Us Bank Mortgage Exposure

Document Sample
Us Bank Mortgage Exposure Powered By Docstoc
					            The Impacts of Securitization on US Bank Holding Companies


                                                          By


                                           Wenying Jiangli and Matt Pritsker 1

                                                    March 2008




                                                   ABSTRACT



We use data from 2001-2007 to assess the impact of mortgage and other forms of asset
securitization on the insolvency risk, profitability, and leverage ratios of US bank holding
companies. Using 3 different estimation techniques, we find that banks use mortgage securitization
to reduce insolvency risk and increase leverage. We also find that securitization techniques
increase bank profitability. Our results suggest that securitization techniques have played a
positive role. This suggests that the current turmoil in mortgage credit and securitization markets
is related to recent excesses in those markets, and that securitization activity will resume after
those excesses are cleared up.

Keywords: Banking, Securitization


JEL Classification: G21, G32




Wenying Jiangli is from Federal Deposit Insurance Corporation, 202 898 6537 and wjiangli@fdic.gov. Matt Pritsker is
from the Board of Governors of the Federal Reserve System, 202 452 3534 and mpritsker@frb.gov. The views
expressed in this paper are those of the authors but not necessarily those of the Federal Deposit Insurance Corporation,
the Board of Governors of the Federal Reserve System. We are grateful for comments from Paul Kupiec, Bill Lang,
Robert Merton, Dan Nuxoll, Paul Povel, Haluk Unal, Wolf Wagner, and members of the Basel Research Task Force
on the Integration of Market and Credit Risk, as well as seminar participants at the FDIC, the European Finance
Association, the Riksbank, and North Carolina State University. Philip Lee and Paul Reverdy provided excellent
research assistance. All errors are our own. 1


                                                           1

                        Electronic copy available at: http://ssrn.com/abstract=1102284
1. Introduction



       Asset securitizations have been an important and expanding part of banking practice since

the early 1990s. By the end of 2006, the outstanding amount of asset backed securities stood at

$US 2.1 trillion, and global total issuance of CDOs (collateralized debt obligations) reached $US

550 billion (SIFMA, 2008). Many banks have reported record earnings from securitization until

the second quarter of 2007. Since then, however, the ABS market has dried up due to the recent

credit crisis triggered by sub-prime loan defaults. Many banks which used to tap the securitization

market face near zero demand for some securitized assets, and banks have reported huge loan write

downs, and many banks face increased insolvency risk that warrant regulatory concerns

(www.securitization.net).



       Given the growth of the markets for securitization, and recent events in those markets, it is

important to understand how securitization activities affect the health of the banking sector. In this

paper, we help to address this question by using bank holding company data from 2001 to 2007 to

empirically quantify the impact of securitization on banks’ insolvency risk, profitability, and

leverage.



       Prior to the advent of securitization (or other forms of credit risk transfer), a bank's

decision to extend a loan was a bundled decision to originate the loan, hold it on balance sheet, and

fund it with the bank's debt and equity. Securitization un-bundles these activities, allowing the

bank to separately choose which loans to originate, and which loans to hold and fund on balance

sheet. For example, in a balance sheet CDO, the bank funds some of its loans that are held on




                                                   2

                    Electronic copy available at: http://ssrn.com/abstract=1102284
balance sheet, by selling them to an off-balance sheet special purpose vehicle (SPV). The SPV

funds its loan purchases by issuing securities whose payments are backed by the performance of

the loans. In many CDO securitizations, the securities issued by the SPV are tranched into risk

classes, with the least subordinate (least risky) tranches receiving the highest credit rating, and the

most risky tranche (often the equity tranche) receiving no rating at all. When a bank sells loans to

an SPV, it often retains a tranche of the securities issued by the SPV, or provides other guarantees

to the SPV structure in order to signal the quality of the loans, or to signal the bank’s intent to

monitor the borrowers.



        Securitization has the potential to significantly impact banks’ insolvency risk, leverage, and

profitability. First, for insolvency risk, securitization can be understood as a credit-derivative

transaction that transforms the risk profile of the asset side of the bank's balance sheet. Holding

the liability side of the bank's balance sheet fixed, and focusing on the asset side alone,

securitization can in theory lower insolvency risk since it can serve as insurance against bank

insolvency in some severe adverse states of the economy. This is sensible since in the standard

securitization, the upper tranches are usually sold (through the SPV) to outside investors and the

issuing bank usually holds the most subordinated or the equity tranche. The credit loss to the

equity tranche is truncated by the level of subordination while losses in the most severe states, the

tail loss, are absorbed by outside investors that own the upper tranches. 2 On the other hand,

securitization is a credit derivative that allows the bank to optimally choose its exposures to

different aspects of the credit risk of an underlying pool of loans. Therefore, depending on the




2
  Though the originating banks commonly provides credit enhancement which includes holding the firs-loss (equity)
tranche that retains some of the credit risk, this is more likely to be the mean, not the tail loss.


                                                         3
rewards for different types of risk exposures, and the exact structure of the transaction,

securitization can be used to increase or decrease the bank’s insolvency risk.



        The above analysis on insolvency risk holds the liability side of the balance sheet fixed.

Because securitization changes the risk profile of the asset side of the bank’s balance sheet, in

realistic settings with taxes and bankruptcy costs, the bank will also change its optimal on-balance

sheet capital structure, which will in turn affect its insolvency risk, and on balance sheet leverage

ratio [Leland, (2007), Jiangli, Pritsker and Raupach (2007)]. Securitization may also affect the

bank's insolvency risk and leverage if the bank engages in regulatory capital arbitrage. This

involves selling loans off its balance sheet to avoid regulatory capital charges, but retaining

exposure to the credit risk by retaining an implicit agreement with the SPV that the bank will buy

back its nonperforming loans. Regulatory capital arbitrage is anticipated to increase the bank's

insolvency risk since the bank is anticipated to reduce its regulatory capital while still maintaining

its exposure to the risk.



        Securitization can affect the bank's profit in two ways. The direct effect of securitization is

anticipated to increase bank profits. Simply put, securitization gives the bank more options for

funding its activities, and managing its risk profile. All else equal, expanded opportunities should

lead to greater expected profits. That said, securitization could lower the profits of some banks

through many indirect channels. As one example, if securitization leads to more competition in

originating securitizable loans, it may depress banks' spreads in originating those types of loans

[Instefjord (2005)], and thus reduce banks' profits.




                                                   4
        In sum, the net effect of securitization on banks' insolvency risk, leverage, and profits is

theoretically ambiguous, and needs to be settled empirically. Our empirical analysis uses two

methods. The first method analyzes the effects of securitization by studying how the risk profile,

leverage, and profitability of securitizers would change if they had to take the assets that they

securitize back onto their balance sheets. This method is prescient since in the recent turmoil some

banks have had to purchase back securitized assets, or hold some loans on balance sheet that they

had originally planned to securitize (www.securitization.net).



        The second method takes into account that securitization is an endogenous decision of the

bank. To control for endogeneity, we use a bank's size as an instrument for its securitization

decision. To justify our approach we perform a set of analysis, including semi-parametric

regression to establish that bank size is a valid instrumental variable for securitization. We exploit

size as an instrument in both univariate and multivariate analysis. In our univariate analysis, we

compare the performance of banks that do securitize with the performance of slightly smaller

banks that do not securitize. In our multivariate analysis, we study the performance among the

same banks Instrumental Variable (IV) regression. Using the univariate and multivariate analysis,

we find that securitization reduces banks’ insolvency risk, as measured by the relatively low

spreads that securitizers pay for uninsured time deposits. Securitization is also positively related to

banks' leverage ratios. For the relationship between securitization and profitability, our results

were more mixed, using the univariate analysis we found that securitization increases bank

profitability, but we failed to detective a statistically significant positive relationship using the

multivariate analysis. Our results hold over a set of robustness checks including the propensity




                                                    5
score technique which creates a matching sample of non-securitizers whose properties are

compared with those of securitizers.



       The remainder of the paper contains five sections. Section 2 provides a brief literature

review; Section 3 describes our data; section 4 analyzes the impacts of securitization on banks if

banks have to retain their securitized assets on balance sheet. Section 5 contains our instrumental

variable analysis. Section 6 describes our robustness checks. A final section 7 concludes.



2. Literature Review

       Securitization is one of many forms of credit risk transfer, the most prominent being credit

derivatives. There are three main strands of the related literature. The first strand examines how

securitization issues, such as the tranches of a CDO should be priced. The second and third

strands, which are more relevant for our analysis, pertain to how securitization affects banks'

funding ability, and how it affects banks' risk management and risk profile.



       The theoretical and empirical parts of the funding strand have slightly different focuses.

The theoretical part of the funding strand studies how securitization can be designed to overcome

various information asymmetries that are associated with transferring credit risk [Gorton and

Pennacchi (1995), Gorton and Souleles (2005), Duffie and DeMarzo (1999), DeMarzo (2005)], as

well as how securitization can affect for better or for worse the quality of markets for sharing

credit risk [Duffee and Zhou (2001), Parlour and Plantin (2007), Morrison (2005)].




                                                  6
       The empirical literature on funding mostly studies whether securitization helps to facilitate

credit extension by relaxing constraints on funding securitizable loans. Working with Bankscope

data for 65 banks that issued Collateralized Loan Obligations (CLOs) between 1995 and 2004,

Hirtle (2007), Cebenoyan and Strahan (2004), and Goderis (2007) et al. find that securitization

helped to increase the supply of credit. Loutskina and Strahan (2006) find that it does so by

reducing the influence of the bank's own financial condition on whether a loan is originated. For

our purposes, relaxation of constraints is one of the channels through which securitization should

increase bank profits, but the literature on funding constraints does not establish that securitization

increases profits overall.



       The risk profile strand of the literature studies how securitization alters banks’ overall risk

profile. Jiangli, Pritsker and Raupach (2007) from a theoretical perspective show that optimally

chosen securitization can sometimes reduce bank's risk of insolvency, and sometimes increase it,

and that whether risk increases or decreases depends on the types of securitization opportunities

that are available to the bank, as well as to other factors. Krahnen and Wilde (2006) theoretically

show that securitizations can be structured to substantially increase bank's systemic risk; although

they do not show that an optimizing bank would use securitization in this way. Franke and

Krahnen (2005) perform event study analysis of 73 European CDO securitizations from 1999-

2002, and find these securitizations are associated with an increase in the average equity beta of

securitizing banks. Using data on a small sample of Canadian banks, Dionne and Harchaoui (2003)

measure banks’ insolvency risk based on the banks' ratios of regulatory capital to risk-weighted

assets. They find that securitization is associated with a decrease in banks' capital ratios. This is

interpreted as an increase in insolvency risk. Dionne and Harchoui’s methodology has two




                                                   7
shortcomings. First, regulatory capital ratios are a poor measure of insolvency risk because the

risk-weighted assets in the denominator of such ratios do not properly account for the correlations

among the returns of the assets in each bank’s portfolio. Second, if securitization truly transfers

risks off of a bank's balance sheet, then lower capital ratios could still be consistent with lower

insolvency risk of the bank.



A second part of the risk profile strand addresses why banks use securitization, and which banks

would benefit from doing so. A consistent finding in the empirical literature is that securitization's

most important role is for funding and liquidity management. 3           Many papers also find that

securitization is not related to regulatorory capital arbitrage. Consistent with the liquidity and

funding motive for securitization, and contrary to the regulatory capital arbitrage and risk-sharing

motivations, using data on Spanish banks, Martin-Oliver and Saurina (2007) find that banks tend

to use forms of securitization that do not shift credit-risk and do not provide regulatory capital

relief. Using data on US Banks and Finance Companies, Minton et al (2004) also find that

securitization is not for the purposes of regulatory capital arbitrage since unregulated finance

companies and investment banks are even more likely to securitize than regulated banks. 4



        The results in the literature on whether high or low risk banks tend to securitize is more

mixed. Minton et. al. find that when risk is measured based on capital ratios (book equity / book

assets), banks with low risk are more likely to securitize. Using different risk measures (credit

provisions / net interest income) and data on European banks, Bannier and Hansel (2007) find that

3
 Also see Thomas and Wang (2004) for US Bank Holding Companies, Vickery (2007) for US banks, savings banks,
and finance companies, Martin-Oliver and Saurina (2007) for Spanish Banks, and Bannier and Hansel (2007) for a
broad sample of European banks.




                                                       8
high risk banks are more likely to securitize. 5 A complication in determining how bank's risk

profiles affect securitization is that the risk profile and securitization are both endogenous. It is

therefore difficult to disentangle how one affects the other.



         Our paper makes three contributions to the empirical literature. First, we employ a new

measure of insolvency risk, a bank's time deposit premium which is the interest rate spread

between its uninsured and insured time deposits. The larger is the likelihood of bank insolvency,

the larger is the spread that uninsured depositors will require to deposit money in the bank.

Because the time deposit premium is a forward looking market determined measure of ex-ante

risk, we believe it improves on risk measures such as capital ratios that are based on accounting

data. We also believe that interest-rate spreads are a better measure of insolvency risk than the

beta measures that have been used in event studies because beta measures covariance, which is a

second moment measure or risk, whereas insolvency risk is more related to higher order moments

and tail events. 6



         Our second and third contributions are more methodological. We believe we are the first

paper that has attempted to analyze how banks are affected if they have to place securitized assets

back on their balance sheet. To the best of our knowledge, we are also the first paper to use size

differences among relatively large banks as an instrument to control for the endogeneity of



4
  Calomiris and Mason (2003) also find that securitization does not appear to be related to banks attempts to exploit
underpriced deposit insurance.
5
  An interesting work on the determinants of securitization, Karaoglu (2005) uses US BHC mortgage securitization
data to examine whether securitization and loan sales decisions are made in order to manipulate accounting figures
through the timing of profit realizations.
6
  As an extreme example of the difference between beta and insolvency risk, if a bank could purchase full insurance
against its own insolvency, its value will still fluctuate with the market, and it will thus have a non-zero beta even if its
insolvency risk is zero.


                                                             9
securitization. This approach will help us to produce more concise estimates of how securitization

affects bank insolvency, capital structure, and overall profitability.



3. Empirical Analysis and Data



        We use two methods to empirically analyze the effects of securitization. The first method

asks the counterfactual question: how would the financial well-being of securitizers be affected if

they had to take the assets that they securitized back on to their balance sheets? This is a

particularly topical approach for analyzing securitization given the turmoil in credit markets, and

in particular for how it has affected the balance sheets of securitizers. Our second method is a more

conventional IV approach which corrects the endogeneity of securitization decision. The IV

approach is built on the observation that bank size is strongly related to whether banks’ engage in

securitization, but is unrelated to banks’ performance once banks have grown beyond a moderate

size 7. Based on this observation we use differences in size among relatively large banks that do

and do not securitize as an instrument for bank performance. Before describing these methods in

more detail, we turn to a description of our data.



        We use FR Y9-C US bank holding company (BHC) data from the second quarter of 2001

to the second quarter of 2007 to analyze the impact of asset securitization on banks. We study

BHCs at a consolidated level because securitization within a BHC group may not be subject to the

same informational and agency problems as securitizations outside the BHC. Our data start from




7
 For example, Wheelock and Wilson (2001) estimated that banks experience increasing returns to scale up to $ 500
million of assets, and essentially constant returns to scale thereafter.


                                                        10
2001, when the Y9-C first began reporting securitization by asset type. 8 The reported asset

categories (from schedule HC-S) are 1-4 Family Residential Mortgage Loans (Mortgage), Home

Equity Lines (HEL), Commercial and Industrial loans (C&I), Credit Card, Auto, and Other

Consumer Loans (Other). Securitizations are recorded by their outstanding principal balance of

assets sold and securitized with servicing retained or with recourse or other seller-provided credit

enhancements. It should be noted that the data distinguish between outright loan sales, and sales

for the purposes of securitization. We only focus on the latter 9,10.



         Although securitization activities have grown on global basis, based on FR Y9-C data, the

number of U.S. bank holding companies (BHCs) that engage in these activities is small. For

example, Table 1 shows that less than 7% of BHCs securitize 1-4 Family Residential Mortgage

loans (Mortgage), which are the most commonly securitized asset class. The number of banks that

securitize other type of assets are even smaller. Although the number of BHCs that are securitizers

is small, the importance of these activities is better measured by the size of the banks that engage

in these activities. By this measure securitizers are a significant part of the banking sector. The

Table 2 columns “% Assets” show, for example, BHCs that are Mortgage securitizers account for

67% of all U.S BHC assets. The share of other type asset securitizers also account for more than

40% of all BHC. The columns “% Loans” show that the volume of securitized Mortgage and

8
  We merger adjust and delete observations for which risk weighted capitals, leverage ratio, loan growth rate, return on
assets are more than 100%, or loan to deposit ratio is more than 10.
9
  For example, securitizing mortgage differs from sale of mortgage in that securitized mortgages are sold into a
securitization, while sale of mortgage is sale, but not into a securitization. In the first case of selling into a
securitization the seller may retain the servicing rights for the mortgage. In both cases there may be some recourse or
credit enhancement used to make the sale. Traditional accounting would not allow a seller to record the sale if there
was any chance the seller would have to take the asset back. Now, banks are allowed to get sale treatment even though
there may be some credit enhancement. The goal of tracking securitization and asset sales is to see if any sales leave
the bank exposed to having to take the asset back (reversing the sale). The biggest difference here is that the first one
is for a securitization while the second one is just a sale of a mortgage asset.




                                                           11
Credit Card loans are similar and sometimes exceed the amount retained on balance sheet, For

other asset classes, such as HEL, C&I and Other, the amounts securitized are much smaller than

the amounts held on balance sheet.



         Before turning to the main analysis, we compare BHCs that securitize with those that do

not securitize along a number of dimensions including their size, their tendency to specialize in

originating particular classes of loans, credit risk, profitability and leverage (Table 3).



          Most of the securitizers do not securitize in all quarters. This can happen when a

securitizer is accumulating loans to package for securitization, or if there are periods of time when

securitization is less profitable. To compare the securitizers and non-securitizers, we calculate the

time-series averages of the variables for each BHC and then compare the averages in cross-

sectional tests.



          A BHC is assigned as a mortgage securitizer if we observe its mortgage securitization

activities in any quarter. BHCs assigned as other types of securitizers are defined in a similar way.

Those that ever securitize any type of assets are Ever securitizers, while those who never securitize

any type of assets are Never-securitizers. There is a total of 2231 bank observations with 147

Mortgage, 23 HEL, 30 C&I, 35 Credit card and 46 Other Securitizers. 185 are Ever-securitizers

and 2046 banks are Never-securitizers.




10
  There is also one column that reports the "all other loans, all leases, and all other assets" which is omitted in our
analysis because we cannot separate loans from leases. We combine Auto with Other because we compare the sold
assets with on balance sheet assets which do not separate Auto from Other loans.


                                                            12
        The most robust difference between securitizers and non-securitizers of any asset types is

that securitizers are significantly larger than non-securitizers 11 where size is measured as the

natural logarithm of on balance sheet assets. The large size of securitizers may reflect economies

of scale in underwriting and securitization, or it may reflect diseconomies of scale in funding

through deposits. That is, the investable opportunities of large banks may outstrip their base of

inexpensive deposits, causing them to turn to securitization as an alternative funding source.

Table 3 also shows that loans of securitizers as percentage of total asset value is slightly lower than

for non-securitizers. This is probably because larger BHCs are more likely to engage in business

lines other than loan origination.



        In addition to size, securitizers also hold a much higher average percentage of the type of

loan they securitize on their balance sheets than do non-securitizers. The percentage mean

difference between securitizers' and non-securitizers' holdings are significant and vary from 12.6

% (Mortgage) to 183.6% (Credit Card). Consistent with financial intermediation theory, the data

suggest that securitization allows banks to specialize in loan origination while allowing the banks

to share the risks of their loan originations with others.



         As noted earlier, securitization provides the bank with added flexibility to shape its risk

profile. To contrast the risk profiles of securitizers and non-securitizers, we use three measures of

risk. The first is the time deposit premium, which is the difference between the interest rates on

small (< $100,000 ) insured time deposits and large (> $100,000) uninsured time deposits. We


11
  Minton et al (2004) using US data also find that large commercial banks are more likely to securitize. Using CLO
data from 17 European countries, Bannier and Hansel (2006) report that large banks are more likely to securitized
CLOs. Martin-Oliver and Saurina (2007) use Spanish bank data also find size is positively related to asset
securitization.


                                                         13
view the time deposit premium as a measure of the banks insolvency, or tail risk. As support for

this interpretation, Gilbert, Meyer and Vaughan (2002) list a total of 12 papers with evidence

pointing to risk pricing by large time deposit holders (their Table 2).12               Although we favor the use

of time deposits as a measure of tail risk, there are some caveats associated with this measure. For

example, the above US$100,000 time deposits may be insured if they are held in joint accounts,

but US BHC data do not provide information on whether accounts are joint. Additionally, we do

not have any information on the maturity and liquidity of the time deposits. Our second and third

risk measures are the loan loss provision rate and banks charge-off ratio. We view both of these as

measures of the expected loss rate of the bank’s portfolio.




         A raw comparison of the risk measures suggests that securitizers and non-securitizers have

different risk profiles. Securitizers have a lower time deposit premium than the non-securitizers,

suggesting they have lower tail risk. The difference can be as high as 173% (HEL). However, the

differences in the time-deposit premium between securitizers and non-securitizers for Mortgage,

Credit Card, Other and Ever are statistically insignificant. We suspect this is because of high

standard deviations of the time-deposit premiums among non-securitizers and the raw figures don't

control for other sources of variation of the time deposit premia.




We recognize that there are other possible measures of tail risk. Perhaps the cleanest is the credit spread on the bank’s
subordinated debt. The main problem with this measure is data availability since many non-securitizers do not issue
subordinated debt. Other possible measures are based on capital ratios, but these ratios are backward looking and
based on accounting data. Additionally, their quality erodes as banks approach insolvency since banks are hesitant to
recognize losses on loans when they approach insolvency [(Dahl, O'Keefe, and Hanweck (1998), and Gunther and
Moore (2000)]. Nevertheless, when a bank's condition weakens, its cost of deposit funding often rises, suggesting that
measures of tail risk based on deposits may be a better measure of solvency risk.



                                                           14
         For the measures of expected loss, our results are stronger. We find that both the provision

and charge off ratios are statistically and economically significantly higher for securitizers than

non-securitizers except the provision ratio of Mortgages and Ever. These results could, for

example, imply that securitization enables the securitizers to extend loans that have higher

expected losses. Alternatively, high observed provisions and charge-off may also reflect a size

effect in that large banks are more likely to take positions that have higher expected losses than

small banks 13.



         To study the profitability of securitizers and non-securitizers, we focus on return on equity

(ROE). (We also examined return on assets, ROA and obtained the qualitatively same results). The

ROE for securitizers ranges from around 10.4% (C&I) to 13.01% (Other) which is higher than that

of the non-securitizers at around 10%. However, the large p-values indicate that the difference

between securitizers and non-securitizers are statistically insignificant for HEL, C&I and Credit

Card.



         The mean leverage ratio of securitizers and non-securitizers are all above 89%. Leverage

ratio of Mortgage, HEL and C&I securitizers are higher than non-securitizers but statistically

insignificant, while a reverse pattern exists for Credit Card and Other. The Ever securitizers also

have a low leverage ratio than the Never-securitizers, and the difference is significant. Our analysis

shows that leverage and securitization decisions are inter-related. If there is a third factor (such as,

perhaps, bank size) that is related to leverage and to the reason why some banks do not securitize,

it may also help to explain our leverage results. In our instrumental variable analysis, we revisit the


13
  Bannier and Hansel (2006) use credit risk provision over net interest income as a measure of credit risk also find that
high credit risk banks are more likely to securitize CLOs.


                                                          15
importance of size when we compare securitizers and non-securitizers who are more similar in size

than is the case with the raw comparisons presented here.



4. What if Banks have to Place Securitized Assets Back on Securitizers Balance Sheets-What

if Banks cannot Securitize Assets?



         The univariate comparisons of securitizers and non-securitizers in Table 3 does not control

for other variables that could contribute to their differences. In this section, we rely on a

methodology which analyses the impact of securitization by comparing the averages of the

observed securitizers’ insolvency risk, profitability and leverage ratio with hypothetical values that

are calculated by assuming the securitized loans were put back on balance sheets. This method can

help assess the consequences to banks of not being able to tap the securitization market, and of

having to purchase large amounts of loans from conduits and move them on their balance sheets.



         As one illustration of our approach, we assume that there is a relationship between a bank’s

insolvency risk and the composition of its balance sheet, including the share of each type of loans

held on the balance sheet (loan shares) 14. We use the non-securitizers data to estimate this

relationship. We then use this estimated relationship to compute the bank's insolvency risk if its

securitized loans were placed back on the balance sheet. This method allows us to predict the

bank's insolvency risk if the securitizer did not securitize its loans. We then compare the predicted

values with the average of the observed securitizers' insolvency risk. The average difference




14
  The loan share variables do not sum to 1 since we only include the five types of loans with available securitization
information.


                                                          16
between the observed and predicted values is one measure of the expected quantitative impact of

securitization. We apply the same exercise to bank's profitability, and leverage ratio.



        To estimate the relationship between balance sheet composition and bank performance,

following the literature, we assume for each bank holding company i, our measures of bank

performance, Y, are a linear function of loan shares; other controls, Z; and bank size, as measured

by Ln (Assets) 15:




                                                                       + β Z + β Ln ( Assets ) + ε .
                 Mort        HEL       C&I       CreditCard      Other
(1) Y = β + β          +β         +β         +β             +β
          0    1 Loans    2 Loans    3 Loans    4 Loans        5 Loans    6     7




In equation (1), Y is insolvency risk, profitability, or a bank's leverage ratio; and the other control

variables are the share of loans to assets, and costs of funding as measured by the average deposit

rate, and interest rates on small (insured) and large (uninsured) time deposits. The dependent and

explanatory variables for each BHC in our sample are again the time-series averages. We thus

report the between estimator, which exploits the full panel data set but estimates the regression

using the time-series averages.



        This model specification is based on the empirical models of Wheelock and Wilson (2000),

and Estrella, Park and Peristiani (2000) for bank insolvency risk 16, Cebenoyan and Strahan (2004)




15
  For simplicity, unless they are needed we have chosen to suppress subscripts i for individual BHCs.
16
  These papers also include the equity to asset ratio, an important variable in explaining bank failure, ROA and
problem loans. We do not include equity to asset ratio and ROA since we do not know how these two ratios change if
the securitized loans were put back on balance sheet.


                                                        17
for ROE 17, and Flannery and Rangan (2004), and Gropp and Heider (2007) for the leverage

ratio 18. We estimated five model specifications of equation (1) for the non-securitizers using OLS.

The results are reported in Tables 4A through 4C. Model (1) includes the shares of Mortgage,

HEL, C&I, Credit Card and Other to loans, and uses size as a control. Model (2) adds loans to

assets to Model (1); Model (3) adds the deposit rate to Model (2); and Models (4) and (5) replace

the deposit rate with insured and uninsured deposit rates respectively.



         It is important to emphasize that each coefficient of the loan share variable in the

regressions for the time deposit premium, ROE and leverage ratio cannot be interpreted in

isolation since if the share of one type of loan is increased, the others share must necessarily

decrease.



         Across the five models, Table 4 A shows that the share of Mortgage, HEL and Other

loans, the deposit rate, and the uninsured deposit rate are positively related to bank's time deposit

premium, while size, and the share of C&I, Credit Card loans, the insured deposit rate, and the

share of loans over assets are negatively related to the time deposit premium. To control for the

effect of problem loans on insolvency risk, we also added the share loans that are past due 30-89

days (3m-pastdue/Loans), 90 days plus (3m+pastdue/Loans), and we controlled for nonaccrual and

charge off loan ratios (NoAcc+Chargeoff/Loans). The 3m+pastdue/Loans and

NoAcc+Chargeoff/Loans rates are positively related to the time-deposit premium, while 3m-

pastdue/Loans are negatively related. The F-statistics reject that the three coefficients are jointly

17
  The additional variables are capital asset ratio and dummies variables which indicate whether or not a bank is a loan
seller, buyer or both, whether or not a bank belong to a multi-BHC or multi-state BHC.




                                                          18
insignificant across four models except for Model (5). The adjusted R-squares vary from 21.2% to

91.2%. Adding the insured and uninsured deposit rates significantly increases the model's

explanatory power.



         ROE is negatively related to the share of Mortgage, HEL, C&I and Other loans, deposit

rate, and insured deposit rate, while positively related to size, share of Credit Card loans, share of

loans to assets, and the uninsured deposit rate. The adjusted R-squares vary from 6.8% to 9.8%.



         The leverage ratio are negatively related to the share of Mortgage, C&I, Credit Card, Other

loans, deposit rate and uninsured deposit rate, and positively rated to size, share of HEL, C&I

loans, loans to assets, insured deposit rate. The adjusted R-squares vary from 3.7% to 11.3%.



         Next, we apply the estimated relationship between time deposit premium, ROE and

leverage ratio to securitizers assuming that the securitizers put the securitized loans back on

balance sheet. Adding back these securitized loans will change total assets, total loans, the share of

all type of assets and the share of problem loans. We recalculated the time series average of these

new variables labeled with “A-” for augmentation. Table 5 reports the mean, and standard

deviation of securitizers’ time series average. The columns labeled with “%Δ” are the percentage

change in observed values when securitized assets are added back on balance sheet.



         Mathematically, adding back any type of securitized asset will increase total assets, loans,

and the share of loan to assets. In particular, adding back securitized mortgages will increase the

18
  Flannery and Rangan (2004) , Gropp and Heider (2007) both regress leverage ratio on market to book asset ratio,
profitability, risk, and size. Most of our no-securitizers are not publicly traded BHCs. Thus, we cannot construct the



                                                           19
share of mortgage loans and decrease shares of other type of loans. However, whether share of

problem loans increases when securitized assets are retained on the balance sheet is less clear cut

because it depends on whether higher quality assets are securitized or retained on balance sheet 19.

When Mortgage, HEL and C&I are moved on balance sheet, average loan quality in each category

increases by some measure but decreases by other measures. For example, Mortgage securitizers

have an increased 3m-pastdue/Loans, 3m+pastdue/Loans ratio but a decreased

NoAcc+Chargeoff/Loans ratio. However, for C&I and Other, all these three measures of problem

loan consistently show that the average quality of securitized C&I and Other loans are lower than

the quality of the loans retained on balance sheets 20.



         Our method of calculating the average difference between the predicted and observed value

of insolvency risk, ROE and leverage ratio involves three steps. First, we create a quarterly

predicted time-series sample for each bank by applying its quarterly augmented variables to the

predicted relationships from equation (1). This produces estimates of quarterly predicted

insolvency risk, ROE and leverage ratios for each bank when its assets are moved back on balance

sheet. In the second and third steps we average these predicted values over time, and then across

banks. The resulting predicted averages are then compared to securitizers' performance when

averaged across time and across securitizing banks.



market to book ratio. However, our other variables can approximate the profitability and risk.
19
   The share of added back problem loans is calculated, for example, for mortgage as the ratio of the sum of on balance
sheet problem loans plus the securitized problem mortgages divided by the sum of on balance sheet loans and
securitized mortgages.
20
   However, we realize that to compare the credit quality of the securitized versus the on balance sheets assets poses a
number of data challenges. First, with aggregate level data, it is impossible to separate the truly kept on book loans
from the temporarily kept on book loans which are intend to be put into a securitization pool. Second, the vintage of
the securitized loans can be quite different from the on balance sheet loans. For example, unseasoned credit cards have
a larger credit risk than the seasoned ones. Without adjusting seasonality, the comparison can be problematic. Third,
there maybe significant heterogeneity of both securitized and on balance sheet assets. For now, we just report what we
observe in the data and leave these difficulties for the future research.


                                                          20
          Table 6 presents our results on how securitizers are affected when securitized assets are

moved back on balance sheet. In the table, sample averages of securitizers performance as

measured by insolvency risk, ROE and leverage ratios are presented in bold face. Estimates of

how performance would change if assets are moved back on balance sheet are presented for five

models. For example, the first cell in the first column indicates that using model (1), the predicted

time deposit premium if securitized mortgages are moved back on balance sheet is 2.75%, while

securitizers observed average time deposit premium is 2.26%. This suggests that putting the

securitized mortgage back on balance would cost the securitizers an additional 49 basis points

when issuing large time deposits.



         Overall, our results on insolvency risk, as measured by the time deposit premium, suggest

that securitization of Mortgages, Home Equity Loans (HEL), and C&I loans lower banks time

deposit premium about 50 bps in most model specifications. These numbers are within the range

reported by Elghanayan (2006) who documented that the rated banks can save 20-80 bps in

funding costs by issuing securitization rather than debt, while the saving for unrated banks can be

well above 100 bps. The predicted values for Credit Card and Other indicate that securitization of

those types of assets may increase banks’ insolvency risk. Our findings that different types of

securitization have different effects on tail risk is consistent with Chen et. al. (2007), who note that

risk retention by the securitization sponsor varies by type of securitization, and is relatively low in

the case of Mortgages, while relatively high for revolving loans such as credit cards. 21


21
  The net effect on tail risk also depends on how securitization affects the bank's capital structure. Lang et. al. (2005)
note that credit card banks tend to hold more economic capital against their off-balance sheet positions. They do not
measure the net effect of credit card securitization on tail risk. For more on credit card securitization and implicit
recourse see also Gorton and Souleles (2005).


                                                           21
        Our result on profitability based on ROE suggest that securitization of Mortgages and the

Other category of assets increases banks' profitability by a range of 34 to 150 bps depending on the

model. By contrast, we find that HEL, C&I and Credit Card securitizations lower banks’ profits.

These results should be interpreted with caution because they have multiple interpretations. For

example, our findings on credit card securitization are consistent with the possibility that

securitization reduces profitability by increasing competition. If this is the case, it does not imply

that an individual credit card bank would be more profitable if it did not use securitization. 22



        In the case of bank's on balance sheet leverage ratios, we found that securitization increases

banks’ on-balance sheet leverage ratio from 43 to 153 bps except for securitizations of the Other

category of assets where we found that securitization is associated with reduced leverage.



        We also experimented by putting all of the securitized assets back on the balance sheet at

the same time. The results are reported in the last column of Table 6 (labeled "All"). Our results

are very mixed. Three out of five models indicate that securitization reduces time deposit

premiums, three out of five models shows that securitization increases ROE and four out of five

models shows that securitization leads to increased leverage ratios. One reason for the mixed

results on tail risk is that securitization of different asset types has different implications for tail

risk. Therefore, the sign of the change in tail risk when moving several asset types back on

balance sheet together is ambiguous. We suspect this may also help to explain why some of our

other results on moving all assets back on balance sheet are mixed. Another reason for the mixed


22
  Similarly, if mortgage securitization has high fixed costs, then banks with low amounts of mortgage lending activity
that do not securitize may be less profitable if they did securitize.


                                                         22
results is that our estimates of how changes in balance sheet composition affects bank performance

(Table 4) should be understood as local linear approximations that will be less accurate when used

to study the effects of large changes in balance sheet composition. The last three columns in Table

5 shows that the impact of putting back all securitized assets creates relatively large percentage

changes in some balance sheet components, suggesting that our approximations may perform less

well in these cases.



5. Instrumental Variables Estimation



       Our second approach for studying the impact of securitization on banks by using regression

analysis with instrumental variables (IV). Our IV approach is motivated by the empirical

observation that securitizers tend to be large banks, presumably because those are the banks that

can overcome the fixed costs of setting up a securitization program. A few studies have

documented the factors that determine bank securitizations [Karaoglu (2005), Bannier and Hansel

(2006), and Martin-Oliver and Saurina (2007)]. All these studies find that large banks are more

likely to securitize assets. The following section formally examines whether or not size is a valid

instrumental variable for securitization. Currently, our analysis forces on mortgage securitizers and

non-securitizers only. For other types of securitization, our sample of securitizers is too low for

statistical inference. In addition, when banks securitize multiple types of assets, as many do, our

IV methodology cannot identify which type of securitization generates our results. Therefore, for

the IV analysis, we chose to focus on the form of securitization which we believe is the most

important and in the most widespread use, which is mortgage securitization.




                                                 23
5.1 Size as a valid instrument



       To establish the bank size is a valid instrument we need to show that size is related to

bank's decision to securitize, but is not related to our performance variables. To first study the size

and the securitization decision we sorted the BHCs in the sample into seven size-buckets based on

the quantiles of Ln(Assets), with bucket seven containing the 22 largest bank holding companies,

and each succeeding size-bucket containing increasingly smaller BHCs. The groupings are

provided in Table 7. The most important feature of the table is an upward jump in the fraction of

securitizers as the size of BHCs increases. For example, among the smallest BHCs, only 1% are

securitizers. This fraction grows to 10% in bucket 5, and then jumps to 41% in bucket 6, and 100%

in bucket 7. Based on this data, we treat the banks in buckets 6 and 7 as likely to securitize based

on size, while banks in smaller buckets are less likely to securitize.



       Although securitization is related to BHC size, as noted earlier, we claim that size is only

weakly related to performance, and principally among smaller banks. To verify this claim , for

banks that do not use securitization we estimated both linear and semi-parametric variants of the

regressions in equation (1) in which all variables other than bank size enter linearly, while bank

size is allowed to enter either linearly or semi-parametrically via the unspecified function

G[Ln(Assets)]:

(2)

                               C&I
                                                               + β Z + G[ Ln( Assets )] + ε
         Mort        HEL                 CreditCard      Other
Y =β +β        +β         +β         +β             +β
    0  1 Loans    2 Loans    3 Loans    4 Loans        5 Loans    6


This equation can be more compactly written as:


                                                  24
                        Yi = X i β + G[Ln(Assetsi )] + ε i ,                 i = 1, .... N



Table 8 reports the results from regressing measures of bank performance on linearly specified

bank size and other controls, following the specifications from equation (2), but using a narrowing

window of size buckets. 23 To save space, after the regressions for the first window of size

buckets, we only report the coefficients on Ln(Assets). The analysis shows that once the smallest

1000 BHCs are omitted from the analysis (and these are all very small banks), size fails to explain

differences in performance among banks that do not securitize.




       To verify that our linear regression results are not due to the choice of linear functional

form, we next approximated the function G[.] semi-parametrically using a cubic spline. Under this

specification, the parameter vector β and the function G[.] are chosen to minimize the objective

function:

            N                                            Ln ( Assets )

            ∑ wi {Yi − X i β − G[ Ln( Assetsi )]}2 + λ        ∫ {G ''[ Ln( Assets)]}
                                                                                       2
                                                                                           d [ Ln( Assets )]
            i =1                                         Ln ( Assets )



where λ > 0, the penalty for roughness of G[.] is chosen using least squares cross-validation, the

weights wi are chosen using a GLS to procedure to account for the heteroskedasticity of ε i

conditional on bank size, and where Ln( Assets ) and Ln( Assets ) are the upper and lower bounds

of the bank-size variable in-sample. The cubic spline estimate of G[.] is asymptotically equivalent

to a kernel regression of Y − X β on Ln(Assets) with a variable bandwidth [Silverman (1984)].

Using this principle, we report asymptotic 95% confidence intervals for G[.] following the




                                                         25
approach that is outlined in Silverman (1985). In the analysis, for each of three Y variables, we

estimated four variants of equation (2): one with a constant and Ln(Assets), a second with a

constant, balance sheet composition variables, and Ln(Assets), a third with a constant, portfolio

performance variables and Ln(Assets), and a fourth with a constant, portfolio composition

variables, portfolio performance variables, and Ln(Assets). 24 In the interest of brevity, we only

report results for the fourth specification.



        Our principal results from the semi-parametric estimates of the effect of bank size on

performance verify our results with parametric analysis. More specifically, beyond the smallest

banks, for non-securitizers, increases in size alone do not appreciably alter bank performance.

This is shown by the flatness of the G[.] function for non-securitizing banks with Ln (Assets) >

14.58, which correspond to size buckets 5 and 6 for the non-securitizers.25 In the case of banks'

leverage (Figure 2), and ROE (Figure 3), size has little or no affect on performance along the

entire range of Ln(Assets). In the case of insolvency risk, a larger bank size is clearly associated

with a smaller time-deposit premium up to a point, but it is clear that for relatively large banks (Ln

(Assets) > 14.58), the relationship of size to performance appears statistically indistinguishable

from a flat line (Figure 1).



        Our parametric and semi-parametric analysis suggests that size is a valid instrument for

identifying the effects of securitization on bank performance, especially when studying banks in


23
   We exclude the top size bucket from this analysis because all banks within this bucket securitize.
24
   The balance sheet composition variables are Mortgages/Loans, HEL/Loans, C&I/Loans, Other Loans/Loans, and
Loans/Assets. The portfolio performance variables are 1 - 3 month past due loans / Loans, 3 months or more past due
loans / Loans, and Non-Accruing and Charged-off Loans / Loans.
25
   Because an intercept is included in the regresssion, the average level of the G[.] function is not identified. We
normalized the G[.] function so that its average level, evaluated at the data observations, is 0. Only changes in the
G[.] function as Ln(Assets) vary can identify whether Ln(Assets) affects bank performance.


                                                         26
groups 5, 6, and 7. We use size in two ways. First, we perform univariate analysis in which we

compare banks that are similar in size (within the same size bucket), but one set of banks

securitizes while the other does not (Table 9, panel A). In panel B of Table 9 we compare

securitizers in one group of size buckets with non-securitizers in other size-buckets. The within

bucket results are mostly insignificant. When comparing across size buckets, the results are more

frequently significant. The ideal comparison is the close size buckets in the neighborhood of sizes

6 and 7 because the fraction of securitizers jumps there, but hopefully other bank characteristics

that may be correlated with size do not. Thus, the ideal comparison should be between buckets 6

and 7 with bucket 5. The ideal comparison shows that securitizers have lower insolvency risk, and

a higher ROE and leverage ratio than non-securitizers. Comparing the securitizers in groups 6 and

7 with the other size groupings produces similar results on the effects of securitization.



5.3 Instrumental variables results



        Our second approach for using size involves multivariate IV regressions in which size is

used as an instrument for securitization in a first stage OLS and then second-stage regressions are

estimated using predicted values for securitization. The results from the first stage regression are

reported in Table 10. The second stage regressions for insolvency risk, ROE and leverage ratio are

presented in Tables 11 A, 11 B and 11 C. In all regressions, we only use data from size groups 5, 6

and 7, or 6 and 7. Our securitization variable is mortsec, which is an indicator for whether the

BHC is a mortgage securitizer. As Table 10 shows, across all model specifications and sample

sizes Ln(Assets) is positive and significantly correlated with mortsec. Tables 11-A, 11-B, and 11-

C present our results for the time deposit premium, ROE, and Leverage. Panel A of each of the




                                                  27
tables presents the two-stage OLS results with mortsec being the instrumented variable based on

size buckets 5, 6 and 7. Panel B repeats this exercise using only buckets 6 and 7. For comparison

with the IV approach, Panels C and D of the Tables report the simple OLS regression results based

on size buckets 5, 6 and 7, and 6 and 7 respectively. To save space, for panels B, C, and D of each

table, we only report the estimates for the instrumented variable, mortsec, sample size, and R-

squared.



        In the case of insolvency risk, the IV and OLS regressions produce significantly negative

coefficients on mortsec, suggesting that mortgage securitization is associated with a reduction in

BHC insolvency risk. Our point estimates suggest that mortgage securitization can reduce the time

deposit premium by 50bps to 380bps [Table 11, panel B, model (5); and Table 11, panel A, model

(1)]. These reductions are economically substantial and statistically significant.



        The IV results on bank profitability fail to detect a statistically significant relationship

between mortgage securitization and bank profitability as measured by ROE. We plan to analyze

these results in more detail in future revisions. The results for mortsec on leverage ratio, reported

in Table 11 C, are positive and statistically different from zero under most model specifications.

Such results indicate that securitizing mortgage allows banks to increase leverage ratio by 1.5% to

4.6%.



6. Robustness Checks




                                                   28
        We performed a series of robustness checks to verify our results sections 4 and 5. In

section 4 equation (1), we dropped the three measures of loan quality, and replaced insured

(uninsured) deposit rates by total interest expense on insured (uninsured) deposits divided by total

insured (uninsured) deposits; our results still hold. In section 5, we performed the same exercise in

our OLS and IV regressions and produce the same results. We also used a propensity score

technique to compare mortgage securitizers and non-securitizes in section 5, and our results still

hold (not reported but available upon request) 26.



        One question is whether our finding that securitization reduces time deposit premia is due

to securitization or due to (other) scale economies in risk-taking that are difficult to detect without

accounting for the endogeneity of risk-taking [Hughes, Mester and Moon (2001)]. Disentangling

these possibilities will be difficult since securitization is itself an important source of scale

economies in risk-taking. Nevertheless, we believe that our analysis extends in controlling for

these scale effects since much of our work relies on comparisons of securitizers and non-

securitizers that are similar in size. By making these comparisons, we automatically control for

other sources of endogenous risk-taking that are related to size, and we have potential to add other

controls such as measures of portfolio diversification in our future analysis. Additionally, it is


26
  The method is to compare the outcome of two samples. One sample receives treatment
(mortsec=1) and the other sample does not receive treatment (mortsec=0). We first run the logit or
probit to get the predicted value of securitization. We create a set of matching sample criteria that
choose untreated (mortsec=0) banks which have a high potential to be a treated banks (mortsec=1).
This sample of untreated banks is our matching banks to the treated banks. This method relies on
the assumption that conditional on other controls in the logit or probit regression, the treatment is
random. It is essentially a conditional comparison of the outcomes of two samples and it corrects
the bias of the selection of the treatment. In our case, we compare insolvency risk, profitability and
leverage ratio of the mortsec=1 banks with that of the mortsec=0 banks by conditional on a set of
other control variables.




                                                     29
important to emphasize that most researchers find that scale economies disappear once bank asset

sizes reach about $10 billion (see the survey by Amel et al (2004)), which is about the size of

banks in group 5 in our sample This suggests that scale economies from non-securitization

activities are essentially exhausted for groups 5, 6, and 7, which are the banks where most of our

analysis is focused.



7. Conclusion



       Using bank holding company data from 2001-2007, we have conducted an empirical

analysis of the effect that securitization has on BHCs. The bulk of our analysis focused on

mortgage securitizations, and analyzed them using three methodologies. The first is an

unconventional but very timely approach that assesses what would happen if mortgage securitizers

had to take their securitized mortgages back on balance sheet. The second approach compares the

average performance of large banks that securitize with banks of comparable size that do not. Our

final approach uses instrumental variable regression in which we use bank size as an instrument

for securitization since beyond very small banks, bank size has little effect on our performance

measures, but has a significant effect on the likelihood that banks securitize. Our results across the

three methodologies present a fairly consistent picture in the case of mortgage securitizers. Using

our second and third approaches we find that mortgage securitization reduces bank insolvency risk

(5 percent significance level based on second and third approaches), increases bank leverage (5

percent significance using the second approach, 10 percent significance with the third approach),

and increase bank's profitability (5 percent significance second approach, not significant third




                                                 30
approach). Our results have the same signs using approach 1, but we have not yet developed a

methodology to assign statistical significance for that approach.



        We also analyzed non-mortgage securitizers using our first approach. Our results in those

cases were less consistent. We believe that this is partially attributable to the small number of

banks that do securitization for the other asset classes. Because of these small sample sizes, we

did not analyze non-mortgage securitizations using our second and third approaches.



        Our overall results suggest a very positive role for mortgage securitization. This raises the

question of how to interpret our results in light of the current turmoil in credit markets in general,

and mortgage markets in particular. Our interpretation is that the high profitability, high leverage,

and low insolvency risk that are associated with securitization in our analysis are reflective of a

positive history of past experience with securitization in banking. Additionally, the relatively low

time deposit premiums of securitizers suggests that the current turmoil in credit markets was not

anticipated by uninsured depositors because it was not reflective of historical experience, but is

instead reflective of recent excesses in mortgage and securitization markets. 27 If our

interpretation is correct, then we predict that because of the positive effects that we estimate for

securitization, we expect that securitization activity will pick-up again once the current problems

in credit markets are cleared up. Only time will tell if our prediction is correct.




27
  For evidence on recent excesses see Ashcraft and Schuermann (2007), Dell Arricia et al (2007), and Mian and Sufi
(2007).


                                                        31
References:

Amel, Dean, Colleen Barnes, Fabio Panetta and Carmelo Salleo (2004) “Consolidation and
Efficiency in the Financial Sector: A Review of the International Evidence” Journal of Banking
and Finance 28, 2493-2519.

Ashcraft, A., and T. Schuermann, (2007), "Understanding the Securitization of Subprime
Mortgage Credit," Working Paper, The Federal Reserve Bank of New York.

Mian, A. R. and A. Sufi, (2007), "The Consequences of Mortgage Credit Expansion: Evidence
from the 2007 Mortgage Default Crisis,". Available at SSRN.

Bannier, Christina E. and Dennis N. Hansel (2006), “Determines of Banks’ Engagement in
Loan Securitization” Goethe-University Frankfurt, Working Paper N0. 171.

Calomiris, Charles and Joseph Mason (2003), “Credit card Securitization and Regulatory
Arbitrage” Federal Reserve Bank of Philadelphia working Paper No 03-7.

Cebenoyan, A.S., and P.E. Strahan, (2004), “Risk Management, Capital structure and Lending at
Banks” Journal of Banking and Finance, 28, 19-43.

Chen, W., Liu, C.C., and S. G. Ryan, 2007, "Characteristics of Securitization that Determine
Issuers' Retention of the Risks of the Securitized Assets," Working Paper.


Dahl, Drew, John O’Keefe and Gerald Hanweck, (1998), “The Influence of Examiners and
Auditors on Loan-Loss Recognition.” FDIC Banking review 11, no. 4:10-25.

Dell' Ariccia, G., Igan, D., and L. Laeven, (2007), "Credit Booms and Lending Standards:
Evidence from the Subrime Mortgage Market," Working Paper, the International Monetary Fund.

DeMarzo, P, and D. Duffie, (1999), "A Liquidity-Based Model of Security Design,"
Econometrica, 67, 65-99.

DeMarzo, P., (2005), "The Pooling and Tranching of Securities: A Model of Informed
Intermediation," Review of Financial Studies, 18, 1-35.

Dionne, Georges and Tarek Harchaoui, (2003), “Banks’ Capital, Securitization and Credit Risk:
An Empirical Evidence for Canada”, HEC Working Paper Number 03-01.

Duffee, G., and C. Zhou, (2001), "Credit Derivatives in Banking: Useful Tools for Managing
Risk?", Journal of Monetary Economics, 48, 25-54.

Elghanayan, Shahram, (2006), “Retail Banks Economic Capital Using a Credit Portfolio Model to
rate Securitizations: Part 3,” Retail Risk Management.




                                                32
Estrella, Arturo, Sangkyun Park and Stavros Peristiani, (2000), “Capital Ratios as Predictors
of Bank Failure,” Federal Reserve Bank of New York Economic Policy Review, July, 33-52.

Flannery and Rangan (2004), Flannery, Mark and Kasturi Rangan, (2004), “What Caused the Bank
Capital Build-up of the 1990s?” FDIC CFR Working Paper, No. 2004-03.

Franke, G., and J. Krahnen, (2005), “Default Risk Sharing Between Banks and Markets:
The Contribution of Collateralized Debt Obligations,” NBER Working Paper 11741.

Gilbert, Alton, Andrew Meyer and Mark Vaughan, (2002), “Can Feedback from the Jumbo-
CD Market Improve Off-Site Surveillance of Community Banks?” Working Paper, Federal
Reserve Bank of St. Louis.

Goderis, Goderis, Ian Marsh, Judit Castello and Wolf Wagner, (2007), “Bank Behavior with
Access to Credit Risk Transfer Markets” Bank of Finland Research Discussion Papers. 4 -2007

Gorton, G. and G. Pennacchi, (1995), "Banks and Loan Sales: Marketing Nonmarketable Assets,"
The Journal of Monetary Economics, 35, 389-411.

Gorton, G. and N.S. Souleles, (2005), "Special Purpose Vehicles and Securitization," Working
Paper, The University of Pennsylvania.

Gropp and Heider (2007) Gropp, Reit and Florian Heider, (2007), “What can Corporate Finance
Say about Banks’ Capital Structure?”. European Central Bank.

Gunther, Jeffery and Robert Moore, (2000), “Financial Statements and Reality: Do troubled
Banks Tell All?” Federal Reserve Bank of Dallas Economic and Financial Review (3rd Quarter),
30-35.

Hirtle, Beverly (2007), “Credit Derivatives and Bank Credit Supply” Federal Reserve Bank of
New York Staff Reports. No-276.

Hughes, Joseph, Loretta Mester and Choon-Geol Moon (2001), “Are Scale Economies in Banking
Elusive or Illusive? Evidence Obtained by Incorporating Capital Structure and Risk-Taking into
Models of Bank Production”, Journal of Banking and Finance 25, 2169-2208.

Instefjord, N., (2005), "Risk and Hedging: Do Credit Derivatives Increase Bank Risk," Journal of
Banking and Finance, 29, 333-345.

Jiangli, W., M. Pritsker, and P.Raupach, (2007), "Banking and Securitization", Working Paper.

Karaoglu, Emre (2005), “Regulatory Capital and Earnings Management in Banks, the Case of
Loan Sales and Securitizations” FDIC CFR Working Paper




                                                33
Lang, W.W., Mester, L.J., and T.A. Vermilyea, 2005, "Potential Competitive Effects on U.S. Bank
Credit Card Lending from the Proposed Bifurcated Application of Basel II," Federal Bank of
Philadelphia Working Paper 05-29.

Leland, H.E., (2007), "Financial Synergies and the Optimal Scope of the Firm: Implications for
Mergers, Spinoffs, and Structured Finance," Journal of Finance 62, 765-807.

Loutskina, Elena and Ohilip Strahan, (2006), “Securitization and the Declining Impact of Bank
Finance on Loan Supply: Evidence from Mortgage Acceptance Rates”

Martin-Oliver, A. and J. Saurina, (2007), “Why do Banks Securitize Assets?” Bank of Spain.

Mian, Atif and Amir Sufi, (2007), “The Consequences of Mortgage Credit Expansion:
Evidence from the 2007 Mortgage Default Crisis”


Minton, Bernadette A., Anthony B. Sanders, and Philip E. Strahan (2004), “Securitization by
Banks and Finance Companies: Efficient Contracting or Regulatory Arbitrage?," Working Paper,
Ohio State University.

Morrison, A.D., 2005, "Credit Derivatives, Disintermediation, and Investment Decisions," Journal
of Business, 78, 621-647.

Parlour, C.A., and G. Plantin, (2007), "Loan Sales and Relationship Banking," Working Paper.

Silverman, B.W., (1984), "Spline Smoothing: The Equivalent Variable Kernel Method," Annals of
Statistics 12, 898-916.

Silverman, B.W., (1985), "Some Aspects of the Spline Smoothing Approach to Non-Parametric
Regression Curve Fitting," Journal of the Royal Statistical Society, Series B (Methodological) 47,
1-52.


Thomas, Hugh and Zhiqiang Wang, (2004), “Banks Securitization and Risk Management”.



Vickery, J., (2007), "How do Financial Frictions Shape the Product Market? Evidence from
       Mortgage Originations," Working Paper, The Federal Reserve Bank of New York.



Wheelock D.C., and P.W. Wilson, (2000), “Why do Banks Disappear? The Determinants of U.S.
Bank Failure and Acquisitions,” The Review of Economics and Statistics 82, 127-138.


Wheelock, D.C., and P.W. Wilson, 2001, “New Evidence on Returns to Scale and Product


                                                34
Mix Among U.S. Commercial Banks,” Journal of Monetary Economics, 47, 653-74.

Securities Industry for Financial Markets Association (SIFMA), (2008), Global Market Statistics,
http://www.sifma.org/research/statistics/global-sector-statistics.shtml.




                                               35
                                                Table 1

                    Number and Percent of BHCs that Securitize Asset by Type

       N is the total number of asset securitizers in a given quarter. The column “% of BHCs” is
      the percentage of bank holding companies that sold or securitized the reported asset class in
          a given quarter. The asset classes are mortgages, Home equity lines of credit (HEL),
                Commercial and Industrial Loans (C & I), Credit Card loans, and Other.

                Mortgage              HEL                 C&I           Credit Card          Other
                       % of              % of               % of                % of             % of
Quarter      N        BHCs       N      BHCs         N      BHCs      N        BHCs     N       BHCs
Q2:2001      98        7.0%      9       0.6%        15     1.1%      12       0.9%     26       1.8%
Q3:2001      62        4.6%      9       0.7%        15     1.1%      12       0.9%     30       2.2%
Q4:2001      52        4.5%      9       0.8%        15     1.3%      10       0.9%     27       2.3%
Q1:2002      58        3.6%      8       0.5%        14     0.9%      12       0.7%     27       1.7%
Q2:2002      69        4.1%      11      0.7%        18     1.1%      10       0.6%     26       1.6%
Q3:2002      68        4.1%      11      0.7%        19     1.1%      10       0.6%     26       1.6%
Q4:2002      66        3.9%      11      0.6%        18     1.1%       8       0.5%     31       1.8%
Q1:2003      70        3.3%      10      0.5%        15     0.7%      10       0.5%     27       1.3%
Q2:2003      75        3.8%      10      0.5%        13     0.7%      10       0.5%     26       1.3%
Q3:2003      77        3.5%      12      0.5%        14     0.6%      10       0.5%     25       1.1%
Q4:2003      72        3.1%      12      0.5%        11     0.5%       9       0.4%     24       1.0%
Q1:2004      63        2.6%      14      0.6%         9     0.4%      10       0.4%     23       0.9%
Q2:2004      62        2.8%      13      0.6%         7     0.3%      10       0.4%     22       1.0%
Q3:2004      65        2.7%      13      0.5%         7     0.3%      10       0.4%     23       1.0%
Q4:2004      85        3.5%      22      0.9%        27     1.1%      21       0.9%     33       1.4%
Q1:2005      68        2.5%      15      0.6%         9     0.3%      12       0.4%     25       0.9%
Q2:2005      72        3.0%      14      0.6%         9     0.4%      12       0.5%     26       1.1%
Q3:2005      71        3.3%      15      0.7%         9     0.4%      12       0.6%     25       1.2%
Q4:2005      74        3.5%      15      0.7%         8     0.4%      14       0.7%     26       1.2%
Q1:2006      47        5.1%      12      1.3%         8     0.9%      17       1.8%     19       2.1%
Q2:2006      46        5.0%      12      1.3%         9     1.0%      15       1.6%     19       2.1%
Q3:2006      45        5.0%      11      1.2%         9     1.0%      15       1.7%     19       2.1%
Q4:2006      43        4.8%      11      1.2%         9     1.0%      15       1.7%     19       2.1%
Q1:2007      48        5.0%      11      1.1%         8     0.8%      15       1.6%     19       2.0%
Q2:2007      44        4.6%      11      1.2%         8     0.8%      15       1.6%     20       2.1%




                                                36
                                                  Table 2:
          Assets of Securitizers as a Percentage of all US BHC Assets and Securitized Loans as a
                                  Percentage of on Balance Sheet Loans by Type.

              Mortgage               HEL                 C&I            Credit Card            Other
Quarter   % Asset % Loans    % Asset   % Loans   % Asset   % Loans   % Asset   % Loans   % Asset % Loans
Q2:2001    85.6%    107.3%     43.5%      6.8%     34.9%      2.5%     48.6%    114.3%    58.6%      5.0%
Q3:2001    82.2%    126.5%     57.9%     11.3%     32.5%      2.4%     59.6%    132.2%    76.9%      6.1%
Q4:2001    83.0%    118.9%     61.4%     13.1%     35.3%      2.6%     62.7%    128.1%    80.9%      5.9%
Q1:2002    79.0%    116.1%     56.1%     10.5%     32.7%      2.3%     59.0%    124.5%    77.3%      5.7%
Q2:2002    79.8%    125.0%     56.8%     12.5%     33.7%      3.4%     54.7%    118.3%    75.5%      5.3%
Q3:2002    77.6%    120.7%     54.6%     13.5%     67.8%      3.4%     52.2%    108.1%    73.1%      4.9%
Q4:2002    76.9%    117.9%     54.6%     12.0%     67.3%      3.3%     51.7%    105.9%    74.4%      5.3%
Q1:2003    76.3%     67.5%     54.3%      5.2%     55.5%      2.2%     51.8%    120.0%    71.7%      4.7%
Q2:2003    77.3%     62.9%     54.8%      5.0%     53.5%      2.1%     52.4%    120.6%    71.9%      4.4%
Q3:2003    75.9%     58.0%     55.5%      5.1%     54.2%      2.1%     51.0%    114.7%    70.5%      4.3%
Q4:2003    76.1%     72.5%     54.1%      7.4%     54.3%      2.3%     50.9%     88.9%    70.0%      4.2%
Q1:2004    77.6%     53.2%     55.9%      4.2%     51.8%      1.9%     53.6%     93.7%    72.3%      6.4%
Q2:2004    80.2%     52.9%     55.2%      3.1%     50.4%      0.9%     55.7%     90.6%    73.4%      5.7%
Q3:2004    80.3%     72.0%     54.4%      2.7%     49.5%      1.0%     55.0%     83.1%    73.6%      5.7%
Q4:2004    82.4%     58.7%     58.9%      3.1%     54.4%      1.8%     61.8%    106.6%    76.3%      9.4%
Q1:2005    80.9%     59.7%     63.2%      3.7%     50.3%      1.0%     57.0%     85.7%    74.7%      5.8%
Q2:2005    81.3%     62.2%     58.6%      3.6%     50.6%      0.8%     57.4%     84.7%    75.1%      6.2%
Q3:2005    78.7%     50.1%     52.1%      3.5%     42.7%      0.7%     50.8%     69.3%    71.6%      6.2%
Q4:2005    79.0%     52.2%     52.0%      3.1%     39.6%      0.7%     51.0%     64.0%    71.8%      8.2%
Q1:2006    74.1%     47.5%     41.0%      2.5%     39.2%      0.8%     51.0%     98.8%    68.4%      8.0%
Q2:2006    75.0%     49.7%     42.2%      4.9%     41.8%      3.4%     49.6%     96.6%    70.1%      8.4%
Q3:2006    73.4%     52.0%     43.0%      5.7%     42.7%      2.4%     50.1%     97.9%    70.5%      8.4%
Q4:2006    73.2%     49.9%     44.4%      5.6%     44.1%      2.7%     50.2%     93.0%    71.3%      8.7%
Q1:2007    72.0%     70.0%     41.8%      2.0%     41.5%      1.0%     52.0%    102.4%    69.1%      7.9%
Q2:2007    72.5%     71.3%     42.8%      2.3%     42.4%      1.1%     53.0%     99.1%    70.2%      7.4%




                                                      37
                                                 Table 3
                          Comparison of securitizers and non-securitizers
All variables in column “Mean” (“Std. Dev”) are the cross sectional mean (Standard deviation) of
the individual BHC time series average. Ln(Assets) is the natural logarithm of assets in thousand
of U.S. dollars. Provision ratio is the total provision divided by total loans. NoAcc+Chargeoff is
the sum of nonaccrual and charge off loans over total loans. Rate on deposit is the interest expense
on deposit divided by total deposit. Time deposit premium is the spread between the rate on large
(above US$100,000) and small (below US$100,000) time deposits. ROE is the income before tax
and extraordinary item and other adjustments divided by average equity. Leverage ratio is the total
liabilities over assets. Column “p-values” report statistical difference between the means of
securitizers and non-securitizers. % difference of means is the difference of securitizers’ and non-
securitizers’ mean over 0.5 times the sum of securitizers’ and non-securitizers’ mean.

                            Non-Securitizers                  Securitizers
Mortgage                  N    Mean       Std De        N       Mean       Std De   p-values   % difference of means
Ln(Assets)               2084 12.9131 0.9971            147    15.1113 2.4762        <.0001                    15.7%
Loans/Assets             2084   0.6673 0.1298           147     0.6436 0.1291        0.0329                    -3.6%
Mortgage/Loans           2084   0.2436 0.1572           147     0.2764 0.1381        0.0138                    12.6%
Provision ratio          2084   0.0021 0.0046           147     0.0025 0.0021        0.3968                    14.1%
NoAcc+Chargeoff/Loans    2084   0.0085 0.0112           147     0.0108 0.0103        0.0137                    24.2%
Time deposit premium     2082   0.0268 0.0366           147     0.0226 0.0377        0.1836                   -16.8%
ROE                      2084   0.1040 0.0587           147     0.1126 0.0559        0.0836                     8.0%
Leverage ratio           2084   0.9041 0.0371           147     0.9083 0.0219        0.1777                     0.5%
Loans/Deposits           2084   0.8484 0.3444           147     0.9732 0.4474        <.0001                    13.7%
HEL
Ln(Assets)               2217   13.0020   1.1454        23     18.1810    2.2322    <.0001                    33.2%
Loans/Assets             2217    0.6654   0.1305        23      0.6078    0.1652    0.0356                    -9.1%
HEL/Loans                2217    0.0350   0.0393        23      0.0858    0.0496    <.0001                    84.2%
Provision ratio          2217    0.0021   0.0045        23      0.0047    0.0035    0.0056                    75.7%
NoAcc+
Chargeoff/Loans          2217    0.0086   0.0110        23      0.0178    0.0185    <.0001                    70.0%
Time deposit premium     2206    0.0268   0.0365        23      0.0019    0.0408    0.0012                  -173.1%
ROE                      2217    0.1044   0.0579        23      0.1174    0.0976    0.2919                    11.6%
Leverage ratio           2217    0.9043   0.0365        23      0.9115    0.0233    0.3469                     0.8%
Loans/Deposits           2217    0.8517   0.3242        23      1.3336    0.9537    <.0001                    44.1%
C&I
Ln(Assets)               2210   13.0057   1.1559        30     16.6947    3.0702    <.0001                    24.8%
Loans/Assets             2210    0.6651   0.1306        30      0.6451    0.1592    0.4068                    -3.0%
C&I/Loans                2210    0.1589   0.0967        30      0.2265    0.0950    0.0001                    35.1%
Provision ratio          2210    0.0021   0.0045        30      0.0036    0.0025    0.0829                    50.0%
NoAcc+Chargeoff/Loans    2210    0.0086   0.0111        30      0.0135    0.0135    0.0153                    44.8%
Time deposit premium     2199    0.0268   0.0366        30      0.0056    0.0290    0.0016                  -130.6%
ROE                      2210    0.1046   0.0581        30      0.1043    0.0823    0.9794                    -0.3%
Leverage ratio           2210    0.9043   0.0366        30      0.9086    0.0227    0.5187                     0.5%
Loans/Deposits           2210    0.8546   0.3401        30      1.0108    0.2558    0.0123                    16.7%
Credit Card
Ln(Assets)               2205   13.0042   1.1585        35     16.2649    2.9858    <.0001                    22.3%
Loans/Assets             2205    0.6653   0.1311        35      0.6344    0.1248    0.1658                    -4.8%
Credit/Loans             2205    0.0032   0.0234        35      0.0741    0.1405    <.0001                   183.6%



                                                   38
Provision ratio         2205    0.0021   0.0036        35     0.0087   0.0212   <.0001   123.4%
NoAcc+Chargeoff/Loans   2205    0.0085   0.0108        35     0.0175   0.0238   <.0001    69.2%
Time deposit premium    2194    0.0267   0.0366        35     0.0170   0.0385   0.1197   -44.5%
ROE                     2205    0.1045   0.0559        35     0.1123   0.1512   0.4287     7.3%
Leverage ratio          2205    0.9044   0.0362        35     0.8983   0.0508   0.3251    -0.7%
Loans/Deposits          2205    0.8507   0.2873        35     1.2345   1.4478   <.0001    36.8%
Other
Ln(Assets)              2194   12.9631   1.0677        46    17.4442   2.2247   <.0001    29.5%
Loans/Assets            2194    0.6658   0.1303        46     0.6194   0.1573   0.0175    -7.2%
Other/Loans             2194    0.0737   0.0746        46     0.1322   0.0872   <.0001    56.8%
Provision ratio         2194    0.0021   0.0045        46     0.0046   0.0039   0.0002    73.4%
NoAcc+Chargeoff/Loans   2194    0.0086   0.0111        46     0.0127   0.0095   0.0123    39.0%
Time deposit premium    2183    0.0267   0.0366        46     0.0186   0.0360   0.1377   -35.8%
ROE                     2194    0.1040   0.0576        46     0.1304   0.0887   0.0024    22.5%
Leverage ratio          2194    0.9047   0.0343        46     0.8867   0.0911   0.0009    -2.0%
Loans/Deposits          2194    0.8524   0.3390        46     1.0582   0.3099   <.0001    21.5%
                                Never                          Ever
Ln(Assets)              2046   12.7925   0.9100        185   14.9079   2.2918   0.0003    15.3%
Loans/Assets            2046    0.6629   0.1258        185    0.6445   0.1273   0.0195    -2.8%
Mortgage/Loans          2046    0.2536   0.1585        185    0.2638   0.1420   0.0572     3.9%
HEL/Loans               2046    0.0320   0.0376        185    0.0453   0.0374   0.1577    34.5%
C&I/Loans               2046    0.1628   0.0986        185    0.1734   0.0905   0.2973     6.3%
Credit/Loans            2046    0.0031   0.0241        185    0.0182   0.0671   0.3978   142.3%
Other/Loans             2046    0.0748   0.0733        185    0.0879   0.0743   0.5703    16.2%
Provision ratio         2046    0.0020   0.0040        185    0.0033   0.0081   0.9214    47.4%
NoAcc+Chargeoff/Loans   2046    0.0079   0.0093        185    0.0111   0.0122   <.0001    32.9%
Time deposit premium    2044    0.0336   0.0386        185    0.0333   0.0376   <.0001    -0.9%
ROE                     2046    0.1031   0.0521        185    0.1075   0.0803   <.0001     4.2%
Leverage ratio          2046    0.9056   0.0334        185    0.9041   0.0501   <.0001    -0.2%
Loans/Deposits          2046    0.8323   0.2643        185    0.9976   0.7348   <.0001    18.1%




                                                  39
                                             Table 4. A
                          Regression Analysis: Time Deposit Premium
     OLS regressions results of time deposit premium on a set of variables for Never securitizers
Variables                        (1)            (2)           (3)              (4)            (5)
Ln(Assets)                     -0.003         -0.003        -0.003           0.003          -0.003
                             (2.82)***      (2.93)***     (2.89)***       (5.77)***      (7.47)***
Mortgage/Loans                 0.088           0.087         0.087           0.043           0.003
                            (15.72)***     (14.88)***    (14.88)***      (11.06)***        (1.74)*
HEL/Loans                       0.001          0.002         0.002          -0.035           0.024
                               (0.04)         (0.10)        (0.10)         (2.46)**      (3.46)***
C&I/Loans                      -0.047         -0.048        -0.048           0.004          -0.020
                             (4.61)***      (4.68)***     (4.61)***         (0.66)       (4.19)***
Credit Card/Loans              -0.013         -0.013        -0.013           0.003          -0.006
                               (0.35)         (0.33)        (0.34)          (0.14)          (0.43)
Other/Loans                     0.044          0.042         0.042           0.009          0.009
                             (3.47)***      (3.24)***     (3.24)***         (1.23)        (2.34)**
3m-pastdue/Loans               -0.126         -0.127        -0.127          -0.112           0.027
                               (0.93)         (0.94)        (0.93)          (1.31)          (0.57)
3m+pastdue/Loans               0.814           0.800         0.799           0.503          -0.037
                             (2.67)***      (2.61)***     (2.60)***        (2.54)**         (0.40)
NoAcc+Chargeoff/Loans           0.038          0.040         0.041           0.097          -0.050
                               (0.31)         (0.32)        (0.33)          (1.24)          (1.28)
Loans/Assets                                  -0.008        -0.008          -0.006          0.001
                                              (1.14)        (1.14)          (1.42)          (0.47)
Deposit rate                                                -0.002
                                                            (1.49)
Insured deposit rate                                                        -1.974
                                                                         (30.03)***
Uninsured deposit rate                                                                       1.289
                                                                                        (99.85)***
Constant                       0.051           0.058         0.058           0.017          0.010
                             (3.84)***      (4.02)***     (3.98)***        (2.02)**        (1.73)*
Observations                    2044           2044          2044            2044            2044
R-squares                       0.21           0.21          0.21             0.70           0.91




                                                 40
                                              Table 4. B
                                      Regression Analysis: ROE
            OLS regressions results of ROE on a set of variables for Never securitizers
Variables                        (1)            (2)              (3)             (4)          (5)
Ln(Assets)                     0.003           0.003            0.004          0.003         0.003
                              (1.95)*       (2.76)***        (2.78)***       (2.31)**     (2.76)***
Mortgage/Loans                 -0.071         -0.058           -0.057         -0.054        -0.053
                            (8.16)***       (6.61)***        (6.60)***      (5.98)***     (6.01)***
HEL/Loans                      -0.155         -0.169           -0.169         -0.166        -0.170
                            (4.93)***       (5.15)***        (5.15)***      (5.04)***     (5.13)***
C&I/Loans                      -0.073         -0.062           -0.061         -0.066        -0.063
                            (4.31)***       (3.56)***        (3.51)***      (3.79)***     (3.61)***
Credit Card/Loans               0.082          0.077            0.077          0.076         0.079
                             (2.16)**       (2.96)***        (2.95)***      (2.84)***     (3.03)***
Other/Loans                    -0.046         -0.022           -0.022         -0.020        -0.020
                            (3.43)***         (1.64)           (1.64)          (1.41)       (1.46)
Loans/Assets                                   0.074            0.073          0.073         0.073
                                            (6.42)***        (6.41)***      (6.42)***     (6.40)***
Deposit rate                                                   -0.003
                                                              (2.68)**
Insured deposit rate                                                           0.160
                                                                              (1.65)*
Uninsured deposit rate                                                                     -0.073
                                                                                           (1.60)
Constant                       0.109           0.041           0.041           0.045        0.044
                             (6.33)***        (2.14)**        (2.11)**       (2.32)**     (2.26)**
Observations                   2046             2046            2046           2046         2044
R-squares                       0.06            0.09            0.09            0.09         0.09




                                                   41
                                              Table 4. C
                                Regression Analysis: Leverage Ratio
      OLS regressions results of leverage ratio on a set of variables for Never securitizers
Variables                       (1)              (2)               (3)            (4)              (5)
Ln(Assets)                   -0.000            0.001             0.001          0.001            0.001
                              (0.01)           (1.26)           (1.48)         (0.79)           (1.23)
Mortgage/Loans               -0.023            -0.009           -0.008         -0.006           -0.005
                            (2.43)**           (1.06)           (1.03)         (0.72)           (0.54)
HEL/Loans                     0.035             0.021            0.019          0.023            0.020
                              (1.23)           (0.92)           (0.87)         (1.01)           (0.89)
C&I/Loans                     0.001             0.013            0.015          0.010            0.012
                              (0.07)           (0.91)           (1.05)         (0.68)           (0.85)
Credit Card/Loans            -0.128            -0.133           -0.133         -0.134           -0.133
                           (4.09)***         (5.11)***        (5.13)***      (5.21)***        (5.16)***
Other/Loans                  -0.030            -0.005           -0.005         -0.003           -0.004
                            (2.50)**           (0.46)           (0.48)         (0.28)           (0.32)
Loans/Assets                                    0.077            0.077          0.077            0.077
                                             (7.02)***        (6.99)***      (7.03)***        (6.93)***
Deposit rate                                                    -0.012
                                                             (14.13)***
Insured deposit rate                                                            0.121
                                                                              (2.46)**
Uninsured deposit rate                                                                          -0.057
                                                                                                (1.89)
Constant                      0.913             0.842            0.840          0.845            0.844
                          (84.54)***        (62.21)***       (62.79)***     (63.72)***       (61.04)***
Observations                   2046             2046             2046           2046             2044
R-squares                      0.03             0.10              0.11           0.11             0.11
 *** Indicates statistical significance at the 1% level, ** indicates statistical significance at the 5% level, and
 * indicates statistical significance at the 10% level.




                                                        42
                                                  Table 5
     Summary Statistics of Securitizers’ Actual and Hypothetical Balance Sheet Information
  This table reports securitizers balance sheet summary statistics and hypothetical statistics (labeled
  with an A- prefix) that are constructed by moving their securitized assets back on balance sheet.
  All variables in column “Mean” (“Std. Dev”) are the cross sectional mean (Standard deviation) of
  the individual BHC time series averages. Mortgage/Loans are on balance mortgages divided by on
  balance sheet loans. HEL/Loans, C&I/Loans, Credit/Loans, Other/Loans and their corresponding
  hypothetical values are defined similarly. % Δ is the percentage change between the hypothetical
  and observed values. Columns “Mort, HEL, C&I, Credit, and Other” present results when those
  securitized asset classes are individually added to the balance sheet. Column “All” is based on
  putting all types of securitized assets back on BHC balance sheets.
                                   Mortgage                         HEL                         C&I
                                      Std.                           Std.                       Std.
                        Mean          Dev.       %∆       Mean       Dev.    %∆      Mean       Dev.    %∆
Mortgage/Loans          0.2764       0.1381               0.2738    0.1657           0.1994    0.1011
A-Mortgage/A-Loans      0.3201       0.1676     15.8%     0.2585    0.1274   -5.6%   0.1954    0.1024   -2.0%
HEL/Loans               0.0531       0.0433               0.0858    0.0496           0.0616    0.0381
A-HEL/A-Loans           0.0489       0.0399     -7.8%     0.1109    0.0804   29.2%   0.0605    0.0376   -1.8%
C&I/Loans               0.1607       0.0776               0.1900    0.0884           0.2265    0.0950
A-C&I/A-Loans           0.1511       0.0755     -6.0%     0.1877    0.0882   -1.2%   0.2448    0.1158   8.1%
Credit/Loans            0.0117       0.0323               0.0445    0.0627           0.0261    0.0497
A-Credit Card/A-Loans   0.0108       0.0307     -8.1%     0.0445    0.0626   -0.2%   0.0259    0.0491   -0.9%
Other/Loans             0.0824       0.0685               0.0907    0.0582           0.0835    0.0598
A-Other/A-Loans         0.0777       0.0667     -5.7%     0.0901    0.0583   -0.7%   0.0822    0.0598   -1.4%
3m-pastdue/Loans        0.0107       0.0074               0.0122    0.0094           0.0089    0.0037
A-3m-pastdue/A-Loans    0.0115       0.0082     7.7%      0.0119    0.0073   -2.6%   0.0097    0.0043   8.5%
3m+pastdue/Loans        0.0022       0.0026               0.0038    0.0037           0.0024    0.0018
A-3m+pastdue/A-
Loans                   0.0027       0.0037     21.5%     0.0038    0.0028   1.8%    0.0025    0.0016   7.2%
NoAcc+Charge-
off/Loans               0.0108       0.0103               0.0178    0.0185           0.0135    0.0135
A-NoAcc+Charge-
off/A-Loans             0.0093       0.0073     -14.4%    0.0165    0.0140   -7.3%   0.0145    0.0125   7.1%
Loans/Assets             0.6436      0.1291                0.6078   0.1652           0.6451    0.1592
A-Loans/A-Assets        0.6629       0.1246     3.0%       0.6139   0.1588   1.0%    0.6498    0.1557   0.7%
Ln(Assets)              15.1113      2.4762               18.1810   2.2322           16.6947   3.0702
Ln(A-Assets)            15.1700      2.5139     0.4%      18.1936   2.2291   0.1%    16.7091   3.0601   0.1%
Obs                       147                                23                        30
                                  Credit Card                       Other                        All
                                       Std.                          Std.                       Std.
                        Mean          Dev.                Mean       Dev.            Mean       Dev.
Mortgage/Loans          0.2402       0.1245               0.2333    0.1124           0.2638    0.1420
A-Mortgage/A-Loans      0.2318       0.1228     -3.5%     0.2303    0.1125   -1.3%   0.2840    0.1695   7.7%
HEL/Loans               0.0420       0.0345               0.0699    0.0439           0.0453    0.0374
A-HEL/A-Loans           0.0404       0.0337     -4.0%     0.0683    0.0440   -2.3%   0.0487    0.0407   7.5%
C&I/Loans               0.1705       0.0830               0.1933    0.0872           0.1734    0.0905
A-C&I/A-Loans           0.1644       0.0821     -3.5%     0.1897    0.0889   -1.8%   0.1599    0.0903   -7.8%
Credit/Loans            0.0741       0.1405               0.0422    0.0959           0.0182    0.0671
A-Credit/A-Loans        0.1046       0.1700     41.1%     0.0408    0.0935   -3.4%   0.0222    0.0809   21.9%
Other/Loans             0.0980       0.0671               0.1322    0.0872           0.0879    0.0743



                                                         43
A-Other/A-Loans        0.0924    0.0616   -5.7%    0.1500    0.1174   13.5%   0.0870    0.0865   -1.1%
3m-pastdue/Loans       0.0139    0.0154            0.0110    0.0052           0.0116    0.0086
A-3m-pastdue/A-Loans   0.0142    0.0163   2.4%     0.0118    0.0057   7.5%    0.0150    0.0072   28.8%
3m+pastdue/Loans       0.0036    0.0056            0.0030    0.0022           0.0025    0.0032
A-3m+pastdue/A-
Loans                  0.0040    0.0058   11.0%    0.0032    0.0022   7.0%    0.0065    0.0031   163.7%
NoAcc+Charge-
off/Loans              0.0175    0.0238            0.0127    0.0095           0.0111    0.0122
A-NonAcc+Charge-
off/A-Loans            0.0187    0.0256   6.9%     0.0133    0.0093   4.9%    0.0152    0.0087   37.4%
Loans/Assets            0.6344   0.1248             0.6194   0.1573           0.6445    0.1273
A-Loans/A-Assets       0.6434    0.1233   1.4%      0.6235   0.1539   0.7%    0.6632    0.1276   2.9%
Ln(Assets)             16.2649   2.9858            17.4442   2.2247           14.9079   2.2918
Ln(A-Assets)           16.2914   3.0023   0.2%     17.4557   2.2193   0.1%    15.1050   2.3522   1.3%
Obs                       35                          46                        185




                                                  44
                                                Table 6
      The Predicted and Observed Measures of Insolvency risk, ROE and Leverage ratio
This table reports predicted values for securitizers’ time deposit premium, ROE, and leverage if
their securitized assets were moved back on balance sheet. These figures are contrasted with
securitizers’ actual values for these variables, which are provided in bold-face. The predicted
values are calculated by moving securitized assets back on balance sheet to create hypothetical
balance sheet variables (labeled with “A-” in Table 5). Predictions are then generated by applying
the balance sheet variables to the regressions in Table 4. Values in plain text are consistent with
securitization reducing time deposit premia, increasing ROE, and increasing leverage. Values
highlighted in yellow are consistent with securitization having the opposite effects. Results are
presented when each asset class is moved back on balance sheet individually (Columns Mortgage,
HEL, C&I, Credit, and Other) and when all securitized assets are returned to the balance sheet
(All).
Time deposit premium                        Mortgage    HEL       C&I       Credit    Other     All
Predicted Value-Model (1)                     2.75%       1.17%     0.62%     1.47%     0.75%   3.36%
Predicted Value-Model (2)                     2.71%       1.17%     0.60%     1.46%     0.73%   3.32%
Predicted Value-Model (3)                     2.71%       1.17%     0.60%     1.46%     0.73%   3.32%
Predicted Value-Model (4)                     2.88%       0.80%     1.34%     2.22%     2.03%   4.03%
Predicted Value-Model (5)                     2.01%       0.13%     0.02%     1.24%     1.29%   2.87%
Mean of securitizers' Time depot. Premium     2.26%      0.19%     0.56%      1.70%     1.86%   3.33%
ROE
Predicted Value-Model (1)                    10.52%     12.13%    11.78%    12.82%    11.54%    10.54%
Predicted Value-Model (2)                    10.84%     12.10%    11.96%    12.94%    11.77%    10.80%
Predicted Value-Model (3)                    10.85%     12.10%    11.96%    12.94%    11.77%     6.74%
Predicted Value-Model (4)                    10.83%     12.13%    11.93%    12.92%    11.71%    10.73%
Predicted Value-Model (5)                    10.92%     12.22%    12.03%    12.97%    11.72%    10.80%
Mean of securitizers' ROE                    11.26%     11.74%    10.43%    11.23%    13.04%    10.75%
Leverage ratio
Predicted Value-Model (1)                    89.81%     89.51%    90.08%    88.37%    89.57%    90.27%
Predicted Value-Model (2)                    90.13%     89.48%    90.25%    88.49%    89.77%    90.54%
Predicted Value-Model (3)                    90.18%     89.48%    90.25%    88.49%    89.77%    90.58%
Predicted Value-Model (4)                    90.12%     89.53%    90.29%    88.45%    89.75%    90.50%
Predicted Value-Model (5)                    90.19%     89.59%    90.32%    88.53%    89.81%    90.53%
Mean of securitizers' Leverage Ratio         90.83%     91.15%    90.86%    89.83%    88.67%    90.41%




                                                   45
                     Table 7. Asset sizes based on Mortgage securitizers
Quantiles            25%        50%        75%        90%        95%        99%
Ln(Assets)           12.22      12.66      13.41      14.58      15.63      18.41
Mean                 13.06
Minimum              10.59
Maximum              21.59
Std. Dev              1.63
Obs                  2232
                                   Ln(assets) size classes
Size                    1        2            3            4       5          6         7
Quantiles            < 25%    25%~50% 20%~75% 75%~90%          90%~95%     95%~99%   > 99%
Fraction of
securitizers         0.72%     3.78%       5.37%      6.61%     10.53%     41.11%    100.00%
# non-securitizers    554       535         529        311        102        53          0
# securitizers          4        21          30         22         12        37         22
Total                 558       556         559        333        114        90         22




                                             46
                   Table 8: Regression of Targeted Variables on Ln (Assets)
   OLS regression results show that size measured as Ln(Assets) is correlated with Time deposit
                    premium, ROE and leverage ratio for small but large banks
                                    (2)             (3)               (4)
A. Sizes 123456              Time Deposit          ROE       Leverage Ratio
                               Premium
Ln(Assets)                        -0.003          0.004             0.000
                                (3.69)**         (1.98)*           (0.43)
Mortgage/Loans                     0.082          -0.052           -0.001
                               (14.42)**        (5.29)**           (0.09)
HEL/Loans                          0.033          -0.141            0.021
                                  (1.82)        (4.27)**           (0.88)
C&I/Loans                         -0.046         -0.071             0.014
                                (4.73)**        (4.15)**           (0.92)
Credit/Loans                      -0.027           0.206           -0.213
                                  (0.67)         (2.00)*          (3.41)**
Other/Loans                        0.038          -0.006           -0.029
                                (3.20)**          (0.36)           (1.61)
Loans/Assets                      -0.008          0.084             0.083
                                  (1.27)        (6.59)**          (6.29)**
3m-pastdue/Loans                  -0.267
                                 (2.03)*
3m+pastdue/Loans                   0.849
                                 2.72)**
NoAcc+Chargeoff/Loans              0.184          -1.171
                                  (1.79)          (1.47)
Constant                           0.056           0.027            0.854
                                (4.39)**          (1.14)         (65.33)**
Observations                       2082            2084             2084
R-squares                          0.20            0.09              0.13
B. Sizes 23456
       Ln (assets)                -0.004          0.003            -0.001
                                (3.80)**          (1.45)           (0.78)
      Observations                 1528            1530             1530
       R-squares                   0.22            0.11             0.13
C. Sizes 3456
       Ln(Assets)                 -0.003          0.006            -0.000
                                 (2.31)*          (1.87)           (0.30)
      Observations                  994             995              995
       R-squares                   0.25            0.14             0.16
D. Sizes 456
       Ln(Assets)                -0.002           0.007            0.001


                                               47
                                         (0.93)                (1.49)                (0.69)
       Observations                       465                   466                   466
        R-squares                         0.33                  0.12                  0.21


E. Sizes 56
       Ln(Assets)                        -0.004                0.003                 0.005
                                         (1.12)                (0.24)                (1.11)
      Observations                        154                   155                   155
        R-squares                         0.42                  0.15                  0.24
F. Size 6
       Ln(Assets)                 -0.003                 -0.006                 0.013
                                   (0.32)                 (0.26)               (0.89)
     Observations                    53                     53                    53
       R-squares                    0.59                   0.26                 0.46
Robust t-statistics in parentheses. *** Indicates statistical significance at the 1% level, ** indicates statistical
significance at the 5% level, and * indicates statistical significance at the 10% level.




                                                            48
                                     Table 9. Comparison within and across size groups
             Panel A compares securitizers and non-securitizers within the same size bucket. In panel B, the
             means of each grouping of non-securitizers to the right of Size 5 (non-sec) are compared to the
             mean of securitizers in groups 6 and 7. For example, non-securitizers in size groups 1-5 have a
             lower mean provision rate than securitizers in groups 6 and 7, with a p-value of 0.028. Groupings
             to the right of Size 5 (non-sec) are compared with that group. For example, securitizers in groups
             6-7 have a higher provision rate than group 5, with a p-value of 0.001.
             All
A           BHCs                           Size 1                            Size 2                        Size 3
            non-                                                             non-                          non-
             sec       sec                non-sec     sec                     sec       sec                 sec       sec
Variable    Mean      Mean      p-value    Mean      Mean     p-value        Mean      Mean      p-value   Mean      Mean      p-value
Provision
ratio       0.0021    0.0025     0.407     0.0020    0.0025    0.605     0.0023        0.0017     0.667    0.0021    0.0016     0.346
Time
depo
premium     0.0268    0.0226    0.184      0.0291    0.0384    0.656     0.0311        0.0381     0.362    0.0260    0.0422    0.011**
ROE         0.1040    0.1126    0.082*     0.0999    0.0850    0.668     0.1057        0.0815    0.037**   0.1031    0.0954     0.395
Leverage
ratio       0.9041    0.9079     0.214     0.9031    0.8817    0.206     0.9045        0.9037     0.924    0.9071    0.9143     0.244
Obs          2092       148                  556        4                  539           21                  530       30

B           Size 4                         Size 5                            Size 6                                  Size 7
            non-                                                             non-
             sec       sec                non-sec     sec                     sec       sec                           sec
            Mean      Mean      p-value    Mean      Mean     p-value        Mean      Mean      p-value             Mean
Provision
ratio       0.0022    0.0020     0.874     0.0022    0.0020    0.854     0.0030        0.0026     0.618              0.0049
Time
depo
premium     0.0205    0.0295     0.258     0.0200    0.0189    0.945     0.0162        0.0091     0.477              0.0030
ROE         0.1057    0.1163     0.358     0.0097    0.0121   0.023**    0.1172        0.1272     0.636              0.1288
Leverage
ratio       0.9037    0.9091     0.493     0.8994    0.9004    0.905     0.8925        0.9098     0.213              0.9080
Obs           314       22                   100       12                  53            37                            22

            Size                 Size                Size                    Size                 Size     Size                 Size
C            1-5                 2-5                  3-5                     4-5                   5      6& 7                  6
            non-                                     non-                    non-
             sec                non-sec               sec                     sec                non-sec    sec                sec
Variable    Mean     p-value     Mean     p-value    Mean     p-value        Mean     p-value     Mean     Mean     p-value     Mean     p-value
Provision
ratio       0.0021   0.028**    0.0022     0.065*    0.0021   0.026**    0.0022        0.126     0.0022    0.0034   0.001***   0.0026     0.295
Time
depo
premium     0.0271   <.000***   0.0263    <.000***   0.0236   <.000***   0.0204       0.004***   0.0200    0.0046   0.035**    0.0091      0.228
ROE         0.1036   0.001***   0.1050    <.000***   0.1046   <.000***   0.1066       0.002***   0.1097    0.1278   0.014**    0.1272    0.0431**
Leverage
ratio       0.9044    0.306     0.9049     0.369     0.9051    0.379     0.9026        0.167     0.8994    0.9091   0.025**    0.9098    0.0443**
Obs          2039                1483                  944                 414                     100       59                  37
             *** Indicates statistical significance at the 1% level, ** indicates statistical significance at the 5% level, and
             * indicates statistical significance at the 10% level.




                                                                        49
                             Table 10. First Stage OLS regressions
     The dependent is mortsec which equals 1 if a bank is a mortgage securitizer, 0 otherwise.
A: Size 5,6,7                   (1)         (2)           (3)           (4)           (5)           (6)
Ln(Assets)                    0.184       0.194         0.182         0.189         0.189         0.189
                          (11.53)*** (12.44)*** (11.02)*** (11.63)*** (11.64)*** (11.63)***
Mortgage/Loans                -0.109      -0.051       -0.194         -0.135        -0.135        -0.135
                              (0.68)      (0.32)        (1.42)        (0.95)        (0.94)        (0.95)
Hel/Loans                     0.862        0.456        0.878          0.488        0.488         0.488
                              (1.11)      (0.58)        (1.13)        (0.62)        (0.62)        (0.62)
C&I/Loans                     -0.542      -0.493       -0.529         -0.495       -0.498        -0.495
                           (2.88)***     (2.47)**    (2.86)***      (2.47)**      (2.48)**      (2.47)**
Credit/Loans                  -0.587      -0.724       -0.910         -1.102        -1.103        -1.102
                             (1.71)*     (2.17)**     (2.38)**     (2.90)*** (2.90)*** (2.90)***
Other/Loans                    0.178       0.264       -0.148         -0.031        -0.030        -0.031
                              (0.55)      (0.86)        (0.43)        (0.09)        (0.09)        (0.09)
Loan/Assets                                0.417                       0.370        0.371         0.370
                                         (2.25)**                    (1.87)*       (1.88)*       (1.87)*
NoAcc+Chargeoff/Loans                                  -3.419         -2.566        -2.578        -2.569
                                                        (1.56)        (1.10)        (1.11)        (1.10)
3m-pastdue/Loans                                       15.088        12.128        12.116        12.133
                                                     (2.86)***      (2.25)**      (2.25)**      (2.25)**
3m+pastdue/Loans                                        9.359        15.787        15.797        15.785
                                                        (0.62)        (1.07)        (1.07)        (1.07)
Deposit rate                                                           0.007
                                                                      (1.33)
Insured Deposit rate                                                                0.103
                                                                                    (1.53)
Uninsured Deposit rate                                                                             0.065
                                                                                                  (1.31)
Constant                      -2.562      -2.984       -2.601         -2.942        -2.944        -2.942
                          (11.50)*** (10.90)*** (11.33)*** (10.70)*** (10.71)*** (10.70)***
Observations                    226         226          226            226          226            226
R-squared                      0.34        0.35          0.38          0.39          0.39          0.39
B: Size 6,7
Ln(Assets)                    0.176       0.190         0.186         0.221         0.200         0.213
                           (6.89)*** (7.74)*** (6.99)*** (9.03)*** (7.94)*** (8.66)***
Mortgage/Loans                -0.309      -0.335       -0.805         -0.764        -0.541        -0.943
                              (1.03)      (1.17)     (3.02)*** (3.07)***          (2.09)**     (3.15)***
Hel/Loans                     0.615       -0.060        0.076         -0.214        0.148         -0.410
                              (0.56)      (0.05)        (0.07)        (0.19)        (0.14)        (0.34)
C&I/Loans                     -1.053      -1.154       -1.289         -1.050       -1.251        -1.173
                            (2.33)**     (2.50)**    (2.83)***      (2.38)**     (3.08)***      (2.53)**
Credit/Loans                  -0.924      -1.181       -1.774         -1.655        -1.681        -1.831


                                               50
                                    (2.14)**        (2.75)***       (3.99)***        (3.66)***   (3.81)***       (4.08)***
Other/Loans                           0.426           0.487           -0.339           -0.174       0.068          -0.354
                                     (0.61)           (0.80)          (0.47)           (0.25)       (0.10)         (0.50)
Loan/Assets                                           0.630                             0.125       0.302           0.147
                                                     (2.39)*                           (0.42)       (1.03)         (0.49)
NoAcc+Chargeoff/Loans                                                 -8.893           -9.463       -6.900         -8.255
                                                                     (2.62)**        (2.69)***    (1.99)**        (2.36)**
3m-pastdue/Loans                                                      32.455           28.557      22.842          30.907
                                                                    (5.83)***        (4.93)***   (3.49)***       (5.05)***
3m+pastdue/Loans                                                      18.579           29.625      24.618          24.507
                                                                      (1.28)          (1.85)*      (1.78)*         (1.54)
Deposit rate                                                                           39.540
                                                                                     (3.57)***
Insured Deposit rate                                                                               49.333
                                                                                                  (2.56)**
Uninsured Deposit rate                                                                                             40.514
                                                                                                                  (1.91)*
Constant                        -2.260         -2.807            -2.368           -3.497          -3.011           -3.038
                              (5.28)*** (6.32)*** (5.67)*** (7.86)*** (6.73)***                                  (7.13)***
Observations                     112            112                112             112              112             112
R-squared                        0.25           0.28              0.38             0.44            0.42             0.41
Robust t-statistics in parentheses. *** Indicates statistical significance at the 1% level, ** indicates statistical
significance at the 5% level, and * indicates statistical significance at the 10% level.




                                                            51
         Table 11 A. Instrumental variables Approach: Time deposit premium-results
  The dependent variable is Time deposit premium. The instrumented variable is mortsec which
                      equals 1 if bank is a mortgage securitizer, 0 otherwise.
A: size 5,6,7 IV                   (1)           (2)           (3)            (4)       (5)
mortsec                         -0.038         -0.035        -0.034         -0.034    -0.034
                              (3.81)***      (3.81)***     (3.67)***      (3.69)*** (3.64)***
Mortgage/Loans                   0.090          0.095         0.094          0.095     0.094
                              (5.60)***      (5.48)***     (5.48)***      (5.48)*** (5.47)***
Hel/Loans                        0.240          0.209         0.205          0.206     0.204
                              (3.07)***      (2.76)***     (2.71)***      (2.72)*** (2.70)***
C&I/Loans                       -0.094         -0.089        -0.094         -0.093    -0.094
                              (3.54)***      (3.35)***     (3.47)***      (3.43)*** (3.51)***
Credit/Loans                     0.026          0.015         0.014          0.014     0.014
                                 (0.66)        (0.36)        (0.34)         (0.34)    (0.34)
Other/Loans                      0.122          0.131         0.133          0.133    0.134
                              (3.06)***      (3.22)***     (3.26)***      (3.25)*** (3.27)***
3m-pastdue/Loans                -1.210         -1.472        -1.493         -1.492    -1.496
                                (1.71)*       (1.98)**      (2.00)**       (1.99)**  (2.01)*
3m+pastdue/Loans                -0.289          0.166         0.185          0.180     0.186
                                 (0.28)        (0.14)        (0.15)         (0.15)    (0.15)
NoAcc+Chargeoff/Loans            0.131          0.208         0.186          0.190     0.180
                                 (0.62)        (0.90)        (0.79)         (0.81)    (0.76)
Loans/Assets                                    0.028         0.029          0.029     0.029
                                               (1.24)        (1.31)         (1.29)    (1.32)
Deposit rate                                                  0.003
                                                           (3.50)***
                                                                             0.027
Insured deposit rate                                                       (2.46)**
Uninsured deposit rate                                                       0.030
                                                                          (3.49)***
Constant                         0.010         -0.009        -0.009         -0.009    -0.009
                                 (1.00)        (0.45)        (0.46)         (0.46)    (0.46)
Observations                      225            225           225            225       225
R-square                          0.31          0.33          0.33           0.33      0.34
B: Size 6, 7 IV
mortsec                         -0.039         -0.036        -0.028         -0.045    -0.005
                              (2.78)***       2.75)***      (2.33)**      (3.66)***   (0.45)
Observations                      112            112           112            112       112
R-squared                         0.26          0.29          0.35           0.38      0.55
C: Size 5, 6, 7 OLS
mortsec                         -0.016         -0.016        -0.016         -0.016    -0.016
                              (2.85)***      (2.88)***     (2.82)***      (2.83)*** (2.81)***
Observations                      225            225           225            225       225


                                             52
R-square                                 0.35              0.36              0.36          0.36             0.36
D: Size 6, 7 OLS
mortsec                         -0.012            -0.013               -0.014           -0.007             -0.011
                                (1.64)           (1.70)*              (1.89)*           (1.15)            (1.80)*
Observations                     112                112                 112              112                112
R-square                         0.34              0.35                 0.37             0.54               0.56
Robust t-statistics in parentheses. *** Indicates statistical significance at the 1% level, ** indicates statistical
significance at the 5% level, and * indicates statistical significance at the 10% level.




                                                            53
                          Table 11 B. Instrumental variable approach: ROE
   The dependent variable is ROE. The instrumented variable is mortsec which equals 1 if bank is a
                                      mortgage securitizer, 0 otherwise.
A: Size 5, 6, 7 IV                (1)            (2)                 (3)                (4)                (5)
mortsec                        -0.003          0.009               0.009              0.009              0.009
                               (0.15)          (0.56)              (0.55)             (0.55)            (0.54)
Mortgage/Loans                  0.015          0.030               0.030              0.030              0.030
                               (0.42)          (0.77)              (0.77)             (0.77)            (0.78)
Hel/Loans                       0.197          0.093               0.094              0.093              0.094
                               (1.39)          (0.87)              (0.88)             (0.88)            (0.88)
C&I/Loans                      -0.056          -0.038             -0.037             -0.037             -0.036
                              (2.26)**         (1.54)              (1.45)             (1.45)            (1.44)
Credit Card/Loans               0.381           0.357               0.358             0.358              0.358
                              (2.52)**       (2.29)**            (2.29)**           (2.29)**          (2.29)**
Other/Loans                    -0.053          -0.036             -0.036             -0.037            -0.037
                               (0.75)          (0.48)              (0.48)             (0.48)            (0.48)
Loans/Assets                                   0.096               0.095              0.095              0.095
                                              (1.94)*             (1.89)*            (1.89)*           (1.89)*
Deposit rate                                                       -0.001
                                                                   (0.66)
Insured deposit rate                                                                 -0.013
                                                                                      (0.67)
Uninsured deposit rate                                                                                  -0.012
                                                                                                        (0.83)
Constant                        0.114          0.048               0.049              0.049              0.049
                             (6.83)***         (1.06)              (1.05)             (1.05)            (1.06)
Observations                     226             226                 226               226                226
R-square                         0.09           0.13                0.13               0.13               0.13
 B: Size 6, 7 IV
mortsec                        -0.037          -0.023             -0.046             -0.024            -0.047
                               (1.46)          (1.12)              (1.55)             (1.26)            (1.62)
Observations                     112             112                 112               112                112
R-square                         0.11           0.17                0.17               0.17               0.15
C: Size 5, 6, 7 OLS
mortsec                         0.010          0.010               0.010              0.010              0.010
                               (1.30)          (1.36)              (1.34)             (1.34)            (1.34)
Observations                     226             226                 226               226                226
R-square                         0.10           0.13                0.13               0.13               0.13
D: Size 6, 7 OLS
mortsec                         0.002          -0.001              0.002              0.002             -0.004
                               (0.13)          (0.07)              (0.13)             (0.13)            (0.30)
Observations                     112             112                 112               112                  112
R-square                         0.15           0.18                0.23               0.19               0.20
  Robust t-statistics in parentheses. *** Indicates statistical significance at the 1% level, ** indicates statistical
 significance at the 5% level, and * indicates statistical significance at the 10% level.



                                                             54
                  Table 11 C. Instrumental variable approach: Leverage Ratio
  The dependent variable is Leverage ratio. The instrumented variable is mortsec which equals 1 if bank is a
                                        mortgage securitizer, 0 otherwise.
A: Size 5, 6, 7 IV                 (1)               (2)               (3)           (4)              (5)
mortsec                           0.019            0.030              0.027         0.027           0.026
                               (2.13)**         (2.02)**            (1.82)*       (1.82)*          (1.79)*
Mortgage/Loans                    0.023            0.037              0.038         0.038           0.039
                                 (0.99)           (1.22)             (1.27)        (1.25)           (1.29)
HEL/Loans                        -0.080           -0.179             -0.170        -0.171          -0.169
                                 (1.03)           (1.33)             (1.26)        (1.27)           (1.25)
C&I/Loans                         0.013            0.030              0.050         0.050           0.050
                                 (0.41)           (0.80)             (1.56)        (1.55)           (1.55)
Credit Card/Loans                -0.006           -0.029             -0.021        -0.022          -0.022
                                 (0.21)           (0.94)             (0.72)        (0.74)           (0.74)
Other/Loans                      -0.206           -0.189             -0.197        -0.197          -0.196
                                 (1.56)          (1.77)*            (1.84)*       (1.84)*          (1.83)*
Loans/Assets                                       0.092              0.084         0.084           0.085
                                                  (1.30)             (1.19)        (1.20)           (1.20)
Deposit rate                                                         -0.012
                                                                  (20.31)***
Insured deposit rate                                                               -0.152
                                                                                (19.62)***
                                                                                                   -0.115
                                                                                                 (19.29)***
Uninsured deposit rate
Constant                          0.907            0.844              0.847         0.847           0.847
                             (124.66)***       (17.76)***         (17.75)***    (17.75)***       (17.73)***
Observations                       226              226                226           226             226
R-square                          0.13              0.17              0.24          0.24             0.24
B: Size 6, 7 IV
mortsec                           0.032            0.044              0.044         0.044           0.046
                                 (1.35)           (1.48)            (1.67)*        (1.56)           (1.53)
Observations                       112              112                112           112             112
R-square                          0.28              0.32              0.32          0.32             0.31
C: Size 5, 6, 7 OLS
mortsec                           0.015            0.015              0.015         0.015           0.015
                               (2.20)**         (2.25)**           (2.16)**      (2.17)**         (2.14)**
Observations                       226              226                226           226             226
R-squared                         0.13              0.19              0.25          0.25             0.25
D: Size 6, 7 OLS
mortsec                           0.027            0.024              0.024         0.023           0.024
                                (1.73)*          (1.93)*            (1.86)*       (1.73)*          (1.94)*
Observations                       112              112                112           112             112
R-square                          0.28              0.35              0.35          0.35             0.35
Robust t-statistics in parentheses. *** Indicates statistical significance at the 1% level, ** indicates statistical
significance at the 5% level, and * indicates statistical significance at the 10% level.




                                                          55
Figure 1: Time deposit premium as a function of Ln(Assets). The figure presents a cubic
spline estimate (the green line) with 95% confidence bounds (the red lines) of the relationship
between the bank’s time deposit premium (the spread between its insured and uninsured deposits)
and bank size as measured by Ln (On Balance-Sheet Assets) after controlling for balance-sheet
composition and the performance of the bank’s assets. Y-BX is the part of time deposit premium
that is not explained by balance sheet composition, and asset performance.




                                              56
Figure 2: Leverage Ratio as a function of Ln(Assets). The figure presents a cubic spline
estimate (the green line) with 95% confidence bounds (the red lines) of the relationship between
the bank’s on-balance sheet leverage ratio and bank size as measured by Ln (On Balance-Sheet
Assets) after controlling for balance-sheet composition and the performance of the bank’s assets.
Y-BX is the part of the leverage ratio that is not explained by balance sheet composition, and asset
performance.




                                                 57
Figure 3: ROE as a function of Ln(Assets). The figure presents a cubic spline estimate (the
green line) with 95% confidence bounds (the red lines) of the relationship between the bank’s
return on equity (ROE) and bank size as measured by Ln (On Balance-Sheet Assets) after
controlling for balance-sheet composition and the performance of the bank’s assets. Y-BX is the
part of ROE that is not explained by balance sheet composition, and asset performance.




                                               58

				
DOCUMENT INFO
Description: Us Bank Mortgage Exposure document sample