is 0.0164. The economic interpretation of this finding is that an 1% increase in the volume of Ratio of Liabilities of zombie banks raises Unemployment rate by 0.016%. This result coincides with the one when I use the Ratio of Deposits held by zombie banks.

Finally, Table 9 presents an attempt at measuring the validity of our results concerning Quantity index Real GDP. Our robustness check confirms our claim that the existence of zombie banks doesn’t have any effect on the value of Quantity index Real GDP and consequently on the real output of the U.S. states as the coefficient of the Ratio of Liabilities may be negative but it remains statistically insignificant when I control for clustered standard errors. Nevertheless, in column (2) of Table 9 the coefficient of Ratio of Liabilities is statistically significant at 10% significance level. However, the low efficiency of estimation in column (2) in combination with the fact that the coefficient of our independent variable isn’t statistically significant in any other column claim that this result should not be taken as definitive.

35_{The coefficient of the Ratio of Liabilities is significant at 1% significance level in column (5) of Table}

7.

36

The coefficient of the Ratio of Liabilities is statistically significant at 1% significance level in all columns except column (4) in which it is significant at 5% significance level.

(29)

29

7. Summary and Conclusions

Zombie banks have attracted a lot of attention in recent years. In this research, I investigate the influences of the emergence of zombie banks in Individual Income, Unemployment rate and Quantity index Real GDP. Rather than focusing on the U.S. economy as total, I use a cross-state dataset and panel data regressions, in which the entities are the U.S. states. This approach allows us to use more observations and to control better for entity, time fixed effects and omitted variables.

Our key results highlight that the losses generated from zombie banks spill into the real economy. In particular, our findings indicate that “zombies” reduce Individual Income. This result supports theory and implications that zombie banks affect negatively several macroeconomic variables. In addition, the findings of this paper claim that the emergence of these insolvent financial institutions triggers economically increases in Unemployment rate. This outcome is in line with Caballero, Hoshi, and Kashyap (2006) and Zwick (2012) who stress that zombie lending displays more depressed job creation and that zombie banks cause a slower recovery in employment. Moreover, I find no correlation between zombie banks and Quantity index Real GDP which measures the productivity of the U.S. states. Hoshi and Kim (2012) provide a possible explanation of why the findings of this research do not indicate any effect of zombie banks on the real output. The authors point out that the productivity gap between “zombies” and healthy firms rises as the concentration of zombie firms that operate in the industry increases. Further, Ahearne and Shinada (2004) and Caballero et al. (2006) claim that industries with high proportion of “zombies” exhibit lower productivity. Thus, in case of the U.S. states zombie lending may not have turned so many insolvent borrowers into zombie firms in order to affect the real output. Finally, our findings suggest that the effects of zombie banks remain constant at least after a year.

I acknowledge two limitations. The first one is that I can’t detect the exact effect of zombie banks on Individual Income as the Population estimates that I use to divide Aggregate Personal Income in order to find the average amount of money that each

(30)

30

person receives, include individuals such as infants and children who are more likely not receiving any income from any source. Consequently, there are implications that the actual effect of zombie banks on Individual Income is essentially larger than the one of our results. Secondly, the regressions that I employ in this paper leave open the possibility that other control variables, whose inclusion in the equations is unfeasible as there is no data available per state and per year, could drive our results. Nevertheless, several robustness checks that confront alternative explanations and examine the validity of our findings suggest that the results of this research are robust.

It is clear that zombie banks affect adversely several macroeconomic indicators. Their evolution and their horrifically unfair lending to insolvent firms destroy value and cripple the economy. Further, their gamble in the resurrection of insolvent borrowers affects the operations and changes the incentives of other healthy financial institutions creating moral hazard problems. Consequently, the view of Wolf (2009) who supports “killing” zombie banks at once can’t be considered unfounded.

Possible ways to extend this research is to study the effect of zombie banks on other macroeconomic indicators such as the stock market. Further, this research could be conducted in a global level as the emergence of these insolvent financial institutions could have different influences in the macroeconomic variables of different countries. Finally, as the solvency of financial institutions is crucial for the whole economic system, detecting zombie banks by using stress tests and identifying strategies in order to “cure” them appear important prospects for future research.

(31)

31

8. References

Acharya, Viral V., and Nada Mora., 2013. “A crisis of banks as liquidity providers.” Unpublished manuscript. NYU Stern School of Business.

Ahearne, A.G., Shinada, N., 2004. Zombie firms and economic stagnation in Japan. International Economics and Economic Policy 2, 363–381.

Caballero, R.J., Hoshi, T., Kashyap, A.K., 2006. Zombie Lending and Depressed Restructuring in Japan (Working Paper No. 12129). National Bureau of Economic Research.

Calderon, C., Schaeck, K., 2013. Depressing depositors and cheering up borrowers: The effects of bank bailouts on banking competition and the evolution of zombie banks.

Giannetti, M., Simonova, A., 2009. On the real effects of bank bailouts: Micro -evidence from Japan.

Homar, T., van Wijnbergen, S., 2013. On zombie banks and recessions after systemic banking crises: It does matter how governments intervene. University of Amsterdam.

Hoshi, T., Kim, Y., 2012. Macroprudential Policy and Zombie Lending in Korea. T. Hoshi and Y. Kim.

Kane, E.J., 1989. Changing incentives facing financial-services regulators. J Finan Serv Res 2, 265–274.

Kane, E.J., 2000. Capital movements, banking insolvency, and silent runs in the Asian financial crisis. Pacific-Basin Finance Journal 8, 153–175.

Kane, E.J., Rice, T., 2001. Bank Runs and Banking Policies: Lessons for African Policy Makers. J Afr Econ 10, 36–71.

Kollewe, J., 2009. George Soros warns “zombie” banks could suck lifeblood out of economy. The Guardian.

Kroszner, R.S., Strahan, P.E., 1996. Regulatory Incentives and the Thrift Crisis: Dividends, Mutual-To-Stock Conversions, and Financial Distress. The Journal of Finance 51, 1285. doi:10.2307/2329395.

Krugman, P., 2009. Banking on the Brink, The New York Times, 23.

(32)

32

Okamura, K., 2011. “Zombie” Banks Make “Zombie” Firms (SSRN Scholarly Paper No. ID 1786496). Social Science Research Network, Rochester, NY.

Peek, J., Rosengren, E.S., 2005. Unnatural Selection: Perverse Incentives and the Misallocation of Credit in Japan (Working Paper No. 9643). National Bureau of Economic Research.

Tarr, D. G., 2009. Bailouts and deficits or haircuts: how to restore U.S. financial market stability. Available at SSRN 1401555.

Whitney, M.A., 2011. Financial Giants Turning Into “Zombie Banks”, n.d. CNBC.com. URL http://www.cnbc.com/id/44087025 (accessed 4.28.14).

Wolf, M., 2009. Why Obama’s new Tarp will fail to rescue the banks. Financial Times, 11.

Zwick, E., 2012. Regulators vs. Zombies: Loss Overhang and Lending in a Long Slump.

(33)

33

(34)

34 Figure 1:

Notes: This figure documents the average Nominal Gross Domestic Product of the U.S. states through time. Gross Domestic Product is counted in millions.

Figure 2:

Notes: This figure presents the average Quantity index Real GDP of the U.S. states through time. Quantity index Real GDP is 100 for the base year (2005).

0 50000 100000 150000 200000 250000 300000 350000 2004 2005 2006 2007 2008 2009 2010 2011 2012

Average Gross Domestic Product

Average Gross Domestic Product In m illi o n s 92 94 96 98 100 102 104 106 108 110 2004 2005 2006 2007 2008 2009 2010 2011 2012

Average Quantity index Real GDP

(35)

35 Figure 3:

Notes: This graph shows the presence of zombie banks in corelation with Individual Income for some of the U.S. states.

Figure 4:

Notes: This figure represents time series of the average Unemployment rate of the U.S. states. 0 10000 20000 30000 40000 50000 60000 0 5 10 15 20 25

Zombie banks and Individual Income

Ratio of Deposits of zombie banks (%)

Ave rag e In d ividu al In co m e FL _HI IN MD MI MO OH SC TN VA WI 0 1 2 3 4 5 6 7 8 9 10 2004 2005 2006 2007 2008 2009 2010 2011 2012

Average Unemployment rate

(36)

36 Figure 5:

Panel A:

Notes: This graph documents the average Ratio of Deposits held by zombie banks of the U.S. states through time. The ratio is expressed in percent.

Panel B:

Notes: This figure represents the number of U.S. states with at least one zombie bank through time. 0 1 2 3 4 5 6 2004 2005 2006 2007 2008 2009 2010 2011 2012

Deposit Percentage of Zombie Banks

Deposit Percentage of Zombies % % 0 1 2 3 4 5 6 7 8 9 10 2004 2005 2006 2007 2008 2009 2010 2011 2012

Number of "zombie" states

(37)

37 Figure 6:

Notes: Wharton Research Data Services (WRDS) provides data on bank financials at state-level. The frequency of the data is yearly. The detection of “zombies” is based on the definition of Kroszner and Strahan (1996). More specifically, I compute Tangible Capital. Tangible Capital is equal to “common equity minus intangible assets”. If Tangible Capital is negative, then the bank is classified as a “zombie”. The map shows all the states with at least one zombie bank from 2004 to 2012.

(38)

38

Table 1: Analytical Description

Notes: This table provides an analytical description of the variables that I use and the sources from which I obtain all the necessary data.

Variable Description Sources

Bank Deposits The total amount of deposits of Wharton Research Data Services- all banks per state and per year. Computstat Bank

Ratio of Deposits The percentage of deposits

held by zombie banks to the aggregate amount of deposits of all banks.

Nominal Gross Nominal GDP by state is the value added Bureau of Economic Analysis (BEA) Domestic Product in production by the labor and capital

located in a state. BEA derives Nominal GDP for a state as the sum of the gross domestic product by state originating in all industries in a state. For each industry, GDP by state is composed of three

components: Compensation of employees Taxes on production and imports less subsidies, Gross operating surplus.

Real GDP Real GDP represents the real estimates Bureau of Economic Analysis (BEA) of Gross domestic product by state and is

measured in chained (2005) dollars.

Quantity index Real A quantity index is an index number that Bureau of Economic Analysis (BEA) GDP measures the change in the level of

quantity from a base year, apart from any changes in relative prices. The value of the quantity index is 100 for the base year.

Aggregate Personal Aggregate Personal Income is the income Bureau of Economic Analysis (BEA) Income that is received by all persons from all

sources. It is calculated as the sum of wages and salaries, supplements to wages and salaries, proprietors' income with inventory valuation and capital

consumption adjustments, rental income of persons with capital consumption adjustment, personal dividend income personal interest income, and personal

current transfer receipts, less contributions for government social insurance.