the risk management paradox - fuqua school of business

The Risk Management ParadoxWhy Firms Should Hedge and Why Many Don’t

Adriano A. RampiniDuke University

Fuqua Forum – FinanceNovember 13, 2020

Adriano A. Rampini The Risk Management Paradox

The Risk Management Paradox

Why firms should hedge and why many don’t

Motivating evidence on risk management

Why should firms hedge? – Existing theory

Ensure sufficient funds if cash flows or net worth drops

Why do many firms not hedge? – Our theory

Hedging requires collateral that firms rather use elsewhere

Which firms hedge? – Empirical evidence on risk management

Airlines (fuel price); banks (interest rates and foreign exchange)

Basic pattern: firms with limited internal funds hedge less

Paradox: Financing is reason for and obstacle to hedging


Research Agenda on Risk Management

Research agenda with S. “Vish” Viswanathan (Duke) and others

Theory

Rampini/Viswanathan, Collateral, risk management, and thedistribution of debt capacity, Journal of Finance 65 (2010)2293-2322

Rampini/Viswanathan, Collateral and capital structure, Journal ofFinancial Economics 109 (2013) 466-492

Rampini/Viswanathan, Financing insurance, Working paper (2019)

Evidence

Rampini/Sufi/Viswanathan, Dynamic risk management, Journal ofFinancial Economics 111 (2014) 271-296

Rampini/Viswanthan/Vuillemey, Risk management in financialinstitutions, Journal of Finance 75 (2020) 591-637


Motivating Evidence on Risk Management

Substantial variation in fuel price hedging across airlines

model that we take to the data is that in an environment inwhich collateral constraints apply to both external financeand risk management activity, risk management is lowerwhen the marginal value of net worth today is very high.And the marginal value of current net worth is very high

when the level of current net worth is low. The reason forthis in the model is the concavity of the value function ofthe firm, which is induced by the concavity of theproduction function and the limited liability and collateralconstraints. It is as if the firm is risk averse with regard tonet worth, and hence the marginal value of net worth ishigh when net worth is low. The state variable is networth, not the amount of collateral itself, as the amount ofcollateral is endogenously determined by the investmentpolicy, given net worth.

Net worth in our model is the sum of current cash flow,the value of capital net of debt, plus the payoff of anycontingent claims used to hedge. We use a total of fivevariables as the empirical analogs of net worth, the statevariable in the model. Close analogs are the total marketvalue and total book value of net worth, and we use these astwo of our measures of net worth. The former is defined asthe market value of the airline less the book value ofliabilities. The latter is the book value of shareholders' equity.

In the model, airlines all have the same productionfunction. In practice, the production function could differacross airlines, which would justify scaling net worth bysome notion of the scale of the airline. To better capturethe potential cross-sectional differences in the scale ofairlines, we also use the above measures of net worthscaled by the total assets of the airline.

We also use the credit rating of the airline. While creditratings measure more than just the internal resources ofthe firm, a poor credit rating captures situations in whichfinancing is particularly costly and internal resources havea high marginal value. We find consistent results across allfive measures of net worth.

In the model, net worth is the predetermined statevariable and variation in net worth determines variation inhedging decisions. When testing this prediction in the data, aconcern is that omitted variables could be simultaneouslydriving net worth and hedging activity. While the empiricalresults are consistent with our interpretation of this correla-tion, causality could still be a concern. One way to addressthis concern is to isolate a precise source of variation in networth through a two-stage least squares estimation.

In determining an instrument for net worth, ourprimary goal is to stay as close to the theory as possible.In the theory, net worth is determined by two exogenousfactors: input prices and productivity. Input prices, that is,fuel prices, likely affect all airlines in a similar manner, andso we focus on productivity as the main source of idiosyn-cratic variation in net worth. We measure productivityusing observed operating income. In Section 5.3, we reportspecifications in which we instrument for net worth usingoperating income. The two-stage least squares estimationyields qualitatively similar results.

5. Hedging and net worth

Our theory predicts that less constrained firms engagein more risk management. Consistent with our theory, weshow in this section that a strong positive correlationexists between airlines' fuel price hedging and net worthboth in the cross section and within airlines over time.Importantly, our data allow us to study the within-firm

Fig. 2. Fuel expense hedging by airline. This figure presents the averagefraction of next year's fuel expenses hedged for each airline. The averageis computed over all years the airline is in the sample. Panel A includesthe full sample. Panel B excludes any airline that has a fuel pass throughagreement at any point in the sample.

Fig. 3. Fuel expense hedging in the time series. This figure presents theaverage fuel cost per gallon across airlines by year and the averagefraction of next year's fuel expenses hedged across airlines by year. Thelonger dashed line represents the weighted average of next year's fuelexpenses hedged when the weights are total assets of the airline.

A.A. Rampini et al. / Journal of Financial Economics 111 (2014) 271–296282

Airlines asempiricallaboratory

Data: 1996-2009

Airlines hedge≈ 20% on average

Southwest hedgesmost (> 50%)

Many airlineshedge very little

What explainsthis variation?


Why Firms Should Hedge – Existing TheoryFirms should hedge to ensure sufficient funds when cash flows drop

Froot/Scharfstein/Stein, Risk management: coordinating corporateinvestment and financing policies, Journal of Finance 48 (1993)1629-1658

2312 The Journal of Finance R©

Figure 3. Risk management as in Froot, Scharfstein, and Stein (1993). With completemarkets, perfect enforcement at time 1, and no financing need at time 0, the optimal hedgingpolicy equalizes the marginal value of net worth μ1(s) across states at time 1.

E[b1] ≥ 0, (13)

assuming for simplicity that A2(s) is sufficiently high in all states that thecollateral constraints bind at time 1 in all states and that qt(s) = 1,∀s ∈ S, t ∈{1, 2}. Note that investment is a choice at time 1 only (k1(s), s ∈ S), that networth at time 1 w1(s) is exogenous, and that there are frictionless marketsallowing transfers across states (b1(s), s ∈ S). Denoting the multipliers on (12)by π (s)μ1(s), the first-order conditions with respect to b1(s),∀s ∈ S, imply

μ1(s) = μ1(s), ∀s, s ∈ S, (14)

that is, the marginal value of net worth is equalized across states. This equa-tion is equivalent to equation (28) in Froot, Scharfstein, and Stein (1993). Ifinvestment opportunities are nonstochastic, that is, A2(s) = A2,∀s ∈ S, then in-vestment is independent of net worth, that is, k2(s) = E[w1]/(1 − R−1θ ),∀s ∈ S,again paralleling their results. Figure 3 illustrates the prediction of this modelin the two-state case.

In our environment enforcement is limited in all periods, not just in thesecond period. With limited enforcement of payments at time 1, the problemin equations (11) to (13) needs to be solved subject to additional collateralconstraints on time 1 payments, which require that θk0 ≥ Rb1(s),∀s ∈ S. If allcollateral constraints are slack, the marginal value of net worth is equalizedacross states as before. However, in general the marginal value of net worthis not equalized across states, unlike in Froot, Scharfstein, and Stein (1993).Rather, it is the sum of the marginal value of net worth and the multiplieron the collateral constraint that is equalized across states in our model. Forall states for which the collateral constraint binds, the firm pledges as muchas possible and in this sense the model predicts what one might call maximalhedging, which is similar to the prediction in Froot, Scharfstein, and Stein(1993).

Firms risk neutral

Suppose two states

High state plenty of funds

Low state too few funds;forced to downsize

Hedging transfers fundsfrom high to low state

Avoids downsizing

Conclusion: Firms should hedge when concerned about limited funds

Puzzle: Why do firms (especially with limited funds) hedge so little?

Stulz (1996): “The actual corporate use of derivatives, however,does not seem to correspond closely to the theory.”


Why Many Firms Don’t Hedge – Our Theory

Risk management subject to financial constraints

THE JOURNAL OF FINANCE • VOL. LXV, NO. 6 • DECEMBER 2010

Collateral, Risk Management, and theDistribution of Debt Capacity

ADRIANO A. RAMPINI and S. VISWANATHAN∗

ABSTRACT

Collateral constraints imply that financing and risk management are fundamentallylinked. The opportunity cost of engaging in risk management and conserving debtcapacity to hedge future financing needs is forgone current investment, and is higherfor more productive and less well-capitalized firms. More constrained firms engagein less risk management and may exhaust their debt capacity and abstain from riskmanagement, consistent with empirical evidence and in contrast to received theory.When cash flows are low, such firms may be unable to seize investment opportunitiesand be forced to downsize. Consequently, capital may be less productively deployedin downturns.

FINANCING AND RISK MANAGEMENT are fundamentally linked as both involvepromises to pay that are limited by collateral constraints. Engaging in riskmanagement and conserving debt capacity have an opportunity cost—currentinvestment is forgone. This cost is higher for more constrained firms. This in-sight has important implications for the extent of corporate risk management.

We provide a dynamic model of collateralized firm financing in which firmshave access to complete markets, subject to collateral constraints due to lim-ited enforcement, and hence are able to engage in risk management. Firmsmay choose to conserve debt capacity to take advantage of future investmentopportunities. Our model predicts that firms with less internal funds exhausttheir debt capacity rather than conserve it, rendering them unable to seize

∗Adriano A. Rampini and S. Viswanathan are at Duke University. We thank Michael Fishman,Denis Gromb, Jeremy Stein, two referees, and the Acting Editor, as well as Amil Dasgupta, Dou-glas Diamond, Emmanuel Farhi, Alexander Gumbel, Yaron Leitner, Christine Parlour, GuillaumePlantin, David Scharfstein, David Skeie, Rene Stulz, and seminar participants at the FederalReserve Bank of New York, Southern Methodist, Duke, Michigan State, INSEAD, Vienna, Stock-holm School of Economics, Mannheim, Illinois, Zurich, British Columbia, Toronto, Minnesota, ETHZurich, Ohio State, Columbia, Maryland, Washington University in St. Louis/the Federal ReserveBank of St. Louis, Collegio Carlo Alberto, the European University Institute, the Jackson Hole Fi-nance Group, the University of Chicago GSB conference “Beyond Liquidity: Modeling Frictions inFinance and Macroeconomics,” the 2008 Basel Committee/CEPR/JFI conference, the Paul WoolleyCentre conference at LSE, the 2008 North American Summer Meeting of the Econometric Society,the 2008 SED Meeting, the 2008 NBER Summer Institute in Capital Markets and the Economy,the 2008 Minnesota Workshop in Macroeconomic Theory, the 2008 Washington University Con-ference on Corporate Finance, the 2008 Conference of Swiss Economists Abroad, the 2009 AEAMeeting, the European Winter Finance Conference, and the 2009 FIRS Conference for helpfulcomments and Wei Wei for research assistance. This paper was previously circulated under thetitle “Collateral, Financial Intermediation, and the Distribution of Debt Capacity.”

2293

Collateral and capital structure$

Adriano A. Rampini n, S. ViswanathanDuke University, Fuqua School of Business, 100 Fuqua Drive, Durham, NC 27708, USA

a r t i c l e i n f o

Article history:Received 13 February 2012Received in revised form15 January 2013Accepted 16 January 2013Available online 15 March 2013

JEL classification:D24D92E22G31G32G35

Keywords:CollateralCapital structureRisk managementLeasingTangible assets

a b s t r a c t

We develop a dynamic model of investment, capital structure, leasing, and risk manage-ment based on firms' need to collateralize promises to pay with tangible assets. Bothfinancing and risk management involve promises to pay subject to collateral constraints.Leasing is strongly collateralized costly financing and permits greater leverage. Moreconstrained firms hedge less and lease more, both cross-sectionally and dynamically.Mature firms suffering adverse cash flow shocks may cut risk management and sell andlease back assets. Persistence of productivity reduces the benefits to hedging low cashflows and can lead firms not to hedge at all.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

We argue that collateral determines the capital struc-ture and develop a dynamic agency-based model of firm

financing based on the need to collateralize promises topay with tangible assets. We maintain that the enforce-ment of payments is a critical determinant of both firmfinancing and whether asset ownership resides with theuser or the financier, i.e., whether firms purchase or leaseassets. We study a dynamic neoclassical model of the firmin which financing is subject to collateral constraintsderived from limited enforcement and firms choosebetween purchasing and renting assets. Our theory ofoptimal investment, capital structure, leasing, and riskmanagement enables the first dynamic analysis of thefinancing vs. risk management trade-off and of firmfinancing when firms can rent capital.

In the frictionless neoclassical model, asset ownershipis indeterminate and firms are assumed to rent all capital.The recent dynamic agency models of firm financingignore the possibility that firms rent capital. Of course, africtionless rental market for capital would obviate finan-cial constraints. We explicitly consider firms' dynamic

Contents lists available at SciVerse ScienceDirect

journal homepage: www.elsevier.com/locate/jfec

Journal of Financial Economics

0304-405X/$ - see front matter & 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.jfineco.2013.03.002

☆ We thank Michael Brennan, Francesca Cornelli, Andrea Eisfeldt,Michael Fishman, Piero Gottardi, Dmitry Livdan, Ellen McGrattan, LukasSchmid, Ilya Strebulaev, Tan Wang, Stan Zin, an anonymous referee, andseminar participants at Duke University, the Federal Reserve Bank of NewYork, the Toulouse School of Economics, the University of Texas at Austin,New York University, Boston University, MIT, Virginia, UCLA, Michigan,Washington University, McGill, Indiana, Koc, Lausanne, Yale, Amsterdam,Emory, the 2009 Finance Summit, the 2009 NBER Corporate FinanceProgram Meeting, the 2009 SED Annual Meeting, the 2009 CEPREuropean Summer Symposium in Financial Markets, the 2010 AEAAnnual Meeting, the 2010 UBC Winter Finance Conference, the IDCCaesarea Center Conference, the 2010 FIRS Conference, the 2010 Econo-metric Society World Congress, and the 2011 AFA Annual Meeting forhelpful comments and Sophia Zhengzi Li for research assistance.

n Corresponding author. Tel.: þ1 919 660 7797; fax: þ1 919 660 8038.E-mail address: [email protected] (A.A. Rampini).

Journal of Financial Economics 109 (2013) 466–492

Insight: Collateral needed to raise financing and for risk management


Why Many Firms Don’t Hedge – Our Theory

Hedging requires funds (or collateral)

Collateral, Risk Management, and the Distribution of Debt Capacity 2313

Figure 4. Financing and risk management subject to collateral constraints. With com-plete markets subject to collateral constraints and a financing need at time 0, financing needs mayoverride hedging concerns, as the figure illustrates; in this case the optimal hedging policy equal-izes the sum of the marginal value of net worth and the multiplier on the collateral constraint,μ1(s) + λ1(s), across states at time 1, as μ0 = Rμ1(s) + Rλ1(s), ∀s ∈ S, from equation (A4).

The prediction is overturned, however, when the problem is further amendedto allow for investment at time 0, as we do in our model. In this dynamicenvironment there are competing demands on the firm’s ability to promise.The financing needs for investment may override the hedging concerns. Asillustrated in Figure 4 for the case with two states, if the need to shift funds totime 0 is sufficiently strong, then all collateral constraints bind and no fundsare shifted across states at time 1, so there is no risk management. Indeed, byProposition 7, financing needs necessarily override hedging concerns if a firm’snet worth is sufficiently low. Thus, the absence of risk management by severelyconstrained firms should not be considered a puzzle.

C. Implementation with Forwards and Futures

Our model entails state-contingent borrowing and thus has complete marketsfor Arrow securities, subject to collateral constraints. One way to implementthe optimal contract is for firms to make only the minimum down paymentrequirement and hedge net worth by keeping financial slack using completeoptions markets subject to short sale constraints as in Section I.D. This imple-mentation makes the opportunity cost of hedging transparent as the requiredoptions premia are paid in advance.

In contrast, forwards and futures may seem to get around the financing con-straints as they require no payment at time 0. Our model shows that this intu-ition is misleading, however, because forwards and futures involve promises topay at time 1 and such promises are limited by collateral constraints. Hence,

Hedging requires funds

Firms constrained now

Borrow funds from futurestates

Hedging would shift fundsto low state tomorrow

But financing operationstoday more urgent

Constrained firms use limited funds for operations not hedging

Model


Empirical Patters in Risk Management

Evidence on airlines and banks

Dynamic risk management$

Adriano A. Rampini a,n, Amir Sufi b, S. Viswanathan a

a Duke University, Fuqua School of Business, 100 Fuqua Drive, Durham, NC 27708, USAb University of Chicago, Booth School of Business, 5807 South Woodlawn Avenue, Chicago, IL 60637, USA

a r t i c l e i n f o

Available online 16 October 2013

JEL classification:D92E22G32

Keywords:CollateralRisk managementCommodity pricesFinancial distressAirlines

a b s t r a c t

Both financing and risk management involve promises to pay that need to be collater-alized, resulting in a financing versus risk management trade-off. We study this trade-offin a dynamic model of commodity price risk management and show that risk manage-ment is limited and that more financially constrained firms hedge less or not at all. Weshow that these predictions are consistent with the evidence using panel data for fuelprice risk management by airlines. More constrained airlines hedge less both in the crosssection and within airlines over time. Risk management drops substantially as airlinesapproach distress and recovers only slowly after airlines enter distress.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

What determines the extent to which firms engage inrisk management? A central insight from the theoreticalliterature is that firms engage in risk management becausefinancing constraints render them effectively risk averse(see Froot, Scharfstein, and Stein, 1993). This insight hasmotivated a large number of empirical papers. However,the empirical findings do not support the prediction that

firms more likely to face financial constraints, such assmall firms, are more likely to manage risk. The mainrobust pattern that emerges from this literature is thatsmall firms engage in less risk management, leading Stulz(1996) to conclude that “[t]he actual corporate use ofderivatives, however, does not seem to correspond closelyto the theory” (page 8).

In this study, we challenge the notion that financialconstraints and risk management should be positivelycorrelated theoretically and empirically. We provide amodel that predicts that commodity price risk manage-ment should be lower and even absent for firms that aremore financially constrained. The basic theoretical insightis that collateral constraints link the availability of finan-cing and risk management. More specifically, if firms musthave sufficient collateral to cover both future payments tofinanciers and future payments to hedging counterparties,a trade-off emerges between financing and risk manage-ment. Commodity price risk management shifts net worthacross states, and firms are effectively risk averse about networth. When net worth is low and the marginal value ofinternal resources is high, firms optimally choose to usetheir limited net worth to finance investment, or downsizeless, at the expense of hedging. Consistent with our model,

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/jfec

Journal of Financial Economics

0304-405X/$ - see front matter & 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.jfineco.2013.10.003

☆ We thank Peter DeMarzo, Hayne Leland, Andrey Malenko, JeremyStein and seminar participants at the 2011 Finance Summit in Revelstoke,British Columbia, Canada; Duke University; University of California at LosAngeles; University of Queensland; the 2011 National Bureau of EconomicResearch Summer Institute corporate finance workshop; Georgia StateUniversity; Harvard University; University of Kentucky; Imperial College;London School of Economics; London Business School; Stanford Univer-sity; University of California at Berkeley; University of Zurich; Universityof St. Gallen; the 2012 Jackson Hole Finance Conference; Catolica-Lisbonand Nova School of Business and Economics; the 2012 Western FinanceAssociation Conference; the 2012 Centre for Economic Policy ResearchEuropean Summer Symposium in Financial Markets; University of Texasat Dallas; Boston University; and the 2013 Utah Winter FinanceConference for helpful comments.

n Corresponding author. Tel.: þ1 919 660 7797; fax: þ1 919 660 8038.E-mail address: [email protected] (A.A. Rampini).

Journal of Financial Economics 111 (2014) 271–296

THE JOURNAL OF FINANCE • VOL. LXXV, NO. 2 • APRIL 2020

Risk Management in Financial Institutions

ADRIANO A. RAMPINI, S. VISWANATHAN, and GUILLAUME VUILLEMEY∗

ABSTRACT

We study risk management in financial institutions using data on hedging of interestrate and foreign exchange risk. We find strong evidence that institutions with highernet worth hedge more, controlling for risk exposures, across institutions and withininstitutions over time. For identification, we exploit net worth shocks resulting fromloan losses due to declines in house prices. Institutions that sustain such shocksreduce hedging significantly relative to otherwise-similar institutions. The reductionin hedging is differentially larger among institutions with high real estate exposure.The evidence is consistent with the theory that financial constraints impede bothfinancing and hedging.

DESPITE MUCH DEBATE ABOUT BANK risk management and its purported failureduring the financial crisis, the basic patterns of risk management in financialinstitutions are not known and the main determinants of banks’ risk manage-ment are not well understood. Since financial institutions play a key role inthe macroeconomy and in the transmission of monetary policy, understandingtheir exposure to shocks is essential for monetary and macro-prudential policy.1

Financial institutions can manage the risk exposures arising from lending and

∗Adriano A. Rampini and S. Viswanathan are with Duke University. Guillaume Vuillemeyis with HEC Paris. We thank Manuel Adelino; Juliane Begenau; Markus Brunnermeier; MarkFlannery; Dalida Kadyrzhanova; Divya Kirti; Kebin Ma; Justin Murfin; Dimitris Papanikolaou;Roberta Romano; Farzad Saidi; Joao Santos; David Scharfstein; Jeremy Stein; Jason Sturgess;Amir Sufi; Adi Sunderam; Yuri Tserlukevich; James Vickery; seminar participants at Duke, HECParis, Princeton, Georgia State, the Federal Reserve Bank of New York, the NBER Insuranceand Corporate Finance Joint Meeting, the FIRS Conference, the NYU-NY Fed-RFS Conference onFinancial Intermediation, the ETH-NYU Law & Banking/Finance Conference, the Paul WoolleyCentre Conference at LSE, the LBS Summer Finance Symposium, the Barcelona GSE SummerForum, INSEAD, Frankfurt School of Finance & Management, the CEPR ESSFM in Gerzensee,the EFA Annual Meeting, EIEF, the Wharton Conference on Liquidity and Financial Crises, theAFA Annual Meeting, the Jackson Hole Finance Conference, the NY Fed Conference on OTCDerivatives, National University of Singapore, Paris-Dauphine, Mannheim, the Conference ofSwiss Economists Abroad, the Bank of Italy, and Zurich; as well as Amit Seru (the Editor); ananonymous associate editor; and two anonymous referees for helpful comments. Rampini is anNBER Research Associate and a CEPR Research Fellow. Viswanathan is an NBER ResearchAssociate. Vuillemey is a CEPR Research Affiliate. The authors have no conflicts of interest todisclose as identified in The Journal of Finance disclosure policy.

Correspondence: Adriano A. Rampini; e-mail: [email protected] Following Bernanke and Gertler (1989), Holmstrom and Tirole (1997), Gertler and Kiyotaki

(2010), Brunnermeier and Sannikov (2014), and Rampini and Viswanathan (2019), among others,analyze the effects of financial institutions’ net worth on the availability of intermediated finance

DOI: 10.1111/jofi.12868C© 2019 the American Finance Association

591

Basic pattern: financially constrained firms hedge less or not at all


Evidence on Airline Fuel Price Risk Management

Substantial variation in fuel price hedging across airlines

model that we take to the data is that in an environment inwhich collateral constraints apply to both external financeand risk management activity, risk management is lowerwhen the marginal value of net worth today is very high.And the marginal value of current net worth is very high

when the level of current net worth is low. The reason forthis in the model is the concavity of the value function ofthe firm, which is induced by the concavity of theproduction function and the limited liability and collateralconstraints. It is as if the firm is risk averse with regard tonet worth, and hence the marginal value of net worth ishigh when net worth is low. The state variable is networth, not the amount of collateral itself, as the amount ofcollateral is endogenously determined by the investmentpolicy, given net worth.

Net worth in our model is the sum of current cash flow,the value of capital net of debt, plus the payoff of anycontingent claims used to hedge. We use a total of fivevariables as the empirical analogs of net worth, the statevariable in the model. Close analogs are the total marketvalue and total book value of net worth, and we use these astwo of our measures of net worth. The former is defined asthe market value of the airline less the book value ofliabilities. The latter is the book value of shareholders' equity.

In the model, airlines all have the same productionfunction. In practice, the production function could differacross airlines, which would justify scaling net worth bysome notion of the scale of the airline. To better capturethe potential cross-sectional differences in the scale ofairlines, we also use the above measures of net worthscaled by the total assets of the airline.

We also use the credit rating of the airline. While creditratings measure more than just the internal resources ofthe firm, a poor credit rating captures situations in whichfinancing is particularly costly and internal resources havea high marginal value. We find consistent results across allfive measures of net worth.

In the model, net worth is the predetermined statevariable and variation in net worth determines variation inhedging decisions. When testing this prediction in the data, aconcern is that omitted variables could be simultaneouslydriving net worth and hedging activity. While the empiricalresults are consistent with our interpretation of this correla-tion, causality could still be a concern. One way to addressthis concern is to isolate a precise source of variation in networth through a two-stage least squares estimation.

In determining an instrument for net worth, ourprimary goal is to stay as close to the theory as possible.In the theory, net worth is determined by two exogenousfactors: input prices and productivity. Input prices, that is,fuel prices, likely affect all airlines in a similar manner, andso we focus on productivity as the main source of idiosyn-cratic variation in net worth. We measure productivityusing observed operating income. In Section 5.3, we reportspecifications in which we instrument for net worth usingoperating income. The two-stage least squares estimationyields qualitatively similar results.

5. Hedging and net worth

Our theory predicts that less constrained firms engagein more risk management. Consistent with our theory, weshow in this section that a strong positive correlationexists between airlines' fuel price hedging and net worthboth in the cross section and within airlines over time.Importantly, our data allow us to study the within-firm

Fig. 2. Fuel expense hedging by airline. This figure presents the averagefraction of next year's fuel expenses hedged for each airline. The averageis computed over all years the airline is in the sample. Panel A includesthe full sample. Panel B excludes any airline that has a fuel pass throughagreement at any point in the sample.

Fig. 3. Fuel expense hedging in the time series. This figure presents theaverage fuel cost per gallon across airlines by year and the averagefraction of next year's fuel expenses hedged across airlines by year. Thelonger dashed line represents the weighted average of next year's fuelexpenses hedged when the weights are total assets of the airline.

A.A. Rampini et al. / Journal of Financial Economics 111 (2014) 271–296282

Airlines asempiricallaboratory

Data: 1996-2009

Airlines hedge≈ 20% on average

Well-capitalizedSouthwest hedgesmost (> 50%)

Many, especiallysmall airlineshedge very little

Does net worthexplain thevariation?


Evidence on Airline Fuel Price Risk ManagementAirlines with stronger balance sheets hedge more

variation of hedging and net worth separately, in contrastto most previous studies.

5.1. Hedging and net worth: cross-sectional evidence

Fig. 4 presents cross-sectional evidence on the correla-tion between measures of net worth and the fraction offuel expenses hedged. For the cross-sectional analysis, wecollapse the yearly data into airline-level averages, whichis equivalent to a between-regression analysis. Each airlinein the scatter plots in Fig. 4 is weighted by total assets,where the size of the circles reflects the size of the airline.Each scatter plot also includes a regression line in whichthe regression is weighted by total assets of the firm.

We focus on regressions that are weighted by total assetsin our study, which is justified given evidence in Appendix Cthat the predicted error term has a much higher standarddeviation for smaller airlines. In such situations, weightedleast squares (WLS) has efficiency gains above ordinary leastsquares (OLS). For completeness, we also report the OLSestimates, which are similar, in our main tables.

Across all five measures, a strong positive correlationemerges between the measure of net worth and thefraction of next year's expected fuel expenses hedged.What could appear to be outliers in these specificationsare generally regional airlines with code-sharing agree-ments that include fuel pass through agreements, whichwe code as hedging 100%. These airlines tend to have lownet worth, and yet one could interpret their fuel passthrough agreements as a decision to hedge. However, aswe argue above, it is questionable whether the fuel passthrough agreements of regional airlines should be inter-preted as an independent decision to hedge. In Appendix B,we provide support for the view that such agreements arepart of a larger wet lease type transaction and not a separatehedging decision.13

In Panel F of Fig. 4, we show a strong positive correla-tion between hedging and operating income scaled by

Fig. 4. Fuel expense hedging and net worth: cross-sectional evidence. This figure presents cross-sectional scatter plots of the fraction of next year's fuelexpenses hedged and measures of net worth in the current year. All variables are averaged across years for each firm. The size of the circles reflects total assets,and the regression lines are based on (firm-mean) asset-weighted regressions. Panel A: Net worth to assets; Panel B: Net worth (mv) to assets; Panel C: Networth; Panel D: Net worth (mv); Panel E: Credit rating; and Panel F: Operating income to assets.

13 Regressions excluding airlines with such agreements yield similarresults and are reported in Section 7.

A.A. Rampini et al. / Journal of Financial Economics 111 (2014) 271–296 283

Measures of financial constraints

Net worth (market value) (Panel B)

Credit rating (Panel E)

Evidence from cross section: comparing across airlines

Mechanism

Southwest Airlines explicitly pledged aircraft as collateral tocounterparties (2010 10-K)


Evidence on Airline Fuel Price Risk Management

Airlines that approach financial distress cut hedging

hedging. The magnitude of the coefficient is larger in thetwo-stage least squares estimation relative to theweighted least squares estimates. This might be becausethe instrument reduces measurement error in net worthor because sources of variation in net worth unrelated toproductivity variation might not be able to explain hed-ging behavior as well.

While the first-difference specifications are only mar-ginally significant, the point estimates are very similar tothe estimates from the pooled and airline fixed effectsregressions. The first-difference instrumental variablesestimates in Columns 4 and 5 have p-values of less than0.12. This is comforting as shocks to operating income ascaptured in the first-difference specification are morelikely to reflect random productivity shocks as inthe model.

6. Hedging around distress

Our theory predicts that severely constrained firmsmight not hedge at all. In this section, we show thatairlines in distress cut their risk management dramatically.Moreover, we show that airlines facing tighter financialconstraints state in their 10-K SEC filings that they arereducing their fuel price hedging because of collateralconsiderations, which is the basic mechanism in ourmodel. We are the first, to the best of our knowledge, toshow the remarkable dynamics of corporate risk manage-ment around financial distress.

6.1. Evidence on hedging around airlines' distress

Received theory would predict that airlines in distressshould engage in substantial risk management given thesevere financial constraints they face. Along these lines,Morrell and Swan (2006) argue that “when an airline isnear bankruptcy, hedging fuel prices may make sense. …An airline near bankruptcy would like to protect itself fromlosses and thus the expense of becoming bankrupt” (p.715). However, airlines in distress also consistently empha-size the importance of preserving cash and internalresources. In contrast to received theory, our model pre-dicts that when the marginal value of internal resources isextremely high, firms reduce risk management. If riskmanagement is subject to the same collateral constraintsas external financing, then we should see a dramaticreduction in jet fuel price hedging as airlines becomedistressed.

We define a firm as being in distress, in our sample,when it is either rated CCCþ or below or is in bankruptcy.Panel A of Table 7 lists the ten instances of distress in oursample. Both America West Holdings Corp. and US AirwaysGroup Inc. were downgraded to CCCþ or worse in 2001.US Airways Group Inc. was subsequently downgraded toCCCþ or worse again in 2004 before the merger withAmerica West Holdings Corp. Seven other airlines entereddistress during our time period. These instances were notclustered in time—two occur in 2001, three in 2004, two in2008, and one each in 2002, 2003, and 2005.

Panel A of Fig. 5 shows the average fraction of nextyear's fuel expenses hedged in the two years before

through the two years after entering distress for the tenairlines that experience distress in our sample. From twoyears before to the year before distress, there is a slightdrop in the fraction of fuel expenses hedged. But the dropin hedging in the year the firm enters distress is remark-able. Airlines go from hedging about 25% of their expectedfuel expenses in the year before to less than 5% in the yearentering distress. The fraction of fuel expenses hedgedrecovers in the two years after the initial onset of distress,although not to the levels seen two years prior.

Given this very large decline in hedging, it should comeas no surprise that the drop is statistically significant at the1% level. This is shown in Panel B of Table 7, where weregress the fraction of next year's fuel expenses hedged onindicator variables for two years before through two yearsafter entering distress. The estimates are robust to the useof WLS or airline fixed effects.

An alternative to studying firms' hedging behavior inand around distress is to study firms' hedging behavior inand around bankruptcy. We report such alternative resultsusing an approach analogous to the above in Table 8. Ofthe ten instances of distress, only seven result in bank-ruptcy. As firms approach bankruptcy, hedging dropssubstantially and, in fact, a significant drop occurs as earlyas two years before the airline enters bankruptcy. Thissuggests that entering distress, not bankruptcy per se, isrelevant for the drop in hedging.

The evidence in Table 7 is consistent with the idea thathedging becomes too costly for airlines in distress. Giventhe high marginal value of internal resources, companiesfacing collateral constraints on external financing areunwilling to use collateral to hedge future fuel price risk

Fig. 5. Fuel expense hedging around distress. This figure providesevidence on fuel expense hedging around distress, where an airline isdefined to be in distress when it is rated CCCþ or worse or, whenunrated, when it is in bankruptcy. Panel A shows the fraction of nextyear's fuel expenses hedged for airlines that enter distress at t¼0. Eachtime period reflects a year. Panel B shows the fraction of airlinesmentioning collateral or their financial position as a restriction onhedging activities.

A.A. Rampini et al. / Journal of Financial Economics 111 (2014) 271–296 287

American Airlines 2009 10-K: “[a]deterioration of the Company’sfinancial position could negativelyaffect the Company’s ability tohedge fuel in the future.”

Evidence from within variation:comparing same airline over time

Distress: drop in rating to CCC+or below

Airlines in distress cut hedgingalmost completely

Slow recovery after distress

Collateral or financial positionmentioned as obstacle in annualreports (“smoking gun?”)

But is relation between hedgingand net worth causal?


Evidence on Bank Risk Management

For causal inference, need exogenous shocks to net worth

Risk Management in Financial Institutions 601

Net worth index Net worth index (ex size)

SizeMarket capitalization / Assets

42

0-2

-4

1995 2000 2005 2010 2015 1995 2000 2005 2010 2015

1995 2000 2005 2010 2015 1995 2000 2005 2010 2015

0.3

0.2

0.1

0

42

0-2

-418

1716

1514

13

Net

wor

th in

dex

Mar

ket c

apita

lizat

ion

/ Ass

ets

Net

wor

th in

dex

(ex

size

)S

ize

Figure 1. Distribution of main measures of net worth. This figure plots the distribution ofthe four main measures of net worth, with one cross-sectional box plot for each quarter. In eachbox plot, the horizontal dash is the median, the diamond is the mean, the whiskers represent the5th and 95th percentiles, and the gray rectangle represents the 25th and 75th percentiles. In the topleft panel, the dark red bar corresponds to 2009, the treatment year in the difference-in-differencesand triple-differences estimations. The data are at the BHC level and cover the period 1995Q1 to2013Q4. Variables are defined in Table A.I. (Color figure can be viewed at wileyonlinelibrary.com)

where, for institution i at date t, AIRit and LIR

it are assets and liabilities that ma-ture or reprice within one year. The maturity gap is effectively an institution’snet floating-rate assets. The change in an institution’s net interest income fromassets and liabilities in the one-year maturity bucket �NIIit is approximatelyproportional to the maturity gap, that is, �NIIit ≈ Maturity gapit × �rt, where�rt is the change in the one-year interest rate. A positive maturity gap impliesthat increases in the short rate increase an institution’s net interest incomeat the one-year horizon because the institution holds positive net floating-rateassets, that is, holds more interest-rate-sensitive assets than liabilities at thismaturity. The maturity gap is used by risk managers in practice (see Saun-ders and Cornett (2008), Mishkin and Eakins (2009)) and is also used in priorresearch (see Hoffmann et al. (2019)).19

19 For bank-level data, one can also construct a measure of the duration gap. See Section II inthe Internet Appendix.

Interest rate hedging byfinancial institutionslargest such market

Data: 1995-2013

Banks’ net worth dropssubstantially duringfinancial crisis

Why? – Loan losses dueto drop in real estateprices

Idea: exploit variation inexposure across banks

Need to constructtreatment and controlgroup


Evidence on Bank Risk ManagementBanks that sustain net worth shocks cut hedging

618 The Journal of Finance R©

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

0.08

0.06

0.04

0.02

00.

008

0.00

60.

004

0.00

20

Inte

rest

rate

hed

ging

Fore

ign

exch

ange

hed

ging

Control Treated

TreatedControl

Panel A: Interest Rate Hedging

Panel B: Foreign Exchange Hedging

Figure 2. Difference-in-differences: effect of treatment on hedging. This figure plots hedg-ing in the treatment and control groups used in the difference-in-differences estimation. Panel Aplots gross interest rate hedging and Panel B plots gross foreign exchange hedging. The sample isrestricted to institutions that hedge at least once before the treatment year. The data are at theBHC level and cover the period 2005Q1 to 2013Q4. Variables are defined in Table A.I. (Color figurecan be viewed at wileyonlinelibrary.com)

Treatment vs. control group

Treatment: lend where real estateprices drop more than median

Treated banks reduce hedgingsignificantly relative to control

Substantial effect: cut by ≈ 1/2

Both interest rate and foreignexchange hedging reduced

Causal interpretation: banks cuthedging because net worth drops


Aside: Household Risk Management

Similar patterns and trade off in household insurance

Financing Insurance∗

Adriano A. RampiniDuke University

S. ViswanathanDuke University

May 2019

Abstract

Insurance has an intertemporal aspect as insurance premia have to be paid up front.We argue that the financing of insurance is key to understanding basic insurancepatterns and insurers’ balance sheets. Limited enforcement implies that insurance isglobally monotone increasing in household net worth and income, incomplete, andprecautionary. These results hold in economies with income risk, durable goods andcollateral constraints, and durable goods price risk, under quite general conditions.In equilibrium, insurers are financial intermediaries with collateralized loans as as-sets and diversified portfolios of insurance claims as liabilities. Collateral scarcitylowers the interest rate, reduces insurance, and increases inequality.

JEL Classification: D91, E21, G22.

Keywords: Household finance; Collateral; Insurance; Risk management; Financialconstraints

∗We thank Hengjie Ai, Mariacristina De Nardi, Emmanuel Farhi, Nobu Kiyotaki, Ralph Koijen, DavidLaibson, Martin Oehmke, Tomek Piskorski, Alp Simsek, Jeremy Stein, Moto Yogo, George Zanjani, andseminar participants at the AEA Annual Meeting, Duke, the NBER-Oxford Saıd-CFS-EIEF Conferenceon Household Finance, the HBS Finance Unit Research Retreat, the Asian Meeting of the EconometricSociety, MIT, UC Berkeley, Harvard, USC, the WFA Annual Conference, the SED Annual Meeting,the Bank of Canada and Queen’s University Workshop on Real-Financial Linkages, Cheung Kong GSB,Cornell, DePaul, Princeton, BYU, Carnegie Mellon, Indiana, Wharton, Chicago, Amsterdam, UCL, Im-perial College, Warwick, the CEPR European Summer Symposium in Financial Markets, the WashingtonUniversity Conference on Corporate Finance, the AFA Annual Meeting, Houston, Minnesota, Illinois,Virginia, the Conference in Honor of Robert M. Townsend, the FIRS Conference, the NBER SI on CapitalMarkets and the Economy, the FTG Summer School on Liquidity in Financial Markets and Institutions,and UC Santa Cruz for helpful comments. Earlier versions of this paper were circulated under the title“Household risk management.” Rampini is a Research Associate of the NBER and a Research Fellow ofthe CEPR. Viswanathan is a Research Associate of the NBER. Duke University, Fuqua School of Busi-ness, 100 Fuqua Drive, Durham, NC, 27708. Rampini: (919) 660-7797, [email protected]; Viswanathan:(919) 660-7784, [email protected].

Ongoing research

More financially constrainedhouseholds buy less insurance

Life, property & casualty, healthinsurance

Why?

Similar mechanism: insurancepremium needs to be paid up front


The Risk Management Paradox

Paradox: Financing is reason for and obstacle to hedging

Reason for hedging is avoiding financial constraints

Firms use limited funds for operations instead of hedging

Empirical puzzles call for new theory

Theory helps understand facts and provides useful practical guidance


Why Many Firms Don’t Hedge – Our TheoryFirm’s dynamic problem: choose policies for

investment (k), financing (b), hedging (h′), and payout (d)

to maximize value

v(w, s) = max{d,w′,k,b,h′}

d+ βE[v(w′, s′)]

subject to budget and collateral constraints for all states

w + b+R−1E[h′|s′]︸︷︷︸hedging claims

≥ d+ k

A′f(k) + k(1− δ) ≥ Rb+ h′ + w′

θk(1− δ)︸︷︷︸collateral

≥ Rb︸︷︷︸financing

+ h′︸︷︷︸hedging

and limited liability d ≥ 0

Both financing and hedging require collateral (or funds upfront)Back


the risk management paradox - fuqua school of business

Documents