Securities Lending as Wholesale Funding:
Evidence from the U.S. Life Insurance Industry
∗Nathan Foley-Fisher Borghan Narajabad Stéphane Verani †
December 2019
Abstract
The securities lending market for corporate bonds relies on the willingness of institutional
investors to lend their bond holdings. Life insurers are major suppliers of bonds in the
securities lending market. By lending their bonds against cash collateral, insurers create
short-term liabilities that are prone to runs. We show that, controlling for bond borrowers’
demand, the maturity of an insurer’s cash collateral reinvestment portfolio determines its
decision to lend individual bonds. Insurers’ cash collateral reinvestment strategy, as part
of their interest rate risk management, drives supply in the securities lending market. Our
results suggest a new source of financial fragility.
JEL Codes: G11, G22, G23
Keywords: securities lending, wholesale funding, corporate bonds, repo, interest rate risk
management, life insurers
∗
All authors are in the Research and Statistics Division of the Federal Reserve Board. For providing valuable comments, we would like to thank, without implication, Jack Bao; Celso Brunetti; Jon Danielsson; Stefan Gissler; Michael Gordy; Diana Hancock; Yesol Huh; Sebastian Infante; Victoria Ivashina; Anastasia Kartasheva; Frank Keane; Beth Kiser; Andrew Metrick; Stefan Nagel; Florian Nagler; Michael Palumbo; Pedro Saffi; Larry Schmidt; Enrique Schroth; Adi Sunderam; Andreas Uthemann; Emily Williams; Motohiro Yogo; and participants in the SFS Cavalcade 2018; UNC/Duke Corporate Finance Conference 2017, Cass Business School Workshop on Corporate Debt Markets, UConn Risk Management Conference 2017, NBER 2016 Conference on Long-Term Asset Management, the annual meetings of the ESEM 2017, SEM 2016, EFA 2016, and the EEA 2016, as well as seminars at USI Lugano, the Riksbank, Federal Reserve Banks of Chicago, Cleveland and Philadelphia, LSE Systemic Risk Centre, BIS, Swiss National Bank, Graduate Institute Geneva, UCSB, University of Western Australia, Monash University; University of Technology Sydney; University of New South Wales, and the Federal Reserve Board. We are grateful to Della Cummings, Melissa O’Brien and Erin Hart for exceptional research assistance. The views in this paper are solely the authors’ and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or of any other person associated with the Federal Reserve System.
†
Introduction
The securities lending market for corporate bonds relies on the willingness of institutional investors to lend their bond holdings. In a securities lending transaction, securities lenders temporarily transfer economic ownership of their asset in exchange for collateral as part of a sale-and-repurchase contract that is similar to a repo transaction. Financial institutions borrow corporate bonds for a variety of well-studied reasons, including implementing short positions and arbitrage strategies. The volume of corporate bond transactions in the securities lending market is far greater than the volume of corporate bonds traded in the secondary market. The suppliers of corporate bonds to the securities lending market are primarily long-term institutional investors, such as insurers and pension funds, that hold big asset portfolios for asset-liability management or regulatory reasons. Without securities lending, all of these bonds would be locked up on the balance sheets of those long-term institutional investors.
Little is known about the supply side of the corporate bond securities lending market, as the existing literature focuses almost exclusively on the demand side. Understanding the supply side of the market is important because, by lending their bonds against cash collateral, U.S. life insurers create short-term liabilities that are prone to run risk.1 As a consequence, the decisions of corporate bond securities lenders may contribute to financial system instability. A prime example is the run on AIG’s securities lending program during the financial crisis of 2007-2009.2 We fill the gap in the literature by identifying a supply channel in the securities lending market for corporate bonds. We can identify the supply channel because we observe the loan decisions of different life insurers about the same individual corporate bonds in their portfolios. Controlling for bond borrowers’ demand, we show that the maturity of a life insurer’s cash collateral reinvestment portfolio drives its decision to lend individual bonds. We use position-level data on interest rate swaps to estimate variation in interest rate risk management across life insurers. We show that U.S. life insurers’ cash collateral reinvestment strategy, as part of
1
As is widely known, maturity and liquidity transformation are associated with vulnerabilities to runs (Diamond & Dybvig 1983, Goldstein & Pauzner 2005) and roll-over risk (He & Xiong 2012, Foley-Fisher, Narajabad & Verani forthcoming).
2
their interest rate risk management, drives their decisions to lend corporate bonds.
We construct a new annual dataset from 2011 to 2015 that combines nearly one million individual bond holdings reported by U.S. life insurers together with security-level data on their securities lending programs. We collect new annual regulatory data made available from 2011, which report the security-level composition of each insurer’s cash collateral reinvestment portfolio.
We use these data to construct measures of the degree to which individual life insurers are engaged in liquidity and/or maturity transformation through their securities lending programs. The relevant transformation for securities lending is between the overnight cash collateral and riskier long-term reinvestment securities—i.e. not the difference between the lent bond and the reinvestment securities. For example, an insurer that reinvests all of its overnight cash collateral in tri-party repo does not engage in any maturity transformation. In contrast, an insurer that reinvests its overnight cash collateral in private label residential mortgage-backed securities engages in a high degree of maturity and liquidity transformation with its securities lending program.
We combine this information with microdata on individual loan transactions from Markit Securities Finance, the most comprehensive source on the securities lending market. We construct daily market measures for each individual corporate bond: The volume-weighted average price for lending a bond, the amount of each bond that is available for lending, and the amount of each bond that is on loan.
the degree of maturity and liquidity transformation in an insurer’s cash collateral reinvestment portfolio.
We then present our empirical strategy for identifying the supply channel of securities lending. The main challenge is to obtain variation in an insurer’s decision to lend anindividual bond that is independent of borrowers’ demand for that bond. First, the demand for each corporate bond may vary with observable and unobservable time-varying and security-specific characteristics, such as expectations about the future profitability of the bond issuer. Second, the securities lending market price may not be an appropriate control for borrowers’ demand because it is an equilibrium outcome. Insurers, because of their size and concentration of holdings, may be able to influence that equilibrium outcome when they decide whether to lend their bond holding.
We overcome this empirical challenge by exploiting the cross-section of U.S. life insurers that hold the same bonds at the same time in different portfolios. We observe the individual bond lending decisions of 88 life insurers that differ by a number of observable characteristics, most importantly how they choose to reinvest their cash collateral. By including bond–time fixed effects in our specifications, we control for potentially confounding factors in a reduced form that encompasses the cost of cash borrowing as well as the availability and distribution of holdings associated with each corporate bond. We find that, controlling for individual bond demand, the cross-sectional variation in maturity and liquidity transformation in U.S. life insurers’ securities lending programs accounts for about 37 percent of the variation in their bond lending decision. As further evidence that securities lending is a source of wholesale funding for life insurers, we show that they treat securities lending and repo as substitutes. We can identify in the statutory data when each insurer uses an individual bond in a repo transaction separately from when it uses that same bond in a securities lending transaction. We show that there is a robust relationship between an insurer’s choice between the alternatives that depends, in part, on the aggregate demand for the bonds in the insurer’s portfolio. When demand is high, the insurer’s bonds are
strong for insurers with more aggressive cash collateral reinvestment strategies.
Our empirical evidence for the supply channel of securities lending prompts two important questions: Why do some life insurers have larger securities lending programs than others? And why do some reinvest their cash collateral into relatively longer-term assets? We provide an answer to these questions based on life insurer interest rate risk management, which is at the heart of life insurers’ business model. We describe how life insurers’ interest rate risk emerges naturally from the liabilities and assets on their balance sheets.
Life insurers can hedge their exposure to declining interest rates by financing a long-term fixed rate bond with short-term floating rate debt. These positions can be taken either physically or synthetically. In a physical position, for example, a life insurer may fund a ten-year Treasury using the cash collateral obtained from securities lending or repo transactions. The insurer could also take this position synthetically by entering an interest rate swap to replicate the cash flows of funding a ten-year Treasury with short-term wholesale funding.
We exploit a key difference in these hedging strategies to tease out empirical evidence for the effect of interest rate risk management on securities lending. Specifically, a life insurers’ choice of hedging strategy has a differential marginal effect on life insurers’ securities lending decisions when long-term interest rates change. Intuitively, a steeper yield curve increases the returns of life insurers who create more maturity transformation through their securities lending programs.3 By contrast, a flatter yield curve increases the cash flows to life insurers that hold received-fixed interest rate swaps. These additional cash flows affect the insurers’ marginal decision to lend bonds.
Using data on over 63,000 position-level interest rate swaps, we exploit the unusual period (2011-2015) during which short-term interest rates are fixed at the zero lower bound in the U.S. During that time period all the variation in the slope of the yield curve arises from movement in longer-term interest rates. We expand our empirical specifications to include the slope of the yield curve interacted with the overall duration added by life insurer’s swap positions and the maturity transformation created through their cash collateral reinvestment portfolios. A great advantage of our approach is that we can test the interest rate risk management hypothesis
3 As we will explain, these returns may also be related to life insurers’ decisions to reach for yield through
without needing to estimate each insurer’s duration gap between assets and insurance liabilities. We find that life insurers with more aggressive cash collateral reinvestment strategies are, on the margin, less likely to lend an individual corporate bond when longer-term interest rates
increase. This is consistent with Sen (2019), who also shows that as interest rates fall (rise), life insurers’ interest rate hedging increases (decreases). We also find that those life insurers with established larger receive-fixed interest rate swap positions are less likely to lend an individual bond when longer-term ratesdecrease. Lastly, using a triple interaction term, we show that the marginal effect of declining longer-term interest rates on an insurer’s bond lending decision is strongest when an insurer has small swap positions and creates more maturity transformation through its securities lending program.
Our evidence that interest rate risk management is a driver of the supply side of corporate bond securities lending has important financial stability implications. In a low interest rate environment, life insurers with more aggressive cash collateral reinvestment strategies are likely to become more vulnerable to runs. While maturity transformation through securities lending may help an individual insurer to hedge its interest rate risk, the activity creates a systemic risk with two potential consequences for corporate bond spot markets. First, as securities borrowers return the insurer’s corporate bonds and demand the return of their cash collateral, the insurer would likely withdraw their reinvestment of cash collateral from short-term markets, which would reduce funding liquidity and adversely affect corporate bond market liquidity (Brunnermeier & Pedersen 2009). Second, as shown in Foley-Fisher, Gissler & Verani (2019), the return of borrowed corporate bonds reduces market making activity, which could also impair the functioning of corporate bond markets.
markets use variation in security ownership, taking as given the decisions of securities lenders, to study the effects of equity securities lending. See, for example, Nagel (2005), Krishnamurthy, Nagel & Orlov (2014), Aggarwal, Saffi & Sturgess (2015), Porras Prado, Saffi & Sturgess (2016). In addition, our paper contributes to a growing literature studying the broader effects of supply-side frictions in insurance markets. For example, the seminal work of Koijen & Yogo (2015, 2016a) show that financial frictions affect life insurers’ product design, pricing, and capital structure decisions. In addition, Becker & Ivashina (2015) show that statutory capital requirements affect life insurers’ demand for corporate bonds. Chodorow-Reich, Ghent & Haddad (2018) show that life insurance investors are insulated from short-term fluctuations in asset markets during non-crisis times. Foley-Fisher et al. (forthcoming) show that runs on life insurers’ wholesale funding structures during the financial crisis had a self-fulfilling component. Ge (2019) shows that life insurance subsidiaries of an insurance conglomerate change their product prices to reallocate resources to property and casualty subsidiaries that experience weather-related losses. Sen (2019) shows that regulatory constraints restrict life insurers’ interest rate risk hedging and affect their capital structure decisions.
1
Securities lending and life insurance companies
In this section, we provide some institutional details. We first outline the typical structure of a securities lending transaction, together with the motivations of each party to the deal. Then we provide an overview of the securities lending market and the specific role of U.S. life insurers. And, finally, we discuss the distinction between the demand and supply channels of securities lending.
1.1 Securities lending transactions
In a prototypical loan, the security lender transfers full legal and economic ownership of the security to the borrower.4 In exchange, the borrower gives the lender collateral in the form of cash or another security. The term of the loan is usually open-ended, with either party able to
4
terminate the deal at any time by returning the security/collateral.5 The securities lender is free to reinvest the cash and, in some cases, rehypothecate the securities used as collateral. In the case of non-cash collateral, the securities lender earns a fee from the borrower. In the case of cash collateral, the securities lender pays a percentage of the reinvestment income to the securities borrower, called the “rebate.” Both the rebate and fee are equilibrium prices negotiated at the outset of the deal that may reflect the scarcity of the security on loan: A hard-to-find “special” security may command a high fee and a low or negative rebate. Typically, the loan is marked to market daily and is “overcollateralized,” with borrowers providing, for example, $102 in cash for every $100 in notional value of a security. The percentage of overcollateralization is called the “margin,” which serves to insure the securities lender against the cost of replacing the lent security if the borrower defaults. In addition to the loss of collateral, the security borrower is dissuaded from defaulting on the loan by reputational effects: lender–borrower relationships are formed through repeated transactions, and are often governed by a single master agreement. Overall, the structure of cash-collateralized securities lending is closely related to a sale and repurchase transaction, in which the securities borrower is entering a reverse-repo arrangement (Duffie 1996, Garbade 2006).
A securities lending transaction usually involves three or four parties. The ultimate owner of the security is typically an institutional investor such as a pension fund, insurance company, mutual fund, or sovereign wealth fund. Owners of large portfolios will often conduct their own lending programs, while smaller owners execute their programs through agent lenders, such as custodian banks or asset managers, that act as large warehouses for securities made available for lending. The end users of the borrowed securities are typically dealers and hedge funds. These security market participants generally use large financial institutions—for example, broker-dealers and investment banks—as intermediaries that regularly search for securities and have established relationships with lenders.
The ultimate owners decide which securities in their portfolios will be made available to lend and how the cash collateral proceeds of their lending programs will be reinvested. When they choose to employ agent lenders, the owners typically provide guidelines or specific instructions
5 Flexibility is often preserved, even in term loans, by allowing either party to break the terms early in
for the type of lending transactions (for example, minimum fee criteria or hard-to-find securities only) and for the reinvestment of cash collateral. In some cases, these reinvestment strategies are subject to regulatory limits.
If agent lenders are involved, they execute owners’ instructions to lend specific securities and reinvest cash collateral. Because agent lenders often have access to the same securities from many ultimate owners, they typically allocate borrowing requests to securities using an algorithm that ensures no owner receives preferential treatment. The majority of U.S. life insurers use agent lenders for their securities lending programs. This allocation process is helpful for our empirical strategy because it is hard for securities borrowers to systematically direct their requests to a particular ultimate owner. The agents earn a share of the profits associated with lending securities, including fees and/or reinvestment income after rebate. In exchange, agents will customarily provide indemnification against the risk that the non-cash collateral is insufficient to replace the lent securities if the borrower defaults. To be clear, this indemnification does not protect the owner against the risk of losses associated with reinvesting cash collateral.
The borrowing intermediary generally performs three functions as it matches end-user requests for securities with lenders’ availability. First, the intermediary helps to assuage securities lenders’ potential concerns about the credit quality of end users, which may be small and weakly regulated. Second, by establishing relationships with lenders and borrowers, they can lower search costs. In the case of broker-dealers, their securities lending intermediation is often combined with prime brokerage to lower costs further. Third, the intermediary may assume some liquidity risk by establishing open-ended loans with lenders, giving them the freedom to recall the securities as needed, and extending term loans to end users so they can be sure their short positions are covered. In exchange for these services, the borrowing intermediary receives a payment from the end user.6
The end users have a variety of reasons for borrowing a specific security. The most common motivations are to manage inventory (Faulkner 2008); to take a short position or to cover a naked short position (Duffie 1996, Keane 2013); to avoid a settlement/delivery failure (Musto, Nini & Schwarz 2018), possibly as part of market making activity; to combine one security with
6
other securities as part of an arbitrage trading strategy; to obtain collateral for use in other transactions (Dive, Hodge, Jones & Purchase 2011); and to take advantage of tax or regulatory arbitrage (Faulkner 2006). The details of these trading strategies are often complex and we refer the reader to the reference list for further explanation.
Figure 1 summarizes how the securities lending market fits into the broader map of the shadow banking system. The cloud represents the general functioning of securities markets, illustrated with the example of hedge funds taking long and short positions. Securities market participants typically borrow both cash funding and securities using broker-dealers as intermediaries. Broker-dealers obtain cash from several sources, for example, MMFs through short-term funding markets. Securities lenders decide whether to lend assets from their portfolio and whether to invest the cash collateral they receive back into short-term markets or into long-term markets.7 In the latter case, they may invest, for example, in relatively longer-term corporate bonds or asset-backed securities. Duffie, Gârleanu & Pedersen (2002) study the effect of search and bargaining in the securities lending market on pricing in the securities market, abstracting from the reinvestment decisions of securities lenders (Figure 2a). Brunnermeier & Pedersen (2009) and Gorton & Metrick (2012) consider securities market transactions funded through margin accounts and bilateral repurchase agreements (repo), abstracting from the source of securities (Figure 2b). Krishnamurthy et al. (2014) focus on the cash provided to broker-dealers from MMFs and securities lenders through short-term funding markets, taking as given the lending and reinvestment decisions of securities lenders (Figure 2c).
1.2 U.S. life insurers in the securities lending market for corporate bonds
The securities lending market for corporate bonds is large. The volume of corporate bond transactions in the securities lending market is typically several times larger than the volume of corporate bonds traded in the secondary market. The volume of bonds lent against cash collateral by U.S. life insurers in our sample is about $50 billion (Figure 4). In comparison, the average secondary market trade volume of corporate bonds is only about $30 billion (Securities Industry and Financial Markets Association Research). Comparing the volume of lending to
7
the volume of secondary market transactions is appropriate because the loans are effectively rolled-over on a daily basis.
As the largest group of institutional investors in corporate bonds, life insurers play a major role in the securities lending market for corporate bonds. The securities lending programs of other types of U.S. insurance companies, e.g. property and casualty insurers, are negligible. In 2015, U.S. life insurers held about $2.4 trillion of corporate and foreign bonds, equal to roughly one-third of the total amount outstanding (Quarterly Financial Accounts of the United States, Federal Reserve Board). Because life insurers tend to invest in corporate bonds rather than equities, their securities lending programs are heavily biased towards lending those bonds against cash collateral.8
To better understand the supply channel of securities lending requires detailed data on individual loans and cash reinvestment decisions. For this reason, the 2010 adoption by state insurance regulators of the NAIC guidelines for enhanced reporting on securities lending programs presents a golden opportunity to observe new and detailed information about all aspects of securities lending and cash reinvestment activities by U.S. life insurers. We can observe for the first time the individual bonds that are lent by life insurers, the maturity of the collateral they received, and their cash reinvestment portfolios. When combined with security-level data on the broader securities lending market, we can deepen our understanding of the strategic use of securities lending by U.S. life insurers.
2
Data
We combine several data sources to obtain the dataset we use in our analysis, covering insurance company statutory filings, bond-level descriptive statistics, and securities lending transactions. The various sources are described in the subsections below, including summary statistics.
8
2.1 Statutory filings of life insurers
The data on insurance company holdings and securities lending activity come from the NAIC Annual Statutory Filings for the end of each year from 2011 through to 2015.9 Within these filings, Schedule D contains reports of all life insurers’ individual fixed income holdings at the end of each year, together with cross-sectional information about each security, including the CUSIP indentifier and whether the bond was on loan as part of the insurer’s securities lending program or subject to a repurchase agreement (repo). We drew information about the total size and performance of the life insurer’s investment portfolio from the summary balance sheet. We focus on all insurance companies that had a securities lending program at any point during our sample period.10 Our baseline dataset includes information on 88 life insurers, with holdings data on almost one million bonds. The first four columns of Table 1 report descriptive statistics for the baseline sample. The average bond holding is about $9 million with a standard deviation of $27 million. The dummy variable for securities lending indicates that about 3 percent of U.S. life insurers’ bond holdings were on loan during the period.
Schedule DB of the NAIC Statutory Filings reports all open derivative positions, which we use to calculate a measure of duration created from interest rate swaps. We parsed the text of 63,227 individual contract-year observations reported by 42 life insurers from 2011 to 2015 and extracted the receiving leg, notional amount, and residual maturity of each contract.11 Our measure of duration for each individual fixed-to-float swap contract is0.75×the residual maturity of the contract minus 1/4×1/2, assuming that the interest rate reset on the floating leg of the swap occurs every 3 months.12 We then multiply each individual swap contract duration by their respective notional amount and divide this number by the duration of a reference 10-year
9
Historical NAIC Quarterly and Annual Statutory Filings are contained in the NAIC Financial Data Repository, a centralized warehouse of financial data used primarily by state and federal regulators.
10 We do not aggregate the data on insurance companies to the parent level because the decisions about
securities lending programs are taken at the company level.
11 The average swap notional amount is $48 million with a standard deviation of $84 million. The median
swap notional amount is $24 million.
12 The factor 0.75 is a commonly used rule of thumb when the actual swap curve is unavailable. The crude
approximation is sufficient to create variation across insurers that is correlated with more accurate measures of duration. The assumption that the interest rate reset on the floating leg of the swap occurs every 3 months is consistent with the widely used 3-month LIBOR benchmark among life insurers in our sample. It follows that the duration of a fixed-for-float swap is given by Swap durationReceive Fixedit = 0.75×Contract residual maturity−
1/4×1/2. Similarly, we calculate the swap duration of a float-for-fixed swap as Swap durationReceive Floatit =
fixed-for-float swap contract, which is calculated as 0.75×10−1/4×1/2. This calculation yields a notional amount-weighted duration for each individual swap contract that isnormalized
by the duration of a reference 10-year fixed-for-float swap contract. We sum the measures of interest rate swaps to construct the aggregate duration of the swap portfolio for each individual insurance company. Lastly, we divide by the size of the life insurer’s general account to express the aggregate duration as a fraction of the insurer’s total assets. This ratio is between -1 and 1 and is our measure of interest rate risk management. A value of zero indicates that the insurer is not adding positive or negative duration to its portfolio using swaps.
The NAIC Quarterly and Annual Statutory Filings also contain Schedule DL, a relatively new report of individual investments made by life insurers using cash collateral received from securities lending, both on- and off-balance sheets. Schedule DL was introduced in 2010 as one of many changes to the reporting and statutory accounting of securities lending transactions following the 2008-09 financial crisis.13 Figure 3 shows an extract from one life insurer’s filing in 2012 showing a sample of the individual investments made using cash collateral received in exchange for lending securities. In general, the new data allow us to better track the securities lending transactions entered into by an insurer and to observe detailed information about the life insurers’ use of the collateral received. For example, from 2010, if the collateral received from securities lending could “be sold or pledged by custom or contract by the reporting entity or its agent,” then the reinvested collateral should be recorded on the balance sheet.14 We hand-coded data about the maturity of the cash collateral received in the securities lending and repo transactions from the regulatory Note 5(e) to the Financial Statements. Because we rely on the detailed information collected as part of the new reporting requirements, our sample by necessity begins in 2011.
13
The new guidelines stem from a review of the securities lending practices at AIG that contributed to its collapse during the 2008-09 financial crisis. In particular, the guidelines specify that borrowers should post cash in the amount of at least 102 percent of domestic securities borrowed (and at least 105 percent if the securities are foreign), that individual loans should not be more than 5 percent of admitted assets, that cash reinvestment should be “prudent,” and that all cash reinvestment securities (on- and off-balance sheet) are reported in the NAIC Quarterly and Annual Statutory Filing Schedule DL. In addition, each asset financed with cash collateral recorded in the NAIC Quarterly and Annual Statutory Filing Schedule D attracts a risk-based capital charge consistent with its NAIC designation code.
http://www.dfs.ny.gov/insurance/circltr/2010/cl2010_16.htm http://www.naic.org/capital_markets_archive/110708.htm
14 Amendments to SSAP No. 91–R, Accounting for Transfers and Servicing of Financial Assets and
Figure 4 shows that the total amount of cash collateral raised by U.S. life insurers from lending securities hovered around $50 billion between 2011 and 2015. The same figure reports at an aggregated level how that cash collateral was reinvested. Each category is based on classifications determined by state regulators and reported by individual insurers. The reinvestment in relatively illiquid assets is suggested by the significant portion that is reinvested in private label ABS and corporate bonds.
2.2 Bond-level descriptive statistics
We merge life insurers’ statutory filings data with Mergent FISD using the CUSIP identifier. FISD provides a wide range of security-level information for fixed income securities, including corporate, agency, and government bonds, with a geographical focus on the U.S. While approximately one-half of all Schedule D holdings by insurers in our sample appear in FISD, 95 percent of the lent securities in our data are matched. Our interpretation is that almost all securities lent in our sample are non-privately placed fixed income bonds issued by U.S. entities. Excluding the bond holdings that do not appear in FISD reduces the size our data sample to about half a million individual bond holdings across the same set of 88 life insurers. Columns 5 through 8 of Table 1 report the additional descriptive statistics, including amount issued, offering yield, credit rating, and residual maturity. In this merged subsample, the average offering amount of the bonds held is about $800 million (with a standard deviation of $900 million) and a yield at origination of about 6 percent. The average residual maturity across all year-end bond holdings is 11.6 years (with a standard deviation of 10 years). Our numerical rating measure indicates that the average is about 20, equivalent to a Standard & Poor’s bond rating of BBB.15 Lastly, the average total amount outstanding across all bonds held by life insurers is about $800 million.
We can demonstrate further that insurers reinvest their cash collateral in relatively illiquid assets by comparing the bonds they lent (Schedule D) to the securities in which they reinvested the cash collateral they received (Schedule DL). The sample reinvestment portfolio reported in
15
Figure 3 indicates that a large proportion of CUSIP identifiers contain “#” and “@” symbols representing privately placed securities.16 Indeed, if we attempt to merge FISD by CUSIP on the reinvestment portfolios, we can match only 30 percent of individual securities even when excluding cash and cash-like reinvestments.17 Recall, by comparison, that we can match over 95 percent of bonds being lent with FISD. This contrast in match rates hints at the liquidity transformation created by securities lending programs. That said, the lack of information about privately-placed securities makes it difficult to measure the degree of maturity and liquidity transformation accurately.
Our proxy for maturity and liquidity transformation is based on the residual maturity that is reported for all types of securities in the regulatory filings. Specifically, we calculate the fraction of assets in an insurer’s cash reinvestment portfolio that have a residual maturity of more than one year minus the fraction of cash collateral that is received by the life insurer for a duration of more than one year. The one-year threshold is not crucial for the results in the paper. Rather, we choose it so that our variable represents the investment by life insurers in assets that MMFs cannot purchase for regulatory reasons.18 It follows that these assets are likely to offer a higher return than cash instruments. Figure 5 shows that there is considerable variation in the calculated fraction across life insurers and over time.
2.3 Securities lending transactions
Lastly, we add information on the market for securities lending using Markit Securities Finance (MSF). According to Markit, this dataset covers about 85 percent of the global market and more than 90 percent of the U.S. market. The daily transaction-level data include identifiers for individual lent securities, such as CUSIP and ISIN, as well as the value, quantity, duration, lending fee, rebate rate, and collateral of the loan. For each lent security, the total value and quantity of the inventory available to lend is also reported. We cannot observe counterparties
16
https://www.cusip.com/pdf/CUSIP_Intro_03.14.11.pdf
17
We identify cash and cash-like reinvestments by selecting descriptions that contain variations of the words “cash”, “money market”, “MMMF”, “prime money”, and “MMKT.”
18 Amendments to regulation Rule 2a-7, adopted by the SEC in July 2014, imposes a set of constraints on
MMMF investment portfolio, including that every security in the portfolio must have a maturity not exceeding 397 days, and that the dollar-weighted maturity of the entire portfolio cannot exceed 60 days. Thus, our one year threshold is six times the regulatory limit on the overall maturity of a mutual fund’s cash reinvestment portfolio.
to individual loans, nor information on securities lenders’ reinvestment of cash collateral. We construct weighted averages of the available variables, for each security, across all transactions conducted during the 28 days around year-end and merge with our other data using the CUSIP identifier. Roughly three-quarters of all bond holdings that appear in both regulatory filings and FISD can be matched to Markit. Moreover, consistent with the high coverage of the securities lending market, more than 95 percent of the securities that insurers report as being on loan are observed in Markit. The high proportion of bond holdings covered by Markit Securities Finance hints at the enormous potential for securities lending by U.S. life insurers.
Our final three–way merged dataset of 88 life insurers contains information on over 300,000 bond holdings, of which about 24,000 are recorded as being on loan. Columns 9 through 12 of Table 1 report the descriptive statistics for this final dataset. The information from the securities lending market suggests that the transaction-weighted average rebate on the bonds is about zero. On average, life insurers hold about 1 percent of each security’s total lendable amount (with a standard deviation of 12 percent), and a Herfindahl–Hirschman Index (HHI) of the concentration of holdings equal to 0.08. Lastly, we estimate each bond’s market tightness, defined as the ratio of the total amount lent to the total amount that is made available by securities lenders.19 We calculate these totals by aggregating the MSF data for each anonymous securities lender. The measure indicates that, on average, about 1 percent of the available amount of each security is actually lent. The remaining entries in these columns show that the other observable characteristics of the bond holdings do not vary significantly between the baseline and merged datasets.
Simple tabulations show that, conditional on the portfolios they hold, life insurers did not disproportionately lend bonds issued by certain industries. Of the bonds lent by life insurers, roughly 62 percent were issued by industrial companies, 22 percent by financial companies, 11 percent by utilities, 5 percent by government and agencies, and one percent by other institutions. The distribution of insurers’ bond lending across types of issuers is almost the same as the distribution of their bond holdings.20 Table 2 offers more detail on the types of
19The calculation of the HHI is limited by our ability to observe only life insurers’ holdings in our data—i.e.,
by necessity, it assumes atomistic holdings by other institutions.
20 About 58 percent of their bond holdings were issued by industrial companies, 22 percent by financial
bonds used in lending transactions compared with those bonds that are not lent. In general, life insurers tend to lend the bonds that they hold in larger amounts and that have a longer residual maturity in comparison with the rest of their portfolio. Life insurers also tend to lend bonds with a lower rebate (higher fee) and in which there is a greater concentration of holding and market tightness. Of course, these pairwise comparisons of characteristics are only indicative. We explore the relationship between the lending decision and the rebate in more detail in the next section.
3
Four facts about U.S. life insurers’ securities lending activity
This section presents four new facts about the association between life insurers’ decision to lend individual bonds, the bonds’ rebate rates, and the insurers’ market share in these bonds. First, life insurers are more likely to lend corporate bonds with higher equilibrium loan prices, suggesting that the supply of bonds in the securities lending market may not be perfectly elastic. Second, life insurers tend to lend those corporate bonds in which they hold a large fraction of the amount outstanding. Third, the cost of borrowing a corporate bond is a function of the distribution of holdings in that bond across potential lenders. Fourth, a life insurer’s decision to lend an individual bond is correlated with the degree of maturity and liquidity transformation in an insurer’s cash collateral reinvestment portfolio.
We are particularly interested in the third fact because it has important implications for the appropriateness of our empirical tests for the supply channel of securities lending in the next section. We show that the insurer’s market share of an individual bond relates to the insurer’s lending decision about that bond. When an insurer has market power–equivalent to a large share of the potential loan market–for a specific bond, then it would make lending decisions knowing that it can influence the securities lending market price (rebate) for lending that bond (Foley-Fisher, Narajabad & Verani 2019b). The existence of bond-specific market power means that the rebate rate is potentially not exogenous to the lending decision.
market share (Market shareijt) in bondicontrolling for the bond rebate rate as well as insurer, year, and bond issuer fixed effects. Market shareijt is the year-t holding by insurerj in bondi as a share of the total amount of the bond that is made available to securities borrowers by all lenders.21 The association between Loanijt and Market shareijt is positive and statistically significant at less than the 1 percent level. The coefficient on Market shareijt suggests that, on average, a one standard deviation (12 percent) increase in an insurer’s market share in a specific bond is associated with a 1 percent increase in the probability that the bond is loaned. The specification also includes Transformationjt, which is a proxy for the degree of maturity and liquidity transformation in insurer j’s securities lending program at time t. We calculate
Transformationjt as the fraction of assets in insurer j’s cash reinvestment portfolio that has a residual maturity of more than one year minus the fraction of cash collateral that is received by the life insurer for a duration of more than one year. The coefficient on Transformationjt
suggests that, on average, a one standard deviation (13 percent) increase in an insurer’s market share in an individual bond is associated with a 1 percent increase in the probability that the bond is loaned.
Column 2 shows that this association is robust to controlling for bond characteristics, market tightness, and concentration for individual securities in the life insurance industry. The market tightness variable is calculated usingbond-level data on the entire U.S. securities lending market and is defined as the ratio of the total amount of a bond that is on loan to the total amount of that bond that is made available by all securities lenders—i.e., insurers, pension funds, mutual funds and other lenders. Bond characteristics include the issuer, residual maturity, offering yield, offering amount, amount outstanding, and credit rating. Bond concentration is measured by the Herfindahl–Hirschman Index (HHI) computed at the bond-year level using only life insurers’ market shares in each individual bond.
Column 3 reports the results from a regression of the bond’s rebate (Rebateit) on
Market shareijt controlling for insurer, year, and bond issuer fixed effects. The coefficient on Market shareijt is positive and statistically significant at less than the 1 percent level. Column 4 shows that this association is robust to controlling for loan market tightness, holdings
21 We aggregate the inventory of every lender reporting to Markit Securities Finance to calculate the total
concentration, and other bond-level characteristics.
Taken together, the results in Table 3 suggest that lenders of hard-to-find securities get a better price on their transaction—as in Duffie (1996)—and this may affect their decision to lend specific securities. In technical terms, the results in this section show that a bond’s rebate is endogenous to an insurer’s decision to lend that bond. Moreover, it is likely that other potentially unobservable determinants of the lending decision are also correlated with the bond’s rebate. In the next section, we show how we can use our detailed data to overcome this endogeneity problem to identify the supply channel of securities lending.
4
The supply channel of securities lending
We identify the supply channel of securities lending by estimating the correlation between the lending decision measured by Loanijt, which is a binary indicator taking the value1 if insurer
j lends an individual bond i at time t, and Transformationjt, which is a proxy for the degree of maturity and liquidity transformation in insurerj’s securities lending program at time t. We calculate Transformationjt as the fraction of assets in insurer j’s cash reinvestment portfolio that has a residual maturity of more than one year minus the fraction of cash collateral that is received by the life insurer for a duration of more than one year. Figure 5 shows the variation of
Transformationjt across insurers and over time. While the mean—indicated by the horizontal blue line in the shaded rectangular region—is close to zero, there is a wide range and some insurers have almost all their cash reinvested in longer-term assets.
4.1 Identification challenge
The main empirical challenge to identify the supply channel of securities lending is to obtain variation in the decision to lend securities that is independent of demand factors. For example, although a significant correlation between Loanijt and Transformationjt is consistent with the supply channel of securities lending, any estimates of this correlation will be biased if there is unobservable variation in bond-specific demand and insurer-specific heterogeneity.
absorb heterogeneity across securities, life insurers, and report dates:
Loanijt=α1i +α2j +αt3+βTransformationjt+Zitγ+ijt . (1)
That said, the coefficient on Transformationjt when estimating Equation 1 may still be biased if the unobservable bond-specific demand and insurer-specific heterogeneity are time varying.
Moreover, including bond-specific equilibrium lending market variables as time-varying proxies for demand will likely produce inconsistent estimates of the partial correlation between the lending decision and liquidity transformation. Intuitively, either quantities traded or prices could proxy for demand if the lender is sufficiently small relative to the overall market. However, life insurers that have the potential to affect these equilibrium market variables and factors that affect the lending decision may also affect these demand proxies. To see this endogeneity problem more formally, consider the previous regression specification representing the loan decision in conjunction with a specification that represents the equilibrium rebate:
Loanijt = α1i +α2j +α3t +βTransformationjt+δRebateit+Zitγ+ijt (2)
Rebateit = α˜1i + ˜α3t + ˜βTransformationjt+Z˜itγ˜+ ˜ijt .
If any common variable in Z˜it and Zit is omitted from the loan decision specification, the estimate of the β coefficient will be inconsistent (Greene 2012). Although it is difficult to accurately gauge the severity of this endogeneity problem, the regression results presented in Table 3 and discussed in the previous section suggest it is significant.
4.2 Main results
We overcome the endogeneity problem that arises from potentially unobserved and time-varying bond demand factors by exploiting the ability to observe in our data the same bonds at the
holdings:
Loanijt =α1i +α2j +α3t+α1i ×α3t | {z }
demand control
+βTransformationjt+ijt . (3)
Recall that, since we seek only a partial correlation between Loanijt and Transformationjt that is plausibly orthogonal to bond demand, our strategy relies only on the assumption that the interaction term,α1i ×α3t, fully absorbs demand factors. In particular, we are assuming that demand does not directly affectTransformationjt. This assumption is easy to defend because the managers of U.S. life insurers’ securities lending programs are simply endowed with a portfolio of bonds over which they can make lending decisions. In the U.S. life insurance industry, the investment portfolios are determined first and foremost by asset-liability (actuarial) management considerations.22 Only after their investment portfolios have been determined by the type of insurance liabilities they issue do portfolio managers consider lending securities.23 A standard practice is for insurers to use agent lenders, who are instructed only in the management of the securities lending portfolio and are not responsible for the insurer’s asset-liability management or broad investment strategy.24
Columns 1 and 2 of Table 4 summarize our main result. Column 1 reports the baseline results of estimating equation 3 including insurer, bond, year, and bond-year fixed effects and Huber-White heteroskedasticity-robust standard errors. Column 2 reports errors two-way clustered by insurer and bond for the baseline specification as a replacement for the Huber-White standard errors, which has no effect on the statistical significance of the results. Recall that the dependent variableLoanijtis a binary indicator that takes the value1if insurerj lends an individual bond
i at time t. The coefficient on Transformationjt suggests that, controlling for corporate bond demand, life insurers with a more aggressive cash collateral reinvestment strategy is more likely to lend the same bond at the same time. Averaging the lending decision over bond holdings (collapsing by insurer-year), the variation inTransformationjt accounts for about 37 percent of
22
The traditional business of life insurers typically consists of meeting a known liability with unknown timing with a lump sum payment. Life insurers also offer annuity-type contracts that may include life and non-life contingencies. See appendix A for a detailed discussion of life insurers’ asset-liability management process.
23http://www.naic.org/capital_markets_archive/110708.htm 24
the variation in bond lending across insurers.25
The remaining columns of the table provide robustness tests of the main result. Column 3 includes insurer-bond fixed effects to control for potentially unobserved insurer-bond specific variation in the demand for lending. Column 4 includes the size of each insurer’s securities lending program (SL Program sizejt) as a determinant of the lending decision, as well as the share of each individual bond holding in an insurer’s portfolio (Bond shareijt). While these additional variables are often beyond the control of the securities lending manager, they may be correlated with securities borrowers’ demand for an insurer’s bond(s) as well as the decision to lend. In all cases, the relationship between the dependent variable and the degree of maturity transformation in the securities lending program remains statistically and economically significant.
4.3 Securities lending and repo
The evidence above strongly suggests that life insurers tap the securities lending market as a source of wholesale funding for longer-term illiquid assets. We also explored how life insurers use securities lending by contrasting with their use of another source of wholesale funding: repo. For brevity, we present the complete analysis including additional details about the use of repo by life insurers in Appendix C. One key difference between securities lending and repo transactions from an insurer’s perspective is the amount of cash collateral that can be raised using the same bond. With securities lending, insurers typically receive 102 percent of the value of the bond, while they usually receive 95 percent or less with repo. The difference reflects the counterparties in each transaction and which participant needs to be compensated for counterparty risk.
We find that securities lending and repo are substitute sources of wholesale funding for life insurers. Insurers choose between the alternatives based on the cost of funding that depends, in part, on the aggregate demand for the bonds in their portfolios. When there islow demand for an insurer’s bonds–that is, when the fraction of bond on the insurer’s balance sheet that are on “special” is low–that insurer is more likely to use repo transactions. Moreover, we show that an insurer with a more aggressive cash collateral reinvestment strategy whose bond portfolios
25
We obtain this estimate by multiplying the standard deviation of Transformationjt by the estimated
are in low demand is even more likely to use repo. This relationship is robust to controlling for the size of insurers’ securities lending and repo programs, the share of each individual bond in the insurer’s portfolio, and clustering by bond ratings to control for potential unobserved heterogeneity across the bonds used in general collateral repo transactions.
5
What drives the supply channel of securities lending?
The previous section offers overwhelming evidence of a supply channel in the securities lending market. That said, those findings prompt two questions about the supply side of the market: Why do some life insurers have larger securities lending programs than others? And why do some insurers reinvest their cash collateral into relatively longer term assets? In this section, we provide an answer based on life insurers’ management of interest rate risk.
5.1 Interest rate risk management of life insurers
Interest rate risk management is at the heart of life insurers’ business model. Life insurers write long-term liabilities, such as whole life insurance policies and life annuities contracts. These long-term liabilities are illiquid because they are not transferable from one individual to another. Life insurers back their liabilities by funding asset portfolios with the premiums they receive. To preserve their net worth, also known as surplus value or surplus ratio, life insurers try to match the cash flows of their assets and insurance liabilities. Because life insurers write fixed income liabilities, they tend to invest their premiums primarily in fixed income securities. The illiquidity of life insurance liabilities allows insurers to invest in relatively illiquid assets to offer a competitive return to policyholders.
Textbook descriptions of life insurers interest rate risk management emphasize using the duration and convexity of assets and liabilities to “immunize” net worth against small interest rate changes (Redington 1952, Fisher & Weil 1971, Kellison 1991).26 However, different cash flows on typical assets and liabilities mean that insurance companies cannot fully avoid interest rate risk (Milgrom 1985). The duration of relatively illiquid fixed income securities in the
26
U.S. is usually shorter than the duration of life insurers’ insurance liabilities (Sen 2019). For example, the duration of the widely-available single premium immediate annuity for a 65-year old female or male is about 10 years, which is more than twice the duration of the median U.S. corporate bond, which has a maturity of about 5 years (standard deviation about 6 years) (Choi, Hackbarth & Zechner 2018).27 The liability-driven investment strategy of life insurers selling insurance products in competitive markets produces asset portfolios with lower duration than their book of insurance liabilities, which the industry refers to as a negative duration gap.28
A negative duration gap means that life insurers are exposed to declining interest rates. When interest rates fall, the present value of life insurers’ assets increases at a slower rate than the present value of their liabilities. The difference between the present value of their statutory assets and liabilities is an insurer’s net worth, and is also referred to as policyholder surplus. A decrease in the long-term interest rate will erode an insurer’s net worth when life insurers have a negative duration gap, which can lead to insolvency if this risk is not adequately managed.
The life insurer’s net worth absorbs the adverse effects of small temporary declines in the interest rate. In Appendix A, we present a parsimonious model of a life insurer managing its interest rate risk in continuous time by maintaining an optimal level of net worth. To manage larger declines in interest rates and the associated widening duration gap between assets and insurance liabilities, life insurers can preserve their net worth by adding positive duration to their balance sheet. In general, this consists of financing long-term fixed rate bonds with short-term floating rate debt. These positions could be taken either physically or synthetically, but the effects are not identical. To understand the difference, consider the specific case of a life insurer financing a ten-year Treasury with short-term wholesale funding. In a physical trade, the life insurer could, for example, issue unsecured commercial paper to finance the ten-year Treasury holding. Alternatively, the life insurer can fund the ten-year Treasury with cash obtained from securities lending or repo transactions, which are alternative sources of secured borrowing. The position adds positive net duration to the insurer’s balance sheet. At the same time, the stream
27 Even if the insurer could match duration, the difference in the dispersion of cash flows remains. Many life
insurance liabilities offer steady stream of payments with no lump sum on “maturity.” By contrast, life insurers’ assets typically repay the principal investment on maturity. As a consequence, a life insurer’s liabilities usually have greater convexity than its assets.
28The size of the duration gap also depends on the options to lengthen or shorten the cash flows of a life insurer’s
of positive cash flows arising from the difference between the Treasury yield and the cost of funding boosts the life insurer’s net worth overtime. Thus, establishing the physical position evenafter interest rates decline helps the insurer both to manage their duration gap and preserve their net worth. The size of the positive cash flows is greater when insurers earn a larger term premium. So, changes in interest rate risks that drive insurers to preserve their net worth may also drive their decisions to reach for yield (Koijen & Yogo 2016b). Our empirical approach will allow us to investigate the cross-sectional variation in life insurers’ reach for yield by earning a term premium.
The insurer could also take a synthetic position by entering an interest rate swap.29 Entering into a ten-year fixed-for-float swap, the life insurer replicates the cash flows associated with funding a ten-year Treasury with short-term wholesale funding. Importantly, swaps can help to preserve an insurer’s net worth only if they are established before interest rates decline. The balance sheet cost of entering a swap is zero because the present discounted value of the fixed and floating cash flows are identical at inception. In effect, the fixed receiver obtains a positive net cash flow only when interest rates decline. Establishing a swap position after interest rates decline can help the insurer to manage their duration gap, but it will not preserve their net worth. Note that transaction costs prevent life insurers from fully hedging their interest rate risk using interest rate swaps.
In general, a life insurer can choose among alternative strategies to manage its interest rate risk depending on their relative marginal costs.30 We have already shown in Section 4.3 that securities lending and repo transactions are alternative sources of wholesale funding for life insurers. In addition, we show in Appendix D that life insurers tend to substitute between interest rate swaps and short-term wholesale funding. Insurers that create larger amounts of positive duration using interest rate swaps are also those that create smaller amounts of positive duration using activities such as securities lending and repo. We now exploit exogenous variation in the relative marginal cost of interest rate swaps over securities lending transactions to show that life insurers’ interest risk management is one driver of their decision to lend their bonds. This approach allows us to test the interest rate risk management hypothesis without needing
29
Pension funds use the same synthetic strategies to manage their interest rate risk (Shang & Hossen 2019).
30
to estimate each insurer’s duration gap between assets and insurance liabilities.31
5.2 Securities lending and interest rate risk management
To identify the connection between securities lending and interest rate risk management, we start from the observation that securities lending programs with aggressive reinvestment strategies generate additional returns per dollar of cash collateral when the yield curve becomes steeper. Similar to how banks are more profitable when there is a greater difference between their short-term funding rate and longer-short-term investments, life insurers’ securities lending programs perform better when short-term interest rates are low and the yield curve is steeper. As noted explicitly in MetLife’s 2017 SEC Form 10-K regulatory filing, the business of lending securities is boosted by low short-term rates and a steeper yield curve.32
“...there are positive offsets under the Low Interest Rate Scenario as short-term rates are much lower ... and the yield curve is steeper than that of the business plan. For example, our securities lending business performs better ... because it is driven by the slope of the yield curve rather than by the level of interest rates.”
As a consequence, when short-term interest rates are low and the yield curve flattens, securities lenders need to lendmore securities to add further positive duration to their balance sheet and build their net worth.
By contrast, existing fixed-for-float interest rate swap positions benefit the life insurer more when short-term interest rates are low and the yield curve flattens. When an established swap position is receiving a (higher) fixed long-term interest rate and the short-term interest rate remains low, a flatter yield curve implies the positive cash flows from the insurer’s swap position are greater. This relationship means that, when short-term interest rates are low and the yield curve flattens, life insurers with larger swap positions receive greater cash flows offsetting the effect of the declining interest rates on net worth. On the margin, those life insurers have
31
The size of each insurer’s duration gap is not observable because there is sparse information on the duration of insurance liabilities. Existing efforts to measure insurers’ interest rate risk relies on simplifying assumptions about the entire balance sheet (Kirti 2017, Domanski, Shin & Sushko 2017, Berends, McMenamin, Plestis & Rosen 2013).
32
relativelylessneed to use their securities lending programs to create additional positive duration and net worth.
The differential response of securities lenders to a flattening yield curve during the period of zero lower bound as a function of their swap positions offers a natural empirical test of our hypothesis that, in equilibrium, interest rate risk management drives the decision of a life insurer to lend its bonds. In other words, if interest rate risk management is affecting the securities lending decisions of life insurers, then the response of the securities lender’s decisions to a changing yield curve ought to depend on whether the life insurer has established positions with interest rate swaps. We can cleanly identify this effect during the period of zero lower bound
because all the movement in the yield curve came from changes in the upper part of the yield curve.
We can implement the test by interacting the observed slope of the yield curve with the life insurer’s swap position and cash collateral reinvestment maturity transformation. Our empirical specification is:
Loanijt= α1i +α2j+α3t +α1i ×α3t+β1Transformationjt+β2Swap durationjt+ β3Transformationjt×T10Y-3Mt+β4Swap durationjt×T10Y-3Mt+
β5Swap durationjt×Transformationjt×T10Y-3Mt+Zitγ+ijt
whereT10Y-3Mt is the 10yr-3month Treasury yield spread andSwap durationjt is a measure of the positive duration created by the life insurer’s swap position, as described in Section 2. We obtain the effect of a flattening yield curve due to a decrease in the long term rate on the decision to lend bonds by calculating the linear combination of marginal effects. Because we estimate the equation above with a data sample corresponding to the zero lower bound (2011-2015), all the variation in the yield curve arises from movements in longer-term rates. We can therefore be confident that the marginal effects we obtain are consistent with the mechanism described above.
Table 4 to include the interaction betweenTransformationjt andT10Y-3Mt. The coefficient on the interaction term suggests that life insurers whose securities lending programs are adding more positive duration through greater maturity mismatch are less likely to lend an individual bond when the yield curve becomes steeper. This result is consistent with interest rate risk management: A steeper yield curve increases the positive duration created by those securities lending programs, implying that less securities lending is required to manage the duration gap on the life insurer’s balance sheet. Column 2 includes both swap duration and securities lending programs, but excludes the interaction terms with the slope of the yield curve. The estimates chime with the cross-sectional findings reported in Appendix D of substitution between securities lending and swap positions as methods for life insurers to create positive duration.
Column 3 adds the interaction between Swap durationjt and T10Y-3Mt. Column 4 repeats the estimation including the interaction terms from both Columns 1 and 3. The results reveal a robust positive β4 coefficient, indicating that as the yield curve becomes flatter, those life
insurers with greater swap positions are less likely to lend an individual bond. This finding is consistent with the intuition that those life insurers do not need to expand their securities lending program by as much because their larger existing swap positions receive greater positive net cash flows when interest rates decline.
Lastly, Column 5 includes the triple-interaction termSwap durationjt×Transformationjt×
T10Y-3Mt. The marginal effect of a change in interest rates on the decision to lend an individual bond is
dLoanijt
dT10Y-3Mt
= ˆβ3Transformationjt+ ˆβ4Swap durationjt+ ˆβ5Swap durationjt×Transformationjt .
transformation is low, then there is no marginal effect of a flatter yield curve on the decision to lend because the securities lending program is not adding much duration. Taken together, the findings reported in Table 6 are compelling evidence that life insurers’ interest rate risk management is, at least partly, driving the supply channel of corporate bond securities lending.
6
Concluding remarks
The net benefit of more securities lending is ambiguous. On the one hand, more securities lending can improve spot market liquidity during non-crisis times. Instead of using interdealer market trading, with potentially large balance sheet costs, dealers can fulfill client orders by borrowing the bonds through the securities lending market (Foley-Fisher, Gissler & Verani 2019). In addition, most securities lenders reinvest at least some of their cash collateral in the tri-party repo market, increasing the amount of funding that is available to trading markets (Brunnermeier & Pedersen 2009).
On the other hand, more securities lending may create systemic risks. Our empirical evidence indicates that the marginal lenders of corporate bonds are those creating more maturity transformation with run risks that can amplify shocks to financial markets (Adrian, Covitz & Liang 2015).33 As we have shown, these vulnerabilities are likely to be exacerbated in a low interest rate environment.34
In light of this discussion, it should be clear that measures such as the degree of maturity and liquidity transformation in securities lending programs and the extent to which securities lenders provide tri-party repo funding are important financial stability metrics. Although some measures are available in the statutory filings of U.S. life insurers, they are not widely available for all securities lenders.35 Pension funds in particular may create similar vulnerabilities. For
33
In addition to the evidence presented by Foley-Fisher, Gissler & Verani (2019), the data curated by Risk Management Associates—used in Krishnamurthy et al. (2014)—suggest that securities lenders were the main reason that tri-party repo market funding fell. Securities lenders were forced to withdraw from these short-term funding markets when, amid widespread concerns about the quality of cash collateral reinvestment portfolios, securities borrowers ran on securities lenders. Runs occurred contemporaneously elsewhere in the financial system, including repo markets (Gorton & Metrick 2010a,b, 2012), asset-backed commercial paper (Covitz, Liang & Suarez 2013, Schroth, Suarez & Taylor 2014), MMFs (Schmidt, Timmermann & Wermers 2016), and life insurance companies (Foley-Fisher et al. forthcoming).
34 Careful monitoring of securities lenders’ programs is especially important for the functioning of corporate
bond markets, where financial stability concerns are greatest (FSR 2019).
35
example, De Nederlandsche Bank (2009) found that the aggressive securities cash collateral reinvestment strategy of Dutch pension funds contributed to their heavy losses in 2008, which prompted the industry’s exit from securities lending markets.36
References
Adrian, T., Begalle, B., Copeland, A. & Martin, A. (2012), ‘Repo and Securities Lending’,
National Bureau of Economic Research Working Paper No. 18549 .
Adrian, T., Covitz, D. & Liang, N. (2015), ‘Financial stability monitoring’, Annual Review of Financial Economics7, 357–395.
Aggarwal, R., Saffi, P. & Sturgess, J. (2015), ‘The role of institutional investors in voting: Evidence from the securities lending market’, The Journal of Finance 70(5), 2309–2346.
Asquith, P., Au, A. S., Covert, T. & Pathak, P. A. (2013), ‘The market for borrowing corporate bonds’, Journal of Financial Economics 107(1), 155–182.
Bai, J., Massa, M. & Zhang, H. (2018), ‘Security Lending and Corporate Financing: The Case of the Debt Market’, mimeo .
Becker, B. & Ivashina, V. (2015), ‘Reaching for yield in the bond market’, The Journal of Finance 70(5), 1863–1902.
Berends, K., McMenamin, R., Plestis, T. & Rosen, R. J. (2013), ‘The sensitivity of life insurance firms to interest rate changes’, Economic Perspectives 37(2).
Brunnermeier, M. K. & Pedersen, L. H. (2009), ‘Market liquidity and funding liquidity’,Review of Financial Studies22(6), 2201–2238.
Chodorow-Reich, G., Ghent, A. & Haddad, V. (2018), ‘Asset insulators’, NBER Working Paper No. 24973.
Choi, J., Hackbarth, D. & Zechner, J. (2018), ‘Corporate Debt Maturity Profiles’, Journal of Financial Economics130(3), 484–502.
Covitz, D., Liang, N. & Suarez, G. A. (2013), ‘The evolution of a financial crisis: Collapse of the asset-backed commercial paper market’, The Journal of Finance68(3), 815–848.
De Nederlandsche Bank (2009), ‘Quarterly Bulletin’.
insurers’ statutory filings lack details on the international dimensions of life insurers’ securities lending programs.
36 https://www.jpmorgan.com/cm/BlobServer/Analysing_the_Dutch_Pension_Fund_Market_From_a_
