CMBS and Conflicts of Interest: Evidence from Ownership Changes for Servicers

• Published date: 01 October 2018
• DOI: http://doi.org/10.1111/jofi.12690
• Author: MAISY WONG
• Date: 01 October 2018

THE JOURNAL OF FINANCE •VOL. LXXIII, NO. 5 •OCTOBER 2018

CMBS and Conflicts of Interest: Evidence from

Ownership Changes for Servicers

MAISY WONG∗

ABSTRACT

Self-dealing is potentially important but difficult to measure. In this paper, I study

special servicers in commercial mortgage-backed securities (CMBS), which sell dis-

tressed assets on behalf of bondholders. Around 2010, ownership changes of four

major servicers raised concerns that they may direct benefits to new owners’ affili-

ates (buyers and service providers). Loans liquidated after ownership changes have

greater loss rates than before (8 percentage points (p.p.), $2.3 billion in losses), rela-

tive to other (placebo) servicers. Together with a case study that tracks self-dealing

purchases, the findings point to potential steering conflicts that could incentivize

tunneling through fees to service providers.

WHILE SELF-DEALING HAS BEEN ALLEGED to harm investors, it is difficult to mea-

sure (Shleifer and Vishny (1997)). The wave of foreclosures of securitized assets

has put a spotlight on intermediaries that manage securitized assets on behalf

of bondholders. Anecdotal evidence suggests that some intermediaries tunnel

private benefits to their affiliates at the expense of distant bondholders (Lee

(2014)). However, it is challenging to quantify the extent of self-dealing for

securitized assets because of the difficulty in tracking them after securitiza-

tion. Moreover, self-dealing incentives are endogenous by nature and often

confounded by omitted variables.

The commercial mortgage-backed securities (CMBS) market provides a use-

ful context to address the empirical challenges faced by the self-dealing lit-

erature. With total assets of $623 billion, it is the second most important

∗Maisy Wong is at The Wharton School, University of Pennsylvania. I am grateful to Fer-

nando Ferreira, Joe Gyourko, and Todd Sinai for their advice. I thank Santosh Anagol; Effi Benm-

elech; Serdar Dinc; Mark Garmaise; Kristopher Gerardi; Laurie Goodman; Jean-Francois Houde;

Pamela Lee; Christopher Palmer; Sergey Tsyplakov; and Nikolai Roussanov; seminar participants

at Penn State University,University of Virginia, and Wharton; and conference participants at the

AREUEA-ASSA annual conference, Housing-Urban-Labor-Macro (HULM), NBER Summer Insti-

tute, WFA Summer Real Estate Research Symposium, and the University of Southern California

Research Symposium. Rachel Brill, Ying Chen, Hye Jin Lee, Jeremy Kirk, WillLin, Laura Nugent,

Joon Yup Park, Rebecca Pierson, Xuequan Peng, Dean Udom, Xin Wan, and Justine Wang were

excellent research assistants. I am also grateful to market participants who shared their experi-

ences with me in confidence. I thank the Research Sponsors Program of the Zell/Lurie Real Estate

Center at Wharton, the Jacobs Levy Grant, and the Wharton Dean’s Research Fund for research

support. All errors are my own. I have no relevant or material financial interests that relate to the

research described in the paper.

DOI: 10.1111/jofi.12690

2425

2426 The Journal of Finance R

source of credit in the commercial real estate sector (Federal Reserve Board

(2016)). Each CMBS trust comprises a pool of mortgages that are collater-

alized by nonresidential properties. Crucially, it is relatively easy to track

the ownership of CMBS assets because real estate transactions are recorded

publicly.

In this paper, I study self-dealing concerns involving four large special ser-

vicers in the CMBS market that service around $500 billion of CMBS debt

during the sample period. Special servicers are firms that manage distressed

mortgages on behalf of bondholders with the goal of maximizing the net present

value of the assets. When a loan is nonperforming, the special servicer decides

whether and how to liquidate it. Loan liquidations typically involve selling

the collateral (nonresidential properties). As sellers, special servicers have to

search for buyers and intermediaries to facilitate the liquidation.

Around 2010, ownership changes of the four special servicers linked them

with new affiliates that present potential self-dealing conflicts. The new owners

are vertically integrated financial institutions with affiliates that can be poten-

tial buyers of CMBS liquidations or potential intermediaries that facilitate real

estate transactions (lenders, brokers, and online auction platforms). The scale

of the servicers can provide the new owners efficiency benefits and comple-

mentarities that help them overcome significant search frictions in commercial

real estate. However, also around 2010, the volume of distressed CMBS assets

started to increase sharply. The increase in the special servicers’ sales together

with the links to the new affiliates raised concerns that these servicers would

face incentives to engage in self-dealing, for example, by selling assets to the

new owners at a discount or steering business opportunities to affiliates to earn

fees. In line with these concerns, Yoon (2012) reports that special servicers ap-

pear to be “burdened by conflicts of interest caused in part by new ownership”

and allegedly “cutting bad deals.”

Motivated by the above concerns surrounding the ownership changes of the

four servicers, I begin by estimating the impact of the ownership changes on

outcomes for liquidated loans. Specifically, I compare the change in loan loss

rates (realized losses divided by loan balance before losses) of the four (treated)

special servicers before and after their ownership changes relative to other

(placebo) servicers. Relative to the treated servicers, the placebo servicers have

fewer affiliates with potential tunneling conflicts.

I find that loans liquidated after the treated servicers changed owners have

8 p.p. higher loss rates than before the ownership change, relative to the placebo

servicers. This translates into aggregate losses of $2.3 billion, or 20% of total

losses from the liquidations by treated servicers under new ownership. The

panel data analysis includes 9,272 loans liquidated from 2003 to 2012, controls

for special servicer fixed effects, month of liquidation fixed effects, and predeter-

mined loan attributes. The key regressor is the interaction between an indicator

for loans liquidated by treated servicers and an indicator for liquidations after

ownership changes. The identification assumption is that unobserved determi-

nants of loan loss rates do not change differentially for treated versus placebo

servicers, conditional on the fixed effects and loan attributes.

CMBS and Conflicts of Interest 2427

Robustness checks address several threats to identification. Since the own-

ership change events happened around 2010, a potential confounding factor is

unobserved market conditions. The placebo servicers serve as counterfactuals

to the extent that treated and placebo servicers face common market conditions

and the effect also remains stable using various controls for market conditions.

In addition, while loans are not randomly assigned across servicers, trends in

the quality of liquidated loans indicate that compositional differences in loans

are unlikely to explain the main effect.

Another major threat relates to the liquidity crises that triggered the own-

ership changes of the treated servicers. Like many debt firms, when credit

spreads widened from 2008 to 2009, balance sheets worsened dramatically for

the previous owners of the servicers, triggering the need for capital infusion.

The treated servicers could have been relatively more capacity constrained and

overwhelmed by their own problems, accumulating distressed debt. The con-

cern is that some of the differences in losses after the ownership changes reflect

differences in stockpiling before the ownership changes, differences that dissi-

pate once the stockpile of debt has been resolved (even while new ownership

remained).

To address concerns that the higher loss rates for treated servicers are con-

founded by a (transitory) stockpiling effect, I exclude the 18 months before and

after the ownership changes and find that the results survive. The patterns re-

main similar if I exclude the three-year window around the ownership change,

using auxiliary data from Bloomberg, which provides a longer postperiod (six

years instead of three years). These findings suggest that the results are not

driven by transitory confounders.

Next, I conduct a bounding exercise to assess how much of the difference in

losses in the postperiod could be driven by the stockpiling effect instead of the

ownership change effect. Unconditional trends in the volume of losses reveal a

bunching pattern, with an “excess mass” in losses after ownership changes that

may reflect new owners’ liquidations of the stockpile of distressed debt. How-

ever, a conservative bounding exercise shows that stockpiling explains at most

21% of the difference in losses. Importantly, conditional differences between

treated and placebo servicers reveal no bunching pattern after controlling for

servicer and month fixed effects. Overall, while self-dealing concerns generally

arise in endogenous settings, the weight of the evidence suggests that the 8 p.p.

higher loss rate is unlikely to be explained away by the confounders discussed

above.

Turning to channels, I explore three self-dealing conflicts raised by market

participants: (i) buying (new owners buying assets sold by special servicers at

a discount), (ii) steering (servicers steering business opportunities to affiliated

service providers to earn fees), and (iii) price discrimination (affiliated service

providers charging bondholders higher fees to sell CMBS assets). The price

discrimination channel suggests that bondholders pay higher liquidation ex-

penses. I find that liquidation expenses are not higher after ownership changes

for treated servicers, relative to placebo servicers, which is inconsistent with

the price discrimination channel.