Bondholder Concentration and Credit Risk: Evidence from a Natural Experiment
Table C.VI Bondholder concentration and CDS spreads
Download 276.61 Kb.
|
- Navigate this page:
- Probit Estimates
- D. Additional figures Figure D.1 Yearly Atlantic hurricane season damages, 1998-2007
- Figure D.2 Credit spreads in the U.S. insurance industry around hurricane Katrina
- Figure D.3 Geographical distribution of bond issuers and outstanding bonds
- A. Distribution of bond issuers
Table C.VI Bondholder concentration and CDS spreads The table reports the estimates of a model: 
where CDS denotes the CDS spread associated with corporate bond issuer i, H is bondholder concentration, and x is a vector of control variables (reported in the table), including the speculative-grade indicator, as well as industry and calendar quarter indicators. The CDS spread is the spread on 5-year CDS contracts, with modified restructuring clause, written over a given issuer. To ensure to focus on the most liquid contracts, a given issuer-quarter observation is included in the sample only if the 10-year CDS contract is also non-missing. All other variables are described in detail in Appendix B. In column (1), the model is estimated with simple OLS. In column (2), to account for the potential endogeneity of the selection of the issuers of bonds that are CDS reference entities, the model is estimated using the Heckman (1979) two-step correction. The first-stage probit estimates associated with this model are reported in column (5). The instrumental variable for the selection equation is log(1+Trades), proxying for the liquidity of the firm’s corporate bonds (CDS reference entities tend to have more liquid bonds). In column (3), the model is estimated with the Heckman two-step procedure combined with instrumental variables for bondholder concentration H, where, as in the previous tables, bondholder concentration H is instrumented by the interaction between the previous-period component of concentration of the Katrina-affected investors and the hurricane Katrina indicator,  , while separately controlling for the levels of these variables. The hurricane Katrina indicator Kat is equal to 1 in 2005Q3, and 0 otherwise. In column (4), the model is estimated again with the Heckman two-step procedure combined with instrumental variables estimation, and the dependent variable is  (i.e., four quarters after t) instead of  , analogously to the approach followed in the previous tables. The first-stage probit estimates associated with the specifications of columns (3) and (4) are reported in column (6). The row labeled (Pseudo) R2 reports the regression R2 in columns (1) and (2), and the probit pseudo-R2 in columns (5) and (6). In all specifications, the t-statistics are based on standard errors clustered around issuers and calendar date, using the procedure described by Thompson (2010). The sample consists of all the issuers of bonds in the Bank of America Merrill Lynch Corporate and High Yield Master Bond Index Database with available bond characteristics in the Mergent Fixed Income Securities Database (FISD), firm characteristics in the Compustat Fundamentals quarterly database, and CDS spreads from the Markit CDS database, over the period between 2003Q1 and 2007Q4. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels.
D. Additional figures Figure D.1 Yearly Atlantic hurricane season damages, 1998-2007 The graph plots the yearly aggregate damages (gray bars) and damages from the costliest hurricane of each Atlantic hurricane season (black bars), for the period 1998-2007. The text above each black bar reports the hurricane name, as well as the month in which it took place. Data on the damages are elaborated from the figures published in the Annual Summaries of North American Storms (1872-2008) by the Monthly Weather Review of the National Hurricane Center, available at the URL: http://www.aoml.noaa.gov/general/lib/lib1/nhclib/mwreviews.html (website accessed on 25 September 2014). The aggregate damages figures are expressed in 2010 constant billions of dollars. 
Figure D.2 Credit spreads in the U.S. insurance industry around hurricane Katrina The graph plots the percentage change in the yield spread on corporate bonds issued by U.S. insurance companies, over the period from 1 August 2005 to 31 October 2005. (Re)insurance companies are first classified into three categories: life insurance (gray solid line), property and casualty with no exposure to hurricane Katrina (gray dashed line), and property and casualty exposed to hurricane Katrina (black solid line). The third category comprises the list of insurance and reinsurance companies described in detail in Appendix A. Daily yield spreads are obtained for each company, from the Bank of America Merrill Lynch U.S. Corporate and High Yield Master index database. If a given insurance company has not issued any outstanding bond covered in the yield data, it is omitted. Next, the average yield spread is computed, for each of the three categories. Finally, on each day t the percentage change of the average yield spread relative to 1 August 2005 is computed, as [Y(t) – Y(1 August 2005)]/Y(1 August 2005), where Y(t) denotes the average yield spread on date t. This is the quantity plotted in the graph, for each (re)insurance company category. The shaded area in the graph corresponds to the hurricane Katrina period 23-30 August 2005. The bonds issued by property and casualty insurance and reinsurance companies that are affected by hurricane Katrina are excluded from the sample in the remainder of the analysis. 
Figure D.3 Geographical distribution of bond issuers and outstanding bonds The graphs describe the geographical distribution of the bond issuers and the outstanding value of bonds associated with the issuers in the sample. Panel A focuses on the issuers’ distribution across US states (Alaska and Hawaii are omitted for convenient spacing): the coloring of each state is proportional to the number of firms (bond issuers) in the sample that are located in the state, with darker areas corresponding to a larger number of firms. Panel B focuses on the distribution of outstanding bonds across US states, as of 2005Q3, i.e. the quarter in which hurricane Katrina occurs. For each state, the total par-amount value of the outstanding bonds issued by firms located in the state is computed. Darker areas correspond to a larger total outstanding value. The figures are elaborated based on a sample corresponding to the intersection of the Bank of America Merrill Lynch U.S. Corporate and High Yield Master Bond Index database, the Mergent Fixed Income Securities Database (FISD), and the Compustat Fundamental Quarterly database, over the period 1998Q1 to 2007Q4. Issuer location data are retrieved from the Compustat database. Data on the par-amount of the outstanding bonds are retrieved from the Mergent FISD database. A. Distribution of bond issuers 
B. Distribution of outstanding bonds as of 2005Q3 
Directory: content -> suppl -> 2015 suppl -> Background Information on Ilvaite suppl -> Torres-Miranda et al. American Journal of Botany suppl -> Assessment of the acoustic telemetry positioning system suppl -> Additional File 1 Text summary suppl -> Additional file 5, Table costs and ccdss process-related outcomes for trials of acute care management a suppl -> Supplementary Notes and Figures 2015 -> Table strains, plasmids, and primers Download 276.61 Kb. Share with your friends:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
)
)
