CMBS and Conflicts of Interest: Evidence from a Natural Experiment on Servicer Ownership Maisy Wong Discussion by Kris Gerardi Federal Reserve Bank of Atlanta
ASSA Meetings San Francisco, CA January 3, 2016
These views expressed are mine and do not necessarily reflect those of the Federal Reserve Bank of Atlanta. Gerardi (Atlanta Fed)
CMBS and Conflicts of Interest: Evidence from a Natural January Experiment 3, 2016 on Servicer 1 /Owne 11
Additional Disclaimer
I know next to nothing about commercial real estate and the CMBS market. But I have studied the issue of agency conflicts in the RMBS market. Part of the literature on the impact of agency conflicts on loan modifications in the non-agency securitization market (Piskorski et al., Agarwal et al., Adelino et al., Kruger, etc.) While some disagreement about identification, close inspection suggests statistically significant but small effects in the aggregate. However, some notable examples where conflicts of interest had sizeable effects.
Gerardi (Atlanta Fed)
CMBS and Conflicts of Interest: Evidence from a Natural January Experiment 3, 2016 on Servicer 2 /Owne 11
Example of Agency Conflicts in RMBS Market Carrington Capital Management Hedge fund that specialized in securitizing subprime mortgages. Poster child for potential unintended consequences of mandatory risk retention policies. Invested in the “credit enhancement” (i.e. equity) tranches of its subprime deals. Acquired servicing rights to many of its deals (via New Century). “They appear to have managed their book for their own personal benefit in a way that screws the investors.” – Laurie Goodman
Gerardi (Atlanta Fed)
CMBS and Conflicts of Interest: Evidence from a Natural January Experiment 3, 2016 on Servicer 3 /Owne 11
Example of Agency Conflicts in RMBS Market Adopted strategy of stopping foreclosures on delinquent mortgages – via repeat “modifications” or perpetual forbearance. Prevented losses from being realized on lower tranches. In some cases generated substantial cash flow to lower tranche investors by helping to trigger “stepdowns” (excess principal payments that occur when deal meets certain health benchmarks) In one deal lowered delinquency rates from 35% to 25% and met the stepdown trigger of 26% ⇒ $18 million in principal and interest payments over 5 month period. At expense of senior tranch investors, who would have benefited from quicker liquidation timelines due to declining property values.
Gerardi (Atlanta Fed)
CMBS and Conflicts of Interest: Evidence from a Natural January Experiment 3, 2016 on Servicer 4 /Owne 11
This Paper
Uses a “natural experiment” to study the importance of agency conflicts in the CMBS market. In 2009–2010 period, 4/5 large special servicers changed ownership. New owners were firms that buy (and sell) commercial properties. Potential conflict of interest since primary job of special servicer is to liquidate distressed properties at lowest possible cost to CMBS investors. ⇒ Anectdotal evidence of self-dealing or “tunneling” problem that produced significant consternation among investors.
Gerardi (Atlanta Fed)
CMBS and Conflicts of Interest: Evidence from a Natural January Experiment 3, 2016 on Servicer 5 /Owne 11
This Paper Simple event study statistical analysis that compares liquidation volumes and loss rates before and after ownership changes. Average liquidation volumes increased by 220% – $101 million per month per special servicer. Average loss rates increased by 11 percentage points. Similar analysis suggests no change for servicers that didn’t change ownership.
Supplemental case study of sample of affiliated transactions by one special servicer (C-III). Nice descriptive analysis of characteristics of 18 transactions using detailed property-level transactions data.
Gerardi (Atlanta Fed)
CMBS and Conflicts of Interest: Evidence from a Natural January Experiment 3, 2016 on Servicer 6 /Owne 11
General Comments Clever idea to exploit ownership changes to measure impact of agency conflicts in CMBS market. Very careful analysis with lots of robustness checks and a transparent discussion of alternative interpretations. Severe econometric issues make identification difficult. Effect of ownership changes estimated off of time-series variation in liquidation volumes and loss severities. Basically comparing volumes/severities in period before ownership change to period after ownership change. Attributing estimated change to conflicts of interest associated with ownership transitions.
Main issue is the timing of the ownership changes: Clustered between December 2009 and July 2010. In midst of worst real estate bust since Great Depression. Gerardi (Atlanta Fed)
CMBS and Conflicts of Interest: Evidence from a Natural January Experiment 3, 2016 on Servicer 7 /Owne 11
General Comments
190
18
180
16
170
14
160
12
150
10
140
8
130
6
120
4
110
event dates
2
Moody's price indices
Liquidation value (in billion of dollars)
Figure 1: Trends in Liquidation and Commercial Real Estate Prices 20
100
90
0
2000
2001
2002
2003
2004
Yearly Liquidation Value
Gerardi (Atlanta Fed)
2005
2006
2007
2008
2009
2010
2011
2012
Monthly Moody's/RCA Commercial Property Price Index
CMBS and Conflicts of Interest: Evidence from a Natural January Experiment 3, 2016 on Servicer 8 /Owne 11
General Comments Primary way of addressing issue is to conduct “placebo” tests. Check to see if CMBS servicers that did not undergo ownership changes experienced similar increases in liquidation volume and loss severities at the event dates. Similar logic to difference-in-differences estimator. Significant increases in full sample, but not when sample window is restricted to +/- 36 months.
Placebo analysis in paper uses all servicers that did not undergo ownership changes. Should include analysis that uses Midland only, as summary statistics clearly show that the five large special servicers are different than the smaller servicers.
Gerardi (Atlanta Fed)
CMBS and Conflicts of Interest: Evidence from a Natural January Experiment 3, 2016 on Servicer 9 /Owne 11
Additional Comments/Suggestions
Might want to try including some geographic controls in the covariate set. Possible that some servicers focus on specific geographies? Servicer FEs might address this issue unless geographic specialization changed over time.
Should control for current LTV and DSCR in regressions. True that they are not pre-determined variables, but they are arguably the most important determinants of loss severities.
Should try smaller window around events – 36 months seems really long for event study (maybe 12 or 24 months?).
Gerardi (Atlanta Fed)
CMBS and Conflicts of Interest: Evidence from a NaturalJanuary Experiment 3, 2016 on Servicer 10 /Owne 11
Additional Comments/Suggestions Might be interesting to consider additional dimensions of servicer behavior. Renegotiation/modifications of distressed loans, which has been studied in the context of RMBS.
Policy Implications? Conclusion discusses potential adverse effects of mandated risk retention requirements based on empirical results. What (if any) policies to curb self-dealing incentives? To what extent were CMBS investors aware of these issues? Are additional disclosure requirements necessary to make investors better aware of affiliations? Discussion of C-III’s transactions in section 5.2.1 suggests an answer of “no.”
Gerardi (Atlanta Fed)
CMBS and Conflicts of Interest: Evidence from a NaturalJanuary Experiment 3, 2016 on Servicer 11 /Owne 11