Reaching for Yield in the Bond Market

• Published date: 01 October 2015
• Date: 01 October 2015
• Author: VICTORIA IVASHINA,BO BECKER
• DOI: http://doi.org/10.1111/jofi.12199

THE JOURNAL OF FINANCE •VOL. LXX, NO. 5 •OCTOBER 2015

Reaching for Yield in the Bond Market

BO BECKER and VICTORIA IVASHINA∗

ABSTRACT

This paper studies reaching for yield—investors’ propensity to buy riskier assets to

achieve higher yields—in the corporate bond market. We show that insurance com-

panies reach for yield in choosing their investments. Consistent with lower rated

bonds bearing higher capital requirements, insurance firms prefer to hold higher

rated bonds. However, conditional on credit ratings, insurance portfolios are system-

atically biased toward higher yield, higher CDS bonds. This behavior is related to the

business cycle being most pronounced during economic expansions. It is also charac-

teristic of firms with poor corporate governance and for which the regulatory capital

requirement is more binding.

AKEY PRINCIPLE OF FINANCE is that the evaluation of returns is only meaningful

on a risk-adjusted basis. However, risk is often hard to measure. This creates

an important limitation in the delegation of investment decisions. Financial in-

termediaries and investment managers that are evaluated based on imperfect

risk metrics face an incentive to buy assets that comply with a set benchmark

but are risky on other dimensions. In other words, imperfect benchmarks may

create incentives to “reach for yield” in the context of fixed income investing or

to reach for apparent “α” more generally. This could lead to excess risk-taking

in financial institutions, a persistent distortion in investment, and, potentially,

amplification of overall risk in the economy. Indeed, reaching for yield is be-

lieved to be one of the core factors contributing to the buildup of credit that

preceded the recent financial crisis (Rajan (2010) and Yellen1). Yet there is

∗Becker is with Stockholm School of Economics and Ivashina is with Harvard University and

NBER. We appreciate comments and suggestions from Jack Bao, Jeff Brown, Jess Cornaggia,

Hitesh Doshi, Ken Froot, Marcin Kacpercszyk (discussant), Peter Kondor,Chay Ornthanalai, Lasse

Pedersen, George Pennacchi, Andrei Shleifer, Jeremy Stein, Rene Stulz, Erik Stafford, Robert

Stambaugh, Marti Subrahmanyam, participants at the NBER Summer Institute, and seminar

participants at the Harvard Business School, Kansas City Federal Reserve, Michigan, Dartmouth,

Ohio State, New York, Bergen, Toronto, Stockholm, Miami, Tel Aviv Finance Conference, Oxford

(Sa¨

ıd), University of Pittsburgh, the SEC, Fordham, and the SFS Finance Cavalcade. We are

grateful for constructive feedback from Ken Singleton (Editor) and two anonymous referees. Matt

Kurzweil and Keith Osinski from TIAA-CREF and Matti Peltonen from the New York State De-

partment of Financial Services were very helpful in clarifying institutional details of the insurance

industry. We also thank Chris Allen, Lydia Petersen, Jennifer Rhee, and Baker Library Research

Services for assistance with data.

1Vice Chair of the Board of Governors of the Federal Reserve System Janet L. Yellen,Remarks

at International Conference: Real and Financial Linkage and Monetary Policy, Bank of Japan,

June 1, 2011, http://www.federalreserve.gov/newsevents/speech/yellen20110601a.htm.

DOI: 10.1111/jofi.12199

1863

1864 The Journal of Finance R

a lack of clear empirical evidence on this phenomenon. This paper provides

an economically important example of reaching for yield and connects it to

investment decisions and incentives.

The incentive to search for apparent αis a broad phenomenon that is not

limited to any specific asset class, but it is likely be more pronounced for illiquid

and complex securities for which risk measurement is particularly problematic.

In this paper, we study reaching for yield in the context of corporate bonds. The

risk of corporate bonds and fixed income more generally is often assessed using

credit ratings. The advantage of credit ratings as a risk measure is that, due

to standardized and well-established scales, they are comprehensible and have

broad coverage for many fixed income securities. Ratings are also not affected

by liquidity or market conditions. Not surprisingly, ratings are commonly used

for contracting and regulation purposes. However, ratings tend to be updated

slowly (Cornaggia and Cornaggia (2013)), and depend on the accuracy of rating

methodology and the absence of agency conflicts (Becker and Milbourn (2011)).2

In addition, a bond’s credit rating primarily reflects the probability of default,

and losses given default, and does not capture the risk premium associated

with the default losses and other priced factors. More formally, for a bond i

Yi−Rf ≈EDefaultLossesi+RPi+αi

=PDi×LGDi+RPi+αi

=PDCR

i×LGDCR

i+[(PDi×LGDi

−PDCR

i×LGDCR

i)+RPi]+αi,(1)

where Y−Rf is the yield spread over the Treasury bond, PD is the market-

assessed probability of default, LGD is the market-assessed loss given default,

subscript CR indicates the parameter implied by credit ratings, RP is the risk

premium, and αis the abnormal return.3A proper performance benchmark

that would not give managers an incentive to take priced risk should include

true expected default losses and the risk premium. With such a benchmark,

the investment manager would have an incentive to find underpriced, positive

αbonds. But, if investment managers are evaluated and/or regulated based on

credit ratings, in which case the benchmark is PDCR ×LGDCR,they can beat

the benchmark by raising αor reaching for yield by taking on more priced risk

not captured in the benchmark (summarized by the term in brackets).

We focus on insurance companies, the largest institutional holder of corporate

and foreign bonds. According to the U.S. Flow of Funds Accounts, in 2010

insurance companies’ holdings represented $2.3 trillion, or more than bond

2In particular, several issues have been raised about the rating of structured products (e.g.,

Benmelech and Dlugosz (2009)).

3The risk premium could be further decomposed into the risk premium associated with default

losses and an illiquidity premium: Y−Rf ≈PD ×LGD +RP +LP +α.Theriskpremiummay

depend on how loss probabilities and magnitudes are associated with aggregate risk measures

(factors).

Reaching for Yield in the Bond Market 1865

holdings of mutual and pension funds taken together.4Regulation requires

insurance companies to maintain minimum levels of capital on a risk-adjusted

basis, often called “RBC” or risk-based capital. (On average, 91% of all securities

holdings by insurance companies are in fixed income (Nissim (2010)), making

the treatment of fixed income the core component of the RBC calculation.) To

determine the capital requirement for credit risk, corporate bonds are sorted

into six broad categories (National Association of Insurance Commissioners

(NAIC) risk categories 1 through 6) based on their credit ratings, with higher

categories subject to higher capital requirements.

To test for the reaching for yield phenomenon we need an unambiguous risk

benchmark. Due to the portfolio composition, the type of regulation, and the

presence of government guarantees, the insurance industry presents an impor-

tant and clear setting for studying this phenomenon. The central hypothesis

is that insurance companies may attempt to increase the yield in their bond

portfolio by taking on extra priced risk, while leaving capital requirements

unaffected. To test this prediction, we examine promised yield spreads and

spreads on credit default swaps (CDSs)5of insurance companies’ investments,

conditional on NAIC risk categories. Cox (1967) uses the term “reaching for

yield” to describe banks’ tendency to lend to high risk borrowers as a way to

increase the promised yield. Reaching for yield in the context of insurance firms

is similar because promised yields are used to determine earnings for insurance

firms. (We discuss this in more detail in Section I.B.)

Holdings data for the analysis come from Lipper eMAXX and cover the

2004:Q1 to 2010:Q3 period, which we divide into before and after the finan-

cial crisis. The data have comprehensive coverage of quarterly fixed income

holdings by individual insurance companies, mutual funds, and pension funds.

Our basic finding for the precrisis period is illustrated in Figure 1.Inthis

figure, insurance firms’ investment behavior is benchmarked against other in-

vestors in our sample. Whereas insurance companies face capital requirements

imposed by regulators based on credit ratings, this is not the case for mutual

and pension funds. In terms of equation (1), the benchmark for insurance firms

is PDNAIC ×LGDNAIC while that for mutual and pension funds is closer to

(PD ×LGD +RP). Because of this difference in risk assessment for insurance

firms, if insurers reach for yield we should find higher relative propensity to buy

high-yielding assets. (This is not to say that mutual and pension funds do not

reach for yield, but reaching behavior for these institutions would be driven by

factors other than credit ratings and manifest itself differently.)Figure 1, Panel

A shows that, consistent with risk-weighted capital requirements, insurance

companies exhibit a strong preference for safer bonds. However, the opposite

is true within regulatory risk categories (Figure 1, Panel B). For securities

4Investment decisions of insurance companies are also important because, like banks, insurance

companies have liabilities to a broad population base.

5An advantage of using CDS spreads in addition to yields is that they may be affected by

illiquidity in a different way from positive net supply assets like bonds (Bongaerts, De Jong, and

Driessen (2011)).