Risk Retention and Qualified Commercial

Mortgages

Sumit Agarwal∗ Brent W. Ambrose† Yildiary Yildirim‡

Jian Zhang§

Preliminary DraftMarch 28, 2018

Abstract

Regulations arising from the Great Recession and financial crisis havedramatically altered the mortgage market. For example, issuers are nowrequired to retain 5% of the credit risk in CMBS deals (the risk retentionrule). However, the rule does not apply uniformly across all commercialmortgages. For example, the carve-out for CMBS deals comprising mul-tifamily loans guaranteed by the GSEs creates a natural experiment fortesting the impact of the risk retention rule on the commercial loan mar-ket. Since the primary objective of this rule is to require deal sponsorsto participate in the risks associated with the loans backing securitizationdeals, we expect that underwriting standards should tighten following theimplementation of the rule. As a result, the Dodd-Frank Act regulationsmay have a profound impact on capital costs in the commercial real estatemarket. The results of this study will shed light on the implications ofregulations on credit to the commercial real estate market.

∗Georgetown University, (703) 388-6389, ushakri@yahoo.com†Penn State University, (814) 867-0066, bwa10@psu.edu‡Baruch College, (607) 262-1060, Yildiray.Yildirim@baruch.cuny.edu§Hong Kong Baptist University, (852) 3411-5349, jianzhang@hkbu.edu.hk

1 Introduction

The Dodd-Frank Wall Street Reform and Consumer Protection Act of 2010

(Dodd-Frank Act) was one of the signature legislative and regulatory actions

arising from the Great Recession and financial crisis.1 The purpose of the Dodd-

Frank Act was to attack perceived structural deficiencies in the financial markets

that led to the Great Recession. Given the central role that mortgage securiti-

zation played in the period prior to the financial crisis, new regulations covering

mortgage origination, securitization, and investment are central features of the

Dodd-Frank legislation. In particular, the Dodd-Frank Act codified a new “Risk

Retention Rule” for issuers of mortgage-backed securities.2 Under the theory

that the creators of securitization deals should have interests aligned with in-

vestors (Demiroglu and James, 2014), this rule requires that sponsors retain 5%

of the credit risk in a CMBS deal.3

While the majority of the Dodd-Frank Act regulations center on the resi-

dential mortgage market, many of the Act’s provisions are much broader and

target other areas of the capital markets. For example, the credit risk retention

rule also applies to commercial mortgage backed securities (CMBS) and spec-

ifies that the CMBS sponsor can meet the 5% risk retention through holding

a vertical or horizontal interest in the deal or by selling the credit risk inter-

est to the so-called “B-Piece Buyer.” A vertical interest refers to the sponsor

having an ownership position of at least 5% in each of the CMBS deal classes

1The Dodd-Frank Act is available online at http://www.cftc.gov/idc/groups/public/

@swaps/documents/file/hr4173_enrolledbill.pdf.2The risk retention rule became effective on December 24, 2016 as a joint regulation of

the Office of the Comptroller of the Currency, the Board of Governors of the Federal Re-serve System, the Federal Deposit Insurance Corporation, and the Securities and ExchangeCommission as specified in Section 941 (Regulation of credit risk retention), Subtitle D (Im-provements to the Asset-Backed Securitization Process), of Title IX (Investor Protections andImprovements to the Regulation of Securities), which amended the Securities Exchange Actof 1934 (15 U.S.C. 78a et seq.).

3See Lewis (2014) and Sargent and Jewesson (2016) for comprehensive overviews of thecredit risk retention rule. Furthermore, Floros and White (2016) note that risk retention isdesigned to attenuate moral hazard between loan originators and investors.

1

(or tranches) while a horizontal interest refers to the first-loss position for the

CMBS deal. However, regardless of type of interest, the risk-retention rule ef-

fectively requires the CMBS issuer to retain exposure to the underly mortgage

pool.

Interestingly, the risk retention rule created a number of exemptions for

securitization deals backed by “Qualified Mortgages” (QM). While most of the

attention surrounding these exemptions focuses on residential mortgages, certain

commercial mortgages also satisfy the regulations for QM status.4 For example,

the exemption applies to mortgages on multifamily or health care facilities that

are insured or guaranteed by the U.S. government or the government sponsored

enterprises (Fannie Mae and Freddie Mac) while they remain in conservatorship

(see Harris et al., 2015). However, the exemptions are narrowly defined such

that the majority of commercial real estate loans that comprise CMBS deals do

not meet the requirements.5

The carve-out for CMBS deals comprising multifamily loans guaranteed by

the GSEs creates an natural experiment for testing the impact of the Dodd-

Frank Act risk retention rules on the commercial loan market. Since the pri-

mary objective of this rule is to require the deal sponsors to participate in the

risks associated with the loans backing the CMBS, we expect that underwriting

standards should tighten following the implementation of the rule. As a result,

4Section 15G (Credit Risk Retention) of the Securities Exchange Act of 1934 (as amendedby Section 941 of the Dodd-Frank Act) outlines the exemptions and exceptions to the creditrisk retention rule for Qualified Residential Mortgages (see Section 15G.–Credit Risk Reten-tion, paragraph (e)–Exemptions, Exceptions, and Adjustments, subparagraph (B)–QualifiedResidential Mortgage.) Furthermore, Sections 1234.15 “Qualifying commercial loans, com-mercial real estate loans, and automobile loans” and 1234.17 “Underwriting standards forqualifying CRE loans” of Subpart D–Exceptions and Exemptions of Part 1234–Credit RiskRetention (Eff. 2-23-15) of Chapter XII–Federal Housing Finance Agency of Title 12–Banksand Banking of the Code of Federal Regulations outline the exemptions for qualifying com-mercial real estate loans.

5McBride (2014) notes that only between 3.6% to 15.6% (by loan balance) of Trepp’s publicconduit universe of CMBS deals issued between 1997 and 2013 would meet the qualified CREexemption. Thus, the vast majority of CMBS transaction are subject to the risk retentionrules set forth by the Dodd-Frank Act.

2

the Dodd-Frank Act regulations may have a profound impact on capital costs

in the commercial real estate market.

In order to test the impact of the Dodd-Frank risk retention rule on the com-

mercial real estate market, we examine various individual loan risk metrics as

well as changes in issuer actions. For example, one aspect reflecting a change in

risk is the observable ex post default performance for commercial loans. Thus,

if the Dodd-Frank regulation altered underwriting and created incentives for

less risk taking, then we should see this change reflected in lower ex post de-

fault rates following enactment of the regulations. We also examine changes

in loan origination spreads to see the whether the risk retention regulation is

reflected loan pricing. Finally, we look to the repose of the CMBS issuers to

changes in the risk retention rule by focusing on loan warehouse time (time-to-

securitization), deal level pricing as reflected in average deal coupons, and issuer

rating shopping.

2 Empirical Method and Data

We obtained data from Trepp on 25,604 commercial mortgages originated be-

tween January 2014 and September 2017.6 These loans were securitized in 425

CMBS deals.

Figures 3.2 and 3.2 show the frequency distribution of loans and CMBS

deals by year/quarter of origination. Consistent with implementation of the

risk retention rule in December 2016, the number of non-agency (or conduit)

CMBS deals declines 60 percent in the first quarter of 2017 from the previous

quarter. At the same time, issuance of agency CMBS in the first quarter of

2017 is 27 percent lower than the fourth quarter of 2016. We also note a similar

6Trepp is cited as one of the real estate industry’s largest providers of information onsecuritized commercial mortgages. Trepp tracks over 1,500 CMBS deals comprising over100,000 mortgages. More information about Trepp is available at http://www.trepp.com/

about-us.

3

decline in the number of non-agency loans.

Table 1 provides summary statistics for the CMBS deal characteristics (Panel

A) and loan-level characteristics (Panel B).

We estimate the following difference-in-difference regression to examine the

affect of the risk retention rule on commercial mortgages:

yi = α+ β1 ∗ Conduiti + β2 ∗ Posti + β3 ∗ (Posti ∗ Conduiti)

+β4 ∗Announcei + β5 ∗ (Announcei ∗ Conduiti) + θ ∗ Zi + εi, (1)

where Conduiti represents a dummy variable denoting loans originated by a

conduit lender (i.e. loans that are not guarantee by a government agency or

part of an agency CMBS deal); Announcei is a dummy variable indicating

loans originated after the announcement of the risk retention rule, Posti is a

dummy variable indicating loans originated originated after the risk retention

rule (December 2016); and Zi is a set of control variables (loan-to-value ratio,

loan maturity, and cap rate) as well as property, originator, time, and location

(state) fixed effects. The coefficient on the interaction terms (Announcei ∗

Conduiti and Posti ∗ Conduiti) are the diff-in-diff parameters that represent

the differential effect of the risk retention regulation on non-agency (conduit)

loans. If the risk retention regulation effectively reduced asymmetric information

or altered the underwriting standards of commercial mortgages, then we expect

β3 and β5 to be significant.

The dependent variable, yi, represents a set of variables that proxy for either

the riskiness of mortgages or the response of the CMBS issuer. For example, we

examine the log time-to-sale, defined as the time lag between mortgage origi-

nation and CMBS issuance (Titman and Tsyplakov, 2010), as a measure of the

costs associated with securitization. To the extent that the risk retention regu-

lation altered the liquidity of the CMBS market, then this should be reflected

4

in a change in the time-to-securitization. We also estimate equation 1 using

a measure of mortgage performance (probability of default) as the dependent

variable since the intention of the risk retention rule is that it causes originators

to internalize the costs of bad underwriting (since CMBS originators are now

required to hold part of the deal).

3 Results

Tables 2 through 6 report the estimation results for equation (1) focusing on

changes in individual loan risk and issuer actions. We examine the loan interest

rate spread and ex-post default probability to capture changes in loan risk while

changes in loan time-to-sale, deal level coupons, and rating shopping capture

responses by the security issuers. The tables follow the same pattern where

column (1) reports the regression with property fixed-effects only while the re-

gression reported in column (2) contains the full set of property type, originator,

origination month, and state level location fixed effects. The inclusion of time

fixed-effects in column (2) precludes the inclusion of the individual indicator

variables for announcement and post-reform periods. Columns (3) and (4) re-

peat the specifications but focus only on the post-reform period. In tables 2 and

3 we focus on changes in risk at the individual loan level to the new regulation

stemming from changes in loan originator underwriting. Tables 4, 5, and 6 turn

to the response of the mortgage-backed security issuer.

3.1 Risk Retention and Individual Loan Risk

To examine the affect of the new risk retention rule on commercial real estate

loans, we examine changes in loan pricing and performance. We anticipate

that mortgage originators responded to the greater CMBS issuer due diligence

requirements by imposing tighter underwriting standards. We observe the out-

5

come of the changes in underwriting by focusing on changes in ex post mortgage

default and the risk premium (loan spread).

Table 2 reports the regression results for loan pricing. The dependent vari-

able is the mortgage spread, which is defined as the difference between the

mortgage interest rate at origination and the comparable maturity Treasury

rate observed at the mortgage origination date. The spread is thus a proxy for

the loan’s risk premium.

As expected, the positive and statistically significant coefficient for Conduit

indicates that non-agency mortgages have higher average loan spreads (risk

premiums). This is consistent with the credit enhancement provided by the

GSE’s. The difference-in-difference parameters (Conduit*Announcement and

Conduit*Post-Reform) are negative and statistically significant, revealing the

expected decline in relative risk premiums on affected mortgages following the

introduction of the risk retention rule. The significant decline in the risk pre-

miums, relative to the non-treated agency mortgages, is consistent with lenders

imposing tighter underwriting standards on commercial mortgages.

Table 3 reports the estimated coefficients for the model of the probability

of loan default (90+ days delinquency) over the 24-months following securiti-

zation. The results confirm the view that lenders tightened underwriting stan-

dards following the announcement and enactment of the risk-retention rule. For

example, the difference-in-difference parameters (Conduit*Announcement and

Conduit*Post-Reform) are negative and statistically significant indicating that

the relative probability of default on non-agency mortgages declined following

the risk retention rule announcement and enactment.

6

3.2 Risk Retention and Issuer Response

In this section, we explore issuer response to changes in the risk retention rule.

We focus on the time between loan origination and securitization (otherwise

known as time-to-sale or warehouse risk) as an observable proxy for the level of

due diligence the CMBS issuer conducts on the mortgage pool prior to securi-

tization. Since the purpose of the risk retention rule is to require the security

issuer to hold a risk position in the security (i.e. have ‘skin-in-the-game’), we

expect a positive shift in the time-to-sale for affected loans following the rule an-

nouncement as issuers subject individual loans to greater due diligence scrutiny,

necessitating a longer warehouse period. As a result, the risk retention rule

should mitigate the natural tendency for issuers to quickly securitize higher-

risk loans. Table 4 presents the estimation results of equation (1) for the full

sample where the dependent variable is the time between loan origination and

securitization (time-to-sale).

Turing first to the control variables, we see that LTV and the capitalization

rate (NOI/Property Vale) have the expected signs. Consistent with incentives

to quickly securitize higher risk loans, we note that the negative and statistically

significant coefficient for LTV implies that mortgages with higher loan-to-value

ratios (or less equity on the part of the borrower) are securitized faster than loans

with lower LTV ratios. Furthermore, the positive and significant coefficient for

the cap rate suggests that, all else equal, loans collateralized by properties that

have higher cap rates are viewed as less risky and thus have longer times-to-

securitization. The intuition is that as borrowers pay higher values per dollar

of cash flow (resulting in lower cap rates), the loans underlying those properties

have greater risk. In contrast, when borrowers pay less per dollar of cash flow

(i.e. a higher cap rate), then the loan underlying the property has less risk.

The negative coefficients for Conduit indicate that loans in non-agency CMBS

7

deals have a significantly shorter period between origination and securitization

than multi-family GSE loans, on average. Again, this is consistent with the

view that time-to-securitization is a proxy for risk appetite on the part of the

issuer since GSE-backed loans are less risky than non-GSE backed loans due

to the credit enhancement provided by the GSEs. As a result, originators are

willing to take on greater warehouse risk for GSE-backed loans as the default

risk exposure on these loans is minimal.

Turning to the impact of the new risk retention rule, we see that the negative

coefficients for Annoucnement and Post-Reform indicate that loans originated in

the announcement period and after the regulation enactment have significantly

shorter time-to-securitization than loans originated prior to the discussion of the

regulation. In order to assess the differential effect of the regulation on mort-

gages subject to the new regulation, we note that the positive and significant

coefficient for Conduit*Announcement implies that the time-to-securitization

for non-agency loans significantly lengthened relative to agency-backed mort-

gages following the announcement of the risk-retention rules. This is consistent

with issuers requiring longer periods for due diligence while creating securities

comprising loans subject to the new regulation. Furthermore, the significant

positive coefficient for Conduit*Post-Reform confirms a permanent increase in

securitization time for non-agency loans relative to agency-backed mortgages.

In table 5 we turn to differences in deal level pricing following the introduc-

tion of the risk retention rule. As expected, the positive and statistically signif-

icant coefficient for Conduit indicates that non-agency CMBS deals have higher

weighted-average coupons than CMBS backed by agency loans. However, the

negative and statistically significant coefficient for Conduit*Post-Reform sug-

gests that following the enactment of the risk-retention rule, relative pricing

premium demanded for non-agency CMBS deals declined. Again, this is con-

8

sistent with the impact of the regulation reducing the overall risk profile of

mortgage-backed securities.

Finally, table 6 turns to the actions of the issuer in securing ratings for the

CMBS deal. In this regression, we exploit a unique feature of the structured

finance product. We note that rating a CMBS tranche that lies below others

in the security waterfall requires an analysis of the interest that the waterfall

promises to the tranches above it (Flynn and Ghent, 0). Thus, the dependent

variable takes a value of one if the deal has a tranche with a missing rating but

the tranche below it in the capital stack is rated. As expected, the positive and

significant coefficient for Conduit indicates that non-agency CMBS deals are

more likely to be associated with issuer rating shopping. However, consistent

with the risk retention regulation altering the incentives to originate risky deals,

we see that the coefficient on the interaction term (Conduit*Post-Reform) is

negative and significant. This indicates that following enactment of the risk

retention rule, issuers were less likely to engage in rating shopping.

9

References

Demiroglu, C. and James, C. M. (2014). The Dodd-Frank Act and the Reg-

ulation of Risk Retention in Mortgage-Backed Securities. In Schultz, PH,

editor, Perspectives on Dodd-Frank and Finance, pages 201–216. Conference

on Dodd-Frank and the Future of Finance, Washington, DC, JUN 13-14,

2013.

Floros, I. and White, J. T. (2016). Qualified residential mortgages and default

risk. JOURNAL OF BANKING & FINANCE, 70:86–104.

Flynn, S. and Ghent, A. (0). Competition and credit ratings after the fall.

Management Science, 0(0):null.

Harris, M., Simonds, R., and Liebherr, L. (2015). Risk Retention and RMBS.

Lewis, M. (2014). A Guide to the Credit Risk Retention Rules for Securitiza-

tions.

McBride, J. (2014). What Qualifies? Risk Retention in

CMBS. http://info.trepp.com/TreppTalk/bid/333936/

What-Qualifies-Risk-Retention-in-CMBS.

Sargent, P. C. and Jewesson, M. D. (2016). The Dawn of CMBS 4.0: Changes

and Challenges in a New Regulatory Regime.

Titman, S. and Tsyplakov, S. (2010). Originator Performance, CMBS Struc-

tures, and the Risk of Commercial Mortgages. REVIEW OF FINANCIAL

STUDIES, 23(9):3558–3594.

10

Figure 1: Frequency Distribution of Agency and Non-Agency CMBS Deals byOrigination Date

11

Figure 2: Frequency Distribution of Agency and Non-Agency Loans by Origi-nation Date

12

Tab

le1:

Des

crip

tive

Sta

tist

ics

Var

iab

leN

mea

nsd

p25

p50

p75

Pan

elA

Dea

l-L

evel

Ch

ara

cter

isti

csC

ond

uit

(Tre

ated

Dea

l)425

0.4

61

0.4

99

00

1D

eal

Cou

pon

(%)

425

4.2

49

0.6

63

4.0

02

4.3

19

4.6

46

Dea

lS

pre

ad(%

)425

2.1

80

0.7

27

1.7

90

2.2

03

2.5

22

D(R

atin

gS

hop

pin

g)425

0.3

36

0.4

73

00

1W

eigh

ted

Ave

rage

Loa

n-t

o-V

alu

e(%

)425

65.2

78

5.9

02

61.4

50

66.0

20

69.3

50

Wei

ghte

dA

vera

geM

atu

rity

(Month

s)425

113.0

80

34.0

54

104

114

117

Sh

are

ofT

op10

Loa

ns

425

46.3

38

16.5

67

35.8

89

45.7

25

55.0

27

Sh

ares

ofM

ult

ifam

ily

Loan

s425

58.0

01

43.1

41

12.0

30

90.5

91

100.0

00

Sh

are

ofR

etai

lP

rop

erty

Loan

425

12.2

07

14.3

63

00

24.8

85

Sh

are

ofO

ffice

Pro

per

tyL

oan

425

11.9

28

15.0

99

00

24.0

94

Sh

are

ofIn

du

stry

Loa

ns

425

2.2

85

3.9

58

00

3.2

80

Sh

are

ofH

otel

Loa

ns

425

7.1

08

8.5

96

00

15.2

45

Sh

are

ofSel

f-Sto

rage

Loan

s425

1.8

40

3.1

52

00

2.5

50

Sh

are

ofO

ther

Typ

eL

oan

425

6.6

31

7.8

85

03.9

70

10.8

91

Pan

elB

Loan

-Lev

elC

hara

cter

isti

csM

ortg

age

Rat

e(%

)23684

4.2

89

0.7

19

3.9

00

4.3

80

4.7

40

Mor

tgag

eS

pre

ad(%

)23684

2.1

47

0.7

46

1.7

50

2.2

00

2.5

80

Tim

e-to

-Sal

e(d

ays)

23684

129.7

55

125.1

49

58.0

00

98.0

00

148.0

00

Del

inqu

ency

(lat

ep

aym

ent,

%)

23684

7.1

61

25.7

85

0.0

00

0.0

00

0.0

00

Del

inqu

ency

(30

day

s+

,%

)23684

0.4

26

6.5

16

0.0

00

0.0

00

0.0

00

Del

inqu

ency

(60

day

s+

,%

)23684

0.1

77

4.2

07

0.0

00

0.0

00

0.0

00

Del

inqu

ency

(90

day

s+

,%

)23684

0.1

10

3.3

12

0.0

00

0.0

00

0.0

00

Loa

n-t

o-V

alu

e(%

)23684

66.0

32

12.3

21

62.2

70

69.1

00

74.2

00

Loa

nM

atu

rity

(mon

ths)

23684

123.4

07

44.7

15

120.0

00

120.0

00

120.0

00

NO

I/(P

rop

erty

valu

e)23684

0.1

38

0.2

06

0.0

82

0.0

94

0.1

14

D(M

ult

ifam

ily

Loa

ns)

23684

0.5

61

0.4

96

0.0

00

1.0

00

1.0

00

D(R

etai

lP

rop

erty

Loa

n)

23684

0.1

48

0.3

55

0.0

00

0.0

00

0.0

00

D(O

ffice

Pro

per

tyL

oan

)23684

0.0

74

0.2

62

0.0

00

0.0

00

0.0

00

D(I

nd

ust

ryL

oan

s)23684

0.0

24

0.1

54

0.0

00

0.0

00

0.0

00

D(H

otel

Loa

n)

23684

0.0

68

0.2

52

0.0

00

0.0

00

0.0

00

D(S

elf-

Sto

rage

Loa

n)

23684

0.0

40

0.1

95

0.0

00

0.0

00

0.0

00

D(O

ther

Typ

eL

oan

)23684

0.0

86

0.2

80

0.0

00

0.0

00

0.0

00

13

Table 2: Risk Retention Rule and Mortgage Rate Spread(1) (2) (3) (4)

VARIABLES Mortgage Spread

Conduit (Treated) 0.393*** 0.100 0.254*** 0.066(8.02) (0.57) (6.26) (0.36)

Announcement 0.312***(7.98)

Conduit*Announcement -0.136*** -0.049(-3.10) (-1.03)

Post-Reform 0.122** -0.136***(2.57) (-4.18)

Conduit*Post-Reform -0.421*** -0.228*** -0.288*** -0.185***(-6.40) (-3.28) (-6.11) (-4.27)

LTV -0.053*** -0.051*** -0.053*** -0.050***(-14.18) (-12.88) (-15.14) (-14.00)

Maturity -0.004*** -0.004*** -0.004*** -0.004***(-8.76) (-10.91) (-7.95) (-10.85)

NOI/(Property Value) -0.004 0.228* 0.003 0.229*(-0.03) (1.89) (0.02) (1.89)

Constant 5.782*** 5.940*** 6.071*** 5.920***(23.77) (17.91) (26.06) (18.35)

Property type Fixed effects Yes Yes Yes YesOriginator Fixed effects NO Yes NO YesOrigination Month Fixed Effects NO Yes NO YesOrigination State Fixed Effects NO Yes NO YesObservations 23,684 23,684 23,684 23,684R-squared 0.31 0.47 0.29 0.47

Regression estimates of the impact of Risk-Retention-Rule on mortgage rate at the loan level.Sample includes loans of all CMBS originated from Jan 2014 to Sep 2017. Dependent variablesincludes: 1.Mortgage rate is the rate on the mortgage; 2. Spread is calculated as the differencebetween the mortgage rate and comparable maturity Treasury rate observed at the mortgageorigination date. Linear interpolation is applied to Treasury rates to obtain the full termstructure.Announcement is a dummy indicating the period of Dec 2014 - Nov 2016. Post-Reformis a dummy variable for CMBS originated after the reform, Dec 2016. Conduit is an indicatorvariable that equals to one for Conduit CMBS and zero for Agency CMBS. Controls includeloan-to-value, mortgage maturity (in months) and NOI/(Property value), the ratio of netoperating income divided by property value at origination, and property type dummies.Originator fixed effects is included in Column (1),(3),(5) and (7) while mortgageoriginator,origination state and quarter are included in Column (2),(4),(6) and (8). Mortgagebalance is used as the weight in Column (3) and (4). Robust T-statistics are reported inparenthesis and are based on standard errors clustered by originators and state. *, **, and ***indicate significance levels of 10%, 5%, and 1%, respectively.

14

Table 3: Risk-Retention Rule and Ex-post Loan-level Performance(1) (2) (3) (4)

VARIABLES D(Default) - 90 days +Conduit (Treated) 0.005*** 0.002* 0.003*** 0.000

(3.73) (1.90) (4.14) (0.42)Announcement -0.000

(-0.06)Conduit*Announcement -0.003** -0.002*

(-2.07) (-1.90)Post-Reform -0.000 -0.000

(-1.06) (-1.09)Conduit*Post-Reform -0.004*** -0.003** -0.002*** -0.001

(-2.81) (-2.47) (-2.75) (-1.23)LTV 0.000 0.000 0.000** 0.000

(1.27) (0.59) (2.32) (1.23)Maturity 0.000 0.000 0.000 0.000

(0.55) (0.21) (0.60) (0.31)NOI/(Property Value) 0.000 0.000 0.000 0.000

(0.24) (0.65) (0.13) (0.68)Constant -0.005 -0.007 -0.008** -0.008*

(-1.28) (-1.51) (-2.29) (-1.77)Property type Fixed effects Yes Yes Yes YesOriginator Fixed effects NO Yes NO YesOrigination Month Fixed Effects NO Yes NO YesOrigination State Fixed Effects NO Yes NO YesObservations 23,684 23,684 23,684 23,684R-squared 0.00 0.04 0.00 0.04

Regression estimates of the impact of Risk-Retention-Rule on ex post performance at the loanlevel. Sample includes loans of all CMBS originated from Jan 2014 to Sep 2017. Dependentvariable, D(Default) is defined as dummy variable that equals to one if the mortgage is classifiedeither as deliquent,foreclosed or ROE. Four versions are created depending on deliquency status.For example, ”D(Default) - late payment +” denotes any type of late payment while ”D(Default) -90s +” represents only late payment with at least 90 days. Announcement is a dummy indicatingthe period of Dec 2014 - Dec 2016. Post-Reform is a dummy variable for CMBS originated afterthe reform, Dec 2016. Conduit is an indicator variable that equals to one for loans of ConduitCMBS and zero for loans of Agency CMBS. Controls include loan-to-value, mortgage maturity (inmonths) and NOI/(Property value), the ratio of net operating income divided by property value atorigination, and property type dummies. Originator fixed effects is included in Column (1),(3),(5)and (7) while mortgage originator,origination state and quarter are included in Column (2),(4),(6)and (8). Robust T-statistics are reported in parenthesis and are based on standard errorsclustered by originators and state. *, **, and *** indicate significance levels of 10%, 5%, and 1%,respectively.

15

Table 4: Risk-Retention Rule and Unobservable Loan Quality - Loan Level(1) (2) (3) (4)

VARIABLES Time-to-SaleFull Sample

Conduit (Treated) -219.230*** -214.635*** -154.716*** -172.686***(-9.56) (-10.34) (-12.56) (-11.53)

Announcement -90.062***(-5.24)

Conduit*Announcement 83.513*** 62.832***(5.29) (4.97)

Post-Reform -146.162*** -71.929***(-6.56) (-7.47)

Conduit*Post-Reform 94.071*** 91.975*** 20.113*** 36.843***(5.35) (7.70) (3.21) (9.57)

LTV -4.864*** -4.253*** -5.535*** -5.137***(-5.92) (-11.48) (-5.92) (-13.37)

Maturity 0.076 0.030 0.022 0.019(0.66) (0.75) (0.18) (0.47)

NOI/(Property Value) 103.957*** 131.224*** 102.700*** 130.624***(4.46) (4.55) (4.40) (4.53)

Constant 582.097*** 530.116*** 562.091*** 564.683***(8.75) (11.18) (8.81) (10.88)

Property type Fixed effects Yes Yes Yes YesOriginator Fixed effects NO Yes NO YesOrigination Month Fixed Effects NO Yes NO YesOrigination State Fixed Effects NO Yes NO YesObservations 23,684 23,684 23,684 23,684R-squared 0.31 0.50 0.27 0.50

Regression estimates of the impact of Risk-Retention-Rule on time-to-sale(shelf risk) at the loanlevel. Sample includes loans of all CMBS originated from Jan 2014 to Sep 2017. Dependentvariable, time-to-sale, is defined as the time lag between mortgage origination and CMBS issuance(Titman and Tsyplakov, 2010). Announcement is a dummy indicating the period of Dec 2014 -Dec 2016. Post-Reform is a dummy variable for CMBS originated after the reform, Dec 2016.Conduit is an indicator variable that equals to one for Conduit CMBS and zero for Agency CMBS.Controls include loan-to-value, mortgage maturity (in months) and NOI/(Property value), theratio of net operating income divided by property value at origination, and property typedummies. Originator fixed effects is included in Column (1) and (3) while mortgageoriginator,origination state and quarter are included. Mortgage balance is used as the weight inColumn (3) and (4). Robust T-statistics are reported in parenthesis and are based on standarderrors clustered by originators and state. *, **, and *** indicate significance levels of 10%, 5%,and 1%, respectively.

16

Table 5: Risk-Retention Rule and CMBS Pricing(1) (2) (3) (4)

VARIABLES Deal Coupon

Conduit (Treated) 0.470** 0.246 0.747*** 0.154(2.14) (0.84) (3.10) (0.62)

Post-Announcement -0.317*(-2.06)

Conduit*Post-Announcement -0.138 -0.200(-1.01) (-1.25)

Post-Reform -0.148 0.117(-1.04) (1.35)

Conduit*Post-Reform -0.435*** -0.519*** -0.325*** -0.339***(-3.11) (-3.25) (-3.08) (-4.07)

Weighted Average LTV -0.049*** -0.052*** -0.046*** -0.050***(-5.41) (-7.69) (-4.81) (-6.70)

Weighted Average Maturity -0.000 -0.000 -0.001 0.000(-0.05) (-0.04) (-0.40) (0.07)

Share of Top 10 Loans -0.009*** -0.011*** -0.010*** -0.010***(-2.99) (-3.88) (-3.19) (-3.67)

Property Type Shares (vs other use type)Shares of Multifamily Loans -0.003 -0.001 -0.003 -0.001

(-0.93) (-0.34) (-0.78) (-0.33)Share of Retail Property Loan -0.006* -0.003 -0.008* -0.003

(-1.90) (-0.44) (-1.85) (-0.46)Share of Office Property Loan -0.006 0.000 -0.010* -0.000

(-1.36) (0.04) (-1.97) (-0.13)Share of Industry Loans -0.001 -0.001 -0.009 -0.002

(-0.16) (-0.09) (-1.04) (-0.28)Share of Hotel Loans 0.003 0.009 -0.005 0.007

(0.34) (0.91) (-0.60) (0.78)Share of Self-Storage Loans 0.005 0.017 -0.001 0.015

(0.62) (1.52) (-0.12) (1.33)Constant 8.251*** 8.455*** 7.874*** 8.302***

(9.27) (9.95) (8.70) (9.44)Securitization Month Fixed Effects NO Yes NO YesLeading Underwriter Fixed Effects NO Yes NO YesObservations 425 425 425 425R-squared 0.33 0.45 0.29 0.45

Regression estimates of the impact of Risk-Retention-Rule on CMBS pricing. Sample includesCMBS originated from Jan 2014 to Sep 2017. Dependent variables includes: 1.Deal Coupon is theCMBS deal weighted average coupon(WAC) paid t investor; 2. Deal Spread is calculated as theDeal Coupon inus comparable maturity Treasury rate observed at the secrutization date. Linearinterpolation is applied to Treasury rates to obtain the full term structure. Post-Reform is adummy variable for CMBS originated after the reform, Dec 2016. Conduit is an indicator variablethat equals to one for Conduit CMBS and zero for Agency CMBS. Controls include weightedaverage loan-to-value, weighted average mortgage maturity (in months), shares of top 10 loans, aswell as shares of loans with different property type. Deal balance is used as the weight in Column(5) to (8). Robust T-statistics are reported in parenthesis and are based on standard errorsclustered by leading-underwriter. *, **, and *** indicate significance levels of 10%, 5%, and 1%,respectively.

17

Table 6: Risk-Retention Rule and Rating ShoppingD(Rating Shopping)

VARIABLES

Conduit (Treated) 0.863* 0.886** 0.920** 0.829**(1.95) (2.61) (2.37) (2.72)

Post-Announcement -0.064(-0.48)

Conduit*Post-Announcement -0.006 -0.125(-0.04) (-0.81)

Post-Reform -0.085 -0.031(-0.58) (-0.61)

Conduit*Post-Reform -0.393*** -0.558*** -0.390*** -0.445***(-3.34) (-3.76) (-3.50) (-3.64)

Weighted Average LTV 0.004 0.003 0.004 0.004*(1.70) (1.71) (1.57) (2.00)

Weighted Average Maturity -0.001 -0.001 -0.001 -0.001(-1.12) (-0.96) (-1.23) (-0.88)

Share of Top 10 Loans -0.004*** -0.003*** -0.004*** -0.003***(-3.97) (-3.75) (-4.87) (-4.08)

Property Type Shares (vs other use type)Shares of Multifamily Loans 0.006 0.003 0.006 0.003

(1.64) (1.00) (1.63) (1.00)Sahre of Retail Property Loan -0.001 -0.005 -0.002 -0.005

(-0.42) (-1.48) (-0.55) (-1.49)Sahre of Office Property Loan 0.003 0.001 0.002 0.001

(0.84) (0.41) (0.68) (0.28)Sahre of Industry Loans 0.007 0.005 0.006 0.004

(1.27) (0.76) (0.95) (0.68)Sahre of Hotel Loans 0.004 -0.001 0.003 -0.002

(0.54) (-0.16) (0.36) (-0.27)Sahre of Self-Storage Loans 0.006 0.008 0.005 0.007

(0.59) (0.98) (0.53) (0.86)Constant -0.439 -0.293 -0.498 -0.390

(-1.00) (-0.99) (-1.07) (-1.08)Securitization Month Fixed Effects NO Yes NO YesLeading Underwriter Fixed Effects NO Yes NO YesObservations 425 425 425 425R-squared 0.25 0.41 0.25 0.41

We exploit a unique feature of the structured finance product - rating a security that lies belowothers in the waterfall requires an analysis of the interest that the waterfall promises to thetranches above it Flynn and Ghent(2017,MS) - and define the deal-level measure of ratingshopping.Regression estimates of the impact of Risk-Retention-Rule on rating shopping at the deal level.Sample includes CMBS originated from Jan 2014 to Sep 2017. Dependent variable, rating shoping,takes a value of 1 for a CMBS deal for which a tranche in the deal is missing a rating from acredit rating agency but a trache below it in the waterfall has a rating from the same credit ratingagency. Post-Reform is a dummy variable for CMBS originated after the reform, Dec 2016.Conduit is an indicator variable that equals to one for Conduit CMBS and zero for AgencyCMBS.Controls include weighted average loan-to-value, weighted average mortgage maturity (inmonths), shares of top 10 loans, as well as shares of loans with different property type. Dealbalance is used as the weight in Column (3) and (4). Robust T-statistics are reported inparenthesis and are based on standard errors clustered by leading-underwriter. *, **, and ***indicate significance levels of 10%, 5%, and 1%, respectively.

18

Tab

leA

1:R

isk-R

eten

tion

Ru

lean

dU

nob

serv

ab

leL

oan

Qu

ali

ty-

Loan

Lev

el(1

)(2

)(3

)(4

)V

AR

IAB

LE

ST

ime-

to-S

ale

Mu

ltip

le-F

am

ily

Loan

s

Con

du

it(T

reat

ed)

-203.1

18***

-187.5

14***

-140.2

58***

-147.5

34***

(-8.0

2)

(-16.5

5)

(-8.3

4)

(-16.1

0)

An

nou

nce

men

t-8

3.0

45***

(-5.1

5)

Con

du

it*A

nn

oun

cem

ent

87.9

85***

62.7

55***

(6.3

9)

(5.1

3)

Pos

t-R

efor

m-1

32.1

67***

-63.3

82***

(-6.0

0)

(-6.3

2)

Con

du

it*P

ost-

Ref

orm

88.8

41***

98.3

60***

8.1

73

48.6

51***

(6.1

5)

(7.8

3)

(0.9

5)

(6.5

8)

LT

V-4

.641***

-2.3

82***

-6.0

09***

-2.8

20***

(-3.1

0)

(-4.3

7)

(-3.4

1)

(-5.3

5)

Mat

uri

ty0.1

70

0.1

08**

0.1

06

0.1

03**

(1.5

2)

(2.5

0)

(0.9

5)

(2.3

8)

NO

I/(P

rop

erty

Val

ue)

404.1

66***

407.7

51***

400.1

86***

405.7

76***

(8.1

4)

(8.7

9)

(7.9

6)

(8.7

6)

Con

stan

t512.7

98***

304.7

40***

548.1

85***

324.1

31***

(4.3

9)

(6.8

3)

(4.4

0)

(7.2

2)

Ori

gin

ator

Fix

edeff

ects

NO

Yes

NO

Yes

Ori

gin

atio

nM

onth

Fix

edE

ffec

tsN

OY

esN

OY

esO

rigi

nat

ion

Sta

teF

ixed

Eff

ects

NO

Yes

NO

Yes

Ob

serv

atio

ns

13,6

12

13,6

12

13,6

12

13,6

12

R-s

qu

ared

0.2

90.5

20.2

50.5

2

Regre

ssio

nest

imate

sof

the

impact

of

Ris

k-R

ete

nti

on-R

ule

on

tim

e-t

o-s

ale

(shelf

risk

)at

the

loan

level.

Sam

ple

inclu

des

loans

of

all

CM

BS

ori

gin

ate

dfr

om

Jan

2014

toSep

2017.

Dep

endent

vari

able

,ti

me-t

o-s

ale

,is

defined

as

the

tim

ela

gb

etw

een

mort

gage

ori

gin

ati

on

and

CM

BS

issu

ance

(Tit

man

and

Tsy

pla

kov,

2010).

Announcem

ent

isa

dum

my

indic

ati

ng

the

peri

od

of

Dec

2014

-D

ec

2016.

Post

-Refo

rmis

adum

my

vari

able

for

CM

BS

ori

gin

ate

daft

er

the

refo

rm,

Dec

2016.

Conduit

isan

indic

ato

rvari

able

that

equals

toone

for

Conduit

CM

BS

and

zero

for

Agency

CM

BS.

Contr

ols

inclu

de

loan-t

o-v

alu

e,

mort

gage

matu

rity

(in

month

s)and

NO

I/(P

rop

ert

yvalu

e),

the

rati

oof

net

op

era

ting

incom

ediv

ided

by

pro

pert

yvalu

eat

ori

gin

ati

on,

and

pro

pert

yty

pe

dum

mie

s.O

rigin

ato

rfixed

eff

ects

isin

clu

ded

inC

olu

mn

(1)

and

(3)

while

mort

gage

ori

gin

ato

r,ori

gin

ati

on

state

and

quart

er

are

inclu

ded.

Mort

gage

bala

nce

isuse

das

the

weig

ht

inC

olu

mn

(3)

and

(4).

Robust

T-s

tati

stic

sare

rep

ort

ed

inpare

nth

esi

sand

are

base

don

standard

err

ors

clu

stere

dby

ori

gin

ato

rsand

state

.*,

**,

and

***

indic

ate

signifi

cance

levels

of

10%

,5%

,and

1%

,re

specti

vely

.

19