Shadow Banking and the Funding of the Nonfinancial Sector

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Shadow Banking and the Funding of the Nonfinancial Sector Joshua Gallin 2013-50 NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Shadow Banking and the Funding of the Nonfinancial Sector By Joshua Gallin1, Federal Reserve Board May 16, 2013 Abstract: I show how to use data from the Flow of Funds Accounts of the United States to estimate how much funding of nonfinancial businesses, households, and governments is provided by the domestic shadow banking system. I define the shadow banking system as the set of entities and activities that provide short-term funding outside of the traditional commercial banking system, but I do not equate all nonbank funding with shadow banking. My results suggest that at the end of 2008, domestic shadow-bank funding of the nonfinancial sector was an important, but fairly modest source of funding relative to that provided by more traditional funding sources such as commercial banks, insurance companies, and pension funds. However, my results suggest that domestic shadow banking played a large role in the increase of nonfinancial-sector debt in the two years before 2008:Q4 and was, at least in an arithmetic sense, the entire reason for the slowdown in nonfinancial-sector debt growth after 2008. Domestic shadow-bank funding of the nonfinancial sector has increased since 2010, but remains well below the level seen right in late 2008. 1 The opinions expressed in this paper are those of the author only. They do not necessarily reflect those of the Governors of the Federal Reserve System or its staff. I would like to thanks Dan Covitz, Patrick McCabe, Rebecca Zarutskie, and Marshall Reinsdorf. 1

Introduction The financial and economic upheaval of the past few years has provided a harsh reminder of the dangers of overreliance on short-term funding. The financial crisis also revealed how little regulators, supervisors, and market participants themselves know about the extent to which such funding was, and continues to be, provided by what is now commonly known as the shadow banking system. Recent research has improved our understanding the role played by elements of the shadow banking system. Pozsar et al. (2010) gives an overview of the shadow banking system, Pozsar (2011) provides information on investors’ pools of cash, and Ricks (2011) examines the growth of private money claims. Others have examined particular instruments used in shadow banking, such as repurchase agreements (Gorton and Metrick, 2010), assetbacked commercial paper (Covitz, Liang, and Suarez, 2009 and Acharya, Schnabl, and Suarez, 2011), auction rate securities and variable-rate demand notes (Han and Li, 2010), and money market mutual funds (McCabe, 2010). In this paper I describe a way to use data from the Flow of Funds Accounts of the United States (FFA) and other readily available sources to provide rough “top down” measures of the size of the domestic shadow banking system.2 In particular, I estimate the amount of debt financing of the nonfinancial sectors of the U.S. economy that is dependent on the shadow banking system. I define shadow-bank funding of the nonfinancial sector loosely as funding provided to households, nonfinancial businesses, and federal, state, and local governments that have a “runnable” link in their intermediation chain. My definition of a runnable link is that the financial intermediary relies significantly on short-term funding that is not insured by the FDIC and that the intermediary does not have direct access to the Federal Reserve’s Discount Window. I examine shadow-bank funding of the nonfinancial sector rather than the financial sector to focus on the direct to real economic activity. A shadow banking system that is just a network of “side bets” with few direct links to the real economy or that primarily funds the traditional banking system might require very different supervision and regulation than one that is inextricably linked to real economic activity. Although it may be self evident to most that the rise and collapse of shadow banking had dire effects on real economic activity, there is actually 2 For the remainder of the paper I drop the word “domestic” when referring to the shadow banking system unless the distinction with the foreign shadow banking system is explicitly needed. 2

little agreement on how best to measure the size of the shadow banking system. The purpose of this paper is to add to our ability to measuring this hard-to-measure sector. The main results are as follows. In the lead-up to the financial crisis of 2008, the domestic shadow banking system was a significant, but not dominant supplier of funding to the nonfinancial sectors of the economy. For example, nonfinancial-sector debt stood at about $34 trillion in the fourth quarter of 2008. Of that debt, I estimate that about $10 trillion was provided by the traditional banking system (either as direct loans or though holdings of securities) and $12 trillion was provided through traditional nonbank sources such as insurance companies, pension funds, and long-term mutual funds—sources that are not typically thought of as runnable. In contrast, only about $4 trillion was provided through short-term funding outside the traditional banking system. Thus, despite the well-deserved notoriety garnered by the shadow banking system, it did not account for a particularly large portion of nonfinancial-sector funding. My estimate of the size of the shadow banking system is much smaller than that provided by Pozsar et al. (2010). There are two main reasons for the difference. First, I focus on the net debt financing of the nonfinancial sector, and therefore ignore the “grossing up” of shadowbanking liabilities that occurs in long intermediation chains. Second, I do not equate all nonbank intermediation, particularly that provided by the GSEs and issuers of private-label asset-backed securities, with shadow banking. Although I find that the shadow banking system was not large compared to traditional banking in terms of the level of financing extended to the nonfinancial sector, I do find that funding from the shadow banking system dropped significantly after 2008. This contraction was, at least in an arithmetic sense, the entire reason for the slowdown in the growth rate of nonfinancial-sector debt over this period. In other words, the sharp contraction of the shadow banking system had enormous effects on nonfinancial-sector debt, and thus presumably on real economic activity. To estimate the size of the shadow banking system, I begin with the observation that every dollar of credit-market debt provided to the nonfinancial sector represents one end of a financial intermediation chain. My aim is to trace intermediation chains from nonfinancialsector borrowers to what I call terminal funders. These are not the households and foreign entities that are the “ultimate” providers of funding (with the financial system as the intermediator). Rather, these terminal funders are one or two links away from such ultimate 3

funders. I define five categories of terminal funders: Traditional banks, which include commercial banks, credit unions, and thrifts that reside in the U.S.; foreign entities, which are entities that are not domiciled in the U.S.; long-term funders, which are domestic nonbank entities such as insurance companies and pension funds that are typically not runnable; the government, which includes federal, state, and local governments (including the Federal Reserve); and short-term funders, which are domestic runnable nonbank providers of short-term financing. Short-term funders notionally include entities such as money market mutual funds (money funds), unregistered liquidity funds, local government investment pools, and cashcollateral reinvestment pools from securities lending programs. I also define intermediate funders, such as broker-dealers, government-sponsored agencies, finance companies, and private securitizers that are links between terminal funders and nonfinancial-sector borrowers. I then use data from the FFA to estimate each terminal funder’s holdings of nonfinancial-sector debt. The calculations often require “drilling down” through layers of FFA data to determine how various sectors are themselves funded. The decomposition of nonfinancial-sector debt into that which is held by the five terminal funders provides a new perspective on the relative size of the shadow banking system. Because shadow banks can provide funding to traditional banks or foreign entities, my definition of short-term funders is narrower that those that include shadow-bank funding of other terminal funders. For example, money market funds, which are clearly runnable (McCabe, 2010), provide significant funding to traditional commercial banks and foreign entities. To provide a very rough measure of shadow-bank funding of the traditional banking system, I use Call Report data to estimate the share of bank liabilities that are short-term and uninsured, and therefore potentially runnable. Fender and McGuire (2010), McGuire and von Peter (2009), and Baba, McCauley, and Ramaswamy (2009) show that foreign financial institutions, especially those in Europe, faced a short-term dollar funding squeeze during 2008 and 2009, in part because they relied heavily on U.S. money market mutual funds. Their work suggests that an important portion of foreign financing of the domestic nonfinancial sectors should also be attributed to the shadow banking system. However, a decomposition of financing provided by foreign entities to traditional and shadow banking is beyond the scope of this paper. 4

The paper ends with a brief discussion of how even an imperfect measure of the size of the shadow banking system could be useful as a tool for macro-prudential supervision of the financial system. Macro measures could provide a perspective that can complement more micro studies, such as Covitz, Liang, and Suarez (2009) and McCabe (2010), that focus on the instruments and markets that make up the shadow banking system. To use a metaphor proposed by Eichner, Kohn, and Palumbo (2010), a macro measure of the shadow banking system can provide a “grainy satellite photo” that prompts market watchers to take a closer look at particular instruments or structures. For example, evidence that nonfinancial sectors are highly dependent on the shadow banking system for funding should raise warning flags about the risks to economic activity. Indeed, although funding from the shadow banking system to the nonfinancial sectors has dropped significantly since 2008, the fact that the shadow banking system remains an important provider of financing to households, nonfinancial businesses, and governments (not to mention domestic and foreign banks and broker-dealers) should raise warning flags about risks to economic activity that arise from reliance on this inherently fragile source of funding Defining and Measuring the Shadow Banking System in a Model Financial System Figure 1 provides a highly stylized model of a financial system. The nonfinancial sector (the purple box) has borrowers with mortgage liabilities that are ultimately funded by savers in the nonfinancial sector. That is, the nonfinancial sector is the ultimate borrower and the ultimate lender, and the financial system provides the intermediation. The financial sector contains a traditional commercial bank, a mortgage securitizer, a broker-dealer, a pension fund, and a money fund. Arrows indicate financial obligations; the arrow heads indicate the direction of the obligation and the color indicates type. For example, the nonfinancial sector has a mortgage loan that it owes to the traditional bank and to the mortgage securitizer (which need not have originated the mortgage); the broker-dealer has a short-term obligation (a security repurchase agreement, or repo) to the money fund and a long-term obligation (a bond) to the pension fund; and the money fund has a short-term obligation (money-fund shares) to the nonfinancial sector. Note that the use of derivatives is outside the scope of this paper. There are multiple ways to define and measure the shadow banking system, even in this simple model. Shadow banking is often defined as the conduct of maturity transformation 5

outside the traditional banking system (Gorton and Metrick 2010; Gibson, 2010; Ricks, 2011). At least two measures of the shadow banking system could arguably satisfy this definition. First, one could interpret “outside the traditional banking system” as excluding any liabilities issued by a traditional bank. In this case, for the model in figure 1, one would add together the brokerdealer’s repo and money-fund shares outstanding because the broker-dealer funds long-term bonds with short-term repo, the money fund finances short-term commercial paper and repo using potentially shorter-term shares, and neither the broker-dealer nor the money fund is a bank. Second, one could interpret “outside the traditional banking system” to mean excluding insured deposits. In this case, one would add bank commercial paper to the first measure.3 Note that both these approaches involve some degree of double counting because the commercial paper and a portion of the repo back the money-fund shares.4 Others use broader definitions of the shadow banking system. For instance, in measuring the size of the shadow banking system, Pozsar et al. (2010) include all the asset-backed securities issued by the GSEs and private-label securitizers. In the context of the schematic in figure 1, this would entail including in a measure of the shadow banking system all the ABS issued by the mortgage securitizer.5 The approach I take in this paper differs subtly from those in the literature. I am interested in measuring the fraction of nonfinancial-sector debt that is funded by intermediation chains that are runnable. I call an intermediation chain runnable if it involves, at any link, shortterm funding outside the traditional commercial banking system. However, I am not interested (in this paper) in measuring the gross amount of shadow banking liabilities or the total liabilities of all entities that have some connection to the shadow banking system.6 Rather, I seek to measure the amount of funding of the nonfinancial system that depends on a runnable source of 3 This is implicitly the approach Ricks (2011) uses to estimate “gross private money-claims outstanding”. 4 Such double counting is more prevalent in a more complicated financial system (not shown) where, for example, the broker-dealer runs a matched-book in repo. In that case, the total amount of repo is the system would increase without any additional funding of the nonfinancial sector. Indeed, long intermediations chains or significant rehypothecation will increase some measures of the shadow banking system without resulting in more funding to the nonfinancial sector. 5 A paper by the Financial Stability Board (“Shadow Banking: Scoping the Issues”, 2011) points out that there is “no clear commonly agreed definition” of shadow banking. That paper suggests that monitoring of the shadow banking system should start with a very broad definition that includes all nonbank credit intermediation and then narrow the focus to nonbank intermediation that includes maturity or liquidity transformation. 6 Nor am I interested here in every type of nonbank maturity transformation. For example, five-year loans for very long-lived commercial real estate assets are a form of maturity transformation that is subject to significant rollover or renewal risk, but is not runnable. 6

funding. That is, I am interested in measuring the degree to which borrowing of nonfinancial entities depends quite directly on the inherently fragile shadow banking system. If key information about counterparties and loan terms for the model financial system in figure 1 were recorded at issuance and re-sale, we could in principle follow intermediation chains with relative ease from the nonfinancial borrower to the terminal funder and identify the form, prevalence, and degree of maturity transformation in the traditional and shadow banking systems. Of course, such comprehensive data do not exist, and “tagging and tracking” all financial instruments is costly and currently politically infeasible. Suppose instead that we had FFA-like data for the simple financial system in figure 1. The actual FFA is an integrated set of national financial accounts and balance sheets. The accounts include measures of financial assets and liabilities for many broad sectors of the economy which can be classified as either financial or nonfinancial. For each sector, the FFA provides sector tables that show a sector’s financial assets and liabilities broken out by the various financial instrument used. For each financial instrument, the FFA has an instrument table that shows which financial and nonfinancial sectors use that instrument to borrow or lend.7 I start by defining what I call “terminal” and “intermediate” funders. The terminal funders are not the ultimate funders of nonfinancial debt—as mentioned, the ultimate funder is the nonfinancial sector itself. Rather terminal funders are one or two links away from the ultimate funder on the intermediation chain. In this example, there are three types of terminal funders: the traditional bank, the long-term funder, and short-term funders. The traditional bank is this case is simply the commercial bank. The traditional bank has whole-loan mortgages and ABS as assets which it funds with a long-term liability to the nonfinancial sector, insured deposits held by the nonfinancial sector, and commercial paper held by the money fund. The pension fund is the long-term funder. Its assets are the ABS and the (unsecured) corporate bond, and its liabilities are the pension obligations to the nonfinancial sector. Of course, a pension fund may engage in frequent trades and may choose to quickly dump assets that it no longer wants. However, it is not typically thought of as being subject to runs. The short-term funders are defined by activities rather than entities, and are therefore not depicted as a box in the figure. Rather, a short-term funder is any nonbank provider of financing 7 The FFA presents full balance sheets for the household and nonfinancial business sectors (corporate and noncorporate). The accounts do not contain full balance sheets for the financial sectors, and therefore lack estimates of financial-sector net worth or equity. 7

using short-term, uninsured, and therefore runnable, methods. In this example, the financing from the short-term funders is the sum of the direct repo between the Broker-Dealer and the nonfinancial sector and the money fund shares. Alternatively, it can be thought of as the sum of all short-term, uninsured instruments in the financial system (the repo, the commercial paper, and the money-fund shares) netted out to eliminate double counting. The “intermediate” funders in this case are the mortgage securitizer and the brokerdealer. As intermediate funders, the holdings of nonfinancial sector debt by the mortgage securitizer and the broker-dealer are apportioned to the terminal funders as described below based on how these two entities are themselves funded. In practical terms, my approach requires following intermediation chains in figure 1 from the nonfinancial borrower to a terminal funder, and stopping there; thus the name. The model financial system in figure 1 has ten intermediation chains. The individual chains are shown in figure 2, and can be thought of as an unraveling of the intermediation chains shown in figure 1. The dollar amount of mortgage obligations held by the commercial bank is simply allocated to the traditional bank (figure 2, lines 1 through 3). Mortgage obligations held by the mortgage securitizer must be followed further along various intermediation chains. To do so, we would look at FFA data on holders of ABS. The commercial bank’s holdings of ABS are, of course, allocated to the traditional bank (figure 2, lines 4 through 6) and the pension-fund holding of ABS are allocated to the long-term funder (figure 2, line 7). For the ABS held by the broker-dealer, we must continue along the intermediation chains. At this point we would look at the sector table in the FFA for broker-dealers. To the extent that the broker-dealer funds its balance sheet using an (unsecured) corporate bond, we would allocate that amount to the long-term funder (figure 2, line 8). To the extent that the broker-dealer funds itself using repo, we would allocate that amount to the short-term funder (figure 2, lines 9 and 10). Thus, each dollar of nonfinancial debt gets allocated to one (and only one) of the three terminal funders.8 The method described above is designed to estimate how much debt of the nonfinancial sector is funded by each terminal funder regardless of how that terminal funder is, itself, funded. The portion attributed to the short-term funder is one measure of the importance of the shadow banking system, and can be compared directly to the portion attributed to the other terminal 8 Note that the approach abstracts from equity. 8

funders. However, the method ignores the extent to which traditional commercial banks are, themselves, funded by runnable sources. A question, then, is should we include the intermediation chain depicted in line 6 of figure 2 in the traditional banking system or the shadow banking system? There is no clear dividing line between the traditional and shadow banking systems. A traditional bank can raise funds through insured deposits or through non-insured “hot money” which includes short-term funding such as commercial paper and jumbo CDs that could be runnable. Moreover, banks can sponsor supposedly off-balance-sheet entities such as assetbacked conduits and money market mutual funds that are runnable and whose assets and liabilities end up, in the event of a crisis, on the sponsor’s balance sheet (Acharya, Schnabl, and Suarez, 2011; McCabe, 2010).9 By allocating to my measure of the traditional bank funder all financing provided to the nonfinancial sector by the commercial banks, I make the division between shadow and traditional banking at the point where the commercial bank legal entity ends: All funding provided by the traditional bank is considered distinct from the shadow banking system and all funding provided by off-balance sheet entities is considered distinct from the traditional banking system. Although my main focus here is on this narrow definition of the shadow banking system, I present supplemental results on a broader concept of shadow banking that includes hot money funding of traditional banks. The Estimation Method Applied to the Actual Financial System Figure 3 presents a schematic of the actual financial system that has more sectors but less detail. The nonfinancial sectors are households, nonfinancial businesses, and governments, and are represented by the large box in the figure. In addition to the three terminal funders I defined in the previous section (the traditional bank, long-term funder, and short-term funder), I add two more: foreign entities, which includes entities domiciled abroad even if they are subsidiaries of U.S. firms, and the government, which includes the federal and state and local governments and the Federal Reserve.10 9 Indeed, elements of the shadow banking system such asset-backed conduits were arguably a form a regulatory arbitrage that allowed traditional commercial banks to increase their use of short-term funding without affecting how there balance sheet looked (Acharya, Schnabl, and Suarez, 2011). 10 Domestic subsidiaries of foreign-owned firms are not considered foreign entities for the purposes of this paper and Government pension plans are classified as long-term funders. 9

I am interested in allocating all funding of the nonfinancial sector to the five terminal funders. Funding can be direct. For example, a household can owe a mortgage loan to the traditional bank, a nonfinancial firm could issue a long-term bond to a foreign entity or a longterm funder such as an insurance company, or a municipal government could issue a variablerate demand obligation that is purchased by a short-term funder such as a money fund. This direct funding is represented by the thick blue arrows in figure 3. Funding of nonfinancial borrowers can be provided indirectly through intermediate funders (the thin blue arrows and then the thick red arrow). Consider an example in which a bank originates a mortgage and then sells it to a private-label issuer of ABS. The ABS issuer funds the purchase by issuing a bond. Just as in the previous section, the portion of that bond issuance that is purchased by, say, a pension, is then said to come from a long-term funder, but through the intermediate funder.11 The assetbacked bond could also be purchased by another intermediate funder such as a broker-dealer, and funded with a repurchase agreement made with a money fund. In that case, the funding comes from the short-term funder but through two intermediate funders. The terminal funders are defined in table 1. The definitions for the traditional bank, government, and foreign entities are straightforward and are based on the FFA banking sectors, government sectors, and the rest-of-the-world sector. However, choosing the sectors to be defined as long-term funders clearly requires judgment calls. I chose sectors such as insurers and pensions that typically do not reply upon short-term funding and are generally not considered runnable. Note that I included mutual funds (excluding money-market mutual funds), closed end funds, and exchange-traded funds in the long-term funder category. Although these types of funds are highly liquid, their liabilities are not like money claims (Ricks, 2011), and I therefore do not consider them runnable in the same sense as instruments such as commercial paper and repo.12 Note also that I include money funds in the short-term funder category. This does not mean that money funds are the only short-term funder. As I mentioned above, the short-term funder category is largely characterized by activities rather than by the entities themselves. The classification of money funds captures only the direct funding of the nonfinancial sector by the 11 I cannot literally determine which portion of each type of asset is funded by different types of liabilities. In most cases this is not even a sensible question. Rather, I assign shares based on the composition of a sector’s liabilities. 12 Mutual funds, closed-end funds, and exchange-traded funds can employ leverage, some of which might create short-term liabilities, and other long-term funders such as insurance funds invest cash collateral from securities lending programs. I leave a more complete treatment of these sectors to future work. 10

short-term funder. In practice, most financing from the short-term funder category comes indirectly through the runnable financing of intermediate funders. As shown in line 1 of table 2, total credit market debt of the nonfinancial sector was $40 trillion in the fourth quarter of 2012. The upper part of the table shows the debt of the major nonfinancial sectors and the lower part of the table shows the instruments used to borrow funds.13 Given these definitions, the estimation procedure is as follows: A. For each of the identified instruments in table 2, use the appropriate FFA instrument table to calculate the share of the dollar amounts of each instrument to be allocated to each terminal funder and each intermediate funder. Apply those shares to the dollar amounts for that instrument to allocate funding to the terminal funders and the intermediate funders. B. For each intermediate funder, use the appropriate FFA sector table to estimate the share of the dollar amounts identified in (A) that should be allocated to each terminal funder and, if relevant, each intermediate funder. Apply those shares to the dollar amounts identified in (A) for intermediate funders to allocate funding to the intermediate funders and the terminal funders. C. Repeat (B) as necessary. For private-label ABS issuers, REITs, finance companies, broker-dealers, and funding corporations, use the liability structure reported in each sector’s FFA table to allocate funding to the five terminal funders. Appendix A provides a more detailed example for mortgages and the full set data and of calculations are available by request from the author. To identify the extent to which traditional banks are funded using runnable sources, I use Call Reports data to define “short-term money” at banks as the sum of large-time deposits with maturity less than 1 year, federal funds purchased and securities sold under agreements to repurchase, deposits in foreign offices, trading liabilities (excluding revaluation losses on 13 See appendix figures A.1 and A.2 for time series of the subcomponents. 11

derivatives), accounts payable, dividends declared but not yet payable, and other borrowed money with maturity less than 1 year.14 Figure 4 shows uninsured short-term liabilities at traditional banks as a share of their total assets. This share provides an admittedly rough estimate of the share of traditional bank funding that is provided by runnable sources.15 The product of this share and the estimate of traditional bank funding from step 1 provides an estimate of the shadow-bank funding that works through the traditional banking system. The remainder represents an estimate of traditional bank funding that is funded by insured deposits and long-term liabilities—that is, an estimate of the most traditional of traditional banking. Results Table 3 summarizes step A of the estimation method by providing snapshots of total debt of the nonfinancial sectors and the holders of that debt in 2006, 2008, 2010, and 2012.16 For these four years, two thirds to three quarters of nonfinancial-sector debt was held directly by one of the five terminal funders (line 2). The vast majority of that debt was held directly by traditional banks (line 3), foreign entities (line 4), and long-term funders (line 5).17 Short-term funders (line 6) have historically not been important direct holders of debt issued by nonfinancial entities. This is not surprising given that shadow banking is typically characterized by long intermediation chains. To the extent that shadow banking funds nonfinancial-sector debt, we should expect that funding to run through the financial sectors that make up the intermediate funders. Taken together, these intermediate funders held about one third of nonfinancial sector debt (line 8). Of this portion, the majority was held by the GSEs (line 9) and issuers of privatelabel ABS (line 10), two intermediate funders that were implicated in the recent shadow banking debacle. Table 4 shows the results of steps B and C of the estimation method for the GSEs and private-label ABS issuers.18 The total amount of GSE securities outstanding increased 14 Not including advances from Federal Home Loan Banks. 15 Note that even if all such funding were removed from a traditional commercial bank, the bank would still have access to financing from the Discount Window. 16 These dates were chosen to focus on the run-up to the financial crisis and the immediate aftermath. A more complete time series can be found in Appendix figure A.3. 17 These direct holdings mainly took the form of whole loans, corporate bonds, and government securities. 18 See Appendix figures A.4 through A.8 for a time series of the allocation shares. 12

substantially from 2006 through 2010 and edged up through 2012 (line 1). In 2006, most GSE securities were held by long-term funders (line 4), traditional banks (line 2), and foreign entities (line 3). Short-term funders were a decidedly minor source of funding for the GSEs in 2006, but had more than doubled their funding share by the end of 2008. That said, traditional banks, foreign entities, and long-term funders each financed more GSE securities in 2008 than did short-term funders. Following the financial crisis, short-term funders’ role in funding the GSEs collapsed and the government’s role expanded dramatically as the Federal Reserve began its Large Scale Asset Purchase program. The lower panel of table 4 presents terminal funders’ financing of private-label ABS. It is well known that much of the shadow banking system involved the purchase (often with significant leverage) of private-label ABS. My estimates indicate that short-term funders did indeed play a significant role in this sector (line 11). However, the results also indicate that most private-label securities, and therefore the underlying nonfinancial-sector debt, were actually funded by the other terminal funders. In other words, traditional banks, insurance companies, pension funds, and the like held significant quantities of private-label ABS. That short-term funders financed only funded fairly modest portions of securities issued by the GSEs and the issuers of private-label ABS is an important result of this paper. It is fairly common to consider the GSEs and private-label securitizers—in their entirety—as part of the shadow banking system (for example, see Pozsar et al., 2010 and Bakk-Simon et al., 2012). These entities are clearly enormous nonbank intermediaries that deserve enormous scrutiny. Pricing of GSE securities and private-label ABS was in many cases prompted by unjustifiably high confidence about the securities’ safety or by regulatory arbitrage, and, in the event, these securities certainly had dramatic implications for financial stability. However, a significant portion of securities issued by these sectors do not appear to have been used as inputs in the creation of runnable private money claims and therefore do not contribute significantly to my measure of shadow banking. According to this approach, securitization and shadow banking are not synonymous. Figure 5 shows the results of the estimation method applied to all nonfinancial sector debt. Short-term funders (the red portion at the bottom of the stack) have been, and remain, a quite modest source of financing for the nonfinancial sector. As suggested by the results in tables 3 and 4, much more of the funding of the nonfinancial sectors has been provided by 13

traditional banks (the combined light and dark blue areas), foreign entities (green), and long-term funders (yellow). In particular, at their peak in the fourth quarter of 2008, short-term funders provided financing for $3.7 trillion of funding to the nonfinancial sector, while traditional banks provided $10.6 trillion, long-term funders provided $11.8 trillion, and foreign funders provided $6.2 trillion. Thus, despite the justified notoriety garnered by the shadow banking system, it is, by this measure of short-term funders, remarkably small. As mentioned above, my measure of short-term funders does not include the portion of funding for traditional banks that comes from short-term and uninsured debt such as commercial paper and large time deposits. Such hot money is likely much less sticky than traditional insured deposits (and long-term liabilities) and is potentially runnable. However, even if one were to consider this portion of traditional bank funding (the light blue portion of figure 5) as part of the shadow banking system, shadow banks would still provide a quite modest portion of funding for the nonfinancial sectors. Although short-term funders and hot-money funding at banks together were not major sources of funding for the nonfinancial sectors, they played outsized roles in the changes in the debt of the nonfinancial sector. Nonfinancial-sector debt increased a cumulative 15 percent from 2006:Q4 to 2008:Q4 (line 1 of table 5). Of this increase, short-term funders contributed about 4¼ percentage points (line 2), making them the largest single contributor.19 Traditional banks, foreign entities, and long-term funders all contributed importantly to this increase. In the two years following the onset of the financial crisis, the cumulative growth rate of nonfinancial-sector debt was halved (to about 7 percent). The dramatic step-down in the growth rate of nonfinancial-sector debt was driven, at least in an arithmetic sense, by the sharp turnaround in financing from short-term funders: Short-term funders subtracted 3¾ percentage points from the cumulative growth rate over this period. Indeed, the “swing” in the contribution of short-term funders from a strong positive to a strong negative accounts for the entire 8 percentage point decline in the growth rate of nonfinancial-sector debt (the column labeled “difference”). In contrast, the swing for traditional banks (-3 percentage points) was much more modest, and was itself almost entirely driven by the swing in funding provided by uninsured short term liabilities. The swings in the contributions of foreign entities and long-term funders (lines 4 and 5) were essentially offsetting. 19 Appendix table A.3 shows each terminal funder’s cumulative growth rate. 14

Financing provided by the government skyrocketed after 2008 as the U.S. Treasury Department and the Federal Reserve System instituted a wide variety of programs in response to the financial crisis and the recession. These programs greatly boosted government funding of nonfinancial-sector debt, which had been minimal prior to the crisis (line 6). Thus, a key feature of the provision of credit to the nonfinancial sector in the run-up to the 2008 financial crisis and in its aftermath was the rise and decline of financing from the shadow banking system. A second key feature was that government entities stepped in to provide a significant portion of the credit that had been, at least in an adding-up sense, supplied by short-term funders. Table 6 summarizes changes in the funding of nonfinancial-sector debt since 2008. Debt growth has picked up somewhat (line 1). Note that short-term funders have contributed about 4 percentage points to this acceleration of nonfinancial-sector debt. Meanwhile, the contribution of long-term and government funders has dropped (lines 5 and 6). Domestic shadow-bank funding of the nonfinancial sector has increased since 2010, but remains well below the level seen right in late 2008. Comparison to other measures of shadow banking This is the first attempt of which I am aware to estimate the share of nonfinancial-sector debt that is funded by the shadow banking system. However, others have used proxies to measure the growth of the importance of the shadow banking system. For example, Gorton and Metrick (2010) used measures such as the size of broker-dealer balance sheets and the amount of repo outstanding at primary dealers to provide a rough sense of the size of the shadow banking system. The measure of Pozsar, Adrian, Ashcraft, and Boesky (2010) is more closely related to mine. Pozsar et al. use FFA data to estimate the total liabilities of the shadow banking system, defined as the sum of outstanding levels of commercial paper, repurchase agreements, GSE liabilities, GSE pool securities, liabilities of private-label ABS issuers, and shares of money market mutual funds (netted to avoid counting both sides of CP and repo transactions, resecuritizations of GSE securities, and other some sources of double-counting). Figure 6 shows their measure of shadow-banking liabilities (the black line) along with the liabilities of the traditional banking system (the blue line). That their measure of shadow-bank 15

liabilities was above the liabilities of the traditional banking system until 2010 is commonly cited evidence that the shadow banking system was as big or even bigger than the traditional banking system. The red line in the figure, which depicts my estimate of funding provided by short-term funders, suggests that the shadow banking system was (and is) not nearly that large as traditional banking in terms of credit extended to the nonfinancial sector. The vast numerical difference between the two measures stems mainly from two significant conceptual differences. First, the method of Pozsar et al. counts one dollar of funding multiple times if there are multiple observable links in an intermediation chain. For example, imagine a long intermediation chain for a $100,000 home mortgage: Suppose the mortgage is packaged into a GSE-backed mortgage pool security, which is then repackaged into a privatelabel ABS, which is then held on the balance sheet of a broker-dealer and funded through repo with a money fund. Using the method of Pozsar et al., the funding of the underlying mortgage would be counted three times—as a GSE-backed security, as a private-label ABS, and as repo— and one would therefore find $300,000 in shadow-banking liabilities. My method—which is focused on understanding how the $100,000 is funded—would allocate only the $100,000 to the short-term funder and the funding from intermediate institutions would not be counted. Second, the method of Pozsar et al. includes in shadow banking a significant portion of liabilities that I allocate to other terminal funders. In particular, by including all the liabilities of the GSEs and of private-label ABS issuers in their measure of shadow banking, Pozsar et al. attribute to the shadow banking system significant amount of financing that is actually provided by the banks, insurance companies, pension funds, and mutual funds that purchase these securities. The conceptual differences are not a matter of a clear and absolute “right” and “wrong” way to measure shadow banking. Rather, they stem from different views about what one is trying to measure. Consider the first conceptual difference, in which Pozsar et al. “gross up” the funding that occurs via long intermediation chains. If one is interested in measuring the importance of such chains, such grossing up is required. If one is interested in end-use funding of the nonfinancial sectors, one should avoid such grossing up. Both approaches are needed. The second conceptual difference between the two measures reflects the breadth of definitions for shadow banking. The term shadow banking is typically attributed to Paul McCulley (2007). He referred to the shadow banking system as “the whole alphabet soup of 16

levered up non-bank investment conduits, vehicles, and structures” that “fund themselves with un-insured commercial paper” and as such are vulnerable to runs. To McCulley and others such as Gibson (2010), Ricks (2011 and 2012), and myself, the key feature of these entities and activities is that they create something akin to private money, and as such are runnable. Thus, the relevant feature shared by traditional and shadow banks is money creation. The relevant difference is that traditional banks have direct access to the Federal Reserve’s Discount Window and can offer government-insured deposits. Shadow banks do not; that is why they are susceptible to runs. Pozsar et al. and others have defined the shadow banking system more broadly to include many kinds of financial intermediation that occurs outside of banks. Some even define shadow banking as any “credit intermediation involving entities and activities outside the regular banking system.” (FSB, 2011). In this view, the relevant feature shared by traditional and shadow banks is financial intermediation and the key difference is in regulatory regimes. If one favors a broad measure of shadow banking, the measures of Pozsar et al. (and others such as Bakk-Simon et al., 2012) are more appropriate. If one prefers a narrower definition that focuses on the creation of private money and runability, the more narrow definition used in this paper and Rick’s (2011 and 2012) approach to measuring private money claims in more appropriate. Policy makers clearly need to focus on risk taking and regulatory arbitrage conducted by nonbank financial intermediaries. But that does not mean we must call all nonbanks shadow banks. To do so seems wasteful of a new term: Why use “shadow banking” as a synonym (or near synonym) for “nonbanking” when “nonbanking” is a perfectly serviceable term? An overly broad definition of shadow banking risks diffusing the attention of policy makers and economists from the key weakness of shadow banking: its inherent susceptibilities to runs, the resulting collapse of privately issued money, and the implications for asset prices and real economic activity. Data Limitations and Potential Remedies The fundamental limitation of using aggregate data from the FFA is that, as already mentioned, such data fall short of the ideal of comprehensive information about counterparties, security types, and contract terms for all forms of lending. Several specific and salient 17

limitations follow from this fundamental issue. First, for some holders of corporate bonds, the FFA do not separately identify holdings of private-label ABS from holdings of corporate bonds issued by the nonfinancial sectors or issued by foreign entities.20 Private-label ABS holdings for traditional banks and foreign entities can be separately identified. However, estimates of privatelabel ABS holdings of long-term and short-term funders must be based on an assumption about the share of private-label ABS in their total holdings of corporate bonds. Second, the FFA do not have any direct data for unregistered domestic private investment pools such as hedge funds, private equity, and so-called “liquidity” funds.21 Any actual assets or liabilities of such funds are assigned by my method to long-term funders because the household sector in the FFA is the residual holder of most instruments.22 To the extent that these private pools are funded by any of the other terminal funders, my method will misclassify this financing. Third, from the perspective of the FFA, the foreign sector is a black box: I cannot tell what types of foreign entities fund the domestic nonfinancial sectors. I particular, I am cannot tell how much of that debt is held by foreign entities that are themselves runnable. Short of collecting data on every instrument and every counterparty, the remedies for the issues are, broadly speaking, more comprehensive and detailed data on balance sheets of financial firms. Various government agencies are already working toward this goal. The SEC recently began collecting more detailed data on the holdings of U.S. money market mutual funds, which could help identify the extent to which foreign entities are themselves runnable.23 The SEC has also begun phasing in a new data collection of balance-sheet information for hedge funds and other private funds; these data could improve our ability to monitor the shadow 20 This is actually true for all bonds issued by the financial sectors, but is particularly important for the private-label ABS sector because of their size and importance in financial intermediation. 21 Unregistered liquidity funds are similar to registered money market mutual funds but are not required to comply with rule 2a-7 and may only sell to qualified investors. 22 For example, suppose there are only two holders of United States Treasury bonds, households and banks. The methodology of the FFA is to use reported bonds outstanding from the Treasury Department and bank holdings from the Call report, and allocate the residual to the household sector. Holdings of any unmeasured sector will therefore be assigned to the household. 23 Foreign banks, especially those in Europe, faced a short-term dollar funding squeeze during 2008 and 2009, in part because they relied heavily on U.S. money market mutual funds (Fender and McGuire, 2010; McGuire and von Peter, 2009; Baba, McCauley, and Ramaswamy, 2009). It is difficult to distinguish MMFs financing of entities domiciled abroad (which are included in my measure of foreign funding) from financing of domestic entities with foreign parents (which would be excluded from foreign funding). The SEC data could potentially be combined with foreign flow of funds and banking data to better determine what portion of foreign funding is runnable. I leave this for future work. 18

banking system.24 In addition, the Office of Financial Research (OFR) was created by Congress to, among other things, improve the quality of financial-market data so that policymakers and market participants will be better able to evaluate firm-specific and market risks. In particular, the OFR intends to collect data on financial transactions and positions and create a “catalog of financial entities and instruments” (OFR, 2012). These efforts are a promising start toward improving the quality of financial statistics. Conclusion In this paper I describe a way to use data from the Flow of Funds Accounts of the United States and other readily available sources to provide a “top down” measure of how much debt financing of the nonfinancial sectors of the U.S. economy is dependent on financial intermediation chains that contain at least one runnable link. I find that in the lead-up to the financial crisis of 2008, such “shadow banking” was a significant, but not dominant supplier of funding to the nonfinancial sectors of the economy: Despite the well-deserved notoriety garnered by the shadow banking system, this portion of the financial system did not account for a particularly large portion of nonfinancial-sector funding when compared to traditional bank funded and other nonbank institutions. However, I do find that funding from the shadow banking system dropped significantly after 2008. This contraction was, at least in an arithmetic sense, the entire reason for the slowdown in the growth rate of nonfinancial-sector debt over this period. In other words, the sharp contraction of the shadow banking system had enormous effects on nonfinancial-sector debt, and thus presumably on real economic activity. Of course, this contraction did not occur in isolation. Runs on short-term funding drove asset fire sales that damaged the ability and desire of all sorts of entities to lend. In addition, shadow banking entities such as asset backed commercial conduits had recourse to traditional commercial banks and thus shadow banking losses became traditional banking losses; securitization and off-balance sheet funding had not resulted in the transfer of risk (Acharya, Schnabl, and Suarez, 2011). From a policy perspective, the approach presented in this paper offers a way to use aggregate data to track the reliance of the nonfinancial sectors on inherently fragile short-term funding markets. A high or rapidly growing reliance on such markets is suggestive evidence of 24 See SEC release: http://www.sec.gov/rules/final/2012/ia-3308-secg.htm. 19

systemic fragility that should raise warning flags for market participants and policy makers. Using the metaphor of Eichner, Kohn, and Palumbo (2010), aggregate short-term funding of the nonfinancial sectors provides a “grainy satellite photo” of the shadow banking system which should be augmented with stepped-up monitoring of specific markets, entities, and instruments. Indeed, such an approach toward financial-market monitoring has already been proposed by the Financial Stability Board (FSB, 2011). The measures in this paper will be improved by ongoing efforts to improve the collection of financial-market statistics. For example, the SEC has improved its collection of data for money market mutual funds and in the process of collecting balance sheet and other information from private investment pools such as hedge funds. In addition, the Office of Financial Research was created by Congress to, among other things, improve the quality of financial-market data so that policymakers and market participants will be better able to evaluate firm-specific and market risks. Such improved data collections are an important element in improving our understanding of the risks to financial markets and the real economy. 20

References Acharya, V.V., P. Schnabl, and G. Suarez, (2013). “Securitization without risk transfer”. Journal of Financial Economics 107, 515-536. Baba, McCauley, and Ramaswamy, (2009). “US Dollar Money Market Funds and Non-US Banks”, BIS Quarterly Review, March 2009, pp. 65-81. Bertaut, Pounder DeMarco, Kamin, and Tryon. (2011). “ABS Inflows to the United States and the Global Financial Crisis”. International Finance Discussion Papers, Number 1028, August 2011. Board of Governors of the Federal Reserve System. Covitz, Liang, and Suarez, (2009). “The Evolution of a Financial Crisis: Panic in the Asset- Backed Commercial Paper Market,” Finance and Economics Discussion Series 2009-36. Washington: Board of Governors of the Federal Reserve System. Eichner, Matthew J., Kohn, Donald L., Palumbo, Michael G., (2010). “Financial statistics for the United States and the crisis: what did they get right, what did they miss, and how should they change?”, Finance and Economics Discussion Series: 2010-20. Board of Governors of the Federal Reserve System. Fender and McGuire, (2010). “European Banks' US Dollar Funding Pressures”, BIS Quarterly Review, June 2010, pp. 57-64. Financial Stability Board, 2011. “Shadow Banking: Scoping the Issues”, www.financialstabilityboard.org/publications/r_110412a.pdf Office of Financial Research. (2012). “Strategic Framework, FY2012 - FY2014” March 2012. Gibson, Michael, (2010). “How the shadow banking system helped cause the financial crisis and how to avoid suffering the same crisis again”. Manuscript. Gorton, Gary, (2010). “E-Coli, Repo Madness, and the Financial Crisis”, Business Economics, July 2010, v. 45, iss. 3, pp. 164-73. Gorton, Gary, (2009). “Information, Liquidity, and the (Ongoing) Panic of 2007”, American Economic Review, May 2009, v. 99, iss. 2, pp. 567-72. 21

Gorton, Gary; Metrick, Andrew, (2010). “Regulating the Shadow Banking System”, Brookings Papers on Economic Activity, Fall 2010, pp. 261-97. Han and Li, (2010). “The Fragility of Discretionary Liquidity Provision: Lessons from the Collapse of the Auction Rate Securities Market”, Finance and Economics Discussion Series 2010-50. Washington: Board of Governors of the Federal Reserve System, 2010. Keohane, David (2012). “The decline of US shadow banking, charted”. FT Alphaville. “http://ftalphaville.ft.com/blog/2012/05/29/1021101/the-decline-of-us-shadow-banking-charted/ McCabe, Patrick, (2010). “The cross section of money market fund risks and financial crises”, Finance and Economics Discussion Series: 2010-51. Board of Governors of the Federal Reserve System. McCulley, Paul, (2007). “Teton Reflections”. http://www.pimco.com/Pages/GCBF%20August- %20September%202007.aspx McGuire and von Peter, (2009). “The US Dollar Shortage in Global Banking”, BIS Quarterly Review, March 2009, pp. 47-63. Pozsar, Adrian, Ashcraft, and Boesky, (2010). “Shadow Banking”, Federal Reserve Bank of New York, Staff Reports: 458. Pozsar, Zoltan, (2011). “Institutional Cash Pools and the Triffin Dilemma of the U.S. Banking System”, IMF Working Paper WP/11/190. Ricks, Morgan, (2011). “A Regulatory Design for Monetary Stability”, Harvard John M. Olin Discussion Paper No. 706. 22

Appendix A: Details on Measuring the Size of the Shadow Banking System Background on the Flow of Funds Accounts of the United States The FFA depend on a variety of data sources, including regulatory filings, public reports from government agencies such as the Bureau of Economic Analysis and the Department of the Treasury, and private data vendors. The quality and detail of the balance-sheet data varies by sector. The best data are for the government sectors, including the monetary authority (the Federal Reserve). Generally speaking, balance-sheet data for commercial banks and insurance companies are also of high quality because these institutions are required to report to various government agencies significant detail about the types of assets they hold. Banks and thrifts must file quarterly Call Reports that include fairly detailed information on assets, including loans and securities such as Treasuries, agencies, municipal debt, a wide variety of ABS categories, and structured financial products (including synthetics). Beyond this fairly detailed set of securities, banks need only report “other debt securities”.25 Insurance companies also must make fairly detailed regulatory filings. Balance-sheet data for most other financial sectors is available, but more limited. Private pension funds are a good example. The main data source for the FFA is schedule H of Form 5500.26 This form has entries for assets such as interest-bearing cash, U.S. government securities, and corporate debt instruments. However, a significant fraction of private pension fund assets are held in the form of trusts and pooled separate accounts, for which the pensions funds currently provide no additional detail. The FFA assumes that the asset allocation in these accounts is identical to that held outside the accounts. Source data for other financial sectors such as broker-dealers, mutual funds, and finance companies have similar shortcomings that prevent sufficiently detailed breakdowns of assets and liabilities. More information on the sources and methods used in the FFA can be found at the Flow of Funds Online Guide (http://www.federalreserve.gov/apps/fof/). 25 However, this catch-all category is split between foreign and domestic sectors. 26 These filings are made with the IRS, the department of Labor, the Employee Benefit Security Administration, and the Pension Benefit Guaranty Corporation. 23

An example using home mortgages A. Estimate the share of funding by each instrument to be allocated directly to each funder: For each of the nine instruments listed in table 2, the FFA has a table that shows who holds the instrument. For example, table A.1 shows FFA data on holders of home mortgages, which totaled almost $10 trillion in 2012:Q4. The bolded lines in the table show direct holdings of the terminal funders (line 2, 7, 10, 11, and 18) and the intermediate funders (line 19). Indented under each of these categories are the FFA sectors that I have assigned to each funding category. Note that most mortgages are not held directly by the terminal funders. Indeed, line 19 shows that the intermediate funders hold almost 70 percent of mortgages. Of those, most are held by the GSEs (either at the actual GSE entity or in off-balance sheet pools) and to a lesser extent at private-label ABS issuers. In some cases the total amount outstanding for an instrument will not equal the amount outstanding from the nonfinancial sector because financial and foreign sectors issue that security. For example, REITs can issue mortgage debt and foreign entities can issue dollar denominated corporate bonds. In these cases, we do not have always have estimates of who holds the security that had been issued by the nonfinancial sector. In those cases, I typically assume that all funders hold equal proportions of the financial, nonfinancial, and foreign issuance. B. Estimate funding of intermediate funders. Estimating the funding of the intermediate funders requires the most assumptions. I treat GSEs (including mortgage pools) separately from the other intermediate funders because the data for GSE are of higher quality. Table A.2 shows the terminal and intermediate funders of GSEs, which is done through agency- and GSE-backed securities. Using these data, I treat GSEs almost the same as I treat the nonfinancial sectors. The one difference is that GSEs own some GSE debt, so I must gross up all the other categories to estimate the amount of funding for the GSE sector that comes from outside the sector. Thus I am implicitly assuming that all GSEs hold other GSE debt in equal proportions. What remains is to estimate how the other five intermediate funders fund themselves. The data gaps are widest here because we do not have high-quality data on what instruments 24

these intermediaries use to fund themselves and to the extent we do know the instruments, we do not have good data on who holds them. My assumptions are as follows: Private-label ABS issuers. The FFA identifies only two sources of funding for this sector, commercial paper and bonds. Unfortunately, the FFA generally does not indentify holders of ABS separately from total corporate and foreign bonds. For depository institutions and credit unions the FFA does identify holdings of private-label MBS. I supplement these data with data from the Call Report to calculate bank holdings of nonmortgage ABS. I estimate foreign funders’ holdings of ABS using data from the Treasury International Capital System. I do not have a good estimate of holdings of private-label ABS by long-term funders. This is an area of ongoing research. One starting point is to assume that long-term funders hold private-label ABS in proportion to their holdings of all corporate bonds. Instead, I calculated the proportion of all corporate and foreign bonds held by long-term funders and scaled down by 40 percent. This likely creates an upward bias to my estimate of the share financed by the short-term funders, which is calculated as the residual. I made this scaling assumption to ensure that the short-term share was positive in all periods (figure A.4, upper panel). Indeed, in the extreme I could assume that long-term funders hold no private label ABS. Even in this extreme (and false) case, shortterm funders would remain a fairly small terminal funder of nonfinancial debt. REITs. The short-term funder share equals the share of REIT credit market debt that is in the form of either repurchase agreements or commercial paper; the traditional bank share equals the share of REIT credit market debt that is bank loans; the foreign entity share is set to zero, and the long-term funder share is the residual (figure A.5, middle panel). Finance companies. The short-term funder share equals the share of finance company credit market debt that is in the form of repurchase agreements; the traditional bank share equals the share of finance company credit market debt that is bank loans; the foreign entity share is set to zero, and the long-term funder share is the residual (figure A.6, lower panel). Broker-dealers. The long-term funder share equals the share of broker-dealer credit market debt that is in the form of corporate bonds or government funding facilities; the traditional bank share 25

and the foreign entity share are set to zero; and the short-term funder share is the residual (figure A.7, upper panel).27 Funding corporations. The long-term funder share equals the share of funding corporation credit market debt that is in the form of corporate bonds or government funding facilities28; the traditional bank and foreign entity shares are set to zero, and the short-term funder share is the residual (figure A.8, lower panel). 27 The government facilities include the Federal Reserve’s Primary Dealer Credit Facility and Asset-Backed Commercial Paper Money Market Mutual Fund Liquidity Facility 28 This includes loans extended by the Federal Reserve to Maiden Lane LLC, Maiden Lane II LLC, Maiden Lane III LLC, the Commercial Paper Funding Facility LLC, American International Group (AIG) and loans extended by the federal government to the Term Asset-Backed Securities Loan Facility and to funds associated with PPIP. 26

Table 1 Definitions of Terminal and Intermediate Funders (for use in estimating direct funding in step A) Funding source Flow of Funds Sector Terminal funders Traditional banks Commercial banks Savings institutions Credit unions Government Federal governemt Monetary authority Foreign entities Rest of the world Long-term funders Households and nonprofits Nonfinancial businesses Property-Casualty insurance companies Life insurance companies Private pension funds State and local government employee retirement funds Federal governement retirement funds Mutual funds Closed-end and exchange traded funds State and local governments 1 Short-term funders Money market mutual funds Intermediate funders Governement sponsored enterprises Agency- and GSE-backed mortgage pools Private-label issuers of asset-backed securities Finance companies Real estate investment trusts Security brokers and dealers Funding corporations 1. This designation is for the purpose of identifying direct funding in step A. It does not mean that money market mutual funds are the only kind of Short-Term Funder.

Table 2 Credit Market Debt Owed by Domestic Nonfinancial Sectors (by debtor sector, end of period, 2012:Q4) billions of dollars percent Total 40,098 --by sector Households 12,831 32.0 Nonfinancial business 12,694 31.7 State and local governments 2,980 7.4 Federal government 11,594 28.9 100.0 by instrument Commercial paper 130 0.3 Treasury securities 11,569 28.9 Agency- and GSE-backed securities 25 0.1 Municipal securities 3,714 9.3 Corporate bonds 5,795 14.5 Depository loans n.e.c. 1,751 4.4 Other loans and advances 1,385 3.5 Mortgages 12,949 32.3 Nonmortgage consumer credit 2,779 6.9

Table 3 Holders of Nonfinancial Sector Debt (end of period, 2012:Q4) 2006 2008 2010 2012 1. Grand total (billions of dollars) 3 0,059 3 4,528 3 6,913 4 0,098 Contributions (percent) 2. Direct from a Terminal Funder 6 5.3 6 5.6 7 1.3 7 4.1 3. Traditional bank 2 5.9 2 4.3 2 2.3 2 1.8 4. Foreign 1 0.4 1 2.6 1 5.5 1 7.2 5. Long-Term 2 0.7 1 9.6 2 4.5 2 4.8 6. Short-Term 2 .4 3 .8 2 .4 2 .4 7. Government 6 .0 5 .3 6 .6 7 .9 8. From an Intermediate Funder 3 4.7 3 4.4 2 8.7 2 5.9 9. GSE 1 5.7 1 7.2 1 7.4 1 6.2 10. Private-label ABS 1 2.3 1 0.6 5 .8 4 .2 11. REIT 0 .5 0 .2 0 .1 0 .1 12. Broker-dealer 0 .5 1 .0 0 .7 0 .9 13. Finance company 5 .6 4 .7 3 .8 3 .3 14. Funding corporation 0 .1 0 .8 0 .9 1 .1 Source: The Flow of Funds Accounts of the United States.

Table 4 Terminal Funders' Holdings of GSE and Private-Lable Securities 2006 2008 2010 2012 1. GSE securities (billions of dollars) 4,717 5,923 6,437 6,511 Percent allocation 2. Traditional banks 24.9 20.7 24.1 26.7 3. Foreign entities 23.2 20.2 15.0 14.9 4. Long-term funders 38.5 37.8 30.2 31.7 5. Short-term funders 6.0 14.3 7.5 8.4 6. Government 7.3 7.0 23.2 18.4 100 100 100 100 7. Private-Label Securities (billions of dollars) 3,703 3,661 2,150 1,673 Percent allocation 8. Traditional banks 16.9 22.5 20.7 25.1 9. Foreign entities 22.4 17.8 20.5 21.4 10. Long-term funders 31.2 32.0 34.7 36.2 11. Short-term funders 29.6 27.7 24.1 17.3 12. Government 0.0 0.0 0.0 0.0 Source: The Flow of Funds Accounts of the United States.

Table 5 A Decomposition of the Growth Rate of Nonfinancial-Sector Debt 2006:Q4-2008:Q4 2008:Q4-2010:Q4 Difference Percent change 1. Total 14.9 6.9 -8.0 Percentage point contributions 2. Short-term funders 4.3 -3.7 -8.0 3. Traditional banks 3.0 -0.8 -3.8 Funded by uninsured, short-term liabilities 0.8 -2.2 -3.0 Funded by insured deposits and long-term liabilities 2.2 1.4 -0.7 4. Foreign entities 3.8 2.7 -1.1 5. Long-term funders 3.2 4.0 0.8 6. Government 0.6 4.7 4.1 Source: The Flow of Funds Accounts of the United States. 14.9 6.9 -8.0

Table 6 A Decomposition of the Growth Rate of Nonfinancial-Sector Debt 2008:Q4-2010:Q4 2010:Q4-2012:Q4 Difference Percent change 1. Total 6.9 8.6 1.7 Percentage point contributions 2. Short-term funders -3.7 0.4 4.1 3. Traditional banks -0.8 1.8 2.6 Funded by uninsured, short-term liabilities -2.2 -1.2 1.0 Funded by insured deposits and long-term liabilities 1.4 3.0 1.6 4. Foreign entities 2.7 3.0 0.3 5. Long-term funders 4.0 2.3 -1.6 6. Government 4.7 1.1 -3.6 Source: The Flow of Funds Accounts of the United States.

Table A.1 Home Mortgages Outstanding (end of period, 2012:Q4) billions of dollars percent 1. Total 9,924 - 2. Traditional banks 2,836 28.6 3. U.S. chartered depository institutions 2,488 25.1 4. Foreign banking offices in U.S. 2 0.0 5. Banks in U.S.-affiliated areas 20 0.2 6. Credit unions 326 3.3 7. Government 104 1.0 8. State and local governments 78 0.8 9. Federal government 26 0.3 10. Foreign entities - 0.0 11. Long-term funders 103 1.0 12. Household sector 59 0.6 13. Nonfinancial corporate business 31 0.3 14. Property-casualty insurance companies - 0.0 15. Life insurance companies 7 0.1 16. Private pension funds 2 0.0 17. State and local govt. retirement funds 4 0.0 18. Short-term funders - 0.0 19. Intermediate funders 6,880 69.3 20. GSEs and Agency- and GSE-backed mortgage pools 5,811 58.6 21. ABS issuers 924 9.3 22. Finance companies 133 1.3 23. REITs 12 0.1 Source: The Flow of Funds Accounts of the United States.

Table A.2 Agency- and GSE-backed Securities (end of period, 2012:Q4) billions of dollars percent 1. Total 7 ,544 --- 2. Traditional banks 1 ,926 25.5 3. U.S. chartered depository institutions 1 ,668 22.1 4. Foreign banking offices in U.S. 3 2 0.4 5. Banks in U.S.-affiliated areas 3 0.0 6. Credit unions 1 98 2.6 7. Holding companies 2 5 0.3 8. Government 1 ,329 17.6 9. Federal government 0 0.0 10. Monetary authority 1 ,003 13.3 11. State and local governments 3 25 4.3 12. Foreign entities 1 ,077 14.3 13. Long-term funders 2 ,031 26.9 14. Household sector 7 3 1.0 15. Nonfinancial corporate business 2 0 0.3 16. Property-casualty insurance companies 1 24 1.6 17. Life insurance companies 3 48 4.6 18. Private pension funds 2 23 3.0 19. State and local government retirement funds 2 01 2.7 20. Federal government retirement funds 7 0.1 21. Mutual funds 1 ,035 13.7 22. Short-term funders 3 44 4.6 23. Money market mutual funds 3 44 4.6 24. Intermediate funders 8 38 11.1 25. Government-sponsored enterprises 3 15 4.2 26. ABS issuers 1 0.0 27. REITs 3 52 4.7 28. Brokers and dealers 1 70 2.2 Source: The Flow of Funds Accounts of the United States.

Table A.3 A Decomposition of the Growth Rate of Private Nonfinancial-Sector Debt 2006:Q4-2008:Q4 2008:Q4-2010:Q4 Difference Percent change 1. Total 12.1 -3.6 -15.7 2. Short-term funders 25.6 -34.5 -60.0 3. Traditional banks 9.2 -5.1 -14.3 Funded by uninsured, short-term liabilities 7.9 -24.9 -32.9 Funded by insured deposits and long-term liabilities 9.8 4.0 -5.8 4. Foreign entites 0.5 -10.1 -10.6 5. Long-term funders 11.3 -3.9 -15.2 6. Government 62.2 81.8 19.6 2008:Q4-2010:Q4 2010:Q4-2012:Q4 Difference Percent change 7. Total -3.6 4.4 8.0 8. Short-term funders -34.5 -2.3 32.2 9. Traditional banks -5.1 5.5 10.6 Funded by uninsured, short-term liabilities -24.9 -17.7 7.2 Funded by insured deposits and long-term liabilities 4.0 13.2 9.2 10.Foreign entites -10.1 1.3 11.4 11.Long-term funders -3.9 8.7 12.6 12.Government 81.8 -6.6 -88.4 Source: The Flow of Funds Accounts of the United States.

Figure 1 A model financial system Mortgage obligation Long-term debt Insured deposits Uninsured short-term liability Long-term debt Obligation to the nonfin. sector Commercial Bank Insured deposits Nonfinancial sector Money ABS Liability: Fund Mortgage Broker- Dealer MF Assets: Bond Shares Insured deposits Long-term debt MF Shares Mortgage Repo Repo Securitizer Pension obl. Pension Pension obligation

Figure 2 Taxonomy of Intermediation Chains for the Model Financial System Ultimate Intermediation Chains Ultimate Terminal Borrower Funder Funder 1. mortgage Traditional bank long-term liability N N o o 2. mortgage Traditional bank insured deposits n n f f 3. mortgage Traditional bank Commercial paper Money fund money fund shares i i Traditional bank n n 4. mortgage Mortgage securitizer ABS Traditional bank insured deposits a a 5. c mortgage Mortgage securitizer ABS Traditional bank insured deposits c i i 6. a mortgage Mortgage securitizer ABS Traditional bank commercial paper Money fund money fund shares a l l 7. mortgage Mortgage securitizer ABS Pension pension obligation S S Long-term funder 8. e mortgage Mortgage securitizer ABS Broker-Dealer bond Pension pension obligation e c c 9. mortgage Mortgage securitizer ABS Broker-Dealer repo Money fund money fund shares t t Short-term funder o o 10. mortgage Mortgage securitizer ABS Broker-Dealer repo r r Initial debt obligation Long-term obligation Uninsured short-term obligation

Figure 3 Funding of nonfinancial-sector debt Nonfinancial sectors: Households, nonfinancial businesses, governments Intermediate Traditional Foreign Long-term Short-term funders Government banks entities funder funder

Figure 4 Uninsured Short-Term Liabilities as a Share of Bank Assets Percent 36 Quarterly 32 28 24 20 Q4 16 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. Call Reports. Note. Uninsured short-term liabilities is the sum of: large-time deposits with maturity less than 1 year, federal funds purchased and securities sold under agreements to repurchase, deposits in foreign offices, trading liabilities (excluding revaluation losses on derivatives), accounts payable, dividends declared but not yet payable, and other borrowed money with maturity less than 1 year (not including FHLB advances).

Figure 5 Debt of the Nonfinancial Sector, by Terminal Funder Trillions of dollars 40 Quarterly 35 Government 30 Long-term funders 25 Q4 20 Foreign entities 15 Traditional banks (through insured deposits and long term liabilities) 10 5 Traditional banks (through uninsured short-term liabilities) Short-term funders 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. Flow of Funds Accounts of the United States.

Figure 6 Measures of Shadow Banking Trillions of Dollars 18 Quarterly Shadow bankling liabilities (method of Pozsar, Adrian, Ashcraft, Boesky, 2010) 16 14 Traditional bank liabilities Q4 12 10 8 6 Short-term funders 4 2 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. Flow of Funds Accounts of the United States.

Figure A.1 Credit Market Debt Owed by Nonfinancial Sectors (by sector) Trillions of dollars 20 Quarterly Household Nonfinancial businses State and local governments Federal government 16 Q4 12 8 4 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. United States Flow of Funds Accounts.

Figure A.2 Credit Market Debt Owed by Nonfinancial Sectors (by instrument) Trillions of dollars 20 Quarterly Mortgages Treasury securities Corporate bonds Municipal securities Consumer credit Depository loans not elsewhere classified Other 16 Q4 12 8 4 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. United States Flow of Funds Accounts.

Figure A.3 Debt of the Nonfinancial Sector, by Terminal Funder (Directly) and Intermediate Funder Trillions of dollars 40 Quarterly Short-term funders (direct) traditional banks (direct) Foreign entities (direct) Long-term funders (direct) Government (direct) 35 GSEs Private-label ABS REITs Broker-dealers Finance companies Funding corporations 30 25 20 Q4 15 10 5 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. Flow of Funds Accounts of the United States.

Figure A.4 Estimated Allocation Shares: Private ABS Percent of Total Quarterly 100 Long-term funders 80 Foreign entities 60 Q4 Traditional banks 40 Short-term funders 20 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. Flow of Funds Accounts of the United States.

Figure A.5 Estimated Allocation Shares: REIT Percent of Total Quarterly 100 Long-term funders 80 Q4 60 40 20 Short-term funders 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. Flow of Funds Accounts of the United States.

Figure A.6 Estimated Allocation Shares: Finance Companies Percent of Total Quarterly 100 80 Long-term funders 60 Q4 40 20 Traditional banks Short-term funders 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. Flow of Funds Accounts of the United States.

Figure A.7 Estimated Allocation Shares: Brokers and Dealers Percent of Total Quarterly Government 100 Long-term funders Q4 80 60 Short-term funders 40 20 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. Flow of Funds Accounts of the United States.

Figure A.8 Estimated Allocation Shares: Funding Corporations Percent of Total Quarterly 100 Government Q4 Long-term funders 80 60 Short-term funders 40 20 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. Flow of Funds Accounts of the United States.

Figure A.9 Debt of the Private Nonfinancial Sector, by Terminal Funder Trillions of dollars Quarterly 25 Government 20 Long-term funders 15 Foreign entities Q4 Traditional banks (through insured deposits and long term liabilities) 10 5 Traditional banks (through uninsured short-term liabilities) Short-term funders 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Source. Flow of Funds Accounts of the United States.