WPS6667 Policy Research Working Paper 6667 The Connection between Wall Street and Main Street Measurement and Implications for Monetary Policy Alessandro Barattieri Maya Eden Dalibor Stevanovic The World Bank Development Research Group Macroeconomics and Growth Team October 2013 Policy Research Working Paper 6667 Abstract This paper proposes a measure of the extent to which real economy to monetary shocks. They present a stylized a financial sector is connected to the real economy. model that illustrates how interbank trading can reduce The Measure of Connectedness is a measure of the the sensitivity of lending to the entrepreneur’s net worth, composition of assets, namely the share of credit to thereby dampening the credit channel transmission the non-financial sectors over the total credit market of monetary policy. The Measure of Connectedness is instruments. The aggregate Measure of Connectedness interacted with both a structural vector autoregressive for the United States declines by about 27 percent in the model and a factor-augmented vector autoregressive period 1952–2009. The authors suggest that this increase model for the United States economy. The analysis in disconnectedness between the financial sector and the establishes that the impulse responses to monetary policy real economy may have dampened the sensitivity of the shocks are dampened as the level of connection declines. This paper is a product of the Macroeconomics and Growth Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at meden@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team The Connection Between Wall Street and Main Street: Measurement and Implications for Monetary Policy∗ Alessandro Barattieri† Maya Eden ‡ Dalibor Stevanovic § October 16, 2013 JEL classification: E44, E52, G20 Keywords: Connection, financial sector, real economy, monetary policy transmission mechanism ∗ We would like to thank Steve Ambler, Susanto Basu, Ben Eden, Patrick F` eve, Jeffrey Frieden, Marc ephane M´ Giannoni, Alain Guay, Michel Juillard, Sydney C. Ludvigson, Jean-St´ esonnier, Kevin Moran, Louis Phaneuf, Thomas Philippon, Franck Portier, Federico Ravenna, Luis Serven, and seminar participants at ESG UQAM, ` Bank of France, EUI, Atelier en macro´ ´ economie CIRPEE-DEEP-TSE 2013 and 2013 CEA annual conference for useful comments and suggestions. Elisa Martin provided outstanding research assis- tance. This paper does not necessarily reflect the views of the World Bank, its Executive Directors or the countries they represent. † Corresponding Author ESG UQAM and CIRPEE. ´ Mail: Case Postale 8888, succursale Centre-ville Montreal (Quebec) H3C 3P8. Tel: +1-514-987-3000 (0850#). Fax: +1-514-987-8494. E-mail: barat- tieri.alessandro@uqam.ca ‡ The World Bank. Macroeconomics and Growth Research Group. 1818 H st. NW. Washington, DC. Tel: +1-857-246-9722. E-mail: meden@worldbank.org § ESG UQAM, CIRPEE ´ and CIRANO. Mail: Case Postale 8888, succursale Centre-ville Montreal (Que- bec) H3C 3P8. Tel: +1-514-987-3000 (8374#). Fax: +1-514-987-8494. E-mail: dstevanovic.econ@gmail.com 1 1 Introduction Two facts constitute the background of this paper. First, the U.S. financial system underwent a radical transformation during the last decades. The complexity and the nature of the process of financial intermediation changed substantially.1 Figure 1 confirms this well known phenomenon by reporting the evolution of the share of total assets in the U.S. economy held by three major groups of actors: i) the traditional actors (commercial banks, savings institutions and credit unions), ii) the insurance, pension and mutual funds, and iii) the so called “shadow banking system” (Government Sponsored Enterprises (GSE), Assets-backed- securities issuers, GSE mortgage pools, finance companies, brokers and dealers).2 While the share of assets held by the traditional actors declined from about 60% to roughly 30% from 1952 to 2010, the share of assets held by the “new” actors increased from almost zero to more than 40% in 2006. Second, a well known result in the economic literature is that in more recent samples, the sensitivity of real variables to monetary policy shocks has declined. A common explanation for this empirical finding is an increase in the effectiveness of monetary policy, as for example proposed by Boivin and Giannoni (2006). Another frequently conjectured (but less studied) hypothesis is that structural changes in the financial sector contributed to the changing nature of the monetary policy transmission mechanism.3 Arguably, part of the difficulty in addressing this hypothesis is the lack of a suitable measure of the structural transformation that affected the U.S. financial system.4 In this paper, we propose such a measure and study its implications for monetary policy. The Measure of Connectedness is a measure of composition of the assets, namely the share of the credit to the non-financial sectors (households, non-financial firms and government) over the total credit market instruments. The aggregate U.S. Measure of Connectedness 1 See Gorton and Metrik (2012). 2 See Adrian and Shin (2010), Poznar et al (2012) and references therein for a comprehensive explanation of the concept of Shadow Banking. 3 A notable exception is Dynan et al (2006), who analyses the impact of monetary policy on real activity before and after relevant regulatory changes. 4 Contributions in this literature analyze the different responsiveness across different sub-samples of the data. 2 declines by about 27% in the period 1952-2009. The increase in the disconnection between the financial sector and the real economy implies a larger interconnectedness within the financial sector. We suggest that this increase in interconnectedness within the financial sector may have dampened the sensitivity of the real economy to monetary shocks. We present a stylized model that illustrates how interbank trading can reduce the sensitivity of lending to the entrepreneur’s net worth, thereby dampening the credit channel transmission of monetary policy. We then present some empirical evidence consistent with the predictions of the model. Importantly, we take here an entirely positive perspective. We are not addressing the question of whether a change in the connection between the financial sector and the real economy is good or bad.5 We proceed in three steps. First, we propose as measure of the structural transformation of the financial sector, the extent to which it is connected to the real economy. The Measure of Connectedness is the share of the credit market instruments represented by claims whose direct counterpart belongs to the non-financial sectors (households, non-financial firms and government). We compute the measure of connectedness for each major player in the U.S. financial system for the period 1952-2009, as well as an aggregate measure, using data from the Flows of Funds. The 27% drop in the aggregate measure over the period 1952-2009 is mostly driven by a composition effect, namely the rising importance of financial institutions that are relatively “disconnected” from the real economy (such as the issuers of Assets- Backed-Securities).6 Second, we propose a stylized model to illustrate how interbank trading, by increasing the liquidity of investment projects, reduces the sensitivity of lending to monetary policy shocks. In particular, we focus on the balance sheet transmission mechanism of monetary policy. According to this channel, changes in the nominal rate affect the net present value 5 We do think that this is a very interesting question and we will explore it in our future research. 6 We also investigate the relation between our measure and the share of finance in U.S. GDP reported by Philippon (2012). While they capture very different concepts, we interestingly find a very high correlation between a series equal to one minus our measure of connectedness and the share of finance in non-defence U.S. value added. 3 of the borrower’s debt obligations and hence his net worth; through various contractual fric- tions, the change in net worth influences the expected return to lending (for more on this transmission mechanism, see Bernanke and Gertler, 1995). We show that this transmission mechanism is mitigated as investment projects become more liquid. The key to this result is as follows: absent an interbank market for investment projects, banks face a tradeoff between investment and liquidity. Consequently, changes in the expected return to investment (for example, through changes in the net worth of the borrower) may change the tradeoff be- tween investment and liquidity, thus affecting the equilibrium amount of investment. In the presence of a liquid interbank market for investment projects, there is effectively no tradeoff between investment and liquidity: since all projects have positive net present values (NPV), all projects are implemented, regardless of small changes in their NPV induced by changes in monetary policy. Finally, we interact our aggregate measure of connectedness with a structural vector auto-regression (VAR) for the U.S. economy, and produce impulse responses to a monetary policy shock conditional on different levels of connection. We check the robustness of our results also using a Factor-Augmented VAR (FAVAR) model. We find that the responses of the real variables to a monetary policy shock are dampened as the level of the connection between the financial sector and the real economy decreases. This paper is linked to several strands of the literature. First, it is related to the litera- ture dealing with measurement of financial intermediation and its characteristics. Philippon (2012) provides evidence on the quantitative importance and the cost of financial interme- diation in the U.S. in the last 130 years. Greenwood and Scharfstein (2013) analyze the growth of the share of finance on GDP in the U.S. while Philippon and Reshef (2013) ana- lyze the growth of the share of finance for several developed countries. Philippon and Reshef (forthcoming) propose evidence on the evolution of the wages in the financial industry for the period 1909-2006.7 Second, this paper is related to the analysis of the relation between financial firms and the real economy. Mesonnier and Stevanovic (2012) use micro-level financial data to construct 7 See also the survey on Financial Intermediation by Gordon and Winton (2003). 4 an aggregate leverage shock and explore its implications for the real variables. Diebold and Yilmaz (2012, 2013) measure the connectedness within financial firms and its contribution to shock transmission. Brunnermeier et al. (2012) analyze the banks’ non-interest income and how this affects their contribution to systemic risk. Hahm, Shin and Shin (forthcoming) analyze the non-core liabilities of Korean banks and their implications for financial vulner- ability.8 We see this paper as complementary to that line of research, since we focus our attention on the composition of assets. Our Measure of Connectedness, in fact, could be interpreted as a measure of the core assets of financial institutions. Finally, the paper is related to the literature on the monetary policy transmission mecha- nism. Boivin and Giannoni (2006) report evidence that the effects of monetary policy shocks on real variables are muted in the post-1980 period, and show how this finding can be ex- plained by an increase in the effectiveness of monetary policy. Boivin et al (2011) report FAVAR evidence as well as evidence from DSGE modeling on the change over time of the monetary transmission mechanism. Confirming the results by Boivin and Giannoni (2006), they also find muted responses of real variables to monetary policy innovations in more re- cent times, and argue that this is mostly accounted for by changes in policy behavior and the effect of these changes on expectations. Adrian and Shin (2011b) consider more in general the role of financial intermediaries in monetary economics.9 Closer to our spirit, Dynan et al. (2006) present evidence of the reduced responsiveness of several economic aggregates to shocks, dividing the sample before and after important regulatory changes. We contribute to this literature by providing a measure that can capture the structural transformation that the U.S. financial sector underwent and propose a model that rationalizes the decreased sensitivity of real activity to monetary policy due to this structural transformation. The paper is organized as follows. In Section 2 we introduce our measure of connectedness between the financial sector and the real economy and document its evolution in the U.S. In Section 3 we outline our stylized theoretical model. In Section 4 we show some empirical 8 See also the discussion of non-core liabilities contained in Adrian and Shin (2011a) 9 A recently proposed complementary channel through which changes in the financial conditions can affect the transmission mechanism of monetary policy is the “risk taking channels”, proposed by Borio and Zhu (2012). See also Bruno and Shin (2013). 5 evidence consistent with the theory. Section 5 concludes with several suggestions for future research. 2 The measure of connectedness Our proposed measure of the connection of a financial sector (or a financial institution) to the real economy is based on the composition of its assets. While the exact formula depends on the type of financial institution analyzed, the Measure of Connectedness is conceptually the share of the credit market instruments represented by claims whose direct counterparts belong to the non-financial sectors (households, non-financial firms and government): CREDIT REALit CON Nit = (1) CREDITit where CREDITit represents the total amount of credit market instruments of a given insti- tution i at time t, while CREDIT REALit represents the credit market instruments whose direct counterparts are households (i.e. mortgages and consumer loans), non-financial firms (i.e. commercial loans) or the government (i.e. treasuries). Such a measure can be computed for each player in the U.S. financial system using data from the Flows of Funds. Probably the most important drawback of using flow of funds data is that we are not able to say much about non-balance sheet items, such as derivatives. Since derivatives are typically used as the main example of the disconnectedness between the financial sector and the real economy, we are aware that we are missing an important piece of information, and we therefore consider our results suggestive of an upper-bound to the level of connectedness. However, an advantage of using flow of funds data is that the measure can be computed for a long time series and hypothetically for different financial systems. Moreover, a conceptually similar measure could be computed also for a single financial institution using balance sheet data. We stress that we try to measure the share of credit market instruments whose direct counterparts are in the non-financial sector. As many commentators observed, the transfor- 6 mation of the financial intermediation process from the “traditional” banking model to the “originate and distribute” model produced a lengthening of the chain linking the ultimate lender to the ultimate borrower (Adrian and Shin, 2011a). This lengthening may have had several consequences, including the erosion of lending standards (Keys et al, 2010), which likely played an important role in the crisis that started in 2007.10 In what follows we describe the construction of the measure for each element of the U.S. financial system for the period 1952Q1-2009Q4.11 We use data from the release Z1 of March 8, 2012. A Data Appendix contains details on the composition of the measure for different players within the financial system. Traditional financial institutions. Traditional financial institutions include Commercial Banks, Savings Institutions and Credit Unions. For commercial banks, at the numerator of our measure we use the sum of Open Market Papers, Treasury Securities, Non-securitized GSE-backed securities, Municipal securities, Non-securitized Corporate Bonds, Bank Loans, Mortgages, Security Credit and Consumer Credit. At the denominator, we use the total amount of credit market instruments12 . For Savings Institutions, we use a similar measure.13 For Credit Unions, the Flows of Funds table does not distinguish between securitized and not-securitized bonds. By including all bonds in the numerator of our measure, once again we are conservative and we accept the risk of over-estimating the connectedness with the real economy. Figure 2 reports the results obtained for the measure of connectedness of the traditional players. All the three indicators are falling over time, from values very close to 1 to values between 0.77 (for the Credit Unions) and 0.84 (Commercial Banks). 10 Moreover, this lengthening of the credit chain might have increase the vulnerability of the system to sudden drop in trust, which arguably was an important element in explaining the freeze of credit markets in the fall of 2008. See Gennaioli, Shleifer, and Vishny (2012) for an example of a model where trust between actors play an important role in the intermediation process. 11 A change in the accounting rule governing the GSE at the beginning of 2010 would introduce a spurious element in our calculations. 12 Flow of Funds series FL.724.005.005, table L.110 13 See appendix for details. 7 Insurances, mutual funds and pension funds. As in the case of Credit Unions, in the case of insurance, pension funds and mutual funds, the tables in the Flow of Funds do not distinguish between asset backed securities and other corporate bonds. Once again, the results need to be interpreted as an upper bound for the connection. Insurance companies can be divided into Property-Casualty Insurance Companies and Life Insurance Companies. As for the mutual funds, we only consider the two most important categories: mutual funds and money market mutual funds (MMMF). Figure 3 reports the results we obtain. Property and life insurance appear to remain fairly connected to the real economy all along the period considered. In contrast, we witness a more pronounced increase in the disconnectedness in the case of mutual funds. Interestingly, the reduction in the connectedness of the MMMF seems to start later, but to have a large drop in the years preceding the crisis, where the MMMF balance sheets were increasingly filled with complex securities like mortgage backed securities (MBS). Figure 4 shows the evolution of the connectedness over time of Pension Funds. We analyze separately private pension funds, local government pension funds, and federal government pension funds. Here the results are starkly different. While private pension funds appear to have become much more disconnected over time from the real economy (starting from the 1970s), the federal government pension funds barely changed, while the local government pension funds started becoming more disconnected during the 1990s. Shadow banking. We consider the connectedness of Government Sponsored Enterprises (GSEs), Brokers and Dealers, Finance Companies and ABS issuers. Finance companies are fully connected to the real economy, and hence we assign a value of 1 to their measure of connectedness. Figure 5 shows our results for the other actors within shadow banking. Not surprisingly, the Brokers and Dealers appear to be fairly disconnected from the real economy ever since they appear in the 1980s. Even more so is the case for the ABS issuers.14 14 The big drop observed in the series is due to the fact that in the first observation available for ABS issuers in the Flow of Funds (1983q2) the only asset class recorded is trade credit. We consider trade credit as being part of the credit to the real economy, hence the Measure of Connectedness of ABS issuers for that quarter is equal to one. 8 More surprising is probably the evolution of the connectedness of the GSEs, who display a downward trend in the 1990s, when they started to invest also in structured financial products. We exclude from the analysis the Agency and GSE-backed mortgage pools. The reason is to avoid double counting.15 Aggregate measure. In order to have an aggregate picture of the connectedness of the U.S. financial system to the real economy, we take a weighted average of each measure of connectedness, weighting each actor by its time-varying share of credit market instruments: CON NU S,t = ωi,t ∗ CON Ni,t (2) i CREDITi,t where ωi,t = CREDITU S,t and i = {CB, SI, CU, P I, LI, M F, P F, GSE, BRO, F IN, ABS }. Figure 6 reports the result. As the figure shows, the aggregate drop in connectedness is of the order of 27% during the period 1952-2009. It is interesting to notice how the aggregate result we obtain for the measure of connectedness is, to a certain extent, due to a composition effect. Figure 7 shows the shares of credit market instruments held by the different types of financial institutions. Similar to Figure 1, the shadow banking share increased greatly in the last 30 years, while the share of the traditional players dropped significantly. Figure 8 reports the results on the share of credit market instruments for two particularly relevant players: commercial banks and ABS issuers. As the figure shows, the drop in the share of credit market instruments held by commercial banks is coupled with an important increase in the share of the ABS issuers, which then dropped substantially during the crisis. It is also interesting to note how our measure of connectedness shows some relation with key moments in the history of the deregulation of the U.S. financial system. The measure has a clear change in trend in the early 1980s, when several deregulation acts were promoted in 15 The Agency and GSE backed securities, in fact, are among the assets that we consider to contribute to the disconnectedness of all the other players in the financial system. An alternative treatment would have been to include this category and assign arbitrarily to this actor a level of connection equal to zero by definition. Once again, we have been conservative and decided to present an upper bound estimate of the level of connection. 9 the U.S.16 . Moreover, the aggregate easure of connectedness seems to accelerate its downward trend for few years, starting in 1999. The Financial Modernization Act, also known as the Gramm-Leach-Bliley Act, passed in 1999, repealed the Glass-Steagall of 1933 and removed the separation between the activities of commercial banking and investment banking, thus spurring a wave of mergers and acquisitions in the U.S. financial sector and leading to a transformation of the business model in several U.S. financial institutions. Finally, it is important to stress again that here we take an entirely positive perspec- tive, and we are not able to say anything about the welfare implications of the increased disconnectedness between the financial sector and the real economy. Measure of connectedness and the share of finance in GDP. It is instructive to investigate the relation between our measure of connectedness and the share of finance in GDP, constructed by Philippon (2012). In Figure 9, we report both a series equal to one mi- nus our measure of connectedness, which is a measure of disconnection between the financial sector and the real economy, and the share of finance in non-defense value added.17 While the two series are conceptually different, they are interestingly highly correlated.18 Our measure of connectedness is a way of representing the structural transformation that affected the U.S. financial system in the last 50 years. Philippon (2012) measures the share of finance in U.S. GDP. One could conjecture that the structural transformation of the U.S. financial sector captured by our measure might have contributed to a reallocation of resources towards finance, thus implying a greater share of finance in GDP. However other factors, such as a sector-specific technological change or the increasing trend toward financial 19 globalization, might also help explaining Philippon’s findings. 16 For Instance the Depository Institutions Deregulation and Monetary Control Act in 1980 and the Garn- St Germain Depository Institutions Act, in 1982. 17 While Philippon’s data are at annual frequency, we interpolated them to transform them into a quarterly series. 18 The correlation between the two series is 0.98. 19 Another recent explanation of Philippon’s results can be found in Gennaioli, Shleifer and Vishny (2013), who propose a Solow-type growth model augmented with a financial intermediation process featuring a role for trust. In their model, the share of finance over GDP grows over time due to the role of financial intermediation as a tool for wealth management in an environment where the ratio of wealth to GDP grows as the economy approaches its steady state. 10 Since this paper focuses mainly on the implications for monetary policy of the structural transformation that affected the U.S. financial system, we focus in what follows on our measure of connectedness, without taking a strong stance on its contribution to the increase in the share of finance in GDP. 3 Connectedness and monetary policy: Theory We present a stylized model that captures a possible relationship between the connectedness of the financial sector and the sensitivity of real activity to monetary policy. We focus on the credit channel of transmission of monetary policy, and show how a financial sector more disconnected from the real economy implies a lower sensitivity of lending to interest rate changes. The basic environment is similar to Diamond and Dybvig (1983) and Allen and Gale (2000), with some ex-ante uncertainty. There are three periods, t = 0, 1, 2. At time 0, a continuum of banks are endowed with 1 unit of deposits (deposits here are time 0 goods). At that time, a fraction q0 of banks are also endowed with an investment opportunity: an investment of I units of deposits at t = 0 (where 0 ≤ I ≤ 1) yields a return of AI α at t = 2. An investment opportunity can be seen as a request of a loan by an entrepreneur. The parameter A is the recoverable part of output, which in principle changes with the entrepreneur’s net worth. In the background, we can think of an entrepreneur with some initial debt obligations, and some net worth (N ). The parameter A depends on his net worth N and therefore is decreasing with the NPV of his debt obligations (decreasing in r, the interest rate set by the monetary authority).20 We assume that any enforceability problems are already summarized in A, the recoverable part of output. The banks observe the realization of A before making their lending decision. In the background, there is a transmission mechanism of monetary policy: A is increasing in the net worth of the entrepreneur N , and N is decreasing in r. We thus focus on the 20 We do not take a strong stance on the reason why A might depend on the net worth of the entrepreneur. A possible interpretation is that the entrepreneur puts more effort if his net worth is higher. Formally, we are assuming a function A(N (r)) with N (r) < 0 and A > 0. 11 balance sheet channel, as described by Bernanke and Gertler (1995). All banks - those who receive an investment opportunity and those who do not - have access to a storage technology, that transforms one unit of output at time t to one unit of output in time t + 1. For banks without an investment opportunity, this is the only technology. Banks with an investment opportunity choose between saving in storage and investing. The bank’s time 0 objective function is: E (θc1 + c2 ) (3) where c1 is consumption goods at time 1 and c2 is consumption goods at time 2. The ¯ > 1 with probability q and value θ = 1 parameter θ is a liquidity shock, that takes value θ ¯ > 1 can be thought of as whatever might with probability 1 − q . A realization of θ = θ induce banks to prefer liquidity at time 1 as opposed to time 2. We first solve the model under the assumption that there is no interbank market. Given our definition of connectedness, this is equivalent to a situation of full connectedness, as the banks only deal with entrepreneurs. We will then allow for the possibility of an interbank market that can be thought of as a partially connected environment. It is useful to keep in mind, given our definitions, that a greater disconnectedness between the financial sector and the real economy implies a greater interconnectedness within the financial sector. The fully connected environment. Absent trade between banks, banks choose the amount of storage and investment to maximize their objective function. For banks that do not have an investment opportunity, the problem is trivial: they will store all deposits; if, ¯, they will consume all goods at t = 1. Otherwise, they are at period 1, they find that θ = θ indifferent between time 1 and time 2 consumption. For banks with an investment opportunity, the bank’s problem can be written as: max E (θc1 + c2 ) (4) I ∈[0,1] 12 s.t. c1 = 1 − I (5) c2 = AI α (6) ¯, the bank strictly prefers to consume its storage rather than to store it Note that, if θ = θ for period t = 2. If θ = 1, the bank is indifferent between the two options, so we can assume without loss of generality that it consumes all stored goods at time t = 1. Substituting in the constraints, the problem amounts to: max E (θ(1 − I ) + AI α ) (7) I ∈[0,1] Assuming an interior solution, the first order condition is: α α 1 E (θ) = αAI α−1 ⇒ I 1−α = A⇒I=( A) 1−α (8) E (θ) E (θ) Note that equation (7) admits an interior solution only if E (θ) > αA. In other words, if the expected valuation of liquidity at t = 1 is sufficiently high, it is optimal for the bank to store some deposits. Assuming that this parametric restriction is satisfied, the interior solution is increasing in A, so there is some transmission of monetary policy (that changes A(N )). In contrast, at the corner solution small changes in A (or in N , or in r) would not change the bank’s investment decision, and there would be no sensitivity to monetary shocks. Formally, we define the sensitivity of real activity with respect to monetary policy in this model to be the semi-elasticity of aggregate investment to interest rate changes. We decompose this semi-elasticity as follows: ∂lnI ∂lnI ∂A ∂N = (9) ∂r ∂A ∂N ∂r We assume the second term to be positive and the third term to be negative. Then, from equation (8), we get: 13 1 ∂ AGG ∂ln(Ic ) ( α A) 1−α ∂A E (θ) = 1 (10) ∂A ( Eα A) 1−α (θ) AGG where Ic is the aggregate investment in the fully connected environment. The partially connected environment. Assume next that a fraction λ of investment opportunities are tradable at time t = 1. This is a very reduced form way of introducing securitization in the model.21 The securitization is the process through which an illiquid asset can be sold by the bank to a special investment vehicle (SIV). The SIV transform the illiquid asset into securities, by issuing bonds with stratified risk profiles. The banks, at time t = 0, know whether their investment opportunity (if any) is tradable (liquid) or not. If there are enough banks with θ = 1 who hold stored goods, the price of time 2 goods in ¯ terms of time 1 goods is 1.22 In this case, banks holding tradable projects who realize θ = θ can sell their investments at time t = 1 at no loss. We treat λ as a proxy for the extent of the possible interconnectedness within the financial system, and thus as a measure of the disconnectedness between the financial sector and the real economy. The maximization problem of a bank that has a non-tradable project is the same as in the fully connected environment. For banks with a tradable project, the ability to sell the project at t = 1 increases the returns to investment. Since banks always weakly prefer to consume at time t = 1, their maximization problem can be written as if they only consume at time 1: max E (θ(1 − I + AI α )) (11) I ∈[0,1] 21 Hobijn and Ravenna (2010) analyze the influence of securitization on monetary policy in the context of a DSGE model. See Moran and Meh (2013) for a DSGE model featuring a shadow banking sector. 22 For this to be the case, we need to assume enough “liquid demand” for these assets. Formally, a sufficient condition for this to hold for every λ is (1 − q0 )(1 − q ) > q0 qA. The left hand side is the amount of storage goods in the hand of θ = 1 banks who did not receive an investment opportunity at time 0. These banks hold storage goods, and are indifferent between time 1 goods and time 2 goods. The right hand side is the value of investment opportunities in the hand of banks who had an investment opportunity, and received a liquidity shock that made it optimal for them to sell at t = 1 (given I = 1, each project promises AI α = A units of time 2 goods). We assume this is the case, in order to capture the large demand for securitized assets prevailing in the financial markets, especially before 2007. 14 The FOC with respect to I is: αAI α−1 = 1 ⇒ I 1−α = αA (12) For αA > 1, there is no interior solution. We will assume that this is the case. To summarize, we are making the following parametric restriction: E (θ) > αA > 1 (13) Given this assumption, banks with tradable investment projects have a corner solution, in which they invest the maximum amount I = 1. Intuitively, there is no liquidity advantage to storage if the investment project is tradable. Given a corner solution I = 1, the amount of investment does not respond to small changes in A or in r - investment in these banks is unresponsive to monetary policy. At the aggregate level, note that investment is now given by: AGG α 1 Id = λ + (1 − λ)( A) 1−α (14) E (θ) The sensitivity to monetary policy is given by: 1 ∂ AGG ∂ ln(Id ) (1 − λ) ∂A ( Eα (θ) A) 1−α = 1 (15) ∂A λ + (1 − λ)( Eα A) 1−α (θ) Note that λ = 0 corresponds to the fully connected case. The absolute value of the above expression is decreasing in λ. To see this, note that the absolute value of this expression is (1−λ)a of the form (1−λ)b+λ where a > 0 and b < 1. The derivative with respect to λ is: −a((1 − λ)b + λ) − (1 − b)(1 − λ)a <0 (16) ((1 − λ)b + λ)2 This result implies that that the sensitivity of aggregate investment to monetary policy is increasing in connectedness. Note that, in this model, our measure of connectedness 15 corresponds to the share of claims at t = 1 for which the direct counterpart is investment projects. The “disconnected” part is claims held by banks on investment projects initiated by other banks.23 The equilibrium level of connectedness is given by: I (λ) CON N (λ) = (17) λq0 q + I (λ) It is easy to verify that this expression is decreasing in λ. AGG ∂Id Moreover, from equation (12), the aggregate amount of lending is increasing in λ ( ∂λ > 0). Intuitively, securitization makes lenders less careful about investing in illiquid assets if they can be sold in case of need at time t = 1. Altunbas, Gambacorta and Marquez-Ibanez (2009), using detailed information on secu- ritization activities for a sample of European banks, find both that securitization shelters banks’ loans from the effect of monetary policy and that it strengthens the capacity of banks to supply loans. Both results hold in the simple model we presented in this section. 4 Connectedness and monetary policy: Evidence This section presents some time series evidence to explore how connectedness affects the responses of economic variables to monetary policy shocks. We follow both a structural VAR (SVAR) approach as well as a factor-augmented VAR (FAVAR). SVAR. We adapt the approach of Boivin and Giannoni (2006) by including our measure of connectedness as an exogenous variable.24 The model can be written as follows: Yt = Φ(L)Yt−1 + βCON Nt−1 Yt−1 + et (18) 23 Instead of “purchasing” the investment projects directly, we can think of the interbank market more realistically as a loan market, in which banks with tradable investment projects borrow against t = 2 investment income. Here, “tradable” has the interpretation of “collateralizable”. 24 While arguably also the connection might be an endogenous variable, the result obtained in figure 6 indicates how the movements in the connection are long-run smooth movements, and thus we believe it can be considered an exogenous when using business cycle frequency data. In addition, the connection is included with lag one. 16 where Yt is a K × 1 vector of endogenous variable, Φ(L) is a matrix polynomial of order p and CON Nt−1 is exogenous. The reduced form errors, et , are assumed to be linear combinations of structural shocks, εt : et = Hεt with E (εt εt ) = Σ, a diagonal matrix. It is easy to see that the impulse responses to any shock in εt will depend on CON Nt−1 . Without loss of generality, assume p = 2 and develop Φ(L): Yt = Φ1 Yt−1 + Φ2 Yt−2 + βCON Nt−1 Yt−1 + et = (Φ1 + βCON Nt−1 ) Yt−1 + Φ2 Yt−2 + et = Φ1,t−1 Yt−1 + Φ2 Yt−2 + et , where Φ1,t−1 = (Φ1 + βCON Nt−1 ). Hence, the impulse response functions (IRFs) are ob- tained for any level of CON Nt−1 by inverting the previous expression: −1 Yt = I − Φ1,t−1 L − Φ2 L2 Hεt . (19) In practice, the coefficients matrices Φ(L) and β are estimated by OLS regression on (18), and H is deduced by imposing enough identification restrictions. The IRFs are then easily computed using (19). The confidence bands can be constructed using a parametric 25 bootstrap. Following Boivin and Giannoni (2006), Yt contains the deviation of the natural 25 We use the following procedure: ˆt and get bootstrap innovations e∗ 1. Shuffle the time dimension of OLS residuals e t 2. Using [Y1 , . . . , Yp ] as initial values and CON Nt−1 , get the bootstrap endogenous variables from ˆ L)Yt∗ Yt∗ = Φ( ˆ ∗ ∗ −1 + βCON Nt−1 Yt−1 + et . 3. Impose the identification restrictions to get H and calculate impulse responses. 17 logarithm of quarterly real GDP (GDPQ) from a linear deterministic trend, the annualized rate of change in the quarterly GDP deflator (GDPD), the natural logarithm of the quarterly average of the monthly spot market commodity price index (PSCCOM) and the quarterly average of the Federal Funds Rate (FFR). The exogenous variable CON Nt−1 contains our aggregate Measure of Connectedness. The data ranges from 1959Q1 to 2009Q1. Four lags are included in the VAR. The identification of structural shocks is achieved by the following recursive ordering: [ PSCCOM, GDPQ, GDPD, FFR ]. Hence, the unexpected monetary policy shock is ordered last in εt . The rotation matrix H is obtained using Choleski ˆt . The 95% confidence intervals are computed decomposition of the covariance matrix of e using 1000 bootstrap replications. In Figure (10), we compare the impulse responses of elements in Yt to an adverse monetary policy shock when the Measure of Connectedness is high and low, respectively CON N = 0.98 and CON N = 0.82. These are the average values of our connectedness measures for the periods 1959Q1-1983Q4 and 1984Q1-2009Q1. The impulse response with confidence bands for each level of connectedness are presented in Figures (11) and (12). As we can see from Figure (10), at the level of connection of 0.98 the adverse monetary shock generates a decrease in output, which exhibits a hump-shaped response. The price level decreases too, but only after a few quarters (the well known price puzzle phenomenon). When we consider a level of connection of 0.82, instead, we see that the response of the GDP to the same monetary policy shock is now not statistically different from zero. Also the responses of the quarterly GDP deflator and the spot market commodity price index are muted at the lower level of connection. Interestingly, there is no evidence of price puzzle in that case. In order to assess whether the difference in the impulse response we obtained under different levels of connection is statistically significant, we plot the difference in Figure (13), and we include confidence intervals at 90% significance level. As the figure shows, the impulse responses of GDP and GDP deflator are statistically different under the two scenarios, while the impulse responses of the commodity price index and the federal funds rate are not 18 statistically significantly different. The results reported in Figure (13) are robust to the inclusion of a time trend in the model, as well as to a different specification of the lag structure.26 FAVAR. We conduct a more refined exercise inspired by the model from Bernanke et al. (2005). In contrast to standard structural VAR models, factor models have a number of advantages: i) they allow considering large amounts of information potentially observed by agents, and thus minimize the risk of omitted variable bias; ii) they are not sensitive to the choice of a specific data series, which may be arbitrary; iii) they are less likely to be subject to non-fundamentalness issues raised by Forni et al. (2009)27 ; and iv) they allow us to compute the response of a larger set of variables of interest to identified shocks. As in the case of SVAR, we introduce our measure of connectedness through interaction terms, in order to obtain impulse response functions that are conditional on a certain level of connectedness. Formally, we consider the following static factor model with latent and observed factors: Xt = ΛF Ft + ΛR Rt + ut (20)       F F F  t  = Φ(L)  t−1  + βCON Nt−1  t−1  + et (21) Rt Rt−1 Rt−1 where Ft is vector of K latent factors and Rt is the observed factor. In our case, Rt is the the Federal Funds Rate, since the objective here is to identify the monetary policy shock. Xt contains N macroeconomic and financial indicators organized into a block of ‘slow-moving’ variables that are largely predetermined to monetary policy, and another consisting of ‘fast moving’ variables that are sensitive to the FED’s rule. The idiosyncratic errors are assumed to be serially uncorrelated. 26 We omitted the results here, they are available upon request. 27 If the shocks in the VAR model are fundamental, then the dynamic effects implied by the moving average representation can have a meaningful interpretation, i.e. the structural shocks can be recovered from current and past values of observable series. Forni et al. (2009) argue that while non-fundamentalness is generic of small scale models, it is highly unlikely to arise in large dimensional dynamic factor models. 19 In our application, Xt contains N = 108 quarterly time series from Ng and Stevanovic (2012), that run from 1959Q1 to 2009Q1. Data include both macroeconomic variables such as GDP, employment, investment, hours worked, inflation rate as well as financial variables 28 such as credit spreads, loans, etc. The data have been transformed to induce stationarity and are standardized prior to estimation. The ICp2 information criterion from Bai and Ng (2002) suggests K = 3 latent factors. The lag order of Φ(L) is set to 4. The estimation and identification of structural shocks consist of several steps. First, ˆ (Ft , Rt ) be the K one must impose Rt as an observed factor when estimating Ft . Let C principal components of Xt . Let XtS be NS ‘slow’ moving variables, and let the K principal ˆ (Ft , Rt ) − ˆ ˆt = C components of XtS be C (Ft ). The estimate of latent factors is F bR Rt where ˆ bR is obtained by least squares estimation of the regression ˆ (Ft , Rt ) = bC C (Ft ) + bR Rt + ut . C ˆt and Rt : Λ The estimation of the loadings is now straightforward by regressing Xt on F ˆF and ΛˆR . ˆt can be ˆt , we estimate (21) as in the case of the SVAR model. Since F Second, using F correlated with Rt , we identify the monetary policy by ordering Rt last. Finally, we invert (21) to obtain factors’ impulse responses, and multiply them by factor loadings to get the IRFs of any element in Xt . While all the impulse responses are available upon request, we present here only a subset of them. As before, we compare the impulse responses to an adverse monetary policy shock when the connectedness is high and low, respectively CON N = 0.98 and CON N = 0.82. In Figure (14) we report the responses of output, investment and employment to an identified monetary policy shock. In all three cases, the responses of real variables to a monetary innovation are muted at lower levels of connectedness. In order to test whether these differences are statistically significant, we compute the difference between the impulse responses and we compute via bootstrap a 90% confidence internal. In Figure (15) we report the results. The 28 The complete description of the data and their transformation is presented in the appendix. 20 impulse responses of GDP, Investment and Employment are statistically different, at least in the first few quarters. In Figure (16) we report the results for CPI inflation, bank lending and loans and leases. Interestingly, also the response of bank lending and loans and lease are muted in the case of a lower connection between the financial sector and the real economy. This is precisely the mechanism that we highlighted in the previous section. As before, we check the statistical significance of these results by plotting the difference of the impulse responses and the associated confidence intervals. We show the results in Figure (17). We conclude that the inclusion of our measure of connectedness into a SVAR or a FAVAR for the U.S. economy generates statistically different responses of real variables to monetary policy innovations. 5 Conclusions This paper documents a declining trend in the share of financial claims whose direct counter- part is in the non-financial sector. The financial sector’s increased ability to buffer idiosyn- cratic liquidity shocks may have contributed to an increase in investment, and a decrease in the sensitivity of investment to fundamentals such as the borrower’s net worth. In this paper, we illustrate how this may have contributed to the dampening of the responsiveness to monetary policy. However, the implications of this structural change in the financial system may have had implications far beyond this. We outline here several potential avenues for future research that make use of the measure of connectedness. First, it would be interesting to develop a quantitative macroeconomic model embedding the concept of connection explored in this paper. This could also be used to evaluate the relative importance of the policy behavior and the disconnectedness in explaining the muted responses of monetary policy innovations on economic variables found using more recent samples. Second, it would be interesting to explore whether the connection between the financial sector and the real economy affects other dimensions of the macro economy (for example, 21 the availability of credit to firms or the vulnerability to financial crisis). Third, it may be insightful to analyze the dynamics of the connection between the fi- nancial sector and the real economy for additional countries, and study how connection was related to performance during the Great Recession. In this sense, it would be interesting also to address “normative” questions, such as the effects of connectedness on welfare. Finally, and especially for policy purposes, it would be important to go beyond the aggregate perspective we take in this paper and use balance sheet data on single financial institutions to analyze the impact of their connection with the real economy on a range of performance measures. This could also help improve the regulation and monitoring of financial institutions. We plan to pursue these avenues in our future research. 22 A Appendix A.1 Measure of Connectedness The following table report the exact composition of the numerator of Measure of Connect- edness for each U.S. financial actor. At the denominator, we always used the total credit market instruments. U.S. Financial Actor CREDIT REAL Commercial Banks Open Market Papers, Treasury Securities, Non-securitized GSE-backed securities, Municipal securities Non-securitized Corporate Bonds, Bank Loans, Mortgages, Security Credit and Consumer Credit Savings Institutions Open Market Papers, Treasury Securities, Non-securitized GSE-backed securities, Municipal securities, Non-securitized Corporate Bonds, Mortgages, and Consumer Credit Credit Unions Open Market Papers, Treasury Securities, Corporate Bonds, Home Mortgages, and Consumer Credit Property Ins Comp Open Market Papers, Treasury Securities, Municipal Securities, Mortgages and Corporate Bonds Life Ins Comp Open Market Papers, Treasury Securities, Municipal Securities, Mortgages and Corporate Bonds Mutual Funds Open Market Papers, Treasury Securities, Municipal Securities, Corporate Bonds and other loans MMMF Open Market Papers, Treasury Securities, Municipal Securities, Corporate Bonds and other loans Private Pension funds Open Market Papers, Treasury Securities, Municipal Securities, Corporate Bonds and other loans Local Govt pension funds Open Market Papers, Treasury Securities, Municipal Securities, Corporate Bonds and other loans Federal Govt pension funds Open Market Papers, Treasury Securities, Municipal Securities, Corporate Bonds and other loans GSEs Open Market Papers, Treasury Securities, Municipal Securities, Corporate Bonds and Mortgages Brokers and Dealers Open Market Papers, Municipal Securities and Corporate Bonds ABS issuers Treasury Securities, Consumer Credit and Trade Credit A.2 Data Used in the FAVAR The transformation codes are: 1 no transformation; 2 first difference; 4 logarithm; 5 first difference of logarithm; 0 variable not used in the estimation (only used for transforming other variables). A * indicates a series that is deflated with the GDP deflator (series #89). No. Series Code T-Code Series Description 1 DRIINTL:GDPRC@US.Q 5 NIA REAL GROSS DOMESTIC PRODUCT (CHAINED-2000), SA - U.S. 2 USCEN:GDPGDR.Q 5 REAL GDP-GDS,BILLIONS OF CH (2000) $,SAAR-US 3 USCEN:GDPSVR.Q 5 REAL GDP-SVC,BILLIONS OF CH (2000) $,SAAR-US 4 USCEN:GDPSR.Q 5 REAL GDP-STRUC,BILLIONS OF CH (2000) $,SAAR-US 5 BASIC:IPN11.M 5 INDUSTRIAL PRODUCTION INDEX - PRODUCTS, TOTAL 6 BASIC:IPN300.M 5 INDUSTRIAL PRODUCTION INDEX - FINAL PRODUCTS 7 BASIC:IPN12.M 5 INDUSTRIAL PRODUCTION INDEX - CONSUMER GOODS 8 BASIC:IPN13.M 5 INDUSTRIAL PRODUCTION INDEX - DURABLE CONSUMER GOODS 9 BASIC:IPN18.M 5 INDUSTRIAL PRODUCTION INDEX - NONDURABLE CONSUMER GOODS 10 BASIC:IPN25.M 5 INDUSTRIAL PRODUCTION INDEX - BUSINESS EQUIPMENT 11 BASIC:IPN32.M 5 INDUSTRIAL PRODUCTION INDEX - MATERIALS 12 BASIC:IPN34.M 5 INDUSTRIAL PRODUCTION INDEX - DURABLE GOODS MATERIALS 13 BASIC:IPN38.M 5 INDUSTRIAL PRODUCTION INDEX - NONDURABLE GOODS MATERIALS 14 BASIC:IPN10.M 5 INDUSTRIAL PRODUCTION INDEX - TOTAL INDEX 15 USCEN:UTLB00004.M 1 CAPACITY UTILIZ-MFG,SA-US 16 BASIC:PMI.M 1 PURCHASING MANAGERS’ INDEX (SA) 23 17 BASIC:PMP.M 1 NAPM PRODUCTION INDEX (PERCENT) 18 DRIINTL:WS@US.Q 5* NIA NOMINAL TOTAL COMPENSATION OF EMPLOYEES, SA - U.S. 19 USCEN:YPR.M 5 PERS INCOME CH 2000 $,SA-US 20 USCEN:YP@V00C.M 5 PERS INCOME LESS TRSF PMT CH 2000 $,SA-US 21 USCEN:AHPMF.M 5* AHE,PROD WORKERS: MFG,SA-US 22 USCEN:AHPCON.M 5* AHE,PROD WORKERS: CONSTR,SA-US 23 USCEN:HPMF.M 5 AWH,PROD WORKERS: MFG,SA-US 24 USCEN:HOPMD.M 5 AVG WEEKLY OT,PROD WORKERS: DUR,SA-US 25 BASIC:LHEL.M 5 INDEX OF HELP-WANTED ADVERTISING IN NEWSPAPERS (1967=100;SA) 26 BASIC:LHELX.M 1 EMPLOYMENT: RATIO; HELP-WANTED ADS:NO. UNEMPLOYED CLF 27 BASIC:LHEM.M 5 CIVILIAN LABOR FORCE: EMPLOYED, TOTAL (THOUS.,SA) 28 BASIC:LHNAG.M 5 CIVILIAN LABOR FORCE: EMPLOYED, NONAGRIC.INDUSTRIES (THOUS.,SA) 29 BASIC:LHUR.M 1 UNEMPLOYMENT RATE: ALL WORKERS, 16 YEARS & OVER (%,SA) 30 BASIC:LHU680.M 1 UNEMPLOY.BY DURATION: AVERAGE(MEAN)DURATION IN WEEKS (SA) 31 BASIC:LHU5.M 5 UNEMPLOY.BY DURATION: PERSONS UNEMPL.LESS THAN 5 WKS (THOUS.,SA) 32 BASIC:LHU14.M 5 UNEMPLOY.BY DURATION: PERSONS UNEMPL.5 TO 14 WKS (THOUS.,SA) 33 BASIC:LHU15.M 5 UNEMPLOY.BY DURATION: PERSONS UNEMPL.15 WKS + (THOUS.,SA) 34 BASIC:LHU26.M 5 UNEMPLOY.BY DURATION: PERSONS UNEMPL.15 TO 26 WKS (THOUS.,SA) 35 BASIC:CES001.M 5 EMPLOYEES, NONFARM - TOTAL NONFARM 36 BASIC:CES002.M 5 EMPLOYEES, NONFARM - TOTAL PRIVATE 37 BASIC:CES003.M 5 EMPLOYEES, NONFARM - GOODS-PRODUCING 38 USCEN:CR.Q 5 REAL PCE,BILLIONS OF CH (2000) $,SAAR-US 39 USCEN:JQCDR.Q 5 REAL PCE-DUR,QTY INDEX (2000=100),SA,SA-US 40 USCEN:JQCNR.Q 5 REAL PCE-NDUR,QTY INDEX (2000=100),SA,SA-US 41 USCEN:JQCSVR.Q 5 REAL PCE-SVC,QTY INDEX (2000=100),SA,SA-US 42 USCEN:JQCXFAER.Q 5 REAL PCE EX FOOD&ENERGY,QTY INDEX (2000=100),SAAR-US 43 DRIINTL:CGRCUS.Q 5 REAL GOVERNMENT CONS. EXPEND.& GROSS INVESTMENT (CHAINED-2000), SA - U.S. 44 USCEN:I.Q 5* GROSS PRIV DOM INVEST,BILLIONS OF $,SAAR-US 45 USCEN:IF.Q 5* GROSS PRIV DOM INVEST-FIXED,BILLIONS OF $,SAAR-US 46 USCEN:IFNRE.Q 5* GROSS PRIV DOM INVEST-FIXED NONRES,BILLIONS OF $,SAAR-US 47 USCEN:IFRES.Q 5* PRIV FIXED INVEST-RES-STRUC,BILLIONS OF $,SAAR-US 48 USCEN:IFRE.Q 5* GROSS PRIV DOM INVEST-FIXED RES,BILLIONS OF $,SAAR-US 49 USCEN:II.Q 1 GROSS PRIV DOM INVEST-CH IN PRIV INVENT,BILLIONS OF $,SAAR-US 50 USCEN:IIF.Q 1 GROSS PRIV DOM INVEST-CH IN PRIV INVENT-FARM,BILLIONS OF $,SAAR-US 51 BASIC:HSFR.M 4 HOUSING STARTS:NONFARM(1947-58);TOTAL FARM&NONFARM(1959-)(THOUS.,SA 52 BASIC:HMOB.M 4 MOBILE HOMES: MANUFACTURERS’ SHIPMENTS (THOUS.OF UNITS,SAAR) 53 BASIC:PMNV.M 1 NAPM INVENTORIES INDEX (PERCENT) 54 BASIC:PMNO.M 1 NAPM NEW ORDERS INDEX (PERCENT) 55 BASIC:PMDEL.M 1 NAPM VENDOR DELIVERIES INDEX (PERCENT) 56 BASIC:MOCMQ.M 5 NEW ORDERS (NET) - CONSUMER GOODS & MATERIALS, 1996 DOLLARS (BCI) 57 BASIC:MSONDQ.M 5 NEW ORDERS, NONDEFENSE CAPITAL GOODS, IN 1996 DOLLARS (BCI) 58 USCEN:M.Q 5 IMPORTS OF GDS&SVC,BILLIONS OF $,SAAR-US 59 USCEN:X.Q 5 EXPORTS OF GDS&SVC,BILLIONS OF $,SAAR-US 60 BASIC:FSPCOM.M 5 S&P’S COMMON STOCK PRICE INDEX: COMPOSITE (1941-43=10) 61 BASIC:FSPIN.M 5 S&P’S COMMON STOCK PRICE INDEX: INDUSTRIALS (1941-43=10) 62 BASIC:FSDXP.M 1 S&P’S COMPOSITE COMMON STOCK: DIVIDEND YIELD (% PER ANNUM) 63 BASIC:FSPXE.M 1 S&P’S COMPOSITE COMMON STOCK: PRICE-EARNINGS RATIO (%,NSA) 64 BASIC:EXRUK.M 5 FOREIGN EXCHANGE RATE: UNITED KINGDOM (CENTS PER POUND) 65 BASIC:EXRCAN.M 5 FOREIGN EXCHANGE RATE: CANADA (CANADIAN $ PER U.S.$) 66 BASIC:FYGM3.M 1 INTEREST RATE: U.S.TREASURY BILLS,SEC MKT,3-MO.(% PER ANN,NSA) 67 BASIC:FYGM6.M 1 INTEREST RATE: U.S.TREASURY BILLS,SEC MKT,6-MO.(% PER ANN,NSA) 68 BASIC:FYGT1.M 1 INTEREST RATE: U.S.TREASURY CONST MATURITIES,1-YR.(% PER ANN,NSA) 69 BASIC:FYGT5.M 1 INTEREST RATE: U.S.TREASURY CONST MATURITIES,5-YR.(% PER ANN,NSA) 70 BASIC:FYGT10.M 1 INTEREST RATE: U.S.TREASURY CONST MATURITIES,10-YR.(% PER ANN,NSA) 71 BASIC:FYAAAC.M 1 BOND YIELD: MOODY’S AAA CORPORATE (% PER ANNUM) 72 BASIC:FYBAAC.M 1 BOND YIELD: MOODY’S BAA CORPORATE (% PER ANNUM) 73 FYGM6-FYFF 1 74 FYGM3-FYFF 1 24 75 FYGT1-FYFF 1 76 FYGT5-FYFF 1 77 FYGT10-FYFF 1 78 FYAAAC-FYFF 1 79 FYBAAC-FYFF 1 80 BASIC:FM1.M 5 MONEY STOCK: M1(CURR,TRAV.CKS,DEM DEP,OTHER CK’ABLE DEP)(BIL$,SA) 81 BASIC:FM2.M 5 MONEY STOCK:M2(M1+O’NITE RPS,EURO$,G/P&B/D MMMFS&SAV&SM TIME DEP(BIL$) 82 USCEN:MNY2@00.M 5 MONEY SUPPL-M2 IN 2000 $,SA-US 83 BASIC:FMFBA.M 5 MONETARY BASE, ADJ FOR RESERVE REQUIREMENT CHANGES(MIL$,SA) 84 BASIC:FMRRA.M 5 DEPOSITORY INST RESERVES:TOTAL,ADJ FOR RESERVE REQ CHGS(MIL$,SA) 85 BASIC:FMRNBA.M 2 DEPOSITORY INST RESERVES:NONBORROWED,ADJ RES REQ CHGS(MIL$,SA) 86 USCEN:ALCIBL00Z.M 5 COML&IND LOANS OUTST,SA-US 87 BASIC:FCLBMC.M 1 WKLY RP LG COM’L BANKS:NET CHANGE COM’L & INDUS LOANS(BIL$,SAAR) 88 BASIC:CCINRV.M 5 CONSUMER CREDIT OUTSTANDING - NONREVOLVING(G19) 89 DRIINTL:PGDP@US.Q 5 NIA PRICE DEFLATOR - GROSS DOMESTIC PRODUCT, SA - U.S. 90 DRIINTL:PCP@US.Q 5 NIA PRICE DEFLATOR - PRIVATE CONSUMPTION EXPENDITURE, SA - U.S. 91 USCEN:PDII.Q 5 GROSS PRIV DOM INVEST,PRICE DEFLATORS (2000=100),SA,SA-US 92 USCEN:JPCD.Q 5 PCE-DUR,PRICE INDEX (2000=100),SA,SA-US 93 USCEN:JPCN.Q 5 PCE-NDUR,PRICE INDEX (2000=100),SA,SA-US 94 USCEN:JPCSV.Q 5 PCE-SVC,PRICE INDEX (2000=100),SA,SA-US 95 BASIC:PUXM.M 5 CPI-U: ALL ITEMS LESS MEDICAL CARE (82-84=100,SA) 96 BASIC:PUXHS.M 5 CPI-U: ALL ITEMS LESS SHELTER (82-84=100,SA) 97 BASIC:PUXF.M 5 CPI-U: ALL ITEMS LESS FOOD (82-84=100,SA) 98 BASIC:PUS.M 5 CPI-U: SERVICES (82-84=100,SA) 99 BASIC:PUCD.M 5 CPI-U: DURABLES (82-84=100,SA) 100 BASIC:PUC.M 5 CPI-U: COMMODITIES (82-84=100,SA) 101 BASIC:PUNEW.M 5 CPI-U: ALL ITEMS (82-84=100,SA) 102 BASIC:PWFSA.M 5 PRODUCER PRICE INDEX: FINISHED GOODS (82=100,SA) 103 BASIC:PMCP.M 1 NAPM COMMODITY PRICES INDEX (PERCENT) 104 UOMO83 1 COMPONENT INDEX OF CONSUMER EXPECTATIONS, NSA, CONFBOARD AND U.MICH. 105 DRIINTL:JLEAD@US.Q 5 COMPOSITE CYCLICAL INDICATOR (1996) - LEADING, SA - U.S. 106 DRIINTL:JLAG@US.Q 5 COMPOSITE CYCLICAL INDICATOR (1996) - LAGGING, SA - U.S. 107 DRIINTL:JCOIN@US.Q 5 COMPOSITE CYCLICAL INDICATOR (1996) - COINCIDENT, SA - U.S. 108 FYBAAC-FYGT10.M 1 BAA SPREAD 109 USCEN:NC16&Z.M 0 CIVILIAN NONINSTITUTIONAL POP: 16 YEARS&OVER,SA-US 110 BASIC:FYFF.M 1 INTEREST RATE: FEDERAL FUNDS (EFFECTIVE) (% PER ANNUM,NSA) 25 References [1] Adrian, T. and Ashcraft, A.B. (2012) “Shadow Banking Regulation”, Federal Reserve Bank of New York, Staff Report N. 559. [2] Adrian, T. and Shin, H.S. (2011a) “Financial Intermediary Balance Sheet Management” Annual Review of Financial Economics, 3, 289-307. [3] Adrian, T. and Shin, H.S. (2011b) “Financial Intermediaries and Monetary Economics” Ch. 12 in the Handbook of Monetary Economics (eds) Benjamin Friedman and Michael Woodford, Elsevier. 2011 [4] Adrian, T. and Shin, H.S. (2010) “The Changing Nature of Financial Intermediation and the Financial Crisis of 2007-2009”. Annual Review of Economics. 2:603-18. [5] Allen, F. and Gale, D. (2000) “Financial Contagion”. The Journal of Political Economy, 108(1): 1-33. [6] Altunbas, Yener, Gambacorta, Leonardo and Marques-Ibanez, David (2009) “Securitisa- tion and the bank lending channel,” European Economic Review, 53(8): 996-1009. [7] Bai, J., and S. Ng (2002), “Determining the Number of Factors in Approximate Factor Models,” Econometrica 70:1,191-221. [8] Bernanke, B.S. and Gertler, M. (1995) “Inside the Blanck Box: The Credit Channel of Monetary Policy Transmission”. Journal of Economic Perspectives. 9(4). Fall 1995: 27-48 [9] Bernanke, B.S., J. Boivin and P. Eliasz (2005),“Measuring the Effects of Monetary Policy: a Factor-augmented Vector Autoregressive (FAVAR) Approach,” Quarterly Journal of Economics 120: 387–422. [10] Boivin, J., and Giannoni, M.(2006) “Has Monetary Policy Become More Effective?”, The Review of Economics and Statistics 88(3): 445-462. 26 [11] Boivin, J., Kiley M. T. and Mishkin, F.S. (2011) “How has the monetary transmission mechanism evolved over time?” Ch. 8 in the Handbook of Monetary Economics (eds) Benjamin Friedman and Michael Woodford, Elsevier. 2011 [12] Borio, c. and Zhu, H. (2012) “Capital Regulation, risk-taking and monetary policy: a missing link in the transmission mechanism?,” Journal of Financial Stability, 8(4): 236- 251. [13] Brunnermeier, M. Dong,G. and Paliab, D. (2012)“Banks’ Non-Interest Income and Sys- temic Risk,” mimeo [14] Bruno, V. and Shin, H.S. (2013) “Capital Flows and the Risk-taking channel of mone- tary policy”, NBER w.p. 18942. [15] Diamond, D. W. and Dybvig, P. H. (1983) “Bank Runs, Deposit Insurance, and Liq- uidity”. The Journal of Political Economy, 91(3): 401-419. [16] Diebold F. X. and Yilmaz K. (2012) “Better to Give than to Receive: Predictive Di- rectional Measurement of Volatility Spillovers (with discussion),” International Journal of Forecasting, 28, 57-66. [17] Diebold F. X. and Yilmaz K. (2013) “On the Network Topology of Variance Decom- positions: Measuring the Connectedness of Financial Firms,” Journal of Econometrics, forthcoming. [18] Dynan K, Elmendorf D.W. and Sichel D.E. (2006) “Can Financial Innovation Help to Explain the Reduced Volatility of Economic Activity?” Journal of Monetary Economics 53: 123. [19] Gennaioli, N., Shleifer, A. and R. Vishny (2013). “Finance and the Preservation of Wealth”, mimeo [20] Gennaioli, N., Shleifer, A. and R. Vishny (2012). “Money Doctors”, mimeo 27 [21] Greenwood, R. and Scharfstein, D. (2013) “The Growth of Modern Finance”. Journal of Economic Perspectives, 27(2): 3-28. [22] Gorton, G. (2009) “The Panic of 2007,” in Maintaining Stability in a Changing Financial System, Proceedings of the 2008 Jackson Hole Conference, Federal Reserve Bank of Kansas City. [23] Gorton, G., and Metrik, A. (2012) “Securitization” in The Handbook of the Economics of Finance, ed. G. Constantinides, M. Harris, and R. Stulz, forthcoming. [24] Gorton, G., and A. Winton (2003): “Financial Intermediation,” in Handbook of the Economics of Finance, ed. by G. M. Constantinides, M. Harris, and R. Stulz, pp. 431-552, North Holland. Elsevier. [25] Hahm, H.J., Shin, H.S. and Shin, K. (forthcoming) “Non-Core Bank Liabilities and Financial Vulnerability”, Journal of Money, Credit and Banking. [26] Hobjin, B. and Ravenna, F (2010) “Loan Securitization and the Monetary Transmission Mechanism”, mimeo. [27] Keys,B.,Mukherjee, T, Seru, A. and V. Vig (2010) “Did Securitization Lead to Lax Screening? Evidence From Subprime Loans”, The Quarterly Journal of Economics, 125(1): 307-362. [28] Mesonnier J.-S. and Stevanovic D. (2012), “Bank Leverage Shocks and the Macroecon- omy: a New Look in a Data-rich Environment,” Working Paper 394, Banque de France [29] Moran, K. and Meh, C (2013) “Shadow Banking and Regulation: a Quantitative As- sessment”. Mimeo. [30] Ng, S. and D. Stevanovic (2012), “Factor-augmented Autoregressive Distributed Lag Models,” mimeo, Columbia University. [31] Ordonez, G and Gorton, G (2012) “Collateral Crises” NBER Working Paper n.17771. 28 [32] Philippon, T. (2012) “Has the US Finance Industry Become Less Efficient?”, mimeo. [33] Philippon, T. and Reshef, A. (forthcoming) “Wages and Human Capital in the U.S. Financial Industry: 1909-2006”, Quarterly Journal of Economics. [34] Philippon, T. and Reshef, A. (2013) “An international look at the growth of modern finance.” Journal of Economic Perspectives, 27(2): 73-96. [35] Pozsar, Z., Adrian, T. Ashcraft, A. and Boesky, H. (2012) “Shadow Banking”, Federal Reserve Bank of New York, Staff Report N. 458. 29 Figure 1: Asset Shares of Different Actors (source: FED Flow of Funds) Asset Shares − U.S. Financial Sector 60 40 20 0 1950 1960 1970 1980 1990 2000 2010 Year Banks, Sav Inst, Credit Insurances and Funds GSE, Pools, ABS, Bro−Deal ... FED Figure 2: The Measure of Connectedness : U.S. 1952-2009, Traditional Actors Measure of Connectedness, Traditional Actors 1 .8 .6 .4 .2 0 1950 1960 1970 1980 1990 2000 2010 YEAR Commercial Bank Savings Institutions Credit Unions 30 Figure 3: The Measure of Connectedness : U.S. 1952-2009, Insurance Companies and Mutual Funds Measure of Connectedness, Mutual Funds and Ins. Comp. 1 .8 .6 .4 .2 0 1950 1960 1970 1980 1990 2000 2010 YEAR Property Insurance Life Insurance Mutual Funds MMMF Figure 4: The Measure of Connectedness : U.S. 1952-2009, Pension Funds Measure of Connectedness, Pension Funds 1 .8 .6 .4 .2 0 1950 1960 1970 1980 1990 2000 2010 YEAR Private Pension Funds Local Govt Pension Funds Federal Govt Pension Funds 31 Figure 5: The Measure of Connectedness : U.S. 1952-2009, Shadow Banking Measure of Connectedness, Shadow Banking 1 .8 .6 .4 .2 0 1950 1960 1970 1980 1990 2000 2010 YEAR GSEs ABS Issuers Brokers and Dealers Finance Companies Figure 6: The Measure of Connectedness : U.S. 1952-2009 Measure of Connectedness, Aggregate 1 .9 CONNECTION .7 .6 .5.8 1950 1960 1970 1980 1990 2000 2010 YEAR 32 Figure 7: Shares of Credit Market Instruments Credit Market Instruments Share, U.S. Financial Industry .8 .6 .4 .2 0 1950 1960 1970 1980 1990 2000 2010 YEAR Traditional Actors Shadow Banking (except Pools) Insurances and Funds Figure 8: Shares of Credit Market Instruments Credit Market Instruments Share, Selected Institutions .5 .4 .3 .2 .1 0 1950 1960 1970 1980 1990 2000 2010 YEAR Commercial Banks ABS Issuers 33 Figure 9: Measure of Connectedness and Share of Finance in GDP (Philippon 2012) 0.4 0.1 Defense adjusted income share of finance, Philippon (2012) 0.3 0.08 Deconnection 0.2 0.06 0.1 0.04 0 0.02 1950 1960 1970 1980 1990 2000 2010 year Figure 10: Comparison of IRFs to a monetary policy shock conditional on different degrees of Connectedness in SVAR −3 PSCCOM x 10 GDP 0 2 c=0.98 −0.005 c=0.82 −0.01 0 −0.015 −2 −0.02 −0.025 −4 −0.03 −0.035 −6 −0.04 −0.045 −8 0 4 8 12 16 0 4 8 12 16 GDPDEF FFR 0.2 1.2 0.1 1 0.8 0 0.6 −0.1 0.4 −0.2 0.2 −0.3 0 −0.4 −0.2 −0.5 −0.4 0 4 8 12 16 0 4 8 12 16 34 Figure 11: IRFs to a monetary policy shock with C = 0.98 in SVAR −3 PSCCOM x 10 GDP 0.01 2 0 0 −0.01 −2 −0.02 −4 −0.03 −6 −0.04 −8 −0.05 −10 −0.06 −0.07 −12 −0.08 −14 0 4 8 12 16 0 4 8 12 16 GDPDEF FFR 0.6 1.5 0.4 1 0.2 0.5 0 −0.2 0 −0.4 −0.5 −0.6 −0.8 −1 0 4 8 12 16 0 4 8 12 16 Figure 12: IRFs to a monetary policy shock with C = 0.82 in SVAR −3 PSCCOM x 10 GDP 0.04 5 0.02 0 0 −0.02 −0.04 −5 −0.06 −0.08 −0.1 −10 0 4 8 12 16 0 4 8 12 16 GDPDEF FFR 0.2 1.5 0.1 1 0 0.5 −0.1 −0.2 0 −0.3 −0.5 −0.4 −0.5 −1 0 4 8 12 16 0 4 8 12 16 35 Figure 13: Difference between IRFs to a monetary policy shock with different levels of Connectedness PSCCOM x 10 ï3 GDP 0.06 4 2 0.04 0 0.02 ï2 0 ï4 ï6 ï0.02 ï8 ï0.04 ï10 ï0.06 ï12 0 4 8 12 16 0 4 8 12 16 GDPD FFR 0.4 0.6 0.3 0.4 0.2 0.2 0.1 0 0 ï0.1 ï0.2 ï0.2 ï0.4 ï0.3 ï0.6 ï0.4 ï0.5 ï0.8 0 4 8 12 16 0 4 8 12 16 Figure 14: Comparison of IRFs to a monetary policy shock in FAVAR, C = 0.98 and C = 0.82, Real Variables FFR GDP 0.3 0.1 c=0.98 c=0.82 0.25 0 0.2 ï0.1 0.15 ï0.2 0.1 ï0.3 0.05 ï0.4 0 ï0.5 ï0.05 ï0.1 ï0.6 0 4 8 12 16 0 4 8 12 16 INVESTMENT EMPLOYMENT 0.1 0.1 0 0 ï0.1 ï0.1 ï0.2 ï0.2 ï0.3 ï0.3 ï0.4 ï0.5 ï0.4 ï0.6 ï0.5 ï0.7 ï0.6 ï0.8 ï0.7 ï0.9 0 4 8 12 16 0 4 8 12 16 36 Figure 15: Difference between IRFs to a monetary policy shock with different levels of Connectedness, FAVAR Real Variables FFR GDP 0.1 1 0.05 0.5 0 0 −0.05 −0.1 −0.5 −0.15 −1 −0.2 −0.25 −1.5 0 4 8 12 16 0 4 8 12 16 INVESTMENT EMPLOYMENT 1 1 0.5 0.5 0 0 −0.5 −0.5 −1 −1 −1.5 −1.5 −2 0 4 8 12 16 0 4 8 12 16 Figure 16: Comparison of IRFs to a monetary policy shock in FAVAR, C = 0.98 and C = 0.82, Financial Variables FFR CPI INFL 0.3 0.1 c=0.98 c=0.82 0.25 0.05 0.2 0.15 0 0.1 ï0.05 0.05 0 ï0.1 ï0.05 ï0.1 ï0.15 0 4 8 12 16 0 4 8 12 16 BANK CREDIT LOANS AND LEASES 0.1 0.1 0 0 ï0.1 ï0.1 ï0.2 ï0.2 ï0.3 ï0.3 ï0.4 ï0.4 ï0.5 ï0.5 ï0.6 ï0.6 ï0.7 ï0.7 ï0.8 ï0.8 ï0.9 0 4 8 12 16 0 4 8 12 16 37 Figure 17: Difference between IRFs to a monetary policy shock with different levels of Connectedness, FAVAR Financial Variables FFR CPI INFL 0.1 0.05 0.05 0 0 −0.05 −0.05 −0.1 −0.1 −0.15 −0.15 −0.2 −0.2 −0.25 −0.25 −0.3 0 4 8 12 16 0 4 8 12 16 BANK CREDIT LOANS AND LEASES 0.2 1 0 0.5 −0.2 0 −0.4 −0.6 −0.5 −0.8 −1 −1 −1.5 −1.2 −1.4 −2 0 4 8 12 16 0 4 8 12 16 38