(Un)Conventional U.S. monetary policy and bank risk-taking

Master‟s Thesis Economics

International Economics and Globalization

(Un)Conventional U.S. Monetary Policy and Bank Risk-Taking

Author: P.T.H. van Rooijen

Supervisor: dr. D.J.M. Veestraeten

Second reader: dr. K. Mavromatis

Date: 14-07-2016

Statement of Originality

This document is written by Student P.T.H. van Rooijen (6339468) who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Abstract

This paper investigates the relationship between (un)conventional monetary policy and bank risk-taking activity in the United States. After the global financial crisis of 2007-2008, economic literature argued that monetary policy actions have the ability to affect risk-taking behavior by banking institutions, typically done through a softening of banks‟ lending standards. Using information on lending standards reported by the Federal Reserve, this paper estimates the effect of monetary policy on bank risk-taking before and after the global financial crisis of 2007-2008. This paper finds that the monetary policy stance of the Federal Reserve increased bank risk-taking before the global financial crisis of 2007-2008, but actually reduced bank risk-taking afterwards.

Table of Contents

1. Introduction ... 1

2. A Missing Link in the Monetary Transmission Mechanism? ... 3

2.1 Conventional Monetary Policy in the United States ... 3

2.2 The Monetary Transmission Mechanism ... 7

2.3 A Missing Channel? ... 10

2.4 Translating the Risk-Taking Channel into Bank Risk-Taking ... 14

2.5 Did Bank Risk-Taking Cease to Exist after the Global Financial Crisis of 2007-2008? 19 3. Empirical Application ... 21

3.1 The Senior Loan Officer Opinion Survey ... 22

3.2 Results from the Senior Loan Officer Opinion Survey ... 24

3.3 Macroeconomic and Financial Variables ... 28

3.4 Empirical model ... 31

4. Empirical Results ... 32

4.1 Bank Risk-Taking before the Global Financial Crisis of 2007-2008 ... 33

4.2 Bank Risk-Taking after the Global Financial Crisis of 2007-2008 ... 36

5. Conclusion ... 39

1 1. Introduction

“For risks to be successfully managed, they must first be identified and measured”

These are the words of Ben Bernanke, Chairman of the Federal Reserve, spoken at the Federal Reserve Bank of Chicago‟s Annual Conference on Bank Structure and Competition in Chicago, Illinois, on May 15, 2008. At the time of Bernanke‟s speech, the subprime mortgage market had crashed in 2007, the bubble in the housing market had burst, and the global financial crisis of 2007-2008 was developing. Although not much context is provided, these words do reflect a substantial problem. The identification and measurement of risks is typically done after an adverse situation has occurred and the damage has already been incurred. For example, in 2001 the U.S. experienced a relatively mild and short-lived recession. The unemployment rate, however, kept on rising until 2003 when the central bank of the U.S. - the Federal Reserve (the Fed) - decided to lower the short-term policy rate in order to promote economic activity and battle the growing unemployment rate (Labonte, 2016). At the time of his speech, Bernanke most likely did not know that future theoretical and empirical economic literature would criticize this expansionary monetary policy stance of the Fed. Theoretical and empirical literature argued that the global financial crisis of 2007-2008 is, partly but not entirely, caused by policy rates that have been too low for too long between 2001 and 2006 (for example see Taylor (2009) and Maddaloni and Peydró (2011)). Naturally, the decision by the Fed to cut policy rates was clearly communicated to the public and the Fed explained as to why it had set course in this direction (Taylor, 2009). However, later on and with the benefit of hindsight, the problem becomes clear that substantial risks were not identified and measured in time, resulting in an economic crisis that was not (or could not have been) avoided.

2 Traditional monetary policy is conducted by the Fed via the three so-called conventional tools and it starts a monetary transmission mechanism. The monetary transmission mechanism consists of several channels through which monetary policy actions influence real economic variables, like the unemployment rate and aggregate output (Ireland, 2005). After the global financial crisis of 2007-2008 a body of literature emerged that explains that the monetary transmission mechanism, as identified earlier on, was incomplete (Borio and Zhu, 2008). The risk-taking channel was missing. When one looks back at the global financial crisis of 2007-2008 and is asked what may have caused this situation, a common answer would probably be: „banking institutions were at fault‟. Banking institutions are important players in the monetary transmission mechanism, and the newly created risk-taking channel indeed partly explains why banks took the risks that are expected to have contributed to the build-up of the global financial crisis of 2007-2008. In this paper we are interested in these risk-taking activities that were undertaken by banking institutions in the period before and after the global financial crisis of 2007-2008. We empirically examine the relationship between the expansionary monetary policy stance and bank risk-taking activity before the global financial crisis of 2007-2008. expecting to see a positive relationship. Phrased differently, we expect that the expansionary monetary policy stance of the Fed after the short-lived recession of 2001 could have caused banking institutions to take additional risks.

In 2009, after the global financial crisis of 2007-2008, the Fed used not-so traditional ways of conducting monetary policy. Between 2009 and 2014, the Fed conducted three rounds of quantitative easing, also known as Large-Scale Asset Purchases (Labonte, 2016). An effect of quantitative easing is that the monetary base, consisting of the reserve base and currency in circulation, increased significantly (Ireland, 2005; Kandrac and Schlusche, 2015). In this paper we also examine whether possible bank risk-taking activity before the global

3 financial crisis of 2007-2008 is identified and understood. We empirically examine whether unconventional monetary policy potentially causes banking institutions again to increase their risk-taking activities. We expect that unconventional monetary policy also leads to an increase in risk-taking activities by banking institutions.

This paper is organized as follows: chapter 2 contains the literature review, chapter 3 consists of the empirical application and empirical model, chapter 4 shows our results, and chapter 5 concludes this paper.

2. A Missing Link in the Monetary Transmission Mechanism?

This chapter reviews theoretical and empirical economic literature, and starts by explaining the conventional tools the Fed uses to conduct monetary policy and how the Fed is able to influence real economic variables, like aggregate demand and the rate of unemployment. Furthermore, this chapter explains that there is also a risk-taking channel in the monetary transmission mechanism. This channel partly explains the risk-taking activities undertaken by banking institutions that eventually contributed to the build-up of the global financial crisis of 2007-2008.

2.1 Conventional Monetary Policy in the United States

The central bank is responsible for conducting monetary policy. In the United States, this central bank is the Federal Reserve, also known as the Fed. According to Labonte (2016), the responsibilities of the Fed fall into four main categories: monetary policy, acting as a lender of last resort, regulating and supervising banks and other financial firms, and providing payment services to financial firms and the government. Labonte (2016) states that the Fed defines monetary policy as “the actions it undertakes to influence the availability and cost of money and credit to promote the goals mandated by Congress” (p.3). These goals of monetary policy are defined by the United States Congress in section 2A of the Federal

4 Reserve Act as: promoting maximum employment, a stable rate of inflation, and moderate long-term interest rates (Federal Reserve Act, 2013). However, Labonte (2016) remarks that there also exists a broader definition of monetary policy. He argues that the expectations of, for instance, buyers and sellers of capital goods have the potential to affect a major portion of spending in the U.S, and that these expectations are affected by monetary policy-induced changes. As a result, Labonte (2016) states that “a broader definition of monetary policy would include: the directives, policies, statements, economic forecasts, and other Fed actions” (p.3).

In principle, the Fed uses three conventional tools to conduct monetary policy: altering reserve requirements, altering the discount window, and conducting open market operations. In his paper, Labonte (2016) explains that reserve requirements are the amount of funds that a bank must hold in vault cash, or deposit at the Fed when customers make deposits. The intended purpose of reserve requirements is to deal with deposit withdrawals. Moreover, reserve requirements also affect the liquidity that is available in the federal funds market (Labonte, 2016). However, the working of the federal funds market will be explained later and in more detail in this section.

The second monetary policy tool is altering the discount window. The discount rate is the rate at which depository institutions and commercial banks can borrow reserves using a lending facility from their regional Federal Reserve Bank (The Discount Rate, 2016). There are three types of discount windows: primary credit, secondary credit, and seasonal credit, and all of them have different interest rates. These three different discount windows ensure that all depository institutions and commercial banks can obtain the necessary reserves in order to meet their liquid requirements (The Discount Rate, 2016). According to Labonte (2016), depository institutions and commercial banks can discount some of their assets in order to obtain temporary reserves on a short-term basis which is usually overnight. Using

5 assets as collateral ensures that all discount window loans are fully secured (The Discount Rate, 2016).

The third and last conventional instrument of the Fed is the tool that is most frequently used to conduct monetary policy: open market operations. The purpose of open market operations is to influence the monetary base, which consists of the reserve base and currency in circulation (Ireland, 2005). It involves the purchase and sale of U.S. Treasury securities, such as government bonds, in the secondary market. The secondary market is also known as the open market: a market where previously issued securities are bought and sold by private investors (Labonte, 2016; Open Market Operations, 2007). When conducting open market operations, the Fed deals only with primary dealers: dealers that have an established relationship with the Fed, trade in government securities, and have accounts at depository institutions (Open Market Operations, 2007). The Fed purchases existing U.S. Treasury securities with newly issued Federal Reserve notes, or sells U.S. Treasury securities from its assets portfolio. The result is that the Fed has complete control over the reserves created by this instrument; funds are added or withdrawn from the accounts of primary dealer‟s at depository institutions (Open Market Operations, 2007). Consequently, the increase or decrease in reserves can allow depository institutions to issue more or less loans (Labonte, 2016). The three conventional monetary policy tools increase or decrease the amount of liquidity that is available for banks, and thus affects the provision of credit by banks (Labonte, 2016). Labonte (2016) explains that:

Since the great inflation of 1970s, most central banks have preferred to formulate

monetary policy in terms of the cost of money and credit rather than in terms of their

supply. The Fed thus conducts monetary policy by focusing on the cost of money and

credit as proxied by an interest rate. In particular, it targets a very short-term interest

6 The Federal Open Market Committee (FOMC) is responsible for deciding upon open market operations and for the federal funds rate target. The FOMC holds eight scheduled meetings per annum and it is during these meetings that the FOMC decides what the target rate should be (Federal Open Market Committee, 2016). The federal funds rate is determined in a private market for overnight reserves of depository institutions. Labonte (2016) explains that depository institutions calculate the amount of reserves that they are required to hold or may want to hold. As a result, depository institutions may want to increase or decrease their holdings of reserves, and this can be done via the federal funds market. The interest rate that is determined in this market is called the federal funds rate (Labonte, 2016). A higher federal funds rate target set by the FOMC requires the Fed to sell existing U.S. Treasury securities to primary dealers. The result is that funds are withdrawn from depository institutions and from the market, which reduces the ability to lend, and vice versa. Consequently, the increase or decrease in the ability to lend can affect the interest rates that depository institutions ask for the provision of credit. Labonte (2016) explains that the federal funds rate may determine other interest rates that banks and other financial firms charge for loans. When the federal funds rate increases or decreases, the interest rate that financial institution charge for the provision of credit typically moves in the same direction. However, he states that the interest rate on issued loans must not increase or decrease proportionally following a change in the federal funds rate. This is because the federal funds rate also affects, but does not determine, longer-term interest rates. Longer-term market rates are influenced not only by what the Fed decides today, but also what the market expects the Fed to decide in the future and what the expected rate of inflation will be. Therefore, the Fed‟s role of managing expectations and behaving transparently has become increasingly important (Labonte, 2016).

7 2.2 The Monetary Transmission Mechanism

According to Ireland (2005), the monetary transmission mechanism describes how real variables, such as aggregate output and employment, are affected by monetary policy-induced changes. Ireland states that there are two key assumptions for the monetary transmission mechanism to work. The first assumption requires that agents are not able to nullify the adjustments of monetary policy-induced changes. Agents should not possess any form of privately-issued securities that are identical to the components of the monetary base which can be used as substitutes. In other words, disabling the ability to nullify the changes of monetary policy-induced changes ensures that the Fed has complete control over the monetary base. For example, if agents could substitute currency in circulation for their own privately-issued pieces of paper, then the monetary transmission mechanism would not work because the Fed‟s control over the monetary base is non-existent (Ireland, 2005). Second, Ireland states that nominal prices should not be able to respond immediately to changes in monetary policy. If nominal prices were to respond immediately to monetary policy-induced changes, then the real value of the monetary base would not change; hence there would be no change to the economy in real terms. In other words, there is some form of nominal price rigidity present that prevents nominal prices from adjusting immediately to monetary policy-induced changes. As a result, monetary policy-policy-induced changes are able to affect the real economy.

According to Ireland (2005) and Kuttner and Mosser (2002), the monetary transmission mechanism starts when the Fed conducts open market operations. The monetary transmission mechanism consists of several channels and proper understanding of this mechanism is crucial for the implementation of monetary policy. Furthermore, proper timing is required and it should be known what the consequences of policy-changes are (Mishkin, 1995). In their paper, Kuttner and Mosser (2002) identified and summarized the channels that

8 are active in the monetary transmission mechanism (see figure 1). The channels that are identified in Kuttner and Mosser (2002) are: the narrow credit channel, the wealth channel, the broad credit channel, the interest rate channel, the exchange rate channel, and the monetarist channel. We will briefly explain these different channels starting with the narrow credit channel. According to Kuttner and Mosser (2002), the narrow credit channel represents the relationship between the provision of credit by depository institutions and the availability of reserves. In figure 1 we see that open market operations affect the amount of reserves held at depository institutions, which in turn affects the loan supply. Kuttner and Mosser (2002) argue that because firms and households rely, partly or wholly, on the provision of credit by banks, the narrow credit channel is able to influence aggregate demand.

9 Second and third, the wealth channel and the broad credit channel are both related to the level of assets prices, as can be seen from figure 1. Kuttner and Mosser (2002) explain that market interest rates affect the prices of assets. Moreover they explain that when market rates increase in the wake of a monetary policy-induced increase in the federal funds rate, the prices of stocks, bonds, and real estate (long-lived assets) decreases. In his paper, Mishkin (1995) explains the effect of a monetary policy-induced change on the price of equities. He explains that when the money supply falls caused by a tightened monetary policy stance, the result is that the public has less money than it desires and consequently compensates by decreasing spending behavior and vice versa. For example, Mishkin (1995) states that the stock market is one place where the public can spend less, thereby decreasing the demand for equities and equity prices such as stocks, bonds, and long-lived assets. A result is that consumption by households that have these assets in their asset portfolio declines as their wealth decreases, resulting in a decline in aggregate demand. Furthermore, the drop in value of equities also causes a decline in the collateral value of households‟ and firms‟ assets. They state that this, in turn, affects the premium that borrowers have to pay on loans which causes an additional decline in consumption and investment spending.

Fourth, the interest rate channel. In his paper, Mishkin (1995) explains this traditional Keynesian channel as follows: given the assumption of nominal short-term price rigidity, a tightening in the monetary policy stance causes the real interest rate to rise. This increase in the real interest rate causes investment spending to fall and results in a reduction in aggregate demand. Naturally, the opposite is true when one speaks of an expansionary monetary policy stance. The fifth channel is the exchange rate channel. According to Mishkin (1995), when a monetary policy-induced change affects market interest rates, the exchange rate of the country in question is affected. Phrased differently, if we assume that this country is the United States and we assume a tightening of the monetary policy stance, Mishkin (1995)

10 argues that the domestic nominal interest rate will rise. The rise in the domestic nominal interest rate causes U.S. deposit accounts to become relatively more attractive compared to foreign deposit accounts. Naturally, the functioning of this channel assumes that the uncovered interest parity conditions hold. As a result, the dollar becomes more attractive and the domestic currency will appreciate. The stronger dollar leads to a reduction in U.S. exports and consequently in a decrease in aggregate demand (Mishkin, 1995).

The final channel is the monetarist channel. Kuttner and Mosser (2002) state that in asset portfolios, various assets are imperfect substitutes. Therefore, they argue that when monetary policy-induced changes affect the amount of outstanding assets, a result is that their relative price will change. A change in relative prices can have, according to Kuttner and Mosser (2002), real effects. Phrased differently, while assuming imperfect substitutability, a change in relative prices of assets caused by a monetary policy-induced change, results in an increase or decrease in the level of wealth of households, depending on the asset allocation in the household‟s asset portfolio. Changes in wealth may have an effect on, for example, spending on capital-intensive goods or the level of consumption by households. In turn, the changes in consumption or capital-intensive spending are able to influence aggregate demand, and thus the real economy.

2.3 A Missing Channel?

The global financial crisis of 2007-2008 was triggered by the crash of the subprime mortgage market from mid-2007 onwards and the burst of the U.S. housing bubble (Bordo, 2008). The global financial crisis of 2007-2008 is considered to be the worst financial crisis since the Great Depression (Maddaloni and Peydró, 2011). Since this global financial crisis, a growing body of literature emerged which discussed its causes and effects. Taylor (2009) states that loose monetary policy is a classic ingredient for boom-bust financial cycles. Indeed, it is

11 often argued that short-term interest rates were too low for too long, and contributed at least partly, to the build-up of the global financial crisis of 2007-2008 (Labonte, 2016). The concept of „low interest rates‟ is quite vague and may be difficult to grasp. When, for example, is it appropriate to consider an interest rate as too low? Taylor (2009) provides an explanation by making use of the „Taylor‟ rule. In his (1993) paper, Taylor suggests a policy rule which suggests appropriate changes in the federal funds rate in response to changes in the rate of inflation and real GDP. By making use of this rule, Taylor (2009) argues that the federal funds rate target, between 2001 and 2006, was well below the target rate suggested by his policy rule. He argues that there had not been a greater departure from the Taylor rule since the 1970s and that this deviation provides empirical evidence that the monetary policy stance was too loose between 2001 and 2006. The reason for this low rate was that the U.S. economy experienced a relatively mild recession in 2001. However, rising unemployment rates and the fear for a new recession caused the Fed to lower the federal funds rate drastically until mid-2004 (Labonte, 2016). The loose monetary policy stance and the corresponding low interest rates in combination with insufficient regulatory oversight, a softening of lending standards, and a steep increase in financial innovation caused the global financial crisis of 2007-2008 (Bordo, 2008; Labonte, 2016; Taylor 2009).

In his paper (1995), Mishkin states that it should be known what the consequences of policy-induced changes are. In retrospect, the global financial crisis of 2007-2008 suggests that some consequences of loose monetary policy-induced changes were unknown; hence the monetary transmission mechanism in figure 1 appeared to be incomplete. The missing link in the monetary transmission mechanism that explains the behavior that eventually led to the global financial crisis of 2007-2008 is commonly referred to as: the risk-taking channel, first described by Borio and Zhu (2008). They define the risk-taking channel as “the impact of changes in policy rates on either risk perceptions or risk-tolerance and hence on the degree of

12 risk in the portfolios, on the pricing of assets, and on the price and non-price terms of the extension of funding” (p.9). Phrased differently, adjustments in the federal funds rate target affect risk-taking behavior, the pricing of assets, and adjustments in lending standards. We know that changes in the federal funds rate target affects asset price levels, as described by Kuttner and Mosser (2002). However, the inclusion of asset prices in the definition of the risk-taking channel, suggests that it is closely related to risk-taking behavior and lending standards.

Borio and Zhu (2008) argue that there are three mechanisms in how the risk-taking channel works: a valuation effect, a search for yield effect, and the effect of communication policies and the reaction function of the central bank. First the valuation effect, or “the impact of interest rates on valuations, incomes and cash flows” (p.9). Borio and Zhu argue that the idea behind the valuation effect is closely related to that of the financial accelerator, and that it strengthens the financial accelerator‟s effect. In their paper, Bernanke, Gertler, and Gilchrist (1996) defined the financial accelerator as the intensification of shocks that are caused by changes in the credit market. We know that the price of assets (of long maturity), and consequently the value of collateral, increases when market interest rates decrease. Consequently, owners of assets therefore have increased net-worth. Borio and Zhu (2008) argue that when the net-worth of asset owners increases, their demand for investment increases as well. An important characteristic, they state, is that this increased net-worth lowers the expected probability of default by borrowers, resulting in an increase in investment. This additional increase in investment caused by lower interest rates raises asset prices even further. Consequently, Borio and Zhu (2008) state that a multiplier effect is involved in this process, and it is this multiplier effect that is the financial accelerator. The relationship between risk-taking and the financial accelerator results from the fact that risk tolerance is affected when profits, incomes, asset values, and the values of collateral increase.

13 Borio and Zhu (2008) state that there is a positive relationship between risk tolerance and an increased level of wealth, and that as a result risk-taking behavior is encouraged.

The second mechanism of the risk-taking channel is defined by a search for yield effect. Rajan (2005) states that especially a period characterized by high interest rates, followed by a period with low interest rates, encourages a search for yield. Borio and Zhu (2008) explain that when rate of return targets are fixed, a decrease in interest rates leads to an increase in risk-tolerance. Phrased differently, when the interest-rate difference increases between fixed target rates of return and interest rates, the result is that it may become more difficult to meet liabilities that have been predefined at long-term fixed rates, concerning for example pension funds or insurance companies (Borio and Zhu, 2008). Gambacorta (2009) for instance states that investors shifted from low-risk government bonds to riskier corporate and emerging market bonds in 2003-2004, when interest rates were below the rate that follows the Taylor rule. Moreover, Altunbas, Gambacorta, and Marques-Ibanez (2014) argue that it is common for private investors to relate the abilities of bank managers with the obtained short-term return. Comparing the abilities of bank managers induces risk-taking behavior for two reasons according to Rajan (2005). First, Rajan argues that bank managers will take risks that are unobservable by investors. He states that these risks that are not easily observed have the potential for a disastrous outcome. Second, Rajan states that behavior is easily copied; managers will display the same type of behavior to avoid underperforming. He explains that as a result, stock prices will move further away from their intrinsic value.

According to Borio and Zhu (2008), the third mechanism through which the risk-taking channel operates is through “communication policies and the reaction function of the central bank” (p.10). Borio and Zhu (2008) argue that a transparent central bank is important for removing uncertainty about the future and therefore for removing risk premia. Long-term market rates are influenced by today‟s decisions of the Fed, but also by decisions in the future

14 (Labonte, 2016). Therefore, an increase or decrease in market interest rates caused by the level of transparency of a central bank leads to a change in risk-taking behavior (cf. the valuation effect and the search for yield effect). Altunbas et al. (2014) point out that there also exists a moral hazard problem when a central bank behaves predictably. They argue that when the reaction function of the central bank is known, or, when banks know that the central bank intervenes in uncertain economic times, that this will lower banks‟ expected probability of large downside risks. According to Borio and Zhu (2008), the view that a central bank is effective in cutting off large downside risks amplifies risk-taking behavior when policy rates are low, compared to normal or high policy rates.

2.4 Translating the Risk-Taking Channel into Bank Risk-Taking

In the last decade, and especially after the global financial crisis of 2007-2008, the amount of available literature on the relationship between monetary policy and risk-taking has grown considerably. Most of the attention, amplified by their paper on the existence of a risk-taking channel by Borio and Zhu (2008), has been given to the relationship between an expansionary monetary policy and bank risk-taking (see for example: Gambacorta, 2009; Maddaloni and Peydró, 2011). The two mechanisms of the risk-taking channel, the valuation effect and the search for yield effect can be expected to lead to a softening of lending standards by banking institutions, which can lead to an unrestrained increase in loan supply (The Monetary Policy of the ECB, 2011). In other words, when the Fed conducts monetary policy and this subsequently affects the risk-taking channel of the monetary transmission mechanism, the search for yield effect and the valuation affect may typically increase bank risk-taking because banking institutions soften their lending standards. In order to empirically assess the relationship between the monetary policy stance of the Fed and the softening of lending standards by banking institutions before and after the global financial crisis of

2007-15 2008, we construct an empirical model based on the existing theoretical and empirical literature. The text below assists in explaining what our empirical model will look like.

According to Bernanke and Blinder (1992), the federal funds rate is sensitive to the supply shocks of bank reserves and accurately captures the monetary policy stance of the Federal Reserve. Phrased differently, when the Fed conducts monetary policy by using open market operations, the supply of bank reserves is altered and the U.S. interbank lending market for reserves moves to a new equilibrium rate. This equilibrium interest rate is the rate we know as the federal funds rate, and this specific variable is of interest. Most empirical research that analyzes the relationship between U.S. monetary policy and bank risk-taking uses the federal funds rate to approximate the monetary policy stance of the Fed. For instance, Ioannidou, Ongena, and Peydró (2009) study the impact of monetary policy on bank risk-taking and pricing in Bolivia between 1999 and 2003. Ioannidou et al. (2009) approximate the monetary policy stance of almost-fully-dollarized Bolivia by making use of the federal funds rate, and study the impact of this rate on the riskiness and pricing of new bank loans. They explain that the period between 1999 and 2003 is characterized by significant variation in the federal funds rate. Ioannidou et al. (2009) find empirical evidence that loose monetary policy conditions increases the willingness of Bolivian banks to take additional risks. Their findings show that a decrease in the federal funds rate translates into an increase in the hazard rate (the probability at which borrowers are expected to default at a certain time) but that the loans are not priced accordingly. Moreover, they find that observing subprime credit rating loans or loans to riskier borrowers becomes easier when the federal funds rate is low. In other words, Ioannidou et al. (2009) empirically find that loans provided to riskier borrowers, or loans that can be categorized as subprime, are more easily observed when the federal funds rate is considered low. One may assume that there is a relationship between the federal funds

16 rate and lending standards by banking institutions that causes, partly or entirely, the increase in the appearance of subprime credit loans or loans to riskier borrowers.

Maddaloni and Peydró (2011) empirically study the effect of short-term and long-term interest rates on bank lending standards in the Euro-area countries and the United States. Their dependent variables for their U.S. research are obtained using the Senior Loan Officer Opinion Survey, a survey which contains questions about the softening or tightening of bank lending standards. Listening to senior loan officers provides a large amount of information because changes in the lending standards reported by the loan officers can, partly but not entirely, predict changes in bank lending and gross domestic product (Lown, Morgan and Rohatgi, 2000). Maddaloni and Peydró (2011) find empirical evidence that low short-term interest rates indeed soften lending standards while long-term interest rates do not. The search for yield effect and the valuation effect of the risk-taking channel typically translate into the softening of lending standards by banking institutions. The results in the empirical study of Maddaloni and Peydró (2011) therefore indicate that low short-term interest rates lead to a softening of lending standards, and thus affect bank risk-taking activity. Moreover, they empirically show that the softening of lending standards, before the global financial crisis of 2007-2008, is amplified by: a loose monetary policy stance for an extended period of time, financial innovation, and low regulatory oversight.

Taylor (2009), states that the monetary policy stance of the Fed was too loose for too long, as summarized in the previous section. Taylor (1993) suggests a policy rule that suggests values of the federal funds rate in response to changes in the rate of inflation and real GDP. But how does inflation and real GDP relate to the federal funds rate, why does Taylor include the rate of inflation and real GDP in his suggested policy rule, and how is this policy rule related to bank risk-taking? First, Gambacorta (2005) empirically estimates the effectiveness of the bank lending channel, identical to the narrow credit channel defined by

17 Kuttner and Mosser (2002), and monetary policy transmission. Gambacorta (2005) empirically finds that when the Italian monetary authority tightens its monetary policy stance, a result is that Italian banks reduce their provision of credit. Gambacorta includes the rate of inflation and real GDP growth in his empirical model to control for cyclical movements in the demand for credit. In other words, according to Gambacorta (2005) there are cyclical movements in the demand for credit and the rate of inflation and real GDP growth, partly or wholly, explain these movements. For example, the Fisher-equation suggests that the real interest rate is equal to the nominal interest rate minus the expected rate of inflation. Thus when the expected rate of inflation increases, the real interest rate that is paid on loans decreases which in turn may increase the demand for additional credit.

Second, why does Taylor (1993) include the rate of inflation and real GDP in his policy rule? Taylor (1993) explains that suggested monetary policy rules generally seem to work well when the federal funds rate is raised by the Fed if the rate of inflation and real income are above a certain threshold, and vice versa. Phrased differently, the policy rule suggested by Taylor can be used to identify periods of loose or tight monetary policy when compared to the federal funds rate. For example, Taylor (2009) argues that the federal funds rate was far below the rate suggested by his policy rule between 2001 and 2006, and that this prolonged loose monetary policy stance of the Fed contributed to the build-up of the global financial crisis of 2007-2008. However, Taylor (2009) argues that the deviations from the suggested policy rule are made with careful consideration by the Fed, and that transparent language is used to communicate these changes to the public. Moreover, according to Taylor (2009), this low monetary policy stance of the Fed can be explained as a deviation from the regular way of conducting monetary policy to combat, for example, deflation. Third, how is the Taylor rule related to bank risk-taking? Empirical papers, for example Altunbas et al. (2009), Gambacorta (2009) and Maddaloni and Peydró (2011), use the Taylor rule to identify

18 periods characterized by interest rates that are below the rate suggested by the Taylor rule, and examine the effect of low interest rates on risk-taking by banks. These papers find empirical evidence that a loose monetary policy stance for an extended period of time contributes to bank risk-taking (Altunbas et al., 2009), and that it leads to a softening of lending standards by banking institutions (Maddaloni and Peydró, 2011).

Securitization activity and insufficient regulatory oversight are, besides loose monetary policy and the loosening of bank lending standards, both causes of the global financial crisis of 2007-2008 (Bordo, 2008). The relationship between securitization activity and risk-taking activities by banking institutions is explained by Taylor (2009). Taylor explains that the complexity of financial instruments increased before the global financial crisis of 2007-2008. Complex mortgage-backed securities were created, consisting of adjustable-rate subprime mortgages and other mortgages. According to Taylor (2009), the rating agencies responsible for estimating the risk on these securities failed to do so for three reasons: a lack of competition in the market, a lack of accountability, and rating agencies most likely failed to understand the complexity, and thus the riskiness, of these securities they were dealing with. One result of this increase in financial innovation was that securities were created that consisted of worthless mortgages. According to Taylor (2009), banks failed to locate the securities that consist of worthless mortgages, and banks still do not know where they are now. “This risk in the balance sheets of financial institutions has been at the core of the financial crisis from the beginning” (Taylor, 2009, p.8). In their paper, Maddaloni and Peydró (2011) explain that securitization activity leads to the construction of assets that may provide attractive returns and that it increases the capacity of banks to provide credit. They measure securitization activity by making use of the growth rate of a type of collateralized short-term debt, called asset-backed commercial paper (ABCP). Covitz, Liang and Suarez (2012) explain that ABCP debt is traditionally issued by commercial short-term debt lending

19 companies in order to finance obligations owed to nonfinancial companies. However, they explain that over time these companies increased their range of financed assets by including highly rated mortgage- and other asset-backed securities, and thus reflecting the mortgage securities consisting of subprime- and other mortgages Taylor (2009) mentions in his paper. Maddaloni and Peydró (2011) empirically show that securitization activity amplified the softening of lending standards by banking institutions.

2.5 Did Bank Risk-Taking Cease to Exist after the Global Financial Crisis of 2007-2008? In 2008, the Fed decided to cut the federal funds rate target to a range between 0% and 0.25% in order to stimulate the economy. This rate is known as the zero-lower bound and it is at this rate, however, that conventional open market operations are ineffective. It is not possible, using open market operations, to further stimulate the economy for the simple reason that the federal funds rate target has reached its lowest bound. The tools used by the Fed to conduct monetary policy therefore needed to change after the global financial crisis of 2007-2008. As a result, the monetary policy stance of the Fed is not accurately reflected by the federal funds rate anymore.

In 2009, when financial conditions began to normalize, the Fed decided to give the economy an extra push by making use of quantitative easing, also known as Large-Scale Asset Purchases. Between 2009 and 2014, there were three rounds of quantitative easing (Labonte, 2016). One purpose of quantitative easing is to lower long-term interest rates by purchasing, for example, long-term Treasury securities, mortgage-backed securities, and agency debt holdings (Labonte, 2016). According to Labonte (2016), the hope is that lower long-term interest rates stimulate interest-sensitive spending in the economy. However, an increasing body of literature after the global financial crisis of 2007-2008, just argued that there is a risk-taking channel active in the monetary transmission mechanism that works

20 through a search for yield effect and a valuation effect. The question that arises is whether quantitative easing may affect the level of risk-taking activity by banking institutions through a softening of lending standards. For example, remember that pension funds and insurance companies have liabilities predefined at long-term rates (Borio and Zhu, 2008). The lower long-term interest rates then imply that it becomes even harder for institutions to reach their liabilities at predefined long-term fixed rates, because lower long-term interest rates suggest lower rates of return in the medium- or long-run. Therefore, lower long-term interest rates may even result in an additional increase in risk-taking activities by, for example, banking institutions. Moreover, lower long-term interest rates also work through to the valuation effect, changing the value of long-lived assets and collateral. So, did banking institutions learn from the global financial crisis of 2007-2008, or did they continue their risk-taking activities by softening lending standards?

First of all, does quantitative easing really lead to lower long-term interest rates? Krishnamurthy and Vissing-Jorgensen (2011) study the effect of the first two rounds of quantitative easing in the U.S. on long-term interest rates. They find empirical evidence that the first two rounds of quantitative easing indeed significantly lowered long-term nominal interest rates. Moreover, Christensen and Rudebusch (2012) empirically study the effect of the first round of quantitative easing in the U.S. on long-term interest rates, and find a negative effect between 50 and 100 basis points on 10-year yields in the fixed-income market.

Labonte (2016) explains that these Large-Scale Asset Purchases of the Fed are financed by bank reserves. During a round of quantitative easing the Fed purchases assets or makes loans. The proceeds are then credited to the recipients‟ reserve accounts at the Fed (Labonte, 2016). Kandrac and Schlusche (2015) state that this increase in the monetary base, comprising of the reserve base and currency in circulation (Ireland, 2005), is a well-known

21 characteristic of quantitative easing. However, banks must hold all the reserves that are initially created by quantitative easing at the Federal Reserve.

The effect of reserve accumulation on bank risk-taking has not been studied extensively, with the exception of Kandrac and Schlusche (2015). In their paper, Kandrac and Schlusche make use of several sources to explain how increases in the reserve base can affect the composition of assets through asset portfolio substitution effects. We mention one source here. Friedman and Schwartz (1975) describe that when banks have larger reserves than previously thought caused by the sale of securities, they will expand investments and loans at a higher rate. Phrased differently, during each of the three rounds of quantitative easing, the Fed purchases assets or makes loans and the corresponding reserve accounts at the Fed of banking institutions are credited. Therefore, the result is that banks have larger reserves than previously thought; hence leading to an expansion of investment and loans at a greater rate than before, according to Friedman and Schwartz (1975). In their paper, Kandrac and Schlusche (2015) provide empirical evidence of the effect of reserve accumulation caused by quantitative easing in the U.S. on lending growth and risk-taking activity by banks. Their results show that within banks‟ portfolios, reserves created by two of the three quantitative easing rounds of the Fed contributed to total loan growth and increased risk-taking in lending activity by banking institutions.

3. Empirical Application

In this chapter we quantify risk-taking activity by U.S. banking institutions by making use of the Senior Loan Opinion Survey. We start this chapter by explaining this survey, what is it exactly and why is it useful? We then provide descriptive statistics on the Senior Loan Opinion Survey and our macro-economic and financial variables, and conclude this chapter

22 by introducing the empirical model we use to assess the relationship between monetary policy and bank risk-taking.

3.1 The Senior Loan Officer Opinion Survey

The two mechanisms of the risk-taking channel, the valuation effect and the search for yield effect can be expected to lead to a softening of lending standards by banking institutions (The Monetary Policy of the ECB, 2011). In this paper, following Maddaloni and Peydró (2011), we quantify bank risk-taking activity by looking at reported changes in lending standards by banking institutions over the past three months using the Senior Loan Opinion Survey (SLOOS). Since 1967, the Fed conducts a quarterly survey on bank lending practices that is available to the public. In this survey, senior loan officers of around sixty banks, supervised by the Fed, are asked to answer approximately twenty questions on the supply and demand for various types of credit (Lown, Morgan and Rohatgi, 2000). The SLOOS is conducted in the months January, April, July, and October, each representing a quarter of a year. However, the SLOOS asks loan officers about changes in their lending standards over the past three

months. This means that the quarterly SLOOS is lagging by one quarter. Therefore, one can

interpret the SLOOS in, for example, the first quarter of 2002, as information on lending standards in the last quarter of 2001. The SLOOS provides qualitative answers. For example, loan officers are asked whether their bank‟s credit standards have tightened over the past three months for: commercial & industrial (C&I) loans or credit lines, consumer loans, and residential mortgage loans. For instance:

Over the past three months, how have your bank’s credit standards for approving applications for C&I loans or credit lines – other than those to be used to finance mergers

23

considerably, 2. tightened somewhat, 3. remained basically unchanged, 4. eased somewhat, 5.

eased considerably.

It is possible to quantify the changes in bank‟s credit standards as the net percentage

tightening of lending standards. Maddaloni and Peydró (2011) define the net percentage

tightening of lending standards as the “the difference between the percentage of banks

reporting a tightening of lending standards and the percentage of banks reporting a softening of standards” (p.2126). However, two of the three mechanisms of the risk-taking channel, the valuation- and search for yield effect can be expected to lead to a softening of lending standards by banking institutions (The Monetary Policy of the ECB, 2011). Therefore, in this paper we assume that it is easier and more appropriate to define and interpret bank risk-taking activity as the net percentage softening of lending standards, reflecting the difference between the percentage banks reporting a softening and the percentage of banks reporting a tightening of lending standards. As a result, a positive figure on the net percentage softening of lending standards indicates that banking institutions have loosened their lending standards over the past three months, thus reflecting an increase in risk-taking activity. Maddaloni and Peydró (2011) note that the net percentage statistic does not differentiate between the degree of tightening or softening of lending standards. For example, respondents can choose between

eased considerably and eased somewhat when answering certain questions. However,

Maddaloni and Peydró (2011) state that the results using different weights for the degree of tightening or softening, do not vary from the results obtained with net percentages.

In their paper, Lown et al. (2000) question the reliability of the SLOOS. They argue that the SLOOS measures „opinions‟ and that it does not provide quantitative answers. Furthermore, the sample size in the survey is small according to Lown et al. (2000). They explain that there are 8,000 banks in the U.S., and yet the survey only asks the opinions of

24 sixty banks that are supervised by the Fed. Accordingly, they argue that this raises the question whether useful information on aggregate lending can be inferred from the answers of loan officers. Moreover, Lown et al. (2000) also explain that a reporting bias may exist because the respondents, supervised by the Fed, may suspect that their answers will be used, by the Fed, for supervisory purposes. However, despite these concerns they find that changes in credit standards reported by loan officers are linked to aggregate loan growth and loan officers provide information that cannot be inferred from other measures of credit availability. Moreover, Lown et al. (2000) find that senior loan officers‟ actions match their words. Their empirical results show that a reported tightening of lending standards is associated with slower loan growth; hence they conclude that loan officers report accurately and that the SLOOS is a useful survey (Lown et al., 2000).

3.2 Results from the Senior Loan Officer Opinion Survey

Table 1 shows the descriptive statistics on changes in lending standards before and after the global financial crisis of 2007-2008. We see that the mean of the net percentage softening of all loans between 2001 and 2008 is negative, between minus twelve and fifteen percent. This means that senior loan officers reported for this period, on average, a tightening of lending standards by banking institutions. The standard deviations on all loans for this period are roughly twice as large in absolute values. Moreover, the minimum values of all loans are large and negative, indicating that banks tightened, or have tightened, their lending standards greatly at a certain point in time. Figure 2 shows a graphical representation of the first half of table 1, that is, for the period between 2001 and 2008. In chapter two, it was noted that the monetary policy stance of the Fed has been too expansionary between 2001 and 2006 (Taylor, 2009). Moreover, two mechanisms active in the risk-taking channel can be expected to lead

25 to a softening of lending standards by banking institutions (The Monetary Policy of the ECB, 2011).

Table 1. Descriptive statistics on changes in lending standards over the past three months

between 2001-2008 and 2009-2014. It should be noted that the Fed adjusted the SLOOS in 2007, asking questions about changes in lending standards of prime mortgage loans, subprime mortgage loans, and nontraditional mortgage loans, instead of one consolidated mortgage class. Moreover, since the second quarter of 2011, changes in standards for consumer loans and auto loans, excluding credit card loans, are also reported separately. However, we treat these questions as one for the period between 2001 and 2008, summing up the number of banks reporting a net softening of lending standards and divide this by the combined number of responding banks. Source: Senior Loan Officer Opinion Survey

From figure 2, it can indeed easily be seen that lending standards by banking institutions as reported by senior loan officers softened between 2001 and 2006. Commercial and industrialized loan standards softened rapidly between 2001 and 2006, while standards for consumer- and residential mortgage loans softened to a lesser extent. On the other hand, during the global financial crisis of 2007-2008, one can see an enormous increase in the

2001:Q1-2008:Q4 2009:Q1-2014:Q4

Net Percentage

Softening of Mean

Std.

Dev. Min Max Mean

Std.

Dev. Min Max

C&I LOANS LARGE C&I LOANS SMALL

CONS. LOANS MORT. LOANS PRIME MORT. LOANS

NT. MORT. LOANS -13.05 -12.78 -13.13 -14.78 - - 30.05 26.42 20.28 26.71 - - -83.60 -74.50 -67.40 -78.65 - - 24.10 24.10 9.8 9.4 - - 5.18 1.80 2.95 - -2.30 -10.95 14.56 14.18 14.98 - 14.05 17.43 -39.60 -42.30 -47.10 - -49.00 -64.00 21.80 23.10 15.96 - 18.30 6.30

Variables Definition: Changes in credit standards for

C&I LOANS LARGE C&I LOANS SMALL

CONS. LOANS MORT. LOANS PRIME MORT. LOANS

NT. MORT. LOANS

C&I loans of large and middle market banks (annual sales ≥ $50 million) C&I loans of small banks (annual sales < $50 million)

Consumer loans and auto loans excluding credit card loans. All residential mortgage loans

Prime rate mortgage loans e.g., fixed rate, standard adjustable rate mortgages Nontraditional mortgage loans e.g., mortgages with limited income

26 tightening of lending standards by banking institutions as reported by senior loan officers for several quarters.

Figure 2. Net percentage softening of C&I loans for large, middle and small firms, consumer

loans and all residential mortgage loans for the period between 2001:Q1 and 2008:Q4

The second half of table 1 shows the descriptive statistics on changes in lending standards between 2009 and 2014. The reason why we exclude the consolidated mortgage class and include the prime- and nontraditional mortgage class for this period is due to the fact that the SLOOS suffers from missing observations for the subprime mortgages. Therefore, in order to look at residential mortgage lending activity during this period, we look at the softening standards for the remaining two mortgage classes: prime- and nontraditional mortgage loans. The mean of the net percentage softening of large- and small C&I loan standards and consumer loan standards are positive, and approximately five percent, two percent, and three percent respectively. This indicates that senior loan officers reported, for these loans on average, a softening of lending standards between 2009 and 2014. The

27 standard deviations and minimum and maximum values of the net percentage softening of large- and small C&I loans and consumer loans are roughly similar to each other. The average of the net percentage softening of prime- and nontraditional mortgage loans is negative, and roughly minus two percent and minus eleven percent respectively. In other words, loan officers report, on average a tightening of lending standards for residential mortgage loans for the period between 2009 and 2014. Figure 3 graphically shows the softening of lending standards by banking institutions between 2009 and 2014, the period characterized by quantitative easing. One can see that lending standards as reported by loan officers softened dramatically after the global financial crisis of 2007-2008, when bank reserves were created by the Fed, and remained roughly constant afterwards.

Figure 3. Net percentage softening of C&I loans for large, middle and small firms, consumer

loans and prime- and nontraditional mortgage loans for the period between 2009:Q1 and 2014:Q4

28 3.3 Macroeconomic and Financial Variables

Table 2 shows the descriptive statistics on the macroeconomic and financial variables that will be used in our regression model. The variables are: the federal funds rate, the growth rate of reserves held at the Fed, the rate of inflation and real GDP growth, the growth rate of outstanding asset-backed commercial papers, and the number of consecutive quarters characterized by a monetary policy rate that deviates from an estimated Taylor rule. In chapter two we have discussed these variables and their theoretical importance. The federal funds rate approximates the monetary policy stance of the Fed. Moreover, one effect of quantitative easing in the U.S. is that bank reserves maintained at the Fed increases. For example, according to for example Friedman and Schwartz (1975), an increase in the amount of reserves by banking institutions maintained at the Fed may lead to an increase loan growth and the expansion of investment. As a result, we are interested in the relationship between the growth rate of bank reserves maintained at the Fed and the change in lending standards by banking institutions. The rate of inflation and real GDP growth both control for cyclical movements in macroeconomic conditions (Gambacorta, 2005). Moreover, securitization activity leads to the construction of assets that may provide attractive returns and that it increases the capacity of banks to provide credit (Maddaloni and Peydró, 2011). Securitization activity is measured by outstanding asset-backed commercial paper. Finally, we also measure the number of consecutive quarters characterized by a monetary policy rate that deviates from an estimated Taylor rule, as suggested by Maddaloni and Peydró (2011). This is done in order to examine the potential effect of a prolonged expansionary monetary policy stance on the softening of bank lending standards, and consequently on bank risk-taking activity.

The first half of table 2 shows descriptive statistics for the period between 2001 and 2008. The average of the federal funds rate during this period is almost three percent, and the

29 standard deviation is around one-and-a-half percent. Average inflation and real GDP growth rates are around three percent and two percent respectively, with standard deviations equaling one percent and one-and-a-half percent respectively. The average growth rate of asset-backed commercial paper is approximately five percent, and a standard deviation that is more than three times larger.

Table 2. Descriptive statistics on macroeconomic and financial variables for the period

between 2001-2008 and 2009-2014

Taylor (2009) argued that the federal funds rate was too low for too long between 2001 and 2006. In order to evaluate this statement by Taylor (2009), Maddaloni and Peydró (2011) perform an OLS regression of the federal funds rate on real GDP growth and inflation, and look at the residual of this regression. They state that a positive residual of this regression corresponds to a monetary policy stance that is too tight, and that a negative residual

2001:Q1-2008:Q4 2009:Q1-2014:Q4

Variables Mean Std.

Dev. Min Max Mean Std. Dev. Min Max

FFR RBDEP INFL GDP ABCPO LOWMP 2.89 - 2.84 2.11 5.04 2.97 1.65 - 0.98 1.49 16.57 7.06 0.51 - 1.25 -2.77 -32.34 -9 5.59 - 5.30 4.41 29.18 16 - 922.34 1.60 1.25 -16.89 0.54 - 2452.39 1.23 2.03 8.76 3.35 - -6.85 -1.62 -4.06 -35.61 -5 - 8201.21 3.76 3.08 -4.30 7 Variables Definition FFR RBDEP INFL GDP ABCPO LOWMP

Quarterly average of the federal funds rate determined in the federal funds market. Source: Reuters DataStream

Quarterly results for year-on-year changes in the growth rate of bank reserves held at the Fed. Source: Reuters DataStream

Quarterly results for year-on-year changes in the U.S. inflation rate. Source: Reuters DataStream

Quarterly results for year-on-year changes in the real GDP growth rate. Source: Reuters DataStream

Quarterly results for year-on-year changes in growth rate of asset-backed commercial paper outstanding. Source: Reuters DataStream

Number of consecutive quarters of federal funds different than the rate suggested by Taylor rule. Source: Author‟s Calculations

30 corresponds to monetary policy stance that is too loose. In other words, when the residual of the regression of the federal funds rate on real GDP growth and inflation is negative, the monetary policy stance of the Fed is considered to be too expansionary, and vice versa. Using the method suggested by Maddaloni and Peydró (2011), we indeed find evidence that the federal funds rate can be considered as too low between 2001 and 2006 (see figure 4). Maddaloni and Peydró (2011) count the number of consecutive quarters in which the residual of the regression of the federal funds rate on the rate of inflation and real GDP growth deviates from zero. Our variable LOWMP counts a total of 12 quarters in which the residual was positive, and a total of 20 quarters in which the residual was negative for the period between 2001 and 2008. Phrased differently, the monetary policy stance of the Fed was too accommodative for sixteen consecutive quarters between 2001 and 2006, and too tight for nine consecutive quarters between 2006 and 2007 (see table 2 and figure 4).

Figure 4. Author‟s calculations of the periods characterized by a federal funds rate below the

rate suggested by the estimated Taylor rule between 2001 and 2008, inspired by Maddaloni and Peydró (2011)

31 Looking at the period between 2009 and 2014 in table 2, where the variable FFR is excluded since the federal funds rate was at the zero-lower bound during this period, and where the Fed created bank reserves via quantitative easing. We see that the average growth rate of bank reserves held at the Fed during this period is 922.34 percent, with a standard deviation of 2452.39, a minimum value of -6.85 percent and a maximum value of 8201.21 percent. The average inflation and real GDP growth rates are around one-and-a-half and one-and-a-quarter percent. The average growth rate of asset-backed commercial paper outstanding is around minus seventeen percent, with a standard deviation that is almost two times as small in absolute terms. Between 2009 and 2014, the variable LOWMP counts a maximum of five consecutive quarters in which the residual was positive, and a maximum of seven consecutive quarters in which the residual was negative. Indicating that the monetary policy stance of the Fed, suggested by an estimated Taylor rule, was too tight for five consecutive quarters and too accommodative for seven consecutive quarters.

3.4 Empirical model

In this paper we empirically examine the potential relationship between bank risk-taking activity, approximated by the softening of lending standards, and the monetary policy stance of the Fed before and after the global financial crisis of 2007-2008. The pre-crisis period is between 2001 and 2008, capturing a full economic cycle and characterized by a monetary policy stance that is considered too loose for far too long. The period between 2009 and 2014 is characterized by quantitative easing, an unconventional monetary policy tool that caused a rapid expansion of reserves created by the Fed. We examine whether the strong increase in the reserves of depository institutions created by the Fed may contribute to a softening of lending standards by banks, and may consequently lead to an increase in bank risk-taking

32 activity. Our empirical model will be based on the model that is used in the paper by Maddaloni and Peydró (2011), and it takes the following form:

NPSt = α +β*MPt-1 + δ*INFL t-1 + λ*GDP t-1 + η*ABCPOt-1 + θ*LOWMPt-1 + εt

Where the dependent variable NPS is the net percentage of banks that have softened their lending standards for commercial and industrialized, consumer, and residential mortgage loans in quarter t. The variable MP approximates the monetary policy stance of the Fed at time t-1, approximated by the federal funds rate for the period between 2001 and 2008 or the percentage change in the reserves of depository institutions held by the Fed for the period between 2009 and 2014. The variables INFL and GDP denote the inflation rate and the real GDP growth at time t-1, and control for cyclical movements in macroeconomic conditions.

ABCPO is the growth rate of asset-backed commercial paper at time t-1, and is included in

the regression to measure securitization activity. The variable LOWMP quantifies the amount of consecutive quarters in which the federal funds rate is below the rate suggested by an estimated Taylor rule at time t-1, as suggested by Maddaloni and Peydró (2011). We estimate our model using OLS and all of our independent variables are lagged by one quarter. We do this because it presumably takes some to time to adjust lending standards by banking institutions in response to changes in macro-economic conditions.

4. Empirical Results

In this chapter we show our empirical findings, based on the empirical model constructed in the previous chapter. The discussion of the empirical results is divided into two sections, one section shows the results before the global financial crisis of 2007-2008, and one section shows the results after.

33 4.1 Bank Risk-Taking before the Global Financial Crisis of 2007-2008

Table 3 shows our regression results for the period between 2001 and 2008. At first glance table 3 shows that for all loans and ceteris paribus, a one-unit decrease in the federal funds rate leads to a softening of lending standards by banking institutions, our measure of bank risk-taking activity.

Table 3. OLS regression results for the period between 2001 and 2008. Standard errors are

within parentheses and *, **, ***, denote the significance levels at 10%, 5%, and 1% respectively

2001:Q1-2008:Q4

Dependent variables- Net percentage softening of lending standards of:

1 2 3 4

Variables C&I LOANS

LARGE

C&I LOANS

LARGE CONS. LOANS MORT. LOANS

CONSTANT FFRt-1 INFLt-1 GDPt-1 ABCPOt-1 LOWMPt-1 R-Squared -31.38*** (8.79) -8.66** (3.21) 2.27 (3.25) 13.87*** (2.28) 1.15*** (0.22) 0.21 (0.59) 0.88 -19.89*** (7.08) -8.04*** (2.59) 0.30 (2.61) 10.26*** (1.84) 1.19*** (0.18) 0.24 (0.47) 0.90 -20.90*** (5.47) -5.84*** (2.00) 1.52 (2.03) 5.92*** (1.42) 1.12*** (0.14) 0.34 (0.37) 0.90 -7.02 (10.09) -14.80*** (3.69) 7.29* (3.73) 1.43 (2.62) 1.76*** (0.26) 0.48 (0.68) 0.80 Variables Definition FFRt-1 INFLt-1 GDPt-1 ABCPOt-1 LOWMPt-1

Lagged quarterly federal funds rate determined in the federal funds market Lagged quarterly year-on-year changes in the U.S. inflation rate

Lagged quarterly year-on-year changes in the real GDP growth rate Lagged quarterly year-on-year changes in growth rate of asset-backed commercial paper outstanding

Lagged number of consecutive quarters of federal funds rate obtained from an estimated Taylor rule