municipal bond search by amir33

VIEWS: 153 PAGES: 12

									       Natural Disaster Risk and Municipal Bond
                 Pricing in California
                Jacob Fowles, Gao Liu and Cezar Brian Mamaril∗
      Martin School of Public Policy and Administration, University of Kentucky

                                      November 12, 2007


                                               Abstract
           The relative inefficiency of the municipal bond market, as compared to
       other capital markets, is well known. Empirical tests have shown that the
       municipal market is sometimes slow to react to new information and doesn’t
       always integrate existing information appropriately. As Hurricane Katrina
       has revealed, natural disaster risk is one type of unsystematic risk that could
       impact a municipalities’ ability to meet debt obligations. Hurricanes, like
       other natural disasters, threaten municipalities in two ways-by damaging lo-
       cal infrastructure and by impacting the local tax base. While Katrina has
       certainly raised awareness of the potential impact of hurricanes on munici-
       palities along the Gulf Coast, it remains unclear if investors in the municipal
       market consider other types of natural disaster risk in other areas when de-
       termining risk premiums for municipal issues. We attempt to fill this gap in
       the literature by conducting an analysis to determine if underlying geologic
       earthquake risk affects interest costs for municipal bond issuers in Califor-
       nia. We find that underlying earthquake risk does matter in determining the
       interest costs for California municipalities, but only for municipal bonds is-
       sued after Hurricane Katrina. Specifically, our analysis shows that, holding
       all other market, issuer and issue characteristics constant, interest costs for
       California municipalities in the highest risk territory increased by 44 basis
       points after Katrina, compared to only a 13 basis point increase in the lowest
       risk territory. The paper concludes by suggesting some potential explanations
       for this finding in the context of municipal market efficiency.


1     Introduction
It is well known that, relative to most other sectors of the capital market, the munici-
pal bond market is at least somewhat less efficient. Information asymmetry-in which
   ∗ Address: 405 Patterson Office Tower, University of Kentucky, Lexington, Kentucky 40506. E-mail: ja-

cob.fowles@uky.edu. The authors are grateful to J.S. Butler and Dwight Denison for advice and suggestions,
and to the faculty and students of the Martin School for their helpful comments.



                                                    1
investors and issuers have differing access to information-is often cited as the primary
cause of this inefficiency. Many municipal bonds are issued by small municipalities
which have previously issued no or few bonds, making it difficult for investors to ac-
curately gauge issuer quality. This problem is exacerbated by the fact that the munic-
ipal bond market is dominated by individual investors, due to the unique tax benefits
associated with municipal bond ownership. Ultimately, the consequence of market
inefficiency in the municipal bond market is that prices are distorted by investors who
demand a risk premium to compensate for their uncertainty, which drives up the capital
costs for municipal issuers. Many mechanisms have been developed in an attempt to al-
leviate this inefficiency with varying levels of effectiveness, municipal bond insurance
being perhaps the most popular (Peng and Brucato, 2004).
    Additionally, much empirical evidence exists which shows that the bond market
does not always react efficiently to the release of new information, although many
of these studies have focused on market reaction to a specific type of information:
changes in bond ratings (Kliger and Sarig, 2000; Reck and Wilson, 2006). However,
given the endogenous nature of bond ratings, it is argued that markets often anticipate
(or at least should have the ability to anticipate) many bond rating changes in advance
of the actual rating change. A more recent vein of literature addresses the market’s
reaction to unanticipated events which, unlike changes in bond ratings, it is assumed
that investors would be unable to anticipate. For instance, empirical evidence shows
that the municipal market was quick to react in response to both the Orange County
bankruptcy announcement and Hurricane Katrina (Denison, 2000; Denison 2006). To
paraphrase Denison, the bond market sometimes exhibits characteristics of semi-strong
efficiency by anticipating and incorporating new information into prices ex ante, but at
other times is unable to do so, and instead reacts to information ex post (2006). Of
course, implicit to this discussion of efficiency is the premise that market price at a
given time is reflective of all relevant, publicly available information at that time (Fama,
1970).


2    Background and Literature Review
As demonstrated most recently by Hurricane Katrina, natural disasters pose signifi-
cant risk for municipalities; while certainly Katrina represents an extreme example in
terms of scale, the categorical distribution of the damages it inflicted upon municipal-
ities along the Gulf Coast is largely generalizable to the types of damages inflicted by
a broad class of different natural disasters. In the event of natural disaster, municipal
market issues are often tied to concerns of how an affected issuer will meet debt obli-
gations in the face of not only damage to local infrastructure, but also loss of revenue
due to the disaster’s negative impacts on the local tax base. Although predictions for
the bond market following Katrina were initially quite dire (Wells, 2005), the munic-
ipal bond market in the aftermath of Katrina proved to be quite resilient-ratings have
stabilized, insurance claims have been paid, and, perhaps most importantly, generous
amounts of federal dollars have been firmly committed to the rebuilding process (Miller
and Bonifer, 2006; Wildasin, forthcoming). Prior to Katrina, little empirical and theo-
retical attention had been paid to the link between natural geophysical disaster risks and


                                            2
municipal bond issues. However, given the salience of the issue generated by Katrina,
this is changing, although much of this work has focused exclusively on the Gulf Coast
(van Kuller, 2005; Denison, 2006; Marlowe, 2006; Handley, 2006). While it is clear
that Katrina has raised awareness of the importance of considering the potential impact
of a specific type of natural disaster-hurricanes-in a particular geographic market-the
Gulf Coast-it is a priori unclear whether or not Katrina has caused changes perceptions
of risk in other geographic locations and for other types of natural disasters. In this pa-
per, we attempt to fill this gap in the literature through an analysis of natural disasters
and the municipal bond market by looking at another type of natural disaster in another
location: namely, earthquake disaster risk in California.
    Earthquakes are low frequency, high consequence events that create widespread
losses over a large geographic area (FEMA, 2000). As such, they have been widely
recognized as having the greatest potential impact on municipalities and therefore,
by extension, bond repayment. Like other natural disasters, earthquake hazard is not
evenly distributed across geographic locations. A recent study by the Federal Emer-
gency Management Agency estimated that of the estimated $4.4 billion dollar annual
loss due to earthquakes, 84 percent of those losses were sustained by only three states-
California, Oregon and Washington, with California alone accounting for $3.3 billion
of the estimated damage costs (2000). The high risk and high projected losses for
urban areas in California are because of the combination of high seismic hazard and
relatively high economic exposure due to the concentration of municipalities (and pop-
ulation) found in California’s higher risk areas.
    California has had a long history of dealing with earthquakes. The first recorded
earthquake in California was reported by the Gaspar de Portola Expedition in 1769,
almost 75 years before California would become a state. The first recorded casualties
as a result of earthquake occurred in 1812 with a church collapse due to an earthquake
centered near San Gabriel. More recently, data from the National Oceanic and Atmo-
spheric Administration (NOAA) shows that, since the 1906 San Francisco Earthquake,
20 of the 36 U.S. earthquakes resulting in at least $1 million in damages have occurred
in California.
    In attempting to better understand and mitigate earthquake risk, the California state
government has had a long history of earthquake hazard analysis which dates back to
the early 1950s, when the California Division of Mines and Geology issued a report
regarding the impact of the Arvin-Tehachapi earthquakes. Legislative mandates as
early as 1973 directed state geologists to gather and publish data regarding known
active fault lines. These early efforts, though certainly a step in the right direction,
were ultimately less than comprehensive and their utilization by the public was minimal
(Palm, 1981).
    This all changed with the Northridge earthquake in 1994, which occurred along a
previously unknown fault line. This single quake accounted for $12.5 billion dollars in
total damages, making it the costliest earthquake in U.S. history and one of the costliest
natural disasters in U.S. history (Mileti, 1999). As a result of the damages caused by
Northridge, state and local governments passed enhanced construction codes designed
to make all new public buildings, homes, and highways more resistant to earthquake
damage. Additionally, the California Legislature quickly passed a series of laws de-
signed to address the emerging collapse of the private market for residential earthquake

                                            3
insurance. Perhaps the most significant was the Seismic Hazard Mapping Act (SHMA),
which mandated that the Division of Mines and Geology produce and make publicly
available seismic hazard maps for the entire state. The SHMA also mandated that all
real estate transactions in California include a ”Natural Hazard Disclosure Statement”
which discloses the assessed natural disaster hazard associated with the property.
    Further, the state created the California Earthquake Authority (CEA), a publicly
managed, privately funded entity to sell earthquake insurance policies to homeown-
ers through private insurance companies. California state law mandates that all sell-
ers of homeowner insurance offer earthquake insurance. Private insurance firms have
the option of offering earthquake coverage themselves, or, alternately, selling their
clients earthquake insurance policies through the CEA. In determining rates, the CEA
is legislatively mandated to set rates ”based on the best available scientific informa-
tion for assessing the risk of earthquake frequency, severity, and loss (California Code
10089.40(a)).” Additionally, the enabling legislation for the agency also specifically
requires that the CEA set rates that accurately reflect relative earthquake hazard, and
specifically prohibits both setting rates in such a way that high risk areas are subsi-
dized by low risk areas and setting a uniform rate for the entire state (California Code
10089.40(c)).
    Ultimately, these legislative efforts had the combined effect that, compared to other
types of natural disasters in other geographic locations, earthquake risk in California
is not only broadly known, but has been known for a fairly long period. Given this,
and the fact that it has been widely recognized that earthquakes pose significant risk
to municipal infrastructure and revenue streams, we test whether or not geologic es-
timates of relative underlying earthquake risk in California impact the cost of capital
for California municipalities when issuing bonds in the municipal market. Specifically,
we would expect that municipalities located in high earthquake risk areas pay a risk
premium to compensate investors for the additional risk posed by earthquake, resulting
in higher interest costs.
    Second, given the attention that has been paid to the impact of Katrina on Gulf
Coast municipal bond markets, we also test whether or not there is any significant dif-
ference in the impact of earthquake risk on true interest costs for municipal issuers
following Hurricane Katrina. Given the empirical literature which finds the municipal
bond market to be somewhere between weak and semi-strong, we hypothesize that the
relationship between the two should be unaffected by Katrina-an event which is cer-
tainly completely exogenous to geologic earthquake risk, and, we hypothesize, public
knowledge of this risk.


3    Data Sources
Earthquake risk in our analysis is measured using earthquake hazard data from the
California Earthquake Authority (CEA). The rates set by the CEA for residential earth-
quake insurance coverage differ by rating territory, which are calculated by zip code.
The CEA divides California into nineteen territories which represent contiguous geo-
graphic areas. These territories are assigned rates which are based on the territory’s
estimated geologic earthquake risk. Risk is determined according to the results of a


                                           4
comprehensive geological analysis which draws upon data from the United States Ge-
ological Survey, the California Geologic Survey, and incorporates the results of com-
puterized disaster simulations utilizing cutting edge simulation software and the most
current geological data, considering proximity to fault and soil quality, among others.
The CEA’s earthquake hazard estimates are widely considered to be among the best
available estimates of earthquake risk in California. Specifically, the variable included
represents the cost in dollars per $1000 of residential earthquake insurance for a sin-
gle story, frame construction home built in 1991 or later with a 15 percent deductible,
which ranges from $.40 in the lowest risk territory to $2.66 in the highest risk territory.1
    Our sample includes 529 California municipal bonds issued between 2004 and
2006, of which forty percent were issued after Hurricane Katrina occurred. Bond is-
sue data are from the California Office of Revenue, with complementary information
regarding issue characteristics taken from the SDC Platinum database.2 Issues are se-
lected into our sample based on the following criteria: first, we only include California
municipal debts sold between January 1, 2004 and December 31, 2006. Second, bonds
issued by California state government, state agencies, and inter-jurisdictional entities
are excluded, as their earthquake risks are unable to be determined and are incom-
parable with those of other local issues. Third, only long term debts (with years to
maturity of at least one year) are included. Finally, bonds with missing values, mainly
true interest costs or underlying ratings, are excluded. All bond issuer zip codes were
determined by performing an internet search for the bond issuer and identifying the
mailing address for the issuing entity. Issuers are assigned earthquake risk depending
upon the rate assigned to the rating territory in which the bond issuer’s mailing address
resides.


4      Empirical Model
The true interest cost can be viewed as the after-tax risk-free interest rate plus a risk
premium, which in turn is a function of bond risk determinants (Poterba and Rueben
2001):

T IC = (1 − τ )ρ + σ(earthquake risk, bond attributes, market environment)

where TIC is the true interest cost of a municipal bond; τ and ρ are the income tax
rate and the market risk-free interest rate respectively; and σ is the risk premium that
investors require to compensate for the unsystematic risk of a particular issuer. The
risk of a municipal bond, as discussed above, may be affected by the earthquake in the
case of California bonds, and other bond attributes and the market environment.
    To control for the different tax treatments received by municipal bonds, we include
two tax treatment variables: whether a bond is subject to the federal income tax and
whether a bond is subject to alternative minimum tax. The Bond Buyer 20 Index is
    1 The CEA rates differ according to construction characteristics, which act as a multiplier to the territory’s

underlying risk assessment. These multipliers are identical across territories, making choice of construction
characteristics arbitrary as long as they are held constant across territories.
   2 Source: TIC and Other Characteristics of Municipal Bond Issues, SDC Platinum, a Thomson Financial

product. Accessed March, 2007


                                                        5
      used to measure the market interest rate, r. Other relevant bond attributes included as
      controls are principal amount, years to maturity, underlying credit rating, whether or
      not a bond is issued with a call feature, and whether or not a bond is issued to refund
      a prior bond issue. Since bond sale method has also been empirically shown to affect
      interest rates in the primary market, it is also included as a control variable (Robbins,
      2002; Simonsen and Robbins, 1996).
          Most aforementioned independent variables are commonly employed in previous
      analyses of municipal bond interest costs and deserve no further discussion.3 Table 1
      provides a brief summary of the variables and the expected directions of their effects
      on true interest cost for municipalities issuing debt in the municipal market.

                                         TABLE I ABOUT HERE


      5      Empirical Analysis and Estimation
      Two OLS regression models were specified to test the two research hypotheses. Model
      I tests the hypothesis that earthquake risk matters in determining TIC for California
      municipal bond issues. It contains the control variables specified above and also in-
      cludes the variable measuring relative earthquake risk. Model II tests the hypothesis
      that Katrina had no effect on TIC for
5. Empirical Analysis and Estimation California municipal bond issues. Both models
      include an interaction term between general obligation bond and earthquake risk be-
      cause it is reasonable to believe that test the two research hypotheses. obligation bonds
Two OLS regression models were specified toearthquake risk impacts general Model I tests the
      and that earthquake risk matters although a priori for California municipal bond issues. It
hypothesisrevenue bonds differently, in determining TICwe are not sure of the form this differ-
      ing impact would take. The squares and also includes the and years to maturity are also
contains the control variables specified above of principal amount variable measuring relative earthquake
      Model II to reflect the nonlinear nature of no effect variables. The models are specified
risk. included tests the hypothesis that Katrina hadthese two on TIC for California municipal bond issues.
      as follows:
The models are specified as follows:

          Model I:
          TIC = β0+ β1(quake risk) + β2(GO bond dummy)+ β3(GO bond dummy × quake risk) +
          β4(Bond Buyer 20 Index) + β5(bond rating) + β6(bidding dummy) + β7(refunding dummy)
          + β8(taxable dummy) + β9(alt. minimum tax dummy) + β10(principal) + β11(principal2) +
          β12(years to maturity) + β13(years to maturity2) + β14(insured dummy) + ε

          Model II:
          TIC = β0+ β1(quake risk) + β2(post-Katrina dummy) + β3(post-Katrina dummy × quake
          risk) + β4(GO bond dummy) + β5(GO bond dummy × quake risk) + β6(Bond Buyer 20
          Index) + β7(bond rating) + β8(bidding dummy) + β9(refunding dummy) + β10(taxable
          dummy) + β11(alt. minimum tax dummy) + β12(callable dummy) + β13(principal) +
          β14(principal2) + β15(years to maturity) + β16(years to maturity2) + β17(insured dummy) + ε

Both models include an interaction term between general obligation bond and earthquake risk because it is
           Table 2 reports the results of these regressions.
reasonable to believe that earthquake risk impacts general obligation bonds and revenue bonds differently,
although a priori we are not sure of the form this differing impact would take. The squares of principal
                                         TABLE II ABOUT HERE
amount and years to maturity are also included to reflect the nonlinear nature of these two variables.
         3 See Solano (2005) and Peng (2002) for a review of the determinants of municipal bond true interest cost.
Table 2 reports the results of these regressions.

Table 2: Regression Results

                                        Model I  6                                   Model II
                                           Std.
 True Interest Cost              Coef.     Err.    t                        Coef.       Std. Err. t
 Quake risk                       0.051     0.034 1.50                       0.005         0.034 0.14
 Post Katrina dummy                                                          0.069         0.066 1.03
 Post Katrina × quake                                                        0.141**        0.055 2.57
 General obligation bond           0.139**       0.061 2.27                  0.170***       0.058 2.92
 GO bond × quake                  -0.033         0.051 -0.65                -0.050          0.048 -1.05
6    Interpretation of Results
In Model I, all of the control variables are statistically significant and signed as ex-
pected, with two exceptions: the general obligation bond dummy and call feature
dummy. The general obligation bond dummy, while statistically significant, is op-
positely signed from both our expectation and previous empirical analyses. The call
feature dummy, while also signed contrary to expectation, is not statistically signifi-
cant. Additionally, it is worthy of mention that the alternative minimum tax dummy
variable, which has been omitted by much of the municipal bond literature, is posi-
tively signed, highly statistically significant, and has a coefficient much smaller than
that of the taxable dummy, a result which is consistent with the findings of Atwood’s
work on municipal bonds and marginal tax rates (2003). However, surprisingly and
contrary to expectation, earthquake risk, while positively signed, is not statistically
significant. This provides some evidence that investors do not demand a risk premium
from municipalities based on the municipality’s relative underlying earthquake risk.
Given the discussion above regarding both the destructive potential and well-known
nature of earthquake risk in California, the reason for the statistical insignificance of
the earthquake risk variable is unknown. While it could be argued that this is sim-
ply representative of inefficiency in the municipal bond market, one plausible alternate
explanation is that the earthquake risk variable is not significant because our model
fails to control for any earthquake readiness measures taken by municipalities which
might serve to offset the underlying earthquake risk. It is possible that investors believe
that current construction standards for homes, public buildings, highways, and the like
are stringent enough as to offset the underlying earthquake risk, and therefore do not
require a risk premium for bonds issued from high earthquake risk areas, although cer-
tainly this is only one possible explanation. Additionally, the variable representing the
interaction between general obligation bond and quake risk is not statistically signifi-
cant, indicating that earthquake risk does not differentially effect revenue and general
obligation bonds.
    Model II is specified similarly to Model I but also includes the dummy variable
which disaggregates the dataset into bonds issued before and after Hurricane Katrina,
and also includes a variable interacting earthquake risk with the post-Katrina dummy
variable in order to allow earthquake risk to impact bonds issued before and after Hur-
ricane Katrina differently. The control variables included in Model II behave similarly
as in Model I in terms of signs and statistical significance. Interestingly, and again con-
trary to expectation, the variable representing the interaction between the post-Katrina
dummy variable and earthquake risk is positive and statistically significant. This tells
us that, contrary to the results generated by Model I, earthquake risk does matter-but
only for municipal bond issues issued after Hurricane Katrina. Specifically, holding all
bond characteristics and market conditions constant, interest costs for municipalities
in the highest risk territory increase by 44 basis points after Katrina. This is an in-
crease in interest costs of 9.85% over pre-Katrina levels. The increase associated with
municipalities in low risk areas is much less-only 13 basis points, an increase of only
2.7%.
    Also of particular relevance is the fact that, while the post-Katrina interaction term
with earthquake risk is statistically significant, the post-Katrina dummy variable itself


                                            7
is not. This tells us that, while municipalities in California issuing debt are paying
higher capital costs after Katrina (since no area in California has an estimated earth-
quake risk of zero), the effect is more subtle than one might initially assume-our results
show that there was no uniform, across-the-board increase in capital costs due to Kat-
rina; instead, capital costs only increased in proportion to the relative earthquake risk
of the issuing municipality.


7    Conclusion
What accounts for this unexpected finding? There are several plausible explanations.
The first is that the municipal bond market is simply quite inefficient, and never consid-
ered earthquake risk when setting risk premiums for California municipal bond issues
before Katrina. This conclusion suggests that the efficiency of the municipal bond
market has been consistently overestimated by prior research. Given the volume and
relative robustness of the previous empirical analyses, this conclusion seems unlikely.
    However, if the previous literature regarding municipal bond market efficiency is
assumed to be correct, our finding suggests something more interesting-that Katrina re-
vealed some new information that affected municipal bond investors’ perception of the
importance of relative earthquake risks in California. What information could have had
such an effect? First, it is possible that, prior to Katrina, investors thought that the pre-
cautions taken by state and local governments in California (in terms of enhanced con-
struction codes and the like) were sufficient to offset the risk posed by natural disasters.
Much of the popular press surrounding the Katrina rebuilding process has focused on
efforts by Gulf Coast state and local governments to re-examine current building codes
and develop stronger codes based on disaster modeling developed from historical data.
It seems possible that Katrina made investors everywhere realize that state and local
efforts to offset natural disaster risk, while previously seen as sufficient, are, in reality,
inadequate. This is subtly different than the argument posed above-the difference being
whether or not the market should have anticipated the risk before Katrina. Above, it is
argued that the market should have recognized the risk before Katrina; here, it is being
argued that the market reacted to the release of previously-unknown information-that
existing governmental standards and preventive measures are inadequate to offset the
underlying earthquake risk.
    Second, it is possible that the federal reaction to Katrina is the new information to
which investors reacted. Certainly, the United States has experienced natural disasters
before Katrina; however, what set Katrina apart (aside from the sheer magnitude of
the storm and associated flooding) is perhaps the perceived ineptitude and slowness
of the federal government in responding to the disaster. Contrast this with the federal
reaction to earlier disasters, such as the Northridge earthquake or Hurricane Andrew,
in which the federal government responded quickly and decisively to provide aid (both
in terms of assistance and federal dollars). This line of reasoning posits that the federal
response to Katrina caused investors to realize that federal bailout of disaster-stricken
areas might not always be as prompt and complete as it was during Northridge, mak-
ing relative underlying earthquake risk matter more than it did in the pre-Katrina era.
Ultimately, further analysis is needed to evaluate the validity of these potential expla-


                                             8
nations.
    Clearly, Katrina has brought The risk of a municipal bond, as discussed above, interaction
   unsystematic risk of a particular issuer. to the forefront the importance of the may be affectedbe-
   by the earthquake in the case
                                   market and natural disasters. Our and the market environment.
tween the municipal bondof California bonds, and other bond attributes research extends that link-
     by moving different tax treatments received by municipal bonds, we by looking not at
ageTo control for thefrom an ex post to an ex ante framework include two tax treatmentnatural
   variables: whether a bond disaster the and income tax of that risk on the cost of capital
disasters, but at naturalis subject toriskfederal the effectand whether a bond is subject to alternative for
   minimum tax. The Bond Buyer 20 Index is used to measure new dimension to Other relevant
issuers in the municipal market. This provides athe market interest rate, r. the tests of bond
   bond attributes included as controls are principal amount, years to maturity, underlying credit rating,
                             considers not only how markets a bond is issued to refund a prior bond
market efficiency thatissued with a call feature, and whether or not deal with disaster, but also how
   whether or not a bond is
                            and how investor perceptions to affect interest rates in the primary
they anticipate them, method has also been empirically shownof risk can change over time.
   issue. Since bond sale
    market, it is also included as a control variable (Robbins, 20002; Simonsen & Robbins, 1996).

Tables costs andindependent further discussion. Table 1 providesprevioussummary of the variables
  Most aforementioned
  bond interest       deserve no
                                 variables are commonly employed in
                                                                    a brief
                                                                            analyses of municipal
                                                                                      3

  and the expected directions of their effects on true interest cost for municipalities issuing debt in the
Table I: Descriptive Statistics and Expected Signs
  municipal market.

    Table 1: Descriptive Statistics and Expected Signs of Variables

                                                                                                                       Std.   Expected
   Variable                                                                    Description                     Mean    Dev.     sign
   True Interest Cost                                                 True interest cost (percentage)           4.63   0.63

   Quake risk                                         Earthquake risk, measured as dollars per $1,000 of        1.28   0.70      +
                                                    residential coverage as set by the California Earthquake
                                                                           Authority

   Bond Buyer 20 Index                                Bond Buyer Index for 20-year GO bonds with mixed          4.48   0.23      +
                                                               quality, quoted every Thursday

   Bond rating                                                Underlying credit rating coded from 0 to 8*       3.38   1.74      -

   Bidding dummy                                                   Dummy variable, competitive sale=1           0.19             -

   General obligation bond                                   Dummy variable, general obligation bond=1          0.53             -
   dummy

   Refunding dummy                                                 Dummy variable, refunding bond=1             0.41             -

   Taxable                                               Dummy variable, subject to federal income tax=1        0.10             +

   Alt. minimum tax dummy                           Dummy variable, subject to alternative minimum tax=1        0.01             +

   Callable dummy                                                  Dummy variable, with call option=1           0.84             +

   Principal                                                        Principal amount ($100,000,000)             .399   .685      -

   Years                                                                    Years to maturity                  22.92   6.65      +

   Post Katrina dummy       Dummy variable, bond issued after October, 2005=1 0.41
   *Note: 0=below BBB; 1=BBB-, BBB, BBB+; 2=A-; 3=A; 4=A+; 5=AA-; 6=AA; 7=AA+; 8=AAA.


                                                                
    3
        See Solano (2005) and Peng (2002) for a review of the determinants of municipal bond true interest cost.

                                                                                     5




                                                                                     9
Table II: Regression Results
Table 2: Regression Results

                                       Model I                                  Model II
                                          Std.
 True Interest Cost             Coef.     Err.            t            Coef.       Std. Err. t
 Quake risk                      0.051     0.034          1.50          0.005         0.034 0.14
 Post Katrina dummy                                                     0.069         0.066 1.03
 Post Katrina × quake                                                   0.141**        0.055    2.57
                                      **
 General obligation bond         0.139        0.061    2.27             0.170***       0.058    2.92
 GO bond × quake                -0.033        0.051   -0.65            -0.050          0.048   -1.05
 Bond Buyer 20 Index             0.546***     0.068    8.00             0.606***       0.066    9.17
 Bond rating                    -0.029**      0.013   -2.27            -0.029**        0.012   -2.30
 Bidding dummy                  -0.229***     0.045   -5.12            -0.218***       0.041   -5.30
 Refunding dummy                -0.122***     0.043   -2.85            -0.100**        0.040   -2.48
 Taxable bond dummy              1.128***     0.095   11.91             1.110***       0.086   12.93
 Alt. minimum tax dummy          0.396***     0.107    3.71             0.389***       0.088    4.40
 Callable dummy                 -0.010        0.065   -0.15             0.017          0.062    0.28
 Principal                      -0.124***     0.045   -2.77            -0.118***       0.005   -2.58
 Principal2                      0.024**      0.012    2.07            -0.105*         0.000    1.74
                                      ***
 Years to maturity               0.083        0.029    2.88             0.085***       0.027    3.10
 Years2                         -0.001        0.001   -1.63            -0.001*         0.000   -1.88
 Insured dummy                  -0.455***     0.118   -3.85            -0.474***       0.115   -4.14
 constant                        1.243***     0.465    2.67             0.930**        0.447    2.08
                                         N = 529                                 N= 529
                                       R2 = .6052                               R2 = .6469
                                   Adjusted R2 = .594                      Adjusted R2 = .635
 ***
       p<.01, ** p<.05, *p<.1


6. Interpretation of Results

In Model I, all of the control variables are statistically significant and signed as expected, with two
exceptions: the general obligation bond dummy and call feature dummy. The general obligation bond
dummy, while statistically significant, is oppositely signed from both our expectation and previous
empirical analyses. The call feature dummy, while also signed contrary to expectation, is not statistically
significant. Additionally, it is worthy of mention that the alternative minimum tax dummy variable, which
has been omitted by much of the municipal bond literature, is positively signed, highly statistically
significant, and has a coefficient much smaller than that of the taxable dummy, a result which is consistent
with the findings of Atwood’s work on municipal bonds and marginal tax rates (2003). However,
surprisingly and contrary to expectation, earthquake risk, while positively signed, is not statistically
significant. This provides some evidence that investors do not demand a risk premium from municipalities
based on the municipality’s relative underlying earthquake risk. Given the discussion above regarding

                                                      7




                                                 10
References
Atwood, T. J. (2003). ”Implicit taxes: Evidence from taxable, AMT, and Tax-exempt
State and Local Government Bond Yields.” The Journal of the American Taxation As-
sociation 25(1): 1.
Crittenden, M. R. (2005). ”Katrina Squeezes Local Bond Market.” CQ Weekly 63(36):
2464-2465.
Denison, D. (2006). ”Bond Market Reactions to Hurricane Katrina: An Investigation of
Prices and Trading Activity of New Orleans Bonds.” Municipal Finance Journal 27(2):
39-52.
Denison, D. (2000). ”Did bond Fund Investors Anticipate the Financial Crisis of Or-
ange County.” Municipal Finance Journal 21(3): 24-39.
Denison, D. (2003). ”An Empirical Examination of the Determinants of Insured Mu-
nicipal Bond Issues.” Public Budgeting and Finance 23(1): 96-114.
Fama, E. F. (1970). ”Efficient Capital Markets: A Review of Theory and Empirical
Work.” The Journal of Finance 25(2): 383-417.
FEMA (2000). HAZUS99 Estimated Annualized Earthquake Losses for the United
States.
Handley, D. M. (2006). ”Hurricanes on the Alabama Gulf Coast: The Manageable
Impacts of Ivan and Katrina.” Municipal Finance Journal 27(2): 95-111.
Jaffee, D. M. and T. Russell (1997). ”Catastrophe Insurance, Capital Markets, and
Uninsurable Risks.” The Journal of Risk and Insurance 64(2): 205-230.
Kilger, D. and O. Sarig (2000). ”The Informational Value of Bond Ratings.” The Jour-
nal of Finance 55: 2879-2902.
Marlowe, J. (2006). ”Volume, Liquidity, and Investor Risk Perceptions in the Sec-
ondary Market: Lessons from Katrina, Rita, and Wilma.” Municipal Finance Journal
27(2): 1-37.
Mileti, D. S. (1999). Disasters by Design : a Reassessment of Natural Hazards in the
United States. Washington, D.C., Joseph Henry Press.
Miller, J. and D. Bonifer (2006). Hurricane Katrina: One Year Later: The Municipal
Bond Market Remains Resilient. GKST Municipal Research.
Palm, R. and M. Hodgson (1992). ”Earthquake Insurance: Mandated Disclosure and
Homeowner Response in California.” Annals of the Association of American Geogra-
phers 82(2): 207-222.
Palm, R. I. (1981). ”Public Response to Earthquake Hazard Information.” Annals of
the Association of American Geographers 71(3): 389-399.
Peng, J. (2002). ”Do Investors Look Beyond Insured Triple-A Rating? An Analysis of
Standard Poor’s Underlying Ratings.” Public Budgeting Finance 22(3): 115-131.
Peng, J. and P. F. Brucato (2004). ”An Empirical Analysis of Market and Institutional
Mechanisms for Alleviating Information Asymmetry in the Municipal Bond Market.”
Journal of Economics and Finance 28(2): 227-239.

                                         11
Poterba, J. M. and K. S. Rueben (2001). ”Fiscal News, State Budget Rules, and Tax-
Exempt Bond Yields.” Journal of Urban Economics 50(3): 537.
Reck, J. L. and E. R. Wilson (2006). ”Information Transparency and Pricing in the
Municipal Bond Secondary Market.” Journal of Accounting and Public Policy 25(1):
1-31.
Richardson, J. A. (2006). ”Natural Disasters and State and Local Finance in Louisiana:
A Case Study in 2005.” Municipal Finance Journal 27: 53-74.
Robbins, M. D. (2002). ”Testing the Effects of Sale Method Restrictions in Municipal
Bond Issurance: The case of New Jersey.” Public Budgeting Finance 22(2): 40.
Simonsen, W. and M. D. Robbins (1996). ”Does It Make Any Difference Anymore?
Competitive versus Negotiated Municipal Bond Issuance.” Public Administration Re-
view 56(1): 57-64.
Solano, P. L. (2005). ”An Appraisal of the Interest Cost Savings of State Bond Banks.”
Municipal Finance Journal 25(4): 13.
van Kuller, K. (2005). ”No Quick Fix for Bonds in Gulf Disaster area: State, Not
Federal, Debt Relief Plans Emerging.” Municipal Finance Journal 26(3): 45-72.
Wells, R. (2005). White House Opposes Muni-Bond Bailout. Wall Street Journal: Oc-
tober 7, 2005: C4.
Wildasin, D. ”Disaster Policies: Some Implications for Public Finance in the U.S.
Federation.” Public Finance Review, forthcoming.




                                         12

								
To top