The transmission of financial shocks: the case of commercial paper dealers during the 2007–2009 crisis Ethan Cohen-Cole∗ University of Maryland - College Park
Judit Montoriol-Garriga Universitat Autonoma de Barcelona
Gustavo Suarez Federal Reserve Board
Jason Wu Federal Reserve Board
October 16, 2011
Abstract This paper analyzes the unprecedented widening of very short term commercial paper spreads as a result of the financial shocks experienced during the 2007–2009 financial crisis. We find that our proxies for rollover risk explain much of the impact of the financial shocks on commercial paper spreads of financial firms while our proxies for credit risk explain much of the spread increases of nonfinancial firms. We also find that firms that used financial-sector dealers to place the paper in the market were more impacted by rollover risk compared to firms that directly placed the paper. These findings contribute to our understanding of the role of financial firms and financial dealers in transmitting shocks to the broader economy and may provide insight into future crisis management. ∗
Ethan Cohen-Cole: Robert H Smith School of Business. email:
[email protected]. (301) 541-7227. Judit Montoriol-Garriga: Universitat Aut` onoma de Barcelona. email:
[email protected]. Gustavo Suarez: Federal Reserve Board of Governors. email:
[email protected]. (202) 452-3011. Jason Wu: Federal Reserve Board of Governors. email:
[email protected]. (202) 452-2556. The views expressed in this paper are those of the authors and do not necessarily reflect those of the Federal Reserve System or its Board of Governors.
1
1
Introduction
Commercial paper spreads widened to unprecedented levels during the 2007–2009 financial crisis. For example, according to data from the Federal Reserve, the spread over the Fed funds rate paid by A2/P2-rated nonfinancial firms to issue overnight commercial paper increased from 60 to 400 basis points during the week following the bankruptcy of Lehman Brothers. The outstanding total commercial paper balance in July 2007 was $2.1 trillion, including $1.2 trillion of asset-backed commercial paper (ABCP). By the end of June 2009, the outstanding balance had dropped to $1.28 trillion, a collapse of about 40 percent of the market. The rise in spreads and decline in volumes have attracted a great deal of attention because commercial paper was generally regarded as a safe asset due to its short maturity and high credit ratings (Kacperczyk and Schnabl, 2010). The unexpected collapse of this market starting in August 2007 has evidenced the weaknesses of short-term debt instruments and their inherent rollover risk. Furthermore, the collapse of the commercial paper market has highlighted the various channels through which disruptions in financial markets can affect the broader economy.1 Although the unprecedented widening of commercial paper spreads since August 2007 has been extensively documented (e.g., Brunnermeier, 2009; Acharya et al., 2010), there is still a lot to be learned from studying its contributing factors. Our paper has two main objectives. First, we aim to understand the factors driving the changes in the very short term commercial paper spreads around three large shocks to the financial system: (i) investors’ realization of the large scope of subprime-related losses in bank balance sheets and the asset-backed commercial paper market freeze in August 2007 (e.g., Acharya et al., 2010); (ii) the distressed purchase of Bear Stearns in March 2008; and (iii) the bankruptcy of Lehman Brothers in September 2008. 1
On top of first-order effects that impact the cost and availability of borrowing to firms that rely on commercial paper, the collapse of this market may have second-order effects, as it is widely thought that firms with access to commercial paper redistribute the credit they receive to credit-constrained firms via trade credit. For instance, Calomiris, Himmelberg, and Wachtel (1996) show that during downturns trade credit flows from firms having access to the commercial paper market to firms without access to this market.
2
Second, we seek to explain the channels through which shocks to financial firms are transmitted to the broader economy. We exploit cross-sectional variations on issuers’ characteristics to analyze changes in commercial paper spreads across issuers’ types. We distinguish between (i) financial and nonfinancial issuers and (ii) issuers that place the paper through dealers and issuers that place the paper directly to the market. Dealers are financial institutions that market paper for issuers and typically provide secondary-market liquidity.2 Our paper analyzes the role of financial firms and dealers in transmitting shocks from the financial sector to the broader economy. We consider two key factors that affect investors’ perceived probability of prompt repayment in the comercial paper market. The first factor is based on expected default of the issuer and is expected to capture the default component of the spreads. We use 5-year Credit Default Swaps (CDS) spreads as our main proxy for credit risk, and Expected Default Frequency (EDF) from Moody’s KMV to check the robustness of our results. The second factor is based on rollover risk of issuers of short-term debt and is expected to capture the non-default component of the spreads. We measure it as the fraction of commercial paper maturing in the upcoming week. Although it is never completely possible to empirically disentangle solvency and liquidity risk, in this paper we are solely interested in assessing the relative importance of each measure in explaining the widening of spreads after financial shocks. We provide evidence that the widening of spreads in the commercial paper market is a combination of increased rollover risk and credit risk for the average issuer. However, we also find important cross-sectional differences depending on whether the issuer is a financial or a nonfinancial firm, and whether the issuer exclusively uses dealers or it directly places the paper to the market. On the one hand, we find that credit risk is the most important factor explaining the widening of spreads of nonfinancial firms. Rollover risk is not significant in August 2007 and March 2008. Only after the bankruptcy of Lehman Brothers in September 2008 did rollover risk become a significant factor 2
In August 2006, about 80 percent of commercial paper was issued by dealers and the remaining 20 percent was directly placed (Stigum and Crescenzi, 2007).
3
for nonfinancial firms.3 Regarding financial firms, we find that commercial paper spreads increased mainly because of rollover risk during the crisis. On the other hand, the analysis of dealer versus non-dealer users shows that the increase in commercial paper spreads of dealer users is mostly due to rollover concerns, a result consistent with the notion that dealer placement exposes commercial paper investors to additional counterparty and settlement risks of the dealer during rollovers. In contrast, spreads of issuers that have some paper placed directly to the market are generally not affected by rollover risk. These results highlight the important role of financial firms and financial dealers in transmitting financial shocks to the broader economy. The salience of the transmission mechanism in this crisis4 and other crises5 has been widely discussed. Our results draw attention to a transmission channel that has been largely neglected, which is that many commercial paper issuers use dealers to place the paper. The disintermediation of financial-service firms and direct access to capital markets by large companies impliy that those firms that were not dependent on dealers could, and did, avoid a substantial portion of the impact of this crisis. This analysis has interesting implications for the debate on the importance of the financial sector for the real economy. Given that the initial financial shock was concentrated in the banking sector, one could expect that firms that depend on bank financing would be negatively affected. Indeed, the continued economic sluggishness more than three years after the initial subprime shocks of August 2007 has repeatedly been attributed to the lack of debt financing from the banking sector.6 In this 3
Our emphasis here is on the short-term effects of the financial shocks on the funding cost of firms. As the recession continued, many parts of the economy were impacted by a myriad of factors. Our empirical approach described below is designed to isolate the short-term impact of the large financial shocks on commercial paper spreads. 4 The degree of transmission has been widely debated. Chari, Christiano, and Kehoe (2008) argue that there was little evidence of impact on nonfinancial firms in October of 2008. Cohen-Cole et al. (2008) respond that evidence of transmission was indeed available at the time of the crisis. Bates, Kahle, and Stulz (2009) point out that nonfinancial firms held sufficient cash stockpiles prior to the crisis to have paid off existing debt, obviating the need for financing. 5 Peek and Rosengren (2000) find evidence for the Japanese financial crisis spillovers into the US economy; Dell’Ariccia, Detragiache, and Rajan (2008) find evidence in a wide range of crises; Khwaja and Mian (2008) find evidence from an emerging market; Chava and Purnanandam (2011) find evidence fom the Russian crisis. Borenzstein and Lee (2002) discuss Korea. 6 See Cornett et al (2010), Duchin et al. (2010), and the press (e.g., CNBC, May 27, 2010.)
4
paper we present evidence that even firms that are not dependent on traditional bank financing but use financial dealers may be impacted by financial shocks through reduced access to capital markets as these dealers become imperiled. It is important to understand the relative scale of rollover risk versus credit risk in commercial paper spreads for at least two reasons.7 First, it matters for investors’ portfolio-allocation decisions according to their investment horizon. The increase in spreads may represent increased compensation demanded by investors as the default risk on the issuer increases (credit risk), or, alternatively, could represent the premium paid by the issuer because of the risk that it may not be able to refinance its due debt (rollover risk). Second, understanding the roles of credit and rollover risk is important for designing efficient policy responses to market disruptions. If rollover risk is the primary problem, then the general consensus is that policymakers should direct efforts to restoring market confidence to make markets more liquid to allow issuers to refinance their liabilities.8
If
credit risk is the main determinant of spread-widening, then policymakers can either improve the solvency of the counterparties involved or allow market forces to reallocate credit independently. In theory, in the absence of rollover concerns, the scope for intervention is more limited. In practice, the distinction between these two explanations is not clear-cut, since increased rollover risk associated with liquidity shortages may be a by-product of credit problems (the other way around may also be true), and the solution is no longer as simple as restoring confidence. The rest of the paper is organized as follows. Section 2 briefly reviews the related literature. Section 3 provides the relevant institutional background about the commercial paper market, focusing mainly on the unsecured segment of the market.
In section 4 we describe our data in
detail. Section 5 describes our empirical strategy and section 6 presents our main results. Section 7
Our main proxy of credit risk is the CDS spread of the issuer. Even though CDS spreads are an imperfect measure for credit risk, they are widely acknowledged as a decent first-order approximation of the credit risk of an issuer. See Schwartz (2010) for a discussion and a novel method of further decomposing credit spreads into credit and liquidity components. 8 See Allen, Carletti, and Gale (2009) and Krishnamurthy (2010) for examples and literature review.
5
7 discusses robustness checks and section 8 concludes.
2
Related Literature
Our paper contributes to a large and growing literature that analyzes the functioning of financial markets and their inherent risks during the 2007–2009 financial crisis. Our paper focuses on the commercial paper market and we show that the principal risk faced by financial firms was their inability to rollover their short-term debt. Our results complement those of Afonso, Kovner, and Schoar (2010) and Cornett et al. (2010). Afonso, Kovner, and Schoar (2010) document that counterparty risk played a much more important role than liquidity hoarding in the interbank market during the 2008 financial crisis. Their interpretation of the results is that lenders in the Fed funds market were able to screen out the worst-performing borrowers. Our results suggest that investors in the commercial paper market demanded a higher premium to invest in financial commercial paper due to increased rollover risk and not so much to credit risk. However, it is important to note that the difference in results may be due to the fact that commercial paper and Fed funds are different markets, and the financial firms participating in each one may differ. Cornett et al. (2010) find that banks’ efforts to manage liquidity shocks during the crisis resulted in a reduction in credit supply to the real economy. We point to evidence that this reduction may have occurred only after October 2008, and only to a subset of firms. Our results are also related to recent studies about the role of rollover risk in financial markets. Acharya, Gale, and Yorulmazer (2010) study the role of rollover risk on market freezes. He and Xiong (2011) study the link between credit risk and rollover risk using a structural credit risk model. In the empirical front, Hu (2010) investigates the role of the refinancing of long-term debt in explaining changes in CDS spreads during the financial crisis. An extant literature and press reports have emphasized the crucial nature of the financial sector
6
for promoting economic growth. Indeed, the work of Bernanke et al. (1999), among others,9 has contributed to our understanding of financial linkages to the real sector, in ways that were not possible within classical macroeconomic models.
Empirical evidence for prior crises shows that
bank distress is transmitted into the real economy. Additionally, there is a growing literature on the empirical implications of the 2007–2009 financial crisis that suggest a real-economy link.10 However, there are relatively few papers that seek to understand the nature, timing, and size of the transmission.11 This paper contributes to this literature by showing spillover from the financial sector to the broader economy through the sensitivity of commercial paper spreads for issuers that depend on dealers to place their paper to rollover risk.
3
Background on the Commercial Paper Market
Commercial paper is a form of borrowing with a fixed maturity, typically between 1 and 270 days. The paper is issued by banks, large corporations, and special-purpose vehicles to meet short-term financial obligations (including operational needs, such as payroll), or to finance the purchase of assets. Commercial paper is viewed as an inexpensive funding option by banks and corporations alike. Prior to the crisis, it was relatively simple for large corporations to access the commercial paper market, and regular funding could be obtained at rates lower than bank loans. However, many issuers maintain a bank line of credit to provide additional access to capital and as a backstop 9
For recent examples, see Cohen-Cole and Martinez-Garcia (2009), Faia and Monacelli (2007), Curdia and Woodford (2010), De Fiore and Tristani (2009). 10 Ivashina and Scharfstein (2010) find that new loans to large borrowers fell during the crisis and provide some evidence that liquidity-constrained banks may have been more likely to cut lending. Campello, Graham, and Harvey (2010) use a CFO survey to find that many corporations cut spending and investment and attribute those cuts to inability to obtain financing. Duchin, Ozbas, and Sensoy (2010) find that corporate investment declined as a result of the crisis and provide evidence that the decline was a result of reduced access to funding. Additionally, a couple of recent papers analyze commercial paper during the crisis, but come to different conclusions. Kacperczyk and Schnabl (2010) find that the market faced a generalized collapse, and Gao and Yun (2009) find that declines in commercial paper use were concentrated among low credit quality firms. Both analyze commercial paper at a somewhat higher level of aggregation than our data permit. 11 A recent exception is Tong and Wei (2009).
7
liquidity facility to the commercial paper program. Commercial paper can broadly be classified according to the presence of collateral, the type of issuing firm, and the issuance channel.12 Our paper focuses on unsecured (uncollateralized) commercial paper from different types of firms and issued across different channels. By studying unsecured paper (instead of collateralized paper), we are able to isolate the impact of the credit quality of the issuer without explicit modeling of the underlying asset quality.
That is, we can
rely on credit-quality estimates of the corporate commercial paper sponsor as a reasonable proxy for credit-worthiness. A comprehensive overview of the asset-backed commercial paper (ABCP) market and its operation during the crisis can be found in Covitz, Liang, and Suarez (2009). Within the commercial paper markets, we will look at the issuance of commercial paper across both financial and nonfinancial firms. Nonfinancial and financial corporations alike use commercial paper as a typically highly liquid and readily accessible source of short-term funding. Anecdotally, these markets grew in popularity both because borrowing costs were somewhat lower than equivalent corporate loans and because they are more flexible for managing short-term or cyclical funding needs. The use of commercial paper by nonfinancial and financial firms differs substantially. Nonfinancial firms typically use commercial paper to fund on-going cash-flow needs such as payroll and inventories.
As a result, interruption of access to these markets could lead to an inability
to manage regular operations and cause layoffs and supply disruptions, which could potentially lead to severe disruptions to the real economy. Financial firms use commercial paper to manage short-term liquidity needs, including maturity mismatch in assets and liabilities, or in some cases to fund large portions of their balance sheets. With respect to the methods of issuing paper, an issuer can either sell securities directly in the money markets or it can sell them through a dealer, in exchange for a fee. The dealer subsequently sells the paper to ultimate investors, such as money market mutual funds. Dealers are typically 12
See Hahn (1998) and Stigum and Crescenzi (2007) for descriptions of the commercial paper market.
8
large banks, including the investment banks that existed prior to the financial crisis. The benefit of direct issuance is the ability to receive the full market price for the securities by saving on fees. However, working through a dealer institution could facilitate issuance if the institution has strong distribution networks. We will highlight this distinction throughout our empirical analysis. A buyer of corporate commercial paper from a dealer assumes the risk of corporate default during the term of the commercial paper as well as the counterparty risk of the dealer during the settlement period between purchase and final settlements. This risk prior to the crisis was negligible, but during the crisis it became large. As a result, issuers that relied on dealers faced the possibility that liquidity concerns at the dealer banks would impact the spreads on their own issuances. This paper provides empirical support for the importance of this link.
4
Data
Our initial dataset includes 17,408 primary market issues of unsecured commercial paper (CP) in the U.S. market between January 1, 2007, and December 31, 2008. This includes information on 101 programs or issuers.
These data are from the Depository Trust and Clearing Corporation
(DTCC), the agent that electronically clears and settles directly placed and dealer-placed commercial paper. The issues in the sample are discount instruments paying face value at maturity. For each transaction, DTCC provides the identity and industry of the issuer, the face and settlement values of the transaction, and the maturity of the security. Using this information, we calculate yields on new paper based on the assumption of a 360-day year. We also obtain from DTCC a separate weekly file that contains program-level information on the maturity distribution of outstanding paper. Further, we supplement the DTCC data with information on credit ratings from Moody’s Investors Service and Standard and Poor’s; on expected default frequencies from Moody’s KMV; and on CDS prices from Markit Partners.
9
The dependent variable in our analysis is the spread over the Fed funds rate for commercial paper maturing in 1–4–days. We choose the shortest-term commercial paper to ensure that we can evaluate rollover risk and credit risk without concern for yield curve impacts, duration, or other issues like term premia. Although one may argue that only liquidity risk matters for such short maturities, Covitz and Downing (2007) show that, even for very short term spreads, credit risk is important in the commercial paper market. Another practical reason for using the shortestmaturity paper is that issuance of longer-dated commercial paper was most affected during the financial crisis, and only the very high quality firms were able to issue longer-term paper. Therefore, issuer selection is least severe for the shortest-maturity paper. Finally, 1–4–day issuances typically account for more than half of daily aggregate issuance. Our key explanatory variables are issuer-level measures of credit risk and rollover risk.
For
our credit risk measure, we obtain 5-year CDS spreads13 for each commercial paper issuer from Markit Partners. A CDS spread is a fraction of the CDS contract notional paid from the buyer to the seller in return for credit protection on debt obligations, in case the issuer defaults. For each day, we average CDS spreads of senior unsecured debt across four primary contract types.14 CDS spreads are generally regarded as real-time market-based measures of credit risk. While there is some evidence that they contain a liquidity component, its importance appears to be relatively small in magnitude (Tang and Yan, 2007; Lin, Liu, and Wu, 2009).
In section 7 we check the
robustness of our results to the use of an alternate measure of credit risk, the Expected Default Frequency (EDF) from Moody’s KMV. 13
We use 5-year CDS instead of CDS spreads at other maturities because the daily time series of the 5-year contracts is the most complete for the issuers in our sample. 14 The four classes are the standard International Swap and Derivatives Association (ISDA) classifications. Class 1 is “No Restructuring”contracts: this option excludes restructuring altogether from the contract, eliminating the possibility that the protection seller suffers a “soft” credit event that does not necessarily result in losses to the protection buyer. Class 2 are “Full Restructuring” contracts. These allow the protection buyer to deliver bonds based on any debt restructuring. Classes 3 and 4 are “Modified Restructuring” and “Modified Modified Restructuring” contracts, respectively. These are full restructuring contracts that limit the bonds that can be delivered to ¡30 months and ¡60 months, respectively.
10
For our rollover risk measure, we calculate the fraction of commercial paper maturing in the upcoming week using information provided by DTCC. This measure provides a proxy for the issuer’s need for liquidity in the short run that reflects the maturity of its liabilities. exposed to rollover risk should pay a higher spread.
Firms
For example, just after the bankruptcy of
Lehman Brothers, controlling for credit risk, firms without commercial paper maturing in the week following the Lehman bankruptcy had lower commercial paper spreads risk compared to firms with a significant portion of commercial paper due. To account for the role of dealer intermediation in this market, we calculate the percentage of each program’s issuances that is placed through a dealer, using data from DTCC. We also subdivide the sample to evaluate the difference between programs that exclusively place their paper through dealers (or “dealer-dependents”) and those that directly sell at least some of their own paper. Table 1 provides summary statistics of the variables for the full sample period (January 2007 to December 2008) and by the three subperiods of our analysis: January to December 2007 (encompassing the crisis in funding markets that erupted in August 2007); January to June 2008 (encompassing the distressed purchase of Bear Stearns in March 2008); and July to December 2008 (encompassing the bankruptcy of Lehman Brothers in September 2008).
Each panel provides
summary information for the core four variables used in the study. We report the mean and the standard deviation for each variable for each subperiod. We also report the number of commercial paper programs (or issuers) as well as the number of observations (or issuances) that appear in each subperiod. [Insert Table 1 here] We observe that the average commercial paper spread over Fed funds increased over time. In our first subperiod (January–December 2007), spreads averaged 10.6 bps across issuers. In the second (January–June 2008) and third (July–December 2008) subperiods, average spreads increased to 22.1 and 80.4 bps, respectively. As expected, the average credit risk measure increases substantially over 11
the three subperiods, in general accord with the increase in commerical paper spreads. However, the measure of rollover risk, the fraction of maturing commercial paper, remains relatively unchanged for the average issuer. Panel B subdivides the data into financial and nonfinancial firms. We note here that financial firms saw a larger increase in credit risk on an absolute basis than nonfinancial firms.
This
observation conforms with the fact that the losses that accumulated in 2007 and 2008 were mostly in the financial sector. In Panel C, we provide summary statistics by the fraction of dealer-placed paper. We note that the spreads for dealer-dependent firms increased, on average, from 14 bps in first subperiod (January–December 2007) to roughly 117 bps in the period encompassing the bankruptcy of Lehman Brothers (July–December 2008). By contrast, the spreads of firms that did not exclusively rely on dealers to place their paper experienced a smaller increase: from 6 bps in January–December 2007 to 44 bps in July–December 2008.
5
Empirical Strategy
The first objective of this paper is to study the contributions of credit risk and rollover risk in the widening of unsecured commercial paper spreads of maturity of 1 to 4 days during the major shocks to the commercial paper market that occurred in 2007–2008.
We consider three major
shocks: the asset-backed commercial paper (ABCP) market freeze in August 2007, the failure of Bear-Stearns on March 14, 2008, and the bankruptcy of Lehman Brothers on September 15, 2008. These shocks preceded some of the sharpest increases in interbank and short-term funding spreads during the financial crisis. For example, the Libor-OIS spread in Figure 1 jumped from less than 10 bps on August 8 to roughly 40 bps on August 9, 2007, as BNP Paribas halted redemptions on affiliated money market funds, citing inability to value their underlying ABS portfolios. The shock to interbank and money markets in August 2007 led to the equivalent of a bank run on asset-backed commercial paper (ABCP) issuers (Brunnermeier, 2009; Covitz, Liang, and Suarez, 2009), most of
12
which were directly or indirectly supported by large commercial banks. Similarly, the collapse of Bear Stearns and the bankruptcy of Lehman Brothers were periods of sharp increases in the cost of interbank funding. [Insert Figure 1 here] The second objective of this paper is to investigate how the importance of credit risk and rollover risk depends on issuers’ characteristics. We exploit cross-sectional variation across financial and nonfinancial issuer, and across the dealer-dependent and non–dealer-dependent issuers. This analysis enables us to identify new channels through which shocks to financial intermediaries are transmitted to the cost of borrowing from non-bank sources. As an illustration, we show in Figure 2 the stylized patterns of commercial paper issuance. The connecting lines represent an active channel, and the black lines across a channel indicate increased borrowing costs via this channel. Panel A shows the pre-crisis situation, where all channels are operational. Panel B shows the situation in a generalized liquidity crisis, where all channels have been severed.
Panel C shows what one would expect if nonfinancial issuers could continue to
issue; effectively the situation in the absence of a financial sector shock transmission mechanism. Panel D shows a situation where financial issuers suffer liquidity shocks and nonfinancial issuers are impacted via dealer networks, but direct market access remains relatively open. [Insert F igure 2 here] We conduct regressions of commercial paper spreads on controls for issuer-level credit risk and rollover risk. Our baseline specification is the following: Spreadit = α + β1 CDSit + β2 Rolloverit + φi + it
(1)
for all issuers i at date t. The variable Spreadit is the commercial paper spread on 1–4–day issues. CDSit is our baseline measure of credit risk, the average 5-year CDS spread, and Rolloverit is our 13
proxy for rollover risk: the fraction of outstanding commercial paper due in the subsequent week. We also include issuer-fixed effects, φi . We cluster standard errors at the issuer level to allow for correlation within issuers over time. A positive coefficient on the CDS variable suggests increased credit risk premia. Analogously, we interpret a positive coefficient on our measure of rollover risk as a heightened sensitivity to short-term funding needs. We expand our baseline specification and conduct a series of event studies–type regressions around the three major financial shocks discussed previously. We divide our data into three subperiods, each surrounding one of the events. For the August 2007 shock that started the crisis, we include data from January to December of 2007. The full year is included here in order to ensure that we have a sufficient baseline of activity leading up to the crisis. For the Bear Stearns collapse on March 14, 2008, we include data from January to June of 2008. This includes both the period of relative calm at the beginning of the year as well as a return to relative calm in mid-year. Finally, for the Lehman bankruptcy on September 15, 2008, we include data from mid-year 2008 until the end of 2008. We run three regressions, one for each subperiod. For each episode, we define a dummy variable that marks the period after the shock. That is, for each subperiod, we define a dummy variable Dtk that is equal to one after the date that marks the shock. k indexes each one of the three events15 : k = 1 : August 9, 2007; ABCP market freeze k = 2 : March 14, 2008; Bear-Stearns collapse k = 3 : September 15, 2008; Lehman Brothers bankruptcy 15
The selection of these three events is somewhat arbitrary. We believe they constitute three major landmarks that changed the credit standards in the commercial paper market. To the extent that these cut-off points are incorrectly identified, that would bias our results against finding any significance.
14
Our expanded specification includes the dummy for each shock Dtk , and their interaction with the credit risk and rollover risk measures: Spreadit = α + β1 CDSit + β2 Rolloverit + β3 Dtk + β4 Dk × CDSit + β5 Dk × Rolloverit + φi + it (2) The main coefficients of interest in our analysis are β4 and β5 , the interactions between our measures of credit risk and rollover risk with the dummy variables for major financial shocks. Next, we divide the sample of commercial paper issuance along the financial/nonfinancial and dealer-dependent/non–dealer-dependent dimensions of the issuer. The first dimension permits us to analyze the differential impacts of credit risk and rollover risk across financial and nonfinancial issuers, comparing their behavior before and after each of the major shocks. The second dimension allows us to shed additional light on the role of intermediation functions on the impact of the crisis. Specifically, we can disentangle whether rollover premia paid by issuers were the result of the need to use a financial intermediary. The argument for the existence of a transmission mechanism is that financial-service firms provide intermediation services with the capital markets. Thus, nonfinancial issuers that do not rely completely on the financial sector for intermediation should be impacted less than those that do.
This argument is different from one citing the traditional channel of
transmission of financial shocks that operates through bank– (or bank loan–) dependent firms. Even firms that are able to fund directly through capital markets, rather than directly through banks, may be affected by disruptions to the financial system.16
6
Results
6.1
Baseline Results
Table 2 reports the results of estimating variations of equation (1).
Univariate regressions in
columns 1 through 4, estimated for the entire sample, suggest that the existence of increased 16
For example, Almeida, Campello, and Weisbach (2004) use bond and commercial paper ratings to classify firms as financially unconstrained, because firms with debt ratings are typically able to issue public debt as an alternative to bank loans.
15
spreads during the financial crisis is consistent with increases in both credit risk and rollover risk. In fact, both credit risk and rollover risk are statistically significantly related with increases in commercial paper spreads. Variation in credit risk appears to explain a larger share of the variation in commercial paper spreads, as measured by the R2 ’s of the regressions. However, rollover risk explains a nontrivial component of the variation in spreads during the crisis. Our measures of credit risk and rollover risk contain independently relevant information, as shown in columns 5 and 6. Both remain statistically significant when simultaneously included in equation (1). The effects of credit risk and rollover risk are statistically significant and economically important. For example, based on the coefficients from the regression in column 6 (equation (1)) that includes issuer-fixed effects, a one-standard deviation increase in credit risk is associated with an increase of 27 bps in commercial paper spreads, and a one-standard deviation increase in rollover risk is associated with an increase of spreads of 6 bps. [Insert T able 2 here] Table 2 estimates variations of equation (1) including and excluding issuer-fixed effects. In subsequent tables, we will use only the specifications that have issuer-fixed effects and robust standard errors. Columns 7 through 9 in table 2 report the results of regression (1) for the three subperiods using our proxies for credit risk and rollover risk, as well as the event dummies. These regressions strongly indicate that each of the three events had a positive and significant impact on commercial paper spreads. The results in column 7 imply an average increase on commercial paper spreads of 22.1 bps after August 9, 2007, with respect to the pre–August 9, 2007, period. Column 8 implies that spreads increased further after the shock on March 14, 2008, by an average of 15.2 bps. The rise was even higher after September 15, 2008. According to our estimates in column 9, average spread increased by 101.9 basis points after the bankruptcy of Lehman Brothers. If our measures of credit risk and rollover risk are significantly correlated, the coefficients on 16
these proxies may not represent the actual relationship between spreads and credit risk or between spreads and rollover risk.17 Table 3 reports regressions of our measure of rollover risk as a function of our measure of credit risk and issuer-fixed effects, for the entire sample (columns 1 and 2) as well as for the three subsamples (columns 3 to 8). The results suggest that, controlling for issuer-fixed effects, variation in CDS spreads is not significantly correlated with our measure of rollover risk, and therefore the proxies for credit risk and rollover risk are likely explaining different components of commercial paper spreads. These results also highlight the importance of controlling for issuer-level fixed effects. [Insert Table 3 here]
6.2
Results by Subperiods
Table 4 reports the results of estimating equation (2), which includes dummies for each of the three shock events, and their interaction with the credit risk and rollover risk measures. [Insert T able 4 here] As shown in columns 1 and 2, the interaction coefficients between our proxies for credit risk and rollover risk with the event dummies are positive, large, and statistically significant for the shocks on August 2007 and March 2008.
As reported in column 3, the interaction of the rollover risk
variable with the dummy representing the shock that followed the bankruptcy of Lehman Brothers is strongly significant and positive.
The magnitude of the coefficient on the interaction of the
shock dummy with our measure of rollover risk is about ten times larger during the September 2008 shock than the corresponding interaction with the dummy during the shock that followed the collapse of Bear Stearns. This finding suggests that, after the bankruptcy of Lehman Brothers, the commercial paper market had become strongly attuned to rollover concerns, and the coefficients 17
For example, the theoretical results in He and Xiong (2011) suggest that rollover risk and credit risk may be correlated.
17
on the interaction between rollover risk and the shock to financial institutions suggest that the effect of rollover risk became very important in explaining increases in commercial paper spreads. One potential confounding factor is the market interventions by the Federal Reserve through the creation of various liquidity facilities aimed at decreasing the credit spreads of commercial paper (for example, AMLF or CPFF). To the extent that these facilities accomplished their goal and ameliorated the conditions in the commercial paper market, our estimates should be interpreted as a lower bound of the effect without intervention.18 The following sections expand the analysis in equation (2) by studying the differential impact of the crisis on the contributions of credit risk and rollover risk across (i) financial and nonfinancial issuers and (ii) dealer-dependent and non–dealer-dependent issuers.
6.3
The Financial Sector
We can move now to investigate the impact of the three crisis events on spreads across types of issuers. As was widely publicized, the crisis manifested first in the financial sector itself. As a result, a key method of determining whether the liquidity shocks associated with rollover risk were contained to the financial sector or were generalized is to observe differences in spread changes across financial and nonfinancial firms. These differences are important for understanding the role of the financial firms in impacting real economic activity.
Referring to Figure 2 again, we can illustrate a basic interpretation of
coefficient combinations across financial and nonfinancial firms. In addition to interpreting the simple rollover and credit coefficients, the schematic infers from the combination of coefficients the possible interaction between the two sectors. Specifically, sets of positive coefficients for all combinations imply, as in table 4, that the economy faced generalized distress. However, mixed 18
An alternative interpretation is that the Federal Reserve intervention actually aggravated the credit conditions on these markets. Nevertheless, existing evidence does not seem to support this view. See for example Duygan et al. (2010) who analyze the effectiveness of AMLF and conclude that it helped loosening the credit conditions in money markets.
18
sets of coefficients imply differential responses to the crisis. Table 5 reports the results of estimating equation (2) on each of the three subperiods separately for financial and nonfinancial firms. For ease of comparison, table 5 also includes the results of estimating the regressions including all issuers. As in the previous tables, we show the results for each of the three events separately, arranged in chronological order. [Insert T able 5 here] The first interesting finding in table 5 is the difference in the significance of the interaction terms coefficients across the financial and nonfinancial issuer regressions. Rollover risk is generally significant for financial issuers, while credit risk explains most of the increases in spreads for nonfinancial issuers. This implies that the impact of the crisis was different across sectors; that is, the market was able to distinguish the impact of the crisis with some degree of precision across sectors—even during the shock that followed the Lehman bankruptcy. Focusing on financial firms, we observe that the pattern of differentiation provides evidence that financial issuers experienced a shock associated with rollover risk in all three events, although the magnitude is much larger after the bankruptcy of Lehman Brothers.
We have shaded the
coefficients that indicate this pattern in columns 2, 5, and 8 in table 5. Financial issuers also became more sensitive to the changes in CDS spreads after the collapse of Bear Stearns, a phenomenon we attribute to changes in perceptions of credit risk. Regarding nonfinancial firms, in each of the three events, increases in spreads were statistically significantly associated with increases in the credit risk measure. We have shaded the coefficients that indicate this pattern in columns 3, 6, and 9 in table 5. Finally, we also note that increases in rollover risk for nonfinancial issuers only became significant after the Lehman bankruptcy event. Indeed, after this event the illiquidity in financial markets was widespread. Both financial and nonfinancial firms that had a higher fraction of commercial paper maturing after the Lehman event experienced a large increase in spreads. 19
We take the fact that our risk proxies are able to differentiate between credit and rollover components of commercial paper spreads as evidence that market investors were able to differentiate firms when purchasing commercial paper. The distinctions across financial and nonfinancial issuers imply that the disruption in financial markets was more orderly than previously thought. The fact that these findings are not consistent with a scenario of disorderly panic suggests that investors used available information to price differently across issuers even in the midst of one of the largest financial crises in history. This pattern facilitates our understanding of the role of the financial sector in the real economy. Our results suggest that the shocks faced during the crisis led to dramatically increased rollover premia at financial firms, but had little direct liquidity impact on nonfinancials during the initial stages of the crisis. Indeed, the market appears to have segregated nonfinancial issuers according to their credit risk. While low credit quality firms experienced increased spreads, high credit quality firms continued to receive short-term funding at similar spreads. Effectively, the runs that occurred in the repo markets (Gorton and Metrick, 2009) and the ABCP market (Covitz, Liang, and Suarez, 2009) mainly concentrated on financial issuers. In order to further support this interpretation of the results, we will run a robustness check only looking at A1/P1 issuers in section 7.1.
6.4
The Role of Dealers
One heightened concern during the crisis was the financial condition of intermediating banks. Indeed, the condition of financial intermediaries is the basis for concern that a financial-sector shock impacts the nonfinancial parts of the economy. After all, without financial intermediaries, a nonfinancial firm may not be able to obtain funds for otherwise-profitable investments. A firm that issued its commercial paper through a dealer could potentially be perceived as being more dependent on financial institutions for funding and thus at greater risk during times of financial market distress.
We test whether the market used this distinction in penalizing the
funding costs of dealer-dependent issuers. 20
A typical relationship between commercial paper dealers and issuers involves issuers placing its commercial paper through dealers. The dealer may buy the paper at a discount and immediately sell it in the market. On occasion, dealers will hold some paper inventory as a service to issuers; this is often useful when issuers need funds of a particular maturity, different from the maturity desired by investors. During the crisis, if a dealer’s own access to funding shrunk, its ability to offer this service would shrink as well. As a result, one might expect that issuers that were dependent on dealers would have been more susceptible to the shocks we observed above. In table 6, we estimate equation (2) dividing the sample into dealer-dependent and non–dealerdependent issuers. We define an issuer to be dealer-dependent if 100 percent of its issues are sold through a dealer.
Non–dealer-dependent issuers are those that sell at least some of their paper
directly to investors, such as money market mutual funds. Issuers that use dealers to sell all of their papers do not have independent access to the commercial paper market and are thus relatively more exposed to intermediation channel disruptions. [Insert T able 6 here] Columns 3 and 6 of table 6 suggest that during the financial shocks of August 2007 and March 2008, issuers that were not entirely reliant on dealers to place commercial paper did not experience increases in spreads associated to higher rollover risk. However, these non–dealer-dependent firms experienced increased sensitivity to credit risk. These results suggest that issuers that are able to place some of their paper directly with investors and do not rely entirely on dealers are less subject to rollover risk. However, during the period that followed the bankruptcy of Lehman Brothers, rollover risk becomes significant even for non–dealer-dependent firms as shown in column 9 of table 6.
We
attribute this result to the fact that the Lehman Brothers event was characterized by widespread liquidity problems. In this event, all firms, regardless of their use of a dealer, experienced increased spreads due to rollover risk, which is associated with increasing liquidity premia. We will see further 21
evidence to support this pattern, below, when we focus on high credit quality issuers in section 7.1. Dealer-dependent firms experienced a large increase in spreads attributable to both credit and rollover risk in all three events. This is consistent with the idea that firms that relied on dealers to place their own paper were in greater need of intermediation services and assumed part of the risk brought about by those financial intermediaries through heightened borrowing costs. The results in this section highlight the role of financial intermediaries in transmitting the initial financial shock to other firms. Even though the commercial paper market is thought of as arm’slength financing, reserved for investment-grade firms, financial intermediaries still play a significant role in this market, as they act as dealers in many of these transactions.
As a consequence,
severe disruptions for financial firms may impact the funding conditions of even some firms that are considered high credit quality when compared to the overall credit spectrum.
7
Robustness
7.1
Credit Quality
To gain further insight into the patterns uncovered in tables 4, 5, and 6, we restrict our sample to the set of issuers with the highest commercial paper ratings (A1/P1 ratings).19
For these
regressions, we find programs that have an A1/P1 rating at the beginning of each of the three sample periods. Even though we restrict the sample by credit quality, the data contain significant cross-sectional variation in CDS spreads even within the high-quality issuers. This variation allows us to draw inferences within this sample. Table 7 reports the results of estimating separate regressions for financial and nonfinancial firms that are rated A1/P1, and table 8 reports the results of estimating separate regressions for 19
Rule 2a-7 of the Investment Company Act of 1940 limits the credit risk that money market mutual funds may bear by restricting their investments to “eligible” securities. An eligible security must carry one of the two highest ratings (“1” or “2”) for short-term obligations from at least two of the nationally recognized statistical ratings agencies. A tier-1 security is an eligible security rated “1” by at least two of the rating agencies; a tier-2 security is an eligible security that is not a tier-1 security.
22
dealer-dependent and non–dealer-dependent issuers. [Insert T ables 7 and 8 here] In table 7, we find qualitatively similar results to those in table 5; however, when the sample is restricted to high-quality issuers, most of the regression coefficients are smaller in magnitude. That implies that the credit risk premium of nonfinancial firms increased during the crisis even for the highest credit quality firms, although not as much compared to the relatively low quality firms. The differential effects of rollover and credit risk on financial and nonfinancial firms suggest that investors were able to distinguish the quality differences even among the highest-quality issuers in the midst of the market turmoil.20 In table 8, we find that dealer-dependent A1/P1-rated issuers did not experience increased spreads due to rollover risk during the crisis. This result contrasts with those in table 6, where we find that dealer-dependent firms were affected by rollover risk. This finding suggests that commercial paper investors demanded a higher rollover premium to lower-quality dealer-dependent issuers but not to higher-quality firms.21 Similar to results in table 7, we observe that the coefficients in table 8 on credit risk and rollover risk during the three market events are smaller in magnitude compared to the coefficients in table 6, suggesting that the increased credit and rollover premiums in the crisis were smaller for the highest-quality programs than for the lower-quality programs.
7.2
Credit Risk Measurement
We contrast our results using CDS spreads as a credit risk proxy with similar regressions based on Moody’s KMV expected default frequency (EDF). Moody’s states that “a public firm’s proba20
Lower-rated commercial paper issuers received a short-term rating of A2 by S&P or P2 by Moody’s. These ratings correspond to relatively highly rated firms in bond markets. For example, a short-term rating of P-2 is roughly consistent with a bond rating of BBB. 21 Another difference with respect to table 6 is the positive coefficient on rollover risk for non–dealer-dependent firms after August 2007 (column 3 of table 8). According to our hypothesis, this coefficient was expected to be insignificant. We attribute this positive coefficient to potential compositional effects. We will further address this issue in section 7.3.
23
bility of default is calculated from three drivers—the market value of its assets, its volatility, and its current capital structure.”22 While the EDF uses current equity information, because capital structure is a slow-moving component, the EDF may be slow to react to market changes. In the context of the crisis, in which market conditions changed very rapidly, the EDF may not adjust sufficiently quickly to capture market perceptions of credit risk. Our EDF results are shown in table 9. We replace the CDS spread variable with the EDF, both as the independent variable and for the interaction terms. Our results are largely consistent with those obtained with the CDS measure, although less significant in some instances. One explanation for this lost of significance is that EDF moves a bit more slowly than CDS spreads, and it is perhaps unable in a timely way to pick up market perceptions of credit risk during the crisis. This paper does not take a view on the superiority of either measure in predicting default; however, because the CDS measure is not a perfect proxy for credit risk, the similarity in results with another credit measure provides assurance that we are capturing the desired effect. Because we are looking at the short-term reaction of credit risk perceptions, we are unconcerned with the ability of the particular measure to forecast long-term default. [Insert T able 9 here]
7.3
Endogenous Dealer-Dependency
The decision to sell commercial paper directly or through dealers responds to economic incentives and, thus, is likely endogenous.
Nonfinancial issuers as well as issuers with smaller commercial
paper programs (in terms of debt outstanding) or lower ratings are more likely to rely exclusively on dealers to sell their paper, as opposed to directly sell at least some of their paper to final investors, such as money market mutual funds. Financial issuers are more likely to place their own commercial paper directly, as they may use their client networks from non–commercial paper 22
http://www.moodyskmv.com/research/edf benefits.html. Accessed on November 22, 2010.
24
activities. Similarly, issuers with larger amounts of commercial paper debt are more likely to develop their own marketing systems, perhaps because of economies of scale. The possible endogeneity of the use of dealers to place paper raises the concern that the results in table 6 may be explained if issuers with higher rollover risk are more likely to be dealer-dependent, in the sense that they rely on dealers to sell all their issues. To alleviate this concern, we use a propensity score matching–type method to check whether the results in table 6 are robust to selection of dealer-dependency on observables. In particular, we first estimate a probit regression of the probability that an issuer relies exclusively on dealers to place their own paper on three variables: a dummy variable for financial issuers; the log size of the amount of commercial paper outstanding at the beginning of the sample; and dummy variables for the rating of the issuer at the beginning of the sample.23 Table 10 contains the estimates of this Probit regression. We then repeat the regressions of table 6, limiting the sample of non–dealer-dependent issuers to those with more than 50 percent predicted probability of being dealer-dependent. These issuers are the most likely comparable to those that actually rely exclusively on dealers. [Insert Table 10 here] The regressions using this matched sample of non–dealer-dependent issuers are reported in columns 3, 6, and 9 of table 11.
Columns 2, 5, 8 reproduce the results using the sample of
dealer-dependent issuers reported in table 6. The results are qualitatively similar to those in table 6. For example, comparing columns 2 and 3 of table 11 suggests that dealer-dependent issuers experienced larger increases in spreads due to rollover risk after the shock in August 2007, while non–dealer-dependent issuers did not experience increases in spreads that were statistically related to rollover risk around the same shock. Columns 5 and 6 indicate that a similar pattern applies to the shock experienced by money markets after the collapse of Bear Stearns in March 2008. As 23
We include one dummy for A1/P1-rated issuers, one for A2/P2 and split-rated issuers, and omit the dummy variable for issuers rated below A2/P2.
25
suggested by columns 8 and 9, after the bankruptcy of Lehman Brothers in September 2008, even non–dealer-dependent issuers experienced increases in spreads associated with rollover risk. [Insert T able 11 here]
8
Conclusions
This paper provides empirical evidence on the role of credit risk and rollover risk in the widening of unsecured commercial paper spreads of maturity 1 to 4 days during three major events in the 2007–2009 financial crisis: the ABCP market freeze in August 2007; the collapse of Bearn Stearns in March 2008; and the bankruptcy of Lehman Brothers in September 2008. Our findings show that initial financial-sector shocks had little impact on how rollover risks are priced in commercial paper spreads of nonfinancial issuers and that the Lehman Brothers event passed through the dealer channel to nonfinancial firms. These findings suggest a more complex set of relationships between the financial sector and the broader economy than that embedded in the notion of bank(loan)dependent firms that do not access capital markets. Indeed, our conclusion is that the real economy is resilient, but not impervious, to financialsector distress. Nonfinancial firms managed to withstand even the large shocks faced by financial firms in the first year of the crisis, shocks that, at the time, were being described as unprecedented. Had the crisis ended earlier, our paper could have concluded that there are little or no transmission effects, even for the largest shocks. With hindsight, and the presence of the enormous shock triggered by the bankruptcy of Lehman Brothers, we observe that the transmission effect indeed exists. The shocks that impacted the economy did so on the margin of impacting borrowing costs for firms that relied on financial firms acting as commercial paper dealers to access capital markets. It is important to note that commercial paper issuers are typically among the highest-rated firms in the credit spectrum.
Therefore, this paper is not studying the impact of the financial
shock on small business or on bank-financing–dependent firms. The financial crisis of 2007–2009 26
likely had an important effect on firms that did not have access to capital markets, which is beyond the scope of this paper. While this study has found evidence of the transmission channel of the financial sector, we do not explore here some other interesting dimensions of this question. Our study looks at the cost of funding of commercial paper issuers, conditional on their participation in the market. We encourage future study on the determinants of market participation and the potential impact of the financial sector on those decisions.
References [1] Almeida, Heitor, Murillo Campello, and Michael S. Weisbach, 2004, The cash flow Sensitivity of cash. Journal of Finance, 59(4), 1777–1804. [2] Acharya, Viral, Douglas Gale, and Tanju Yorulmazer, 2010, Rollover risk and market freezes. Journal of Finance, forthcoming. [3] Acharya, Viral V., Philipp Schnabl, and Gustavo Suarez, 2010, Securitization without risk transfer. NBER Working Paper Series, w15730. [4] Afonso, Gara M., Anna Kovner, and Antoinette Schoar, 2010, Stressed not frozen: The fed funds market in the financial crisis. NBER Working Paper Series, w15806. [5] Allen, Franklin, Elena Carletti, and Douglas Gale, 2009, Interbank market liquidity and central bank intervention. Journal of Monetary Economics 56, 639–652. [6] Bates, Thomas W., Kathleen M. Kahle, and Rene M. Stulz, 2009, Why do U.S. firms hold so much more cash than they used to? Journal of Finance 64, 1985–2021.
27
[7] Bernanke, Ben S., Mark Gertler, and Simon Gilchrist, 1999, The financial accelerator in a quantitative business cycle framework. In John B. Taylor and Michael Woodford, eds.: Handbook of Macroeconomics, Vol. 1 (North Holland, Amsterdam). [8] Borensztein, Eduardo, and Jong-Wha Lee, 2002, Financial crisis and credit crunch in Korea: Evidence from firm-level data. Journal of Monetary Economics 49, 853–875. [9] Brunnermeier, Markus, 2009, Deciphering the Liquidity and Credit Crunch 2007–08. Journal of Economic Perspectives 23, 77–100. [10] Calomiris, Charles W., Charles P. Himmelberg, and Paul Wachtel, 1995, Commercial paper, corporate finance, and the business cycle: a microeconomic perspective. Carnegie Rochester Conference Series on Public Policy, 42, 203–250. [11] Campello, Murillo, John R. Graham, and Campbell R. Harvey, 2010, The real effects of financial constraints: Evidence from a financial crisis. Journal of Financial Economics 97, 470–487. [12] Chari, V.V., Lawrence J. Christiano and Patrick J. Kehoe, 2008, Facts and myths about the financial crisis of 2008. Federal Reserve Bank of Minneapolis Working paper, 666. [13] Chava, Sudheer, and Amiyatosh Purnanandam, 2011, The effect of banking crisis on bankdependent borrowers. Journal of Financial Economics 99, 116–135. [14] Cohen-Cole, Ethan, Burcu Duygan-Bump, Jose Fillat, and Judit Montoriol-Garriga, 2008, Looking behind the aggregates: a reply to “Facts and myths about the financial crisis of 2008.” Working paper, QAU08-5, Federal Reserve Bank of Boston. [15] Cohen-Cole, Ethan, and Enrique Martinez-Garcia, 2008, The balance sheet channel. Working paper, QAU08-7, The Federal Reserve Bank of Boston.
28
[16] Cornett, Marcia Millon, Jamie John McNutt, Philip E. Strahan, and Hassan Tehranian, 2010, Liquidity risk management and credit supply in the financial crisis. Journal of Financial Economics, forthcoming; Working paper, available on SSRN. [17] Covitz, Dan, and Chris Downing, 2007, Liquidity or credit risk? The determinants of very short-term corporate yield spreads. Journal of Finance 62, 2303–2328. [18] Covitz, Daniel M., Nellie Liang, and Gustavo A. Suarez, 2009, The evolution of a financial crisis: Panic in the asset-backed commercial paper market. Board of Governor of the Federal Reserve System Working paper, 2009-36. [19] Curdia, Vasco, and Michael Woodford, 2010, Credit frictions and optimal monetary policy. Journal of Money, Credit, and Banking 42, 3–35. [20] De Fiore, Fiorella, and Oreste Tristani, 2009, Optimal monetary policy in a model of the credit channel. European Central Bank Working paper, 1043. [21] Dell’Ariccia, Giovanni, Enrica Detragiache, and Raghuram Rajan, 2008, The real effect of banking crises. Journal of Financial Intermediation 17, 89–112. [22] Duca, John V., The money market meltdown of the Great Depression. Journal of Money Credit and Banking, forthcoming. [23] Duchin, Ran, Oguzhan Ozbas, and Berk A. Sensoy, 2010, Costly external finance, corporate investment, and the subprime mortgage credit crisis. Journal of Financial Economics 97, 418– 435. [24] Duygan-Bump, Burcu, Patrick M. Parkinson, Eric S. Rosengren, Gustavo A. Suarez, and Paul S. Willen, 2010. How effective were the Federal Reserve emergency liquidity facilities? Evidence from the Asset-Backed Commercial Paper Money Market Mutual Fund Liquidity Facility. Quantitative Analysis Unit Working Paper QAU10-3, Federal Reserve Bank of Boston. 29
[25] Erlich, Natalie, 2010, May 37, Banks unlikely to ease lending standards: FICO. CNBC.com, available at: http://www.cnbc.com/id/37377093. [26] Faia, Ester, and Tommaso Monacelli, 2007, Optimal interest rate rules, asset prices, and credit frictions. Journal of Economic Dynamics and Control 31, 3228–3254. [27] Gao, Pengjie, and Hayong Yun, 2009, Commercial paper, lines of credit, and the real effects of the financial crisis of 2008: Firm-level evidence from the manufacturing industry. Working paper, SSRN. [28] Gorton, Gary B., and Andrew Metrick, 2009, Securitized banking and the run on repo. NBER Working Paper Series, w15223. [29] Hahn, Thomas K., 1998, Commercial paper. In Federal Reserve Bank of Richmond: Instruments of the money market, Monograph, 105–127. [30] He, Zhiguo, and Wei Xiong, 2011, Rollover Risk and Credit Risk. Working Paper, Princeton University. [31] Holmstrom, Bengt, and Jean Tirole, 1998, Private and public supply of liquidity. Journal of Political Economy 106, 1–40. [32] Hu, Xing, 2010, Rollover risk and credit spreads in the financial crisis of 2008. Working Paper, Princeton University. [33] Ivashina, Victoria, and David Scharfstein, 2010, Bank lending during the financial crisis of 2008. Journal of Financial Economics 97, 319–338. [34] Kacperczyk, Marcin T., and Schnabl, Philipp, 2010, When safe proved risky: Commercial paper during the financial crisis of 2007–2009. Journal of Economic Perspectives 24, 29–50.
30
[35] Khwaja, Asim Ijaz, and Atif Mian, 2008, Tracing the impact of bank liquidity shocks: Evidence from an emerging market. The American Economic Review 98, 1413-1442. [36] Krishnamurthy, Arvind, 2010, Amplification mechanisms in liquidity crises. American Economic Journal: Macroeconomics 2, 1–30. [37] Peek, Joe, and Eric Rosengren, 2000, Collateral damage: Effects of the Japanese bank crisis on real activity in the United States. The American Economic Review 90, 30–45. [38] Schwarz, Krista, 2010, Mind the gap: Disentangling credit and liquidity in risk spreads. Working paper, available on SSRN. [39] Stigum, Marcia, and Anthony Crescenzi, 2007, Stigum’s Money Market. New York, McGrawHill. [40] Tang, Dragon Yongjun, and Hong Yan, 2007, Liquidity and credit default swap spreads. Working paper, available on SSRN. [41] Tong, Hui, and Shang-Jin Wei, 2009, The composition matters: Capital inflows and liquidity crunch during a global economic crisis. IMF Working paper, wp09164.
31
Figure 1: 1‐month Libor‐OIS spread 4
Daily 3.5
3
Bear Stearns collapse
Lehman Brothers Bankruptcy
2
ABCP market freeze, August 9, 2007
1.5
1
0.5
Note: This figure plots the daily spread of 1‐month U.S. Libor over 1‐month OIS from January 2006 to July 2009 . Source: BBA and Prebon.
3‐Jul‐09
3‐May‐09
3‐Mar‐09
3‐Jan‐09
3‐Nov‐08
3‐Sep‐08
3‐Jul‐08
3‐May‐08
3‐Mar‐08
3‐Jan‐08
3‐Nov‐07
3‐Sep‐07
3‐Jul‐07
3‐May‐07
3‐Mar‐07
3‐Jan‐07
3‐Nov‐06
3‐Sep‐06
3‐Jul‐06
3‐May‐06
3‐Mar‐06
0 3‐Jan‐06
Percentage points
2.5
Figure 2: Financial Sector Transmission Mechanism A. Pre‐Crisis Nonfinancial Issuer
B. Liquidity Crisis
Financial Issuer
Dealer
Dealer
Nonfinancial Issuer
Financial Issuer
Dealer
Dealer
Investors
Investors
C. Isolated Financial Sector Crisis: No Impact on Nonfinancial Issuers
D. Transmission of Crisis to Nonfinancial Issuers Through Dealer; Direct Nonfinancial Access Remains
Nonfinancial Issuer
Financial Issuer
Dealer
Dealer
Investors
Nonfinancial Issuer
Financial Issuer
Dealer
Dealer
Investors
Note: Figure shows stylized patterns of commercial paper issuance. Each box represents a type of issuer. Trapezoids represent the presence of a dealer in a transaction. The oval at bottom of each diagram are investors that buy commercial paper. The connecting lines represent an active channel, and the black lines across the channel indicate increased costs via this channel due to market disruptions. Panel A shows the situation pre‐crisis, where all channels are operational. Panel B shows the situation in a generalized liquidity crisis, where all channels have been severed. Panel C shows what one would expect if nonfinancial issuers could continue to issue; effectively the situation in the absence of a financial sector transmission mechanism. Panel D shows a situation where financial issuers suffer liquidity shocks and nonfinancial issuers are impacted via dealer networks, but direct market access remains open.
Table 1: Summary Statistics January 2007-December 2008 Mean Std Dev Progs Obs (1) (2) (3) (4)
January 2007-December 2007 Mean Std Dev Progs Obs (5) (6) (7) (8)
January 2008-June 2008 Mean Std Dev Progs Obs (9) (10) (11) (12)
July 2008-December 2008 Mean Std Dev Progs Obs (13) (14) (15) (16)
PANEL A All firms
0.309 0.736 42.118 80.562
0.725 0.921 28.471 34.563
101 17,408 101 17,408 101 17,408 101 101
0.106 0.268 41.797 80.986
0.254 0.158 29.005 33.703
89 89 89 89
8,543 8,543 8,543 89
0.221 0.812 42.954 79.939
0.335 0.458 28.330 35.144
85 85 85 85
4,557 4,557 4,557 85
0.804 1.241 1.584 1.413 41.872 27.522 80.214 35.159
90 90 90 90
4,308 4,308 4,308 90
0.172 0.900 25.950 69.239
0.604 1.165 15.240 40.435
53 53 53 53
6,916 6,916 6,916 53
0.034 0.234 25.081 67.590
0.174 0.173 13.694 39.907
42 42 42 42
3,171 3,171 3,171 42
0.087 0.919 26.677 66.249
0.274 0.501 16.341 41.883
40 40 40 40
1,878 1,878 1,878 40
0.493 1.041 2.011 1.654 26.696 16.467 65.900 42.453
45 45 45 45
1,867 1,867 1,867 45
0.399 0.628 52.776 93.064
0.781 0.695 30.099 20.701
48 48 48 48
10,492 10,492 10,492 48
0.149 0.288 51.664 92.957
0.283 0.144 31.064 21.055
47 47 47 47
5,372 5,372 5,372 47
0.315 0.737 54.364 92.107
0.341 0.409 29.364 21.939
45 45 45 45
2,679 2,679 2,679 45
1.042 1.327 1.257 1.089 53.480 28.612 94.529 16.702
45 45 45 45
2,441 2,441 2,441 45
59 59 59 59
8,560 8,560 8,560 59
0.141 0.281 48.568 100.0
0.283 0.132 29.888 0.000
55 55 55 55
4,711 4,711 4,711 55
0.310 0.773 51.516 100.0
0.338 0.469 29.517 0.000
54 54 54 54
2,555 2,555 2,555 54
1.166 1.373 1.400 1.313 53.931 28.731 100.0 0.000
54 54 54 54
2,147 2,147 2,147 54
Firms with less than 100% dealer-placed paper (non-dealer-dependent) Overnight CP spread 0.160 0.546 42 8,848 Credit risk 0.788 0.996 42 8,848 Rollover risk 31.502 22.977 42 8,848 Fraction of dealer-placed paper 53.255 40.076 42 42
0.064 0.252 33.473 50.228
0.205 0.183 25.532 38.102
34 34 34 34
3,832 3,832 3,832 34
0.107 0.861 32.026 44.993
0.294 0.438 22.411 38.366
31 31 31 31
2,002 2,002 2,002 31
0.445 0.969 1.767 1.484 29.892 20.053 50.536 40.414
36 36 36 36
2,161 2,161 2,161 36
Overnight CP spread Credit risk Rollover risk Fraction of dealer-placed paper
PANEL B Financial firms Overnight CP spread Credit risk Rollover risk Fraction of dealer-placed paper Nonfinancial firms Overnight CP spread Credit risk Rollover risk Fraction of dealer-placed paper
PANEL C Firms with 100% dealer-placed paper (dealer-dependent) Overnight CP spread 0.463 0.844 Credit risk 0.682 0.832 Rollover risk 53.092 29.426 Fraction of dealer-placed paper 100.0 0.000
Note: Credit risk is defined as the CDS spread of the issuer. Rollover risk is defined as the fraction of paper maturing over the next week. All variables are in percent.
Table 2: Baseline Regression Full Sample (1) Credit
(2)
(3)
(4)
0.286*** 0.295*** [0.061] [0.058] 0.006*** [0.001]
Rollover
0.002** [0.001]
(5)
(6)
January 2007- January 2008July 2008December 2007 June 2008 December 2008 (7) (8) (9)
0.296*** 0.296*** [0.060] [0.058]
‐0.149* [0.087]
0.149*** [0.023]
‐0.032 [0.073]
0.006*** [0.001]
0.001* [0.000]
0.001** [0.000]
0.005* [0.003]
0.002** [0.001]
0.221*** [0.034]
Post Aug 2007
0.152*** [0.018]
Bear
1.019*** [0.187]
Lehman 0.098** [0.042]
0.092** [0.042]
0.076 [0.052]
Issuer-fixed effects?
N
Y
N
Y
Observations Number of programs R-squared
17,412 101 0.132
17,412 101 0.413
17,408 101 0.047
17,408 101 0.307
Constant
0.218*** ‐0.164*** [0.046] [0.045]
‐0.003 [0.061]
0.037 [0.028]
‐0.014 [0.026]
0.061 [0.162]
N
Y
Y
Y
Y
17,408 101 0.188
17,408 101 0.416
8,543 89 0.416
4,557 85 0.76
4,308 90 0.658
Note: The dependent variable is the overnight CP spread. Credit risk is defined as the CDS spread of the issuer. Rollover risk is defined as the fraction of paper maturing over the next week. Post Aug 2007 is a dummy = 1 after August 9, 2007. Bear is a dummy = 1 after March 14, 2008. Lehman is a dummy = 1 after September 15, 2008. Robust standard errors are clustered by issuer.
Table 3: Relationship between rollover risk and credit risk
Credit
(1)
(2)
‐1.669 [1.054]
‐0.109 [0.861]
Post Aug 2007
January 2007-December 2007 (3) (4) 33.203*** [11.593]
1.229 [8.928]
‐5.418* [2.879]
‐0.792 [2.595]
Bear
January 2008-June 2008 (5) (6) ‐9.672** [4.312]
‐1.034 [1.288]
1.522 [1.409]
0.801 [1.385]
Lehman
Constant
43.347*** 42.199*** [3.878] [0.633]
35.001*** [3.807]
41.776*** [1.825]
July 2008-December 2008 (7) (8) ‐2.595* [1.553]
0.186 [1.460]
0.594 [2.684]
‐0.676 [2.270]
49.899*** 43.317*** 45.643*** 41.966*** [5.497] [1.311] [4.932] [2.065]
Issuer-fixed effects?
N
Y
N
Y
N
Y
N
Y
Observations R-squared Number of programs
17,408 0.003 101
17,408 0.641 101
8,543 0.027 89
8,543 0.651 89
4,557 0.025 85
4,557 0.742 85
4,308 0.017 90
4,308 0.691 90
Note: The dependent variable is Rollover risk, defined as the fraction of paper maturing over the next week. Credit risk is defined as the CDS spread of the issuer. Post Aug 2007 is a dummy = 1 after August 9, 2007. Bear is a dummy = 1 after March 14, 2008. Lehman is a dummy = 1 after September 15, 2008. Robust standard errors clustered by issuer.
Table 4: Baseline with interaction terms January 2007- January 2008December 2007 June 2008 (1) (2) Credit Rollover Post Aug 2007 Credit * Post Aug 2007 Rollover * Post Aug 2007
-0.299*** [0.081] 0 [0.000] -0.066 [0.048] 0.549*** [0.112] 0.003*** [0.001]
Bear
0.100*** [0.022] 0 [0.000]
Rollover * Bear Lehman
0.110*** [0.026]
0.068** [0.027]
-0.217 [0.247] 0.295 [0.201] 0.021*** [0.003] 0.844*** [0.269]
Y Y
Y Y
Y Y
8,543 0.455 89
4,557 0.768 85
4,308 0.708 90
Credit * Lehman Rollover * Lehman
Issuer-fixed effects? Robust standard errors clustered by issuer? Observations R-squared Number of programs
-0.311 [0.226] -0.006** [0.003]
-0.011 [0.034] 0.101*** [0.028] 0.002*** [0.001]
Credit * Bear
Constant
July 2008December 2008 (3)
Note: The dependent variable is the overnight CP spread. Credit risk is defined as the CDS spread of the issuer. Rollover risk is defined as the fraction of paper maturing over the next week. Post Aug 2007 is a dummy = 1 after August 9, 2007. Bear is a dummy = 1 after March 14, 2008. Lehman is a dummy = 1 after September 15, 2008.
Table 5: Financial vs Non Financial CP issuers January 2007-December 2007 ALL (1) Credit Rollover Post Aug 2007 Credit * Post Aug 2007 Rollover * Post Aug 2007 Bear Credit * Bear Rollover * Bear Lehman Credit * Lehman Rollover * Lehman Constant
Issuer-fixed effects? Robust standard errors clustered by issuer?
Observations R-squared Number of programs
FIN (2)
NON-FIN (3)
January 2008-June 2008 ALL (4)
FIN (5)
NON-FIN (6)
July 2008-December 2008 ALL (7)
FIN (8)
NON-FIN (9)
-0.299*** [0.081] 0.000 [0.000] -0.066 [0.048] 0.549*** [0.112] 0.003*** [0.001]
-0.005 -0.132 0.100*** 0.081*** 0.119*** -0.311 0.028 -1.251*** [0.055] [0.105] [0.022] [0.021] [0.039] [0.226] [0.092] [0.258] 0.000 0.000 0.000 -0.001 0.000 -0.006** -0.008** -0.005 [0.000] [0.000] [0.000] [0.001] [0.000] [0.003] [0.004] [0.003] -0.096* -0.028 [0.050] [0.087] 0.186 0.701*** [0.129] [0.229] 0.004*** 0.002 [0.001] [0.001] -0.011 -0.106* 0.082** [0.034] [0.059] [0.035] 0.101*** 0.135*** 0.092** [0.028] [0.040] [0.039] 0.002*** 0.003** 0.001 [0.001] [0.001] [0.001] -0.217 -0.235 -0.198 [0.247] [0.201] [0.360] 0.295 0.071 0.942*** [0.201] [0.097] [0.242] 0.021*** 0.022*** 0.017*** [0.003] [0.006] [0.004] 0.110*** 0.016 0.080* 0.068** -0.03 0.112*** 0.844*** 0.287* 1.731*** [0.026] [0.012] [0.040] [0.027] [0.037] [0.034] [0.269] [0.153] [0.274] Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
8,543 0.455 89
3,171 0.291 42
5,372 0.484 47
4,557 0.768 85
1,878 0.689 40
2,679 0.768 45
4,308 0.708 90
1,867 0.653 45
2,441 0.743 45
Note: The dependent variable is the overnight CP spread. Credit risk is defined as the CDS spread of the issuer. Rollover risk is defined as the fraction of paper maturing over the next week. Post Aug 2007 is a dummy = 1 after August 9, 2007. Bear is a dummy = 1 after March 14, 2008. Lehman is a dummy = 1 after September 15, 2008. Financial Dummy = 1 for financial firms. Robust standard errors clustered by issuer. All regressions have issuer-fixed effects.
Table 6: Dealer-dependent vs. Non-dealer dependent issuers January 2007-December 2007
Credit Rollover Post Aug 2007 Credit * Post Aug 2007 Rollover * Post Aug 2007
ALL (1)
Dealerdependent (2)
Non-dealerdependent (3)
-0.299*** [0.081] 0.0000 [0.000] -0.066 [0.048] 0.549*** [0.112] 0.003*** [0.001]
-0.225 [0.142] 0.0000 [0.000] -0.06 [0.088] 0.615** [0.235] 0.003*** [0.001]
-0.281*** [0.079] 0.0000 [0.001] -0.056 [0.053] 0.473*** [0.121] 0.002 [0.001]
Bear Credit * Bear Rollover * Bear
January 2008-June 2008
July 2008-December 2008
ALL (4)
Dealerdependent (5)
Non-dealerdependent (6)
ALL (7)
Dealerdependent (8)
Non-dealerdependent (9)
0.100*** [0.022] 0.0000 [0.000]
0.149*** [0.024] 0.0000 [0.000]
0.062** [0.026] 0.001 [0.001]
-0.311 [0.226] -0.006** [0.003]
-1.288*** [0.377] 0.001 [0.003]
0.034 [0.147] -0.013*** [0.004]
-0.011 [0.034] 0.101*** [0.028] 0.002*** [0.001]
0.061* [0.034] 0.093** [0.040] 0.001* [0.001]
-0.054 [0.058] 0.123*** [0.044] 0.001 [0.001] -0.347 [0.383] 1.239*** [0.346] 0.014*** [0.004] 1.419*** [0.385]
-0.079 [0.265] -0.071 [0.145] 0.027*** [0.006] 0.445** [0.187]
0.110*** [0.026]
0.116** [0.044]
0.057* [0.033]
0.068** [0.027]
0.096*** [0.028]
-0.023 [0.039]
-0.217 [0.247] 0.295 [0.201] 0.021*** [0.003] 0.844*** [0.269]
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
8,543 89 0.455
4,711 55 0.473
3,832 34 0.405
4,557 85 0.768
2,555 54 0.772
2,002 31 0.713
4,308 90 0.708
2,147 54 0.755
2,161 36 0.634
Lehman Credit * Lehman Rollover * Lehman Constant
Issuer-fixed effects? Robust standard errors clustered by issuer?
Observations R-squared Number of programs
Note: The dependent variable is the overnight CP spread. Credit risk is defined as the CDS spread of the issuer. Rollover risk is defined as the fraction of paper maturing over the next week. Post Aug 2007 is a dummy = 1 after August 9, 2007. Bear is a dummy = 1 after March 14, 2008. Lehman is a dummy = 1 after September 15, 2008, Dealer-dependent dummy = 1 for firms that place 100% of their paper through dealers. Robust standard errors clustered by issuer. All regressions have issuer-fixed effects.
Table 7: Financial vs Non Financial CP issuers - A1/P1 only January 2007-December 2007 ALL (1) Credit Rollover Post Aug 2007 Credit * Post Aug 2007 Rollover * Post Aug 2007 Bear Credit * Bear Rollover * Bear
FIN (2)
Credit * Lehman Rollover * Lehman
Issuer-fixed effects? Robust standard errors clustered by issuer?
Observations R-squared Number of programs
ALL (4)
FIN (5)
NON-FIN (6)
-0.134 0.037 -0.133 0.076*** 0.060*** 0.099*** [0.089] [0.045] [0.114] [0.019] [0.020] [0.030] 0.000 0.000 0.000 0.000 0.000 0.000 [0.000] [0.000] [0.000] [0.000] [0.001] [0.001] -0.069* -0.083 -0.058** [0.035] [0.051] [0.021] 0.343*** 0.148 0.511*** [0.127] [0.128] [0.039] 0.001** 0.002*** 0.000 [0.000] [0.001] [0.000] -0.002 -0.073 0.079* [0.038] [0.059] [0.039] 0.077*** 0.109*** 0.058 [0.027] [0.037] [0.049] 0.001 0.002 -0.001 [0.001] [0.001] [0.001]
Lehman
Constant
NON-FIN (3)
January 2008-June 2008
July 2008-December 2008 ALL (7)
-0.312* [0.181] -0.001 [0.002]
FIN (8)
NON-FIN (9)
0.002 -1.109*** [0.089] [0.210] -0.007** 0.000 [0.003] [0.002]
-0.488** -0.347 -0.628*** [0.193] [0.210] [0.152] 0.328** 0.085 0.851*** [0.160] [0.104] [0.179] 0.011*** 0.017*** 0.009** [0.003] [0.006] [0.004] 0.016 -0.076*** -0.081** -0.092*** 0.460** 0.228* 1.038*** [0.015] [0.025] [0.037] [0.021] [0.215] [0.118] [0.195]
0.015 [0.011]
-0.021** [0.009]
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
5,081 0.194 56
2,575 0.098 33
2,506 0.278 23
2,827 0.452 51
1,592 0.482 30
1,235 0.42 21
2,692 0.41 56
1,575 0.361 35
1,117 0.589 21
Note: The dependent variable is the overnight CP spread. Credit risk is defined as the CDS spread of the issuer. Rollover risk is defined as the fraction of paper maturing over the next week. Post Aug 2007 is a dummy = 1 after August 9, 2007. Bear is a dummy = 1 after March 14, 2008. Lehman is a dummy = 1 after September 15, 2008. Financial Dummy = 1 for financial firms. Robust standard errors clustered by issuer. All regressions have issuer-fixed effects.
Table 8: Dealer placement vs Direct placement - A1/P1 only
January 2007-December 2007 DealerNon-dealerdependent dependent ALL (1) (2) (3) Credit Rollover Post Aug 2007 Credit * Post Aug 2007 Rollover * Post Aug 2007
-0.134 [0.089] 0.0000 [0.000] -0.069* [0.035] 0.343*** [0.127] 0.001** [0.000]
-0.127 [0.130] 0.0000 [0.000] -0.099** [0.043] 0.471*** [0.103] 0.001 [0.001]
0.039 [0.067] 0.0000 [0.000] -0.018 [0.046] 0.066 [0.121] 0.001* [0.000]
Bear Credit * Bear Rollover * Bear
January 2008-June 2008 DealerNon-dealerdependent dependent ALL (4) (5) (6)
0.076*** [0.019] 0.0000 [0.000]
0.108*** [0.021] 0.0000 [0.000]
0.061** [0.025] 0.001 [0.001]
-0.002 [0.038] 0.077*** [0.027] 0.001 [0.001]
0.068** [0.031] 0.073** [0.030] 0.000 [0.000]
-0.035 [0.062] 0.088* [0.046] 0.000 [0.001]
-0.008 [0.096] -0.007** [0.003]
-0.680*** [0.171] 1.033*** [0.110] 0.008 [0.005] 0.952*** [0.176]
-0.276 [0.204] 0.051 [0.106] 0.015*** [0.004] 0.224* [0.111]
0.021 [0.029]
-0.012 [0.014]
-0.076*** [0.025]
-0.070** [0.026]
-0.124*** [0.042]
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
5,081 0.194 56
2,209 0.277 32
2,872 0.088 24
2,827 0.452 51
1,271 0.561 30
1,556 0.254 21
2,692 0.41 56
923 0.588 30
1,769 0.328 26
Rollover * Lehman
Observations R-squared Number of programs
-1.128*** [0.177] 0.002 [0.002]
0.015 [0.011]
Credit * Lehman
Issuer-fixed effects? Robust standard errors clustered by issuer?
-0.312* [0.181] -0.001 [0.002]
-0.488** [0.193] 0.328** [0.160] 0.011*** [0.003] 0.460** [0.215]
Lehman
Constant
July 2008-December 2008 DealerNon-dealerdependent dependent ALL (7) (8) (9)
Note: The dependent variable is the overnight CP spread. Credit risk is defined as the CDS spread of the issuer. Rollover risk is defined as the fraction of paper maturing over the next week. Post Aug 2007 is a dummy = 1 after August 9, 2007. Bear is a dummy = 1 after March 14, 2008. Lehman is a dummy = 1 after September 15, 2008, Dealer-dependent dummy = 1 for firms that place 100% of their paper through dealers. Robust standard errors clustered by issuer. All regressions have issuer-fixed effects.
Table 9: Robustness on Credit Measure: EDF Jan 2007-Dec 2007
Jan 2008-Jun 2008
Jul 2008-Dec 2008
Jan 2007-Dec 2007
Jan 2008-Jun 2008
Jul 2008-Dec 2008
FIN (1)
NON-FIN (2)
FIN (3)
NON-FIN (4)
FIN (5)
NON-FIN (6)
Dealerdependent (7)
-2.231** [0.918]
-2.432* [1.363]
-0.158 [0.607]
0.856 [1.520]
-3.725*** [1.355]
-6.313*** [2.214]
-2.639 [1.590]
-1.932** [0.794]
0.985 [1.270]
0.093 [0.593]
-5.762*** [1.338]
-1.508 [1.541]
Rollover
0.000 [0.000]
0.000 [0.001]
-0.001 [0.001]
0.000 [0.000]
-0.009* [0.004]
-0.004 [0.003]
0.000 [0.000]
0.000 [0.001]
-0.001* [0.000]
0.000 [0.001]
0.001 [0.004]
-0.014*** [0.004]
Post Aug 2007
-0.038* [0.023]
0.149** [0.067]
0.051 [0.058]
0.056 [0.049]
Credit * Post Aug 2007
1.429** [0.695]
2.379 [1.863]
3.302** [1.518]
0.264 [0.871]
Rollover * Post Aug 2007
0.003*** [0.001]
0.001 [0.001]
0.003** [0.001]
0.002 [0.001]
-0.259 [0.204] 1.209 [1.526] 0.027*** [0.006] 0.649*** [0.184]
Credit (EDF)
Non-dealerdependent (8)
Dealerdependent (9)
Non-dealerdependent (10)
Dealerdependent (11)
Non-dealerdependent (12)
Bear
-0.007 [0.036]
0.149*** [0.044]
0.105*** [0.034]
0.054 [0.045]
Credit * Bear
0.512 [0.393]
-0.171 [0.789]
0.462 [0.640]
0.061 [0.477]
Rollover * Bear
0.003** [0.001]
0.001 [0.001]
0.001** [0.001]
0.001 [0.001]
0.071*** [0.025]
0.096** [0.039]
0.058 [0.039]
0.176*** [0.064]
-0.461** [0.204] 3.608*** [1.336] 0.024*** [0.006] 0.765*** [0.181]
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
3,171 0.29 42
5,372 0.465 47
1,878 0.654 40
2,679 0.755 45
1,867 0.662 45
2,441 0.722 45
4,711 0.468 55
3,832 0.389 34
2,555 0.751 54
2,002 0.695 31
2,147 0.733 54
2,161 0.641 36
Lehman Credit * Lehman Rollover * Lehman Constant
Issuer-fixed effects? Robust standard errors clustered by issuer?
Observations R-squared Number of programs
0.043 [0.361] 6.069*** [2.083] 0.017*** [0.004] 0.929*** [0.262]
0.124*** [0.046]
0.065** [0.030]
0.181*** [0.052]
0.027 [0.043]
0.208 [0.432] 5.735*** [1.335] 0.015*** [0.005] 0.743** [0.314]
Note: The dependent variable is the overnight CP spread. Credit risk is defined as the CDS spread of the issuer. Rollover risk is defined as the fraction of paper maturing over the next week. Post Aug 2007 is a dummy = 1 after August 9, 2007. Bear is a dummy = 1 after March 14, 2008. Lehman is a dummy = 1 after September 15, 2008. Financial Dummy = 1 for financial firms. Dealer-Dependent Dummy = 1 for issuers that placed 100% of their paper through dealers. Robust standard errors clustered by issuer. All regressions have issuer-fixed effects.
Table 10: First-stage probit regression for matching regression in Table 11 Dependent variable: Probability of placing all paper through dealers (1) Log (CP Outstanding)
-0.070* [0.040]
A-1/P-1 Rating Dummy
0.011 [0.128]
A-2/P-2 Rating Dummy
0.141 [0.201]
Financial Dummy
-0.218* [0.112]
Observations R-squared
100 0.0933
Note: The dependent variable is the probability that the issuer places all its paper through dealers. Log(CP Outstanding) is the log of CP outstanding from Jan 2007 to Dec 2008; Rating dummy = 1 for A-1/P-1 rated issuers; and Financial Dummy = 1 for financial issuers. The table report the marginal effects from a probit model. Robust standard errors clustered by issuer are reported in brackets.
Table 11: Dealer-dependent vs Non-dealer dependent firms. Matched sample of non-dealer-dependent issuers January 2007-December 2007 DealerNon-dealerdependent dependent ALL (1) (2) (3) Credit Rollover Post Aug 2007 Credit * Post Aug 2007 Rollover * Post Aug 2007
-0.192** [0.092] 0 [0.000] -0.011 [0.072] 0.571*** [0.167] 0.002* [0.001]
-0.225 [0.142] 0 [0.000] -0.06 [0.088] 0.615** [0.235] 0.003*** [0.001]
-0.219 [0.134] 0.001 [0.001] 0.082 [0.104] 0.569*** [0.180] -0.001 [0.001]
Bear Credit * Bear Rollover * Bear
ALL (4)
January 2008-June 2008 DealerNon-dealerdependent dependent (5) (6)
0.116*** [0.027] 0 [0.000]
0.149*** [0.024] 0 [0.000]
0.052 [0.034] 0.002 [0.001]
0.052 [0.033] 0.099*** [0.035] 0.001 [0.001]
0.061* [0.034] 0.093** [0.040] 0.001* [0.001]
0.073 [0.093] 0.116 [0.074] -0.001 [0.001]
-0.441 [0.651] -0.011* [0.006]
-0.347 [0.383] 1.239*** [0.346] 0.014*** [0.004] 1.419*** [0.385]
0.341 [0.582] 0.211 [0.561] 0.019* [0.009] 1.066* [0.598]
0.116** [0.044]
0.054 [0.070]
0.089*** [0.029]
0.096*** [0.028]
0.003 [0.066]
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
6,367 0.469 75
4,711 0.473 55
1,656 0.483 20
3,364 0.773 71
2,555 0.772 54
809 0.777 17
3,075 0.727 73
2,147 0.755 54
928 0.702 19
Rollover * Lehman
Observations R-squared Number of programs
-1.288*** [0.377] 0.001 [0.003]
0.091*** [0.034]
Credit * Lehman
Issuer-fixed effects? Robust standard errors clustered by issuer?
-0.993*** [0.234] -0.003 [0.003]
-0.121 [0.317] 0.862*** [0.235] 0.015*** [0.003] 1.390*** [0.265]
Lehman
Constant
ALL (7)
July 2008-December 2008 DealerNon-dealerdependent dependent (8) (9)
Note: The dependent variable is the overnight CP spread. Credit risk is defined as the CDS spread of the issuer. Rollover risk is defined as the fraction of paper maturing over the next week. Post Aug 2007 is a dummy = 1 after August 9, 2007. Bear is a dummy = 1 after March 14, 2008. Lehman is a dummy = 1 after September 15, 2008. Dealer-dependent dummy = 1 for issuers that placed 100 percent of their paper through dealers. The sample of this table is restricted to issuers that place all their commercial paper through dealers or that are very likely to place their paper through dealers, as implied by a probit regression of the probability of placing all paper through dealers on rating dummies, amount of paper outstanding, a dummy for financial issuers (reported in Appendix Table 2). Robust standard errors clustered by issuer. All regressions have issuer-fixed effects.
