University of Amsterdam
Amsterdam Business School
Master in International Finance
Master Thesis
Institutional Lenders in Syndicated Loan and Loan spread
Author:
Zhou, Yanfei
Student number: 11933046
Supervisor:
Dhr. Dr. Tomislav Ladika
September 2018
2
Index
Abstract ... 3
1. Introduction ... 4
2. Literature Review ... 6
2.1. Background of institutional lenders ... 6
2.2. Pricing for syndicated loans ... 8
2.3. Pricing for institutional loans ... 9
2.4. Connections and improvements with regard to the previous studies 12 3. Data and Methodology ... 14
3.1. Overview of the sample ... 14
3.2. Cross-loan test ... 17
3.3. Within-issuance spread gap test ... 21
3.4. Combine the results of cross-loan test and within-issuance test . 24 4. Result and Interpretation ... 24
4.1. Cross-loan test ... 24
4.2. Within-issuance spread gap test ... 27
5. Limitations of the Current Research ... 28
6. Potential Extensions of the Current Research ... 29
Conclusion ... 30
Exhibit 1 ... 32
Exhibit 2 ... 33
Exhibit 3 ... 34
3
Abstract
Non-bank institutional lenders are playing more and more important role in the syndicated loan market in the last two decades. The institutional lenders consistently charge higher interest than banks do. This phenomenon raises the question: why do the institutional lenders charge the higher loan spreads than banks do despite the fact that they are investing in the similar loan facilities and sometimes even loan facilities issued by the same company under the same market condition? In a sample of 5601 syndicated loan facilities issued by US public companies in the period from 2007 to 2017, I find that the spread differences are due to institutional lenders’ information disadvantage. At the same time, the spread differences are also the consequences of the syndication pricing mechanism.
4
1. Introduction
After the introduction of loan ratings by Moody’s and S&P 500 in 1995, there has been an increasing number of institutional investors, including hedge funds, private equity funds, prime funds, hybrid funds, and insurance companies, participating in syndicated loans. In pricing loans to institutional lenders, several previous studies (Ivashina, 2009; Nandy & Shao, 2010; Ivashina & Sun, 2011; Lim, Minton & Weisbach, 2012) indicate that non-bank institutional lenders charge higher all-in-drawn spreads on syndicated loans than banks in the primary market. The purpose of this paper is to understand why there is a historical spread difference between the spreads of institutional syndicated loan facilities and the spreads of bank facilities. One of the possible explanations of the higher spreads is that unlike commercial banks who have on-going relationships with borrowers and better access to internal information, institutional investors are less informed than banks (Petersen and Rajan, 1994). The extra loan spreads are charged as a compensation of the costly information production process (Nandy&Shao, 2010). Another possible explanation can be that the companies which normally have no access to bank funding have to rely on non-bank institutional lenders as the last resorts. So the extra spread is the compensation of the liquidity provided by non-bank lenders. From lenders’ perspective, non-bank institutional lenders such as PE funds, hedge funds and insurance companies have higher return objectives and risk tolerance than commercial banks due to the different investment mandates. The non-bank
5
institutional lenders may selectively lend money to the loans with lower quality and higher interest rates. Other studies like the one Ivashina and Sun conducted in 2011 reveal the opposite effect the non-bank institutional lender’s participation have on loan spread. The participation of the non-bank institutional lenders increases the supply of the credit thus leading to the contraction of loan spread in the primary market. Other technical factors such as liquidity of syndicated loan secondary market and demand and supply of the credit may also contribute to the higher spread charged by the institutional facilities. From a practitioner’s point of view, Standard Poor’s 2007 guide to the loan market says: “In pricing loans to the institutional investors, it’s a matter of the spread loan relative to the credit quality and market based factors. This second category can be divided into liquidity and market technical (i.e., supply/demand), or supply relative to demand, is a matter of simple economics.”
To address the way in which non-bank institutional lenders are involved in syndicated loans, I consider a sample of 5601 loan facilities of syndicated loans from Thomson-one syndicated loan data base, each of which was originated between 2007 to 2017 in the United States. I also download the yields of United States corporate bonds between 2007 and 2017 from Federal Reserve website to reflect the Macroeconomic conditions during that period.
Cross-loan spread test and within-issuance test are conducted to see whether the extra spread charged by institutional lenders can be fully explained by firm specific factors, market condition factors and loan specific factors. If yes, then
6
the extra spread is caused by the adverse selection. This paper finds that the extra spread charged by institutional investors are not entirely caused by adverse selection. Information disadvantage of the institutional lenders and the pricing mechanism of the syndication can the possible explanations.
This paper parallels the research conducted by Lim, Minton and Weisbach (2012) in terms of the two major hypotheses. The innovative work is that the market condition factors which is represented by credit spread is introduced in the cross-loan test. Another extension of the previous research is to include the dummy variable indicating tranche order difference in the within loan test. Unlike Lim, Minton and Weisbach (2012), this paper combine the result of the two major test and give a joint analysis.
2. Literature Review
2.1. Background of institutional lenders
A syndicated loan is a loan where more than two financial institutions are jointly lending money to the borrower. There are two primary types of lenders participate in the syndicated loan market: banks and non-bank institutional investors. Institutional lenders include mutual funds, private equity funds, insurance companies and collateralized debt obligations (CDO). The credit boom before the crisis in 2008 was characterized by the increasing participation of non-bank institutional lenders associated with securitization (Ivashina&Sun,
7
2010). Neuhann and Saidi (2016) conclude that the deregulation of the banking industry facilitated the participation of non-bank institutional lenders by broadening intermediation relationships among universal banks and their borrower firms. Several previous studies (Nandy&Shao,2010; Ivashina&Sun,2011) show that the non-institutional lenders have different appetite of risk when sourcing the borrowers. The institutional loans are broadly characterized by highly leveraged loans with low credit quality. Financing purposes of these loans are mainly for LBOs, stock repurchase, recapitalization and merger and acquisitions.
A typical syndicated loan consists of term loan A, term loan B and revolving line. Term loan is the loan that is repaid in a determined date and at a pre-determined interest rate. Under a term loan facility, the full amount of committed fund should be withdrawn and the facility is terminated as soon as the full amount is repaid. Unlike the term loans, the revolving lines give borrowers the right to withdraw money whenever they need within a committed amount. The revolving line requires an up-front fee which is called commitment fee. Nandy and Shao (2010) and Ivashina and Sun (2011) document that institutional lenders usually participate in term loan B tranches instead of term loan A and revolving lines. Unlike banks which have less costly and relatively stable source of fund, financial institutions have higher cost of fund. The undrawn fund from revolving line is too costly for institutional investors. However, the difference between term loan A and term loan B is unobservable yet. A possible
8
explanation for institutional investor’s preference towards term loan B can be the unobservable covenants differences. Barnish, Miller and Rushmore (1997) argue that term loan B has longer maturity and higher prepayment fees.
2.2. Pricing for syndicated loans
We can think of the “book-building” mechanism of syndication as a two-stage process. The first stage is similar to a traditional sealed-bid auction with a spread ceiling. The difference between traditional sealed-bid auction and the first step of syndication is that the deal cannot be cleared by a single buyer. The interest rate of syndicated loan is expressed as a spread over London interbank offered rate (LIBOR). As soon as the syndication is open, all the investors submit their bids indicating the share they are willing to buy and the spread they are willing to pay. The spread can only be bid down. At the end of the day, a minimum amount of the clearing spread will be determined. If the loan is under-subscribed in the first stage, the syndication will enter the second stage which can be described as “retail” pricing stage where a fixed spread is announced and all the potential lenders submit their bid indicating the share and spread (Lazear, 1986). The fixed price will be broadcasted to all the investors for a period of time before it is discounted and the deal will be closed at this discounted price (A discounted price of a loan is equivalent to an inflated-up spread).
9
2010; Ivashina & Sun, 2011; Lim, Minton & Weisbach, 2012), the all-in-drawn spread is used as a measure of total cost of the loan. All-in-Drawn spread includes an interest spread and an annual fee. In this paper, I used the all-in-drawn spread for each loan facility as an approximation of the spread charged by all the participants in this loan facility because Thompson-one syndicated loan database does not provide the allocation of the fund and fees for each facility.
2.3. Pricing for institutional loans
Commercial banks follow the Asset Liability Management (ALM)1 approach when setting their risk and return objectives while institutional lenders like hedge funds and private equity funds follow the Asset Only rule when determining their risk and return objectives. Due to the different constrains when addressing risks and setting return objectives, private equity funds and hedge funds usually have higher required rate of return than commercial banks do. Another type of non-bank institution lender is insurance company which also follow the Asset Liability Management principle but with much longer liability duration than commercial banks do which leads to an overall higher risk tolerance and required return. There are several studies (Nandy & Shao, 2010;
1
Asset Liability management approach: In the context of determining a strategic asset allocation, an asset liability management approach involves explicitly modeling liabilities and adopting the allocation of asset that is optimal in relationship to funding liabilities. Asset Only management approach: In the context of determining a strategic asset allocation, an approach that focuses on the characteristics of the assets without explicitly modeling the liabilities.
10
Ivashina & Sun, 2011; Lim, Minton & Weisbach, 2012) have the same finding that non-bank institutional lenders charge higher all-in-drawn spreads on syndicated loans than banks do even for an identical loan. One existing explanation of the higher spread for institutional loan facilities is that institutional investors have higher required return (Lim, Minton & Weisbach, 2012). However, this theory provides no explanation for why institutional lenders and banks charge different interest rate even for the loan of the same issuer and with same loan specific characters. Lim, Minton and Weisbach (2012) find that within the same issuance the institutional facilities charge higher spread than the bank facilities. Therefore, the higher required return (or higher risk tolerance) of institutional investor cannot fully explain why institutional investors charge higher spread than banks even for the loan facilities with identical credit risks. It is more logical to say that the institutional investors charge higher spread because they take on extra risks which are not fully explained by credit risk. This phenomenon has also been explained by the moral hazard theory which is in consistency with the information asymmetry theory. Holmstrom (1979) introduces a moral hazard framework whereby informed" investors must conduct due diligence and monitoring before uninformed investors are willing to invest. Sufi (2007) addresses the information asymmetry issue in syndicated loans. His research reveals that the information asymmetry problem is less when the participants of a syndicated loan have previous relationship with the borrower. Petersen and Rajan (1994) state that non-bank institutional investors
11
are relatively less informed than the banks because the institutional lenders do not have the close previous relationships with borrowers as commercial banks do. In Nandy and Shao’s 2010 research, they conclude that non-bank institutional investors receive higher all-in-drawn spread in primary market in compensation of being informationally disadvantaged. They also address the issue by adverse selection theory. Institutional lenders tend to lend to riskier borrowers and lend for riskier purposes. Their findings parallel those of Brophy, Ouimet, and Sialm (2009), who conclude that companies having difficulties raising funds view the non-bank institutional investors as last resort. Thus the higher spreads of the institutional loans can be seen as the return for providing extra liquidity which banks are not able to provide. However, the adverse selection theory only explains why the majority of the institutional loan facilities have riskier risk profile but it does not explain why the loan facilities where non-bank institutional lenders are present provide higher loan spread even with the same company specific characters (credit rating, leverage ratio, industry) and loan specific characters (maturity, collateral and purpose of the loan). Ivashina and Sun (2011) introduce supply-and-demand of credit theory to explain how the increasing institutional lenders change the loan spreads in a market-wide view. They introduce a concept called time-on-the-market(TOM), which is the number of days from the initiation of a syndication to completion of the loan, to measure institutional demand pressure for syndicated loan. They find that TOM is higher for institutional loans and TOM is positively correlated
12
to the loan spreads. This theory is to some extend practical since it is consistent with the two-steps mechanism of the syndication pricing process mentioned in 2.2.
2.4. Connections and improvements with regard to the previous studies This paper is related to the research conducted by Lim, Minton and Weisbach (2012). We both examine the cross-sectional spread difference between institutional facilities and bank facilities. In the cross-sectional spread analysis, we both examine the incremental effect the institutional lenders on loan spread by controlling other factors that could affect the loan spread constant. In their research the factors include facility size, number of participants, the lender’s previous relationship with the borrower, maturity, asset of the borrower, leverage and industry ROA. I extend their research by introducing the new variable that could affect the loan spread, namely the average credit spread at the period when the loan was launched. According to Allen and Saunders (2002), the market’s perception of credit risk is cyclical. Credit spread narrows when the economy strengthens and investors expect the firms’ credit metrics to improve. Conversely, credit spread widens when the economy weakens There are other interrelated factors that collectively determine the market-wide credit spread, such as financial market performance (Norden & Weber, 2004) and general market demand and supply of loans (Ivanshina & Sun, 2011). Credit spreads narrow in strong-performing markets overall including equity
13
market. Credit spreads narrow in times of high demand for loans. I also substitute the firm specific factors, such as asset of the borrower, leverage and industry ROA, with S&P’s credit rating due to the lack of firm specific data. Studies show that loan spread on lower-quality issues tend to be more volatile than spreads on higher-quality issues (Duffie & Singleton, 1999). I apply the following method to reflect the heterogeneity between investment grade syndicated loan and high-yield syndicated loan. When studying the incremental effect the institutional lenders have on the loan spreads, I use the Moody’s AAA bond yield spread to represent the market-wide credit risk.
To further eliminate the unobservable company specific factors and market-wide factors that could affect the pricing, I also conduct the within- issuance spread gap study to observe the spread difference in different tranches within the same issuances. My study parallels the one of Lim, Minton and Weisbach (2012) in the way that we both examine the factors which can affect the spread gaps between the institutional tranches and bank tranches within the same issuance. I extend their research by including the dummy variable of institutional status. I define the dummy as the follows:
Dummy equals to 1 when institutional tranche order is alphabetically lower than the bank tranche. (i.e. Tranche identity difference equals to 1 when tranche Cis institutional and tranche B is bank)
Dummy equals to 0 when institutional tranche order is NOT alphabetically lower than the bank tranche. (i.e. Tranche identity
14
difference equals to 0 when tranche A is institutional facility and tranche B is bank facility)
Although there is no seniority difference between tranche A and tranche B, there are some contractual differences between two types of tranches in practitioner’s perspective. I intent to test whether the contractual difference among different tranches serves as one of the factors that affect the spread gaps between institutional facilities and bank facilities.
3. Data and Methodology
3.1. Overview of the sample
I obtain my sample of syndicated loans from the Thomson one database for the 2007-2017 period. I exclude all the loan facilities which were issued by financial institutions2 and bridge loans in the sample. My sample consists of 5601 loan facilities, each of which is issued by public companies located in the United States. All the loan facilities are denominated in US dollars and the interest rates are expressed as spreads over LIBOR. The commitment fees for the revolvers are added back to the interest rates in order to get the actual financing cost of the loan. I used the “Institutional flag” provided by Thomson one database to distinguish institutional facilities with non-institutional facilities. The flag is “Y” when the loan is associated with at least one institutional lender. The
2
Financial institutions include alternative financial investment firms, asset management companies, banks, brokerage companies, credit institutions, government sponsored financial entities, insurance companies.
15
flag is “N” when the loan is only financed by commercial banks and investment banks. Of the 5601 loan facilities, 1207 facilities are associated with non-bank institutional lenders while the remaining 4394 are only associated with commercial banks and investment banks. The sample consist of 2468 stand-alone term loan facilities and 3133 revolver facilities packed with term loans. In the sample, I document 4227 leveraged loan facilities among which 1191 are institutional facilities and 3036 are bank facilities. According to the Thomson one definition of leveraged loan, a loan with a credit rating of BB or lower is defined as a leveraged loan facility. 98% of the institutional facilities are leveraged which parallel the finding by Nandy and Shao (2010). The institutional lenders tend to lend to the leveraged loans. The average of the principal amount of the institutional loan facilities is 648 million dollars which is much higher than that of the non-institutional facilities. The average principal of institutional facilities is 648 million dollars which is much higher than 360 million dollars for the non-institutional facilities. The average loan spread of institutional loans is much higher than that of the non-institutional loans which parallel the findings from previous researches (Ivanshina, 2009; Nandy & Shao, 2010; Ivashina & Sun, 2011; Lim, Minton & Weisbach, 2012). The mean purpose of this thesis is to find the factors causing the higher spread of the institutional loan facilities (Table 1).
Table 1 Sample Overview
Institutional Facilities Non-institutional Facilities
16
Number of Leveraged facilities 1191 3036
Percentage of leveraged facilities (%) 98.67% 69.09%
Average years of Maturity (year) 4.63 4.73
Average Principal Amount ($ mil) 648 360
Average Number of Participants 3.6 3.8
Average Loan Spread (bps) 429 246
I obtain the S&P credit rating for the loan facilities as an indicator of credit risk. Among all the 5601 loan facilities, 3240 facilities are rated. There are 17.4% of the facilities are not rated for institutional facilities and 43.25% of the bank facilities are not rated (Table 2). Non-bank institutional lenders are considered less informed compared to the banks due to the lack of private information, institutional investors therefore primarily depend on third party rating to make their investment decision (Sufi, 2006).
Table 2 Credit Ratings for Institutional Facilities and Non-Institutional Facilities
Credit Rating Institutional facilities Non-institutional facilities Institutional facilities (%) Non-institutional facilities (%) AAA 0 1 0.00% 0.02% AA 0 4 0.00% 0.09% A 2 65 0.17% 1.48% BBB 23 575 1.91% 13.09% BB 385 1025 31.90% 23.33% B 522 513 43.25% 11.68% CCC 65 60 5.39% 1.37% Not Rated 210 2151 17.40% 48.95% Total 1207 4394 100% 100%
In order to quantify the credit risk, I obtain the Global Corporate Annual Default Rates By Rating Category (Table 3) to represent the credit risk (S&P,2007). To incorporate the general market condition factors (demand and supply of credit or
17
economy conditions, etc.) in the regression, I calculate the monthly credit spreads between S&P AAA rated bonds and US 5-year treasury bonds in the period between 2007 to 2017 (Exhibit 1). I acquire the monthly S&P AAA-rated bond yields from S&P website and the monthly 5-year treasury bonds yields from the US department of treasury website.
Table 3 Global Corporate Annual Default Rates By Rating Category
AAA AA A BBB BB B CCC 2007 0 0 0 0 0.2 0.25 15.24 2008 0 0.38 0.39 0.49 0.81 4.08 27.27 2009 0 0 0.22 0.55 0.75 10.92 49.46 2010 0 0 0 0 0.58 0.86 22.62 2011 0 0 0 0.07 0.09 1.67 16.3 2012 0 0 0 0 0.3 1.56 27.52 2013 0 0 0 0 0.1 1.63 24.34 2014 0 0 0 0 0 0.78 17.13 2015 0 0 0 0 0.16 2.39 25.88 2016 0 0 0 0 0.47 3.68 32.67 2017 0 0 0 0 0.38 1 30
Source: S&P Gobal
3.2. Cross-loan test
18
bps higher than that of the bank facilities for the past 10 years. Information asymmetry can be the possible explanation of the spread difference. Normally, bank lenders have access to more private information of the borrowers and especially the leading arrangers have previous relationship with the borrowers. The higher loan spread can be seen as the compensation for the costly information processing process (Nandy & Shao, 2010). An alternative explanation for the spread difference can be that non-bank institutional lenders tend to invest in the loans with higher risks. As documented in 3.1, only 2.08% institutional facilities invest in loans with rating of BBB or higher while 14.68% of bank facilities invest in loans with rating of BBB or higher.
The purpose of this paper is to understand why there is a historical spread difference between the spreads of institutional syndicated loan facilities and the spreads of bank facilities. To study the factors that can potentially affect the loan spread, I estimate the following equation:
All-in-drawn spread= c +Dummy(Institution)Maturity+Participants +*Principal amount+Credit RatingCredit spread +
I use the all-in-drawn spread to indicate the pricing of the syndicated loans. I group the factors that could contribute to the spread difference into four categories:
A. Institutional status: institutional status is indicated by Dummy (institution) which equals to 1 when the facility is associated with non-bank institutional lenders and equals to 0 when the facility if not associated with non-bank
19
institutional lenders
B. The loan specific factors: Maturities, Number of participants of the loan, Principal amount.
C. The firm specific factors: Credit ratings3.
D. Market wide factors: The economy condition, the demand and supply of credit. I use the spread between the yields of AAA corporate bond and the yield of US 5-year treasury bonds.
I introduce the t variable to indicated the presence of institutional lenders. The dummy variable equals to 1 when there is at least one non-bank institutional lender is present in the underling loan facilities.
Sirri and Tufano (1998) find that the bigger the size of a fund is, the smaller the cost of searching potential investment target is. Therefore, I include the size of the facilities as one of the independent variables. The size of the facilities is represented by the variable “Principal Amount” which is the natural log of the principal amount.
Ivanshina (2009) argues that the spread of a syndicated loan is affected by how the shares of the loan are allocated. The more dispersedly the shares are allocated, the larger the loan spread is due to information asymmetry. In several previous studies, the percentage of share hold by the leading arrangers is adopted to indicate the dispersibility of the loan allocation. Because there is no
3
Strictly speaking, the credit ratings I used in this research are the ratings for the loan facilities instead of the firms. I consider the loan ratings reflect both company specific risks which are determined by the borrowers’ financial conditions and the covenants of the respective loans.
20
information about how the loan facilities are allocated among the participants in Thomson one database, I use the natural log of number of participants as a substitute of leading arranger share. The higher the number of participants, the more diversified the facility is.
In previous researches (Ivashina & Sun, 2011; Lim, Minton & Weisbach, 2012), industrial ROA, numbers of analysts following, Z score, debt to asset ratio, etc. are used as the indicators of company specific risks. Instead I apply credit rating of loans from Thomson one data base to indicate the credit risk. I use the one-year default probability to quantify the credit rating. The credit spread is introduced to represent the demand and supply of credit and the general economy conditions in the periods when the loans were issued.
Because the sample contains 2361 unrated loans for which the credit rating information is not available. I estimate equation (2) for the entire sample without considering the independent variable “Credit Rating”. Then I estimate equation (1) for all the rated borrowers.
All-in-drawn spread= c +Dummy(Institution)Maturity+Participants +*Principal amount+Credit spread +
In case there is heterogeneity among the different credit rating syndicated loan facilities, I estimate equation (3) for facilities over Brated, Brated, BB-rated, B-rated and below B-rated separately.
All-in-drawn spread= c +Dummy(Institution)Maturity+Participants +*Principal amount +Credit spread +
21
If the spreads differences between institutional facilities and bank facilities are entirely due to adverse selection, the coefficient should not be statistically significant. The spread difference would be explained by the firm specific factors, loan specific factors and market wide factors. But there is a possibility that not all factors are included in the estimation. The possibility of missing explanatory variables is very high. If I include all the factors that could have impact on the loan spreads and the coefficient for the institutional dummy is statistically significant, the spread differences can be explained by the information asymmetry theory.
3.3. Within-issuance spread gap test
In order to eliminate the effect of company specific factors and market wide factors, I conduct the within issuance spread difference study. I define a “issuance” as one package of syndicated loan facilities issued by a same borrower on the same day. One issuances may consist of one tranche of loan facility or multiple tranches of loan facilities. Each of the tranche can have different interest rates, different principal sizes, different maturities, different collaterals, different covenants and different lenders. Because all the tranches in the same loan are issued by the same borrower and under the same market condition, I assume the company specific risk and market risk should be the same. Therefore, the spread gaps among the tranches of the same issuance can be the result of different loan specific factors such as maturity, principal size,
22
collateral, number of participants and different loan covenants, etc. It can also be explained by the presence of the institutional lenders who usually require higher returns. Nandy and Shao (2010) and Ivanshina (2009) both state that the companies have difficulties borrowing from banks rely on non-bank institutional investors who are seen as the last resort. The extra spreads charged by the non-bank institutional lenders are considered as the compensate to the extra liquidity they provide to the lower quality loans.
In my sample, there are 1893 issuances which have more than one tranche. I put the tranches within the same issuance in alphabetical order4. According to the pricing process of the syndicated loan mentioned in 2.2, I believe the investors in lower tranches have less information than investors investing in higher tranches. I calculate the spread gaps between the adjacent tranches. The definition of spread gap is as follows.
Spread Gap n-m = LN (spread n /spread m) (4) For two adjacent tranches involving one institutional tranche and one non-institutional tranche, I always put the spread of non-institutional tranche in the numerator and the spread of non-tranche in the denominator. In this way I can identify the institutional status differences. To do so, I introduce a dummy variable to indicate the institutional status. For two adjacent tranches involving one institutional tranche and one non-institutional tranche, the dummy equals to 1. Because I always put institutional tranche before the non-institutional tranche
4
Thomson one database provides tranche ID for each tranche such as tranche A, tranche B, tranche C. The leading arrangers usually invest in tranche A.
23
when make the comparison, the dummy of 1 means tranche n is institutional while tranche m is not. When the two tranches are both institutional or non-institutional, the dummy equals to 0. Then I introduce other four variables to indicate the difference between two adjacent tranches. I estimate the following equation:
Spread Gap n-m= c + Maturity n-m + Principal n-m + No. Participants n-m +
Dummy (tranche order n-m) +Dummy(Institutional Status n-m) + (5)
Maturity n-m: LN (Maturity n /Maturity m)
Principal n-m: LN (Principal n /Principal m)
No. Participants n-m: LN (no.Participants n /no.Participants m)5
Dummy (tranche order n-m): Dummy equals to 1 when tranche n has lower order than tranche m. Dummy equals to 0 when tranche n has higher order that tranche m.
I construct a sample of 2486 observations. I want to know whether the non-bank institutional tranches charge higher spreads than the bank tranches even though they were issued by same borrowers and under the same market conditions. I control other factors that could affect the loan spreads within the issuance such as the maturity difference, principal size difference, tranche order difference. I expect the coefficient for Dummy(Institutional Status) to beta statistically significant greater than 0.
5
Here the definitions of Maturity and Principal are different from the ones Lim, Minton, & Weisbach(2012) give. They use the differences of logs to indicate Maturity and
24
3.4. Combine the results of cross-loan test and within-issuance test
The purpose of the first test, the cross-loan test, is to test whether institutional lenders charge higher loan spread even if holding the loan specific factors, firm specific factors and market conditions constant. In another word, adverse selection cannot be used to explain the extra spread for institutional facilities. To further eliminate the unobservable firm specific factors and market condition factors, the within-issuance test is conducted to further prove the loan spread premium are not caused by adverse selection but by other factors stemmed from the special nature of institutional lenders and the syndication pricing process.
4. Result and Interpretation
4.1. Cross-loan test
Exhibit 2 presents the results of estimates for equation (1), (2) and (3). Column (1) shows the OLS estimate coefficients and p-values of equation (2) for the entire sample of 5601 facilities. In this column, the coefficient for institution status indicator is 207 and is significantly different from zero. This result indicates that holding market conditions and loan specific conditions constant, loan spread of non-bank institutional facilities are 207 basis points higher than that of bank facilities. 207 basis points spread difference is significant given the average loan spread of the bank loan facilities is only 246 basis points. However, this regression does not consider firm specific factors.
25
Column (2) shows the OLS estimate coefficient and p-value of equation (1) for 3241 facilities issued by rated companies. Equation (1) take credit risk into consideration which is represented by default rate probabilities. The coefficient of institutional status indicator is 177 basis points which means keeping firm specific factors, loan specific factors and market factors stable, the loan spreads of institutional loan facilities are 177 basis points higher than that of the bank facilities. This result can be the evidence that institutional investors are less informed compared to banks. The extra spreads are charged to compensate the costly information processing work. The coefficient of natural log of maturity is 38 which is statistically significant different from zero. This is coherent with the term structure of interest rate. The coefficient of natural log of number of participants is -41 which is significant different from zero. The negative coefficient shows that when the more loan participants, the higher the interest rate is. The nature of syndicated loans is to share risks among several borrowers. At the same time the information asymmetry exists between leading arrangers and other participants. I perceive the coefficient of natural of number of participants as the equilibrium of diversification effect and information asymmetry effect. The coefficient of natural log of principal amount is -32 and it is statistically significant. The negative coefficient indicates that larger loans are safer and require less loan spread. As I expected, the coefficient for credit spread which is the indicator of market conditions is 0.63. The loan spreads are positively related to the market average credit spread for AAA-rated corporate loans. The coefficient of default
26
probability is 19.61 which is not significant different from zero.
Column (3), (4), (5), (6) and (7) show the OLS estimates of equation (3) for loans with different ratings separately. The coefficients of the institutional indicator and credit spread for each of those columns are all positive and statistically different from zero. There are only 72 loan facilities are issued by companies with ratings over BBB. For those loan facilities, only the coefficients of the institutional indicator and credit spread are significantly different from zero. This means the spread of high quality syndicated loans are mainly explained by the presence of institutional investors and the macroeconomic conditions. There are 597 loan facilities issued by BBB-rated companies. For BBB-rated loan facilities, the coefficients of institutional indicator, natural log of maturity and credit spread are positive and statistically different from zero. There are 1410 BB-rated loan facilities. All the coefficients of the five independent variables are significantly different from zero. The coefficients of institutional lenders, natural log of maturity and credit spread are positive. The other two variables are negatively correlated with the loan spread. For 1035 B-rated facilities, only the coefficients of institutional indicator, natural log of maturity and credit spread are significantly different from zero. For the 86 loan facilities issued by companies with below B-rating, the coefficients of institutional status, natural log of maturity, the natural log of number of participants and credit spread are significantly different from zero. However, facilities with below B ratings, the loan spread is negatively correlated to maturity because the coefficient of natural log of maturity is -88.
27
Maybe for companies with bad financial situations the short term liquidity is essential and the liquidity spread is higher for short term loans and lower for long term loans.
4.2. Within-issuance spread gap test
Exhibit 3 shows the result of OLS estimates for equation (5). The within issuance spread gap study is conducted to eliminate the firm specific factors and market wide factors when estimate the factors affecting loan spread with the same issuance. The coefficient of dummy variable indicating the difference in institutional status is positive and significantly different from zero. This results indicates that within the same issuance, the tranche financed by at least one institutional lender has higher loan spread than the tranches financed only by banks keeping the loan specific factors constant. The coefficient of Maturity n-m is positive and statistically significantly different from zero. This means within the same issuance, the tranche with longer maturity charges higher interest. The coefficient of Principal n-m is negative and statistically significantly different from zero. This means within the same issuance, the tranche with larger amount of principal has lower loan spread. The coefficient of Number of participants n-m is negative and significantly different from zero indicating that within the same issuance, the tranche financed by larger number of lenders has lower loan spread. The coefficient for Dummy (tranche order n-m) is not significantly different from zero. The tranche order cannot reflect the contractual difference between tranches if there is any.
28
5. Limitations of the Current Research
Data limitation is the main challenge for this research. I use proxy variables that are closely related to the real variable of interest. However, the proxy variable may not be the perfect representatives of certain factors. For example, I use the default probabilities of the respective borrowers provided by Standard Poor to represent the firm specific factor. Firstly, this variable can only be regressed for companies with credit ratings. Secondly, borrowers should price the loan based on ex ante default probabilities but instead the default probabilities are ex post. Another example is maturity. I calculated the maturity by subtracting the dates of maturity by dates of initiation. However, many syndicated loans are bullet which means they are amortized along time. Thus the maturity I used here is not a perfect representative of the duration of the loan. In addition, the number of participants in certain facility is the proxy variable for the concentration of shares with the facility. However, the number of participants does not indicate how the shares of the loans are allocated within certain facility.
Another limitation of this research is that it does not consider the ownership relationship the bank institutional lenders have with the borrowers. The non-bank institutional investors sometimes charge less loan spread when they hold the equity of the borrowers because they are compensated by the private information about the firm.
29
6. Potential Extensions of the Current Research
There are several ways to extent the current research. First of all, the default probabilities can be replaced by other variables such as Z-score and debt ratio to represent the firm specific factors. The firm specific risks may also vary among different industries. The industrial ROA, industrial leverage ratio and industry beta can be introduced as independent variables to explain the spread differences among facilities. Secondly, with regard to the loan specific factors, the covenants like whether the loan is renegotiable or not should be considered. In order to reflect the actual durations of the loans, researchers should consider amortization when calculating the maturities of the loans. If the data is available, the percentage of share hold by institutional investor can serve as one of the variables to represent the loan specific risks. In banking industry, the clients and banks usually have long term relationships and the banks tend to possess more private information about their new clients. Whether the lenders have previous relationship with the borrowers can also be factored into the estimate.
Last but not the least, there are several different types of non-bank institutional lenders such as insurance companies, private equities and hedge funds. Those institutions have different risk appetite and different way of assessing the risks. Future researches can distinguish the different type of institutions when making the regressions.
30
Conclusion
In the paper, 5601 loan facilities issued by US public companies in the period from 2007 to 2017 are studied. Among the 5601 loan facilities, 1207 loan facilities are financed by at least one non-bank institutional lenders. The untraditional players are becoming more and more active in the syndicated loan market. According to my research, on average the institutional facilities have higher loan spread than the bank facilities. Holding firm specific factors, market wide factors and loan specific factors constant, the institutional loan facilities still charge significantly higher spread than the similar bank facilities. This result applies for facilities with different ratings and unrated facilities. This can be explained by the information asymmetry theory. Unlike banks, the institutional investors usually do not have as much private information about the borrowers and previous relationships with the borrowers as banks do. The extra spreads charged by institutional investors are the compensation for the extra due diligence cost. The result of within issuance spread gap study shows that even the facilities are issued under same market condition and by same borrowers, the institutional facilities still charges higher spreads. This result can be explained by the two-stage pricing mechanism of syndicated loans. If in the first two-stage, the loan is undersubscribed, then it enters the “retail biding” stage where the lead arrange will raise the loan spread to attract investors with higher required rate of return and higher risk tolerance.
31
to several questions concerning the spread differences between the institutional loan facilities and bank facilities. On the contrary to some previous researches arguing that the spread differences are caused by the adverse selection, this research finds out that the spread differences are due to institutional lenders’ information disadvantage. At the same time, the spread differences are also the consequences of the syndication pricing mechanism.
32
Exhibit 1
Spread between yield of S&P AAA bond and the yield of US treasury bondYear/Month Credit Spread (bps) Year/Month Credit Spread (bps)
2007/12 97 2013/1 119 2008/1 111 2013/2 111 2008/2 119 2013/3 125 2008/3 132.1 2013/4 104 2008/4 126 2013/5 145 2008/5 111 2013/6 136 2008/6 105 2013/7 117 2008/7 115 2013/8 109 2008/8 123 2013/9 111 2008/9 145 2013/10 112 2008/10 188 2013/11 125 2008/11 189 2013/12 106 2008/12 194 2014/1 83 2009/1 176 2014/2 118 2009/2 158 2014/3 111 2009/3 187 2014/4 90 2009/4 200 2014/5 101 2009/5 156 2014/6 115 2009/6 117 2014/7 103 2009/7 121 2014/8 105 2009/8 99 2014/9 119 2009/9 110 2014/10 104 2009/10 133 2014/11 109 2009/11 114 2014/12 107 2009/12 131 2015/1 98 2010/1 84 2015/2 152 2010/2 110 2015/3 113 2010/3 102 2015/4 123 2010/4 86 2015/5 136 2010/5 73 2015/6 145 2010/6 96 2015/7 117 2010/7 114 2015/8 142 2010/8 83 2015/9 135 2010/9 128 2015/10 134 2010/10 142 2015/11 135 2010/11 138 2015/12 129 2010/12 124 2016/1 125 2011/1 101 2016/2 145 2011/2 102 2016/3 140 2011/3 101 2016/4 128 2011/4 102 2016/5 121 2011/5 95 2016/6 114 2011/6 128 2016/7 130 2011/7 81 2016/8 140 2011/8 76 2016/9 139 2011/9 106 2016/10 139 2011/10 154 2016/11 151 2011/11 121 2016/12 112 2011/12 115 2017/1 103 2012/1 121 2017/2 103 2012/2 125 2017/3 108 2012/3 123 2017/4 104 2012/4 100 2017/5 102 2012/5 108 2017/6 97 2012/6 153 2017/7 97 2012/7 113 2017/8 92 2012/8 127 2017/9 102 2012/9 121 2017/10 87 2012/10 109 2017/11 86 2012/11 104 2017/12 95 2012/12 127
33
Exhibit 2
This table presents the OLS regression coefficient estimates of Equation (1), (2) and (3) and corresponding values. Column (1) shows the OLS estimate coefficients and p-values of equation (2) for the entire sample of 5601 facilities. Column (2) shows the OLS estimate coefficient and p-value of equation (1) for 3241 facilities issued by rated companies. Column (3), (4), (5), (6) and (7) show the OLS estimates of equation (3) for loans with different ratings separately. ***, **, * correspond to statistical significance at the 1%, 5%, and 10% level, respectively.
Dependent variable= All-in-drawn spread (bps)
All facilities Rated facilities Above BBB-rated facilities BBB-rated facilities BB-rated facilities B-rated facilities Below B-rated facilities (1) (2) (3) (4) (5) (6) (7) Coef. (p-value) Coef. (p-value) Coef. (p-value) Coef. (p-value) Coef. (p-value) Coef. (p-value) Coef. (p-value)
Institutional facility flag 207.5990***
(0.0000) 176.9275 *** (0.0000) 202.2139*** (0.0000) 135.5730*** (0.0000) 81.37887*** (0.0000) 166.6887* (0.0553) 55.38980* (0.0985) LN(maturity) 8.130430 (0.2450) 37.67202 * (0.0784) 1.937494 (0.8205) 14.66203*** (0.0059) 11.70929** (0.0703) 152.5498** (0.0602) -88.38449** (0.0265) LN(number of participants) -38.09839*** (0.0086) -40.95259* (0.0982) 3.226161 (0.7322) 0.830287 (0.1426) -17.11898*** (0.0000) -43.19392 (0.3690) -98.88737** (0.0164) LN( principal amount) -31.54011*** (0.0002) -31.88770** (0.0399) 2.796459 (0.5478) 0.521158 (0.8393) -5.321024** (0.0151) -43.95731 (0.2749) 37.61126 (0.1313) Credit spread (bps) 0.479470 * (0.0996) 0.629802 * (0.0562) 0.69927*** (0.0030) 0.452720*** (0.0001) 0.141858* (0.0964) 1.149519** (0.0369) 1.904205* (0.0755) Default probability (%) 19.60528 (0.1187) Number of observations 5601 3241 72 597 1410 1035 86 Adjusted R2 0.39603 0.38614 0.394115 0.238568 0.441936 0.25375 0.135632
34
Exhibit 3
This table shows the OLS regression coefficient estimates for equation (5). ***, **, * correspond to statistical significance at the 1%, 5%, and 10% level, respectively.
Dependent variable=Spread gap =LN(spreadn/Spreadm)
All issuance with multiple tranches (8)
Coef. (p-value) Dummy(Institutional Status n-m) 0.184097 ***
(0.0000) Maturity n-m: LN (Maturity n /Maturity m) 0.160147 ***
(0.0000)
Principal n-m: LN (Principal n /Principal m) -0.016520 ***
(0.0005)
No. Participants n-m:LN (no.Participants n
/no.Participants m)
0.004125 **
(0.0900)
Dummy (tranche order n-m): -0.026090
(0.3234)
Number of observations 2486
35
References
Allen, L., Saunders, A., 2002. A Survey of Cyclical Effects in Credit Risk Measurement Models. NYU Working Paper No. S-CDM-02-04.
Barnish, K., Miller, S., Rushmore, M., 1997. The New Leveraged Loan Syndication Market. Applied Corporate Finance 10, 79-88.
Brophy, D., Ouimet, P., Sialm, C., 2009. Hedge Funds as Investors of Last Resort? The Review of Financial Studies 22, 541-574.
Duffie, D., Singleton, K., 1999. Modeling Term Structures of Defaultable Bonds. The Review of Financial Studies 12, 687–720.
Holmstrom, B., 1979. Moral hazard and observability. Journal of Economics 10 (1), 74-91.
Ivashina, V., 2009. Asymmetric Information Effects on Loan Spreads. Journal of Financial Economics 92, 300-319.
Ivashina, V., Sun, Z., 2011. Institutional demand pressure and the cost of corporate loans. Journal of Financial Economics 97, 500-522.
Lazear, E., 1986. Salaries and Piece Rates. The Journal of Business 59, 405-431.
Lim, J., Minton, B., Weisbach, M, 2012. Syndicated Loan Spreads and the Composition of the Syndicate. Working Paper, National Bureau of Economic Research.
Nandy, D., Shao, P., 2010. Institutional investment in syndicated loans. Working Paper, York University.
36
Neuhann, D., Saidi, F., 2016. Bank deregulation and the rise of institutional lending. Working Paper, University of Texas.
Peterson, M., R. Rajan, 1994, The benefits of lending relationships: Evidence from small business data, Journal of Finance 49, 3-37.
Sufi, A., 2006. The Real Effects of Debt Certification: Evidence from the Introduction of Bank Loan Ratings, Working paper, University of Chicago GSB. Sufi, A., 2007. Information Asymmetry and Financing Arrangements: Evidence from Syndicated Loans 62, 629-668.
