Self-Fulfilling Runs: Evidence from the U.S. Life Insurance Industry Nathan Foley-Fisher, Borghan Narajabad, and Stéphane Verani
Discussion by Lawrence Schmidt University of Chicago
Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion
Motivation
Crisis featured "run-like" behavior in many different markets Two mechanisms can lead to runs: 1 2
Deterioration in fundamentals: bad future returns / delayed repayment Externalities (payoff complementarities) from concerns about others’ actions. Runs → self-fulfilling prophecies (Diamond & Dybvig, 1983)
Big identification challenge: beliefs about other investors’ actions may be correlated with fundamentals Many theoretical models, such as global games models of bank runs (Goldstein & Pauzner, 2005), imply a link between the two Runs are triggered by bad signals about fundamentals Strategic complementarities create negative externalities → additional fragility from self-fulfilling beliefs about other investors’ behavior
Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion
Identifying self-fulfilling beliefs Ideal experiment: find an instrument for other investors’ actions that is independent of fundamentals One approach from literature: use quasi-experimental variation in other investors’ information Hertzberg, Liberti, & Paravasini (2011): policy reform of credit registry forced lenders to reveal their (already known) private information to others Schmidt, Timmermann, & Wermers (2015): exploit variation in investor composition in money market funds after 2008 Lehman failure
This paper: exploits source of quasi-experimental variation in set of potential actions of other investors using data from life insurers Structure of XFABN contracts creates time-series variation in % of other investors who can "run" before I can move again.
Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion
Contribution Exploit a unique feature of XFABN market for U.S. life insurers to empirically identify complementarities: 1
Empirics: primary focus of the paper (and this discussion) Collect data on contractual terms for 54 XFABN conduits (a lot of work!) Use predetermined variation in contractual terms to test for the role of self-fulfilling beliefs Find significant evidence of the self-fulfilling component: self-fulfilling beliefs explain "about 40% of the observed $18 billion withdrawals by investors between the third quarter of 2007 and the end of 2008" Focus on identification: perform lots of robustness exercises
2
Theory: stylized model with complementarities to describe payoff structure of XFABN market Model is similar to He and Xiong (2012) Provides formal justification for empirical identification approach
Overall, paper provides a new way to answer an important question using new data, yielding interesting results. Very interesting and ambitious! Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion
Where are complementarities coming from? "Insolvency is rarely an issue for insurers. In they event that they breach [a] regulatory capital threshold, which happens much sooner than insolvency, life insureres are immediately taken over by their State regulator. Consequently, insurance liability holders can be reasonably certain they will eventually get repaid. However, there could be tremendous uncertainty over when investors will get their money back." No risk of losing principal. Why? Buyers are customers of the life insurance companies → seniority over other debt holders Life insurers are taken over by regulators if capital ratios are too low Source of complementarities ≈ sequential-service constraint. Current spinoffs lower future liquidity for other investors Peculiar timing: if I "run" today, I still don’t get paid for another year. Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion
Basic approach Ideal estimating equation: Dijt = γ0 + γ1 Et [Sij,t+1 ] + γ2 Qijt + xjt0 β + ijt where Dijt : % of XFABN i issued by insurer j at that is converted at time t Sij,t+1 : % of all other (excluding i) XFABN from issuer j that is converted into spinoffs between the current election date t and next election date t + 1 Qijt : % of XFABN i issued by insurer j already converted prior to time t Authors instrument for Sij,t+1 using RE_ex3mij,t+1 ≡ upper bound of the support of Sij,t+1 (maximum % of people in other XFABN who can convert) γ1 > 0 consistent with complementarities Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion
Main result IV coefficient on Sij,t+1 > 0 and highly significant.
Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion
Comment #1: Coefficient stability and sample period
Essentially no spinoffs for 2/3 of the sample (2005-2010) Remainder concentrated in Q3-4 of 2007 First stage (Sij,t+1 ) forecasting errors are highly autocorrelated Fixed effects mostly driven by matching unconditional mean of Dijt Potential solutions: Focus on key period (late 2007), rolling 1st stage regressions, collapse to cross-section? Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion
Comment #2: Test key exclusion restriction Dijt = γ0 + γ1 Et [Sij,t+1 ] + γ2 Qijt + xjt0 β + ijt We want RE_ex3mij,t+1 to be uncorrelated ijt Due to data availability concerns, authors don’t include many time-varying firm characteristics in baseline regressions Partly related to prior comment: issuer characteristics should have low explanatory power in normal times
Can test whether big changes in fund riskiness are uncorrelated with fitted values from first stage regressions during peak run period. An alternative worth ruling out: Investors who value liquidity most highly are likely to choose securities which provide them with the option to spin off assets most frequently These investors are more likely to convert as market liquidity deterioriates Fixed effects don’t solve this problem in current setup Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion
Comment #3: Put magnitudes in more context Discussion in paper suggests that first-order risk is failure of parent insurance companies Identifiable variation in complementarities involves beliefs about other investors in XFABN only Is the XFABN market big enough for the sources of cross-sectional variation to be a first order strategic consideration? XFABN ⊂ FABS ⊂ all short-term liabilities Figures in the paper suggest that XFABN < 10% of FABS market. Context from AIG 2007 10-K (warning: I’m not an accountant) $96 billion book value of equity, $65 billion in short term assets, $35.6 billion in operating cash flow $13.6 total in CP and extendible notes Upper bound on XFABN @ 10% of total = $1.36 bil
What is the right "denominator" to scale the effects by? Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion
Comment #4: What about contemporaneous complementarities?
Investors in the same XFABN move simultaneously Scope for complementarities in these contemporaneous decisions as well Particularly relevant for insurers with a single XFABN, if any exist Current method could not find evidence of the importance of self-fulfilling beliefs in these cases In current tests, Sij,t+1 = 0 always by construction. Is this what we want? Do we think complementarities are weaker in these firms?
Lawrence Schmidt
FFNV Self-fulfilling runs - Discussion