Post OPA-90 Vessel Oil Transfer Spill Prevention: The

Post OPA-90 Vessel Oil Transfer Spill Prevention:

The Effectiveness of Coast Guard Enforcement

WAYNE K. TALLEY1,*, DI JIN2 and HAUKE KITE-POWELL2

1Department of Economics, Old Dominion University, Norfolk, VA 23529, USA;2Marine Policy Center, Woods Hole Oceanographic Institution, Woods Hole, MA 02543, USA;*Author for correspondence (e-mail: [email protected])

Accepted 16 June 2004

Abstract. Although oil spills from tanker accidents receive the most publicity, most vesselspills are not the result of accidents but of oil transfer activities. We investigate determinants

of the size of vessel oil transfer spills as well as the effectiveness of Coast Guard enforcementactivities in reducing their size. Vessel out-of-water and in-water oil transfer spillage functionsare estimated utilizing tobit regression and detailed data on individual vessel oil transfer spills

as well as Coast Guard safety/environmental enforcement activity data for the 1991–1995period. Our estimation results suggest that Coast Guard hull but not machinery inspectionsare effective in reducing both out-of-water and in-water spills; patrols by air, but not by boat,are effective in reducing out-of-water spills; but neither is effective in reducing in-water spills.

The results also show that the type of vessel (oil- and non-oil-cargo), vessel characteristics,vessel operations, weather/visibility conditions, and waterway type are determinants of postOPA-90 vessel oil transfer spills.

Key words: oil spill, oil transfer, vessel, water pollution

JEL classification: Q25, K32, L92, L51

1. Introduction

In March 1989 the Exxon Valdez tank ship ran aground in Prince WilliamSound, Alaska, spilling nearly 37,000 tons of oil. This tanker accident spillranks 34th in size among worldwide vessel oil spills on record, but it was themost expensive in oil spill history (White and Baker 1998). Exxon has paid$2.2 billion for clean-up, $1 billion to settle state and federal lawsuits, and$300 million for lost wages to 11,000 fishermen and business firms. The costto the fisheries of south-central Alaska has been estimated to be $108.1million, the largest component being a $65.4 million reduction in the pinksalmon fishery in the first year following the accident (Cohen 1995). In 1994an Alaska jury awarded an additional $5.3 billion in punitive and compen-satory damages to those harmed by the Exxon Valdez oil spill. Exxon’sappeal was rejected by an Alaska appeals court in March 2000. The Exxon

Environmental & Resource Economics (2005) 30: 93–114 � Springer 2005

Valdez accident was the impetus for the US Congress to pass the OilPollution Act of 1990 (OPA-90), which strengthened accountability for vesseloil spills in US waters.

Although oil spills from tanker accidents receive the most publicity, mostspills are not the result of vessel accidents but of oil transfer activities, i.e.,routine operations that involve the movement (intentional or unintentional)of oil cargo and/or fuel oil to and from vessels. Such activities include loadingand unloading of oil cargoes, fueling, cleaning tanks, bilge pumping,and ballasting.1 Statistics for the 1970–2000 period on worldwide oil spills byoil-cargo vessels (tank ships, tank barges, and combination oil-cargo/non-oil-cargo ships) collected by the International Tanker Owners Pollution Feder-ation (ITOPF 2002) reveal that 53% are transfer spills and only about 21 %are the result of vessel accidents.2

Studies by Epple and Visscher (1984), Cohen (1987), and Grau andGroves (1997) have investigated oil transfer spills from oil-cargo vesselsand the effectiveness of Coast Guard enforcement activities in reducingthis spillage. In general, these studies conclude that enforcement activitieshave been effective in reducing vessel oil transfer spills. For 1973–1975 oiltransfer spills, Epple and Visscher (1984) found a negative relationshipbetween spill size and Coast Guard man-hours per oil transfer activity.For 1973–1977 spills, Cohen (1987) found negative relationships betweenspill size and both Coast Guard man-hours monitoring transfers andpatrol man-hours per transfer activity. Based upon 1984–1987 spills, Grauand Groves (1997) found a negative relationship between spill size and theprobability of oil transfer operations being monitored by the CoastGuard.3

The purpose of this study is to investigate determinants of the size ofvessel oil transfer spills as well as the effectiveness of Coast Guard enforce-ment activities in reducing the size of these spills. It differs from previousstudies of vessel oil transfer spills in several ways. We examine spills from thepost OPA-90 period, extending existing empirical studies of similar spills inthe 1970s and 1980s (Epple and Visscher 1984; Grau and Groves 1997).While previous studies have focused on oil-cargo vessels (i.e., tankers andbarges), we consider transfer spills from non-oil-cargo vessels as well andexplore spill size variation with respect to vessel type. Also, we developseparate functions for in-water and out-of-water spills and analyze theeffectiveness of Coast Guard enforcement on the two types of spills. Inaddition, vessels that were not involved in a detected spill are considered, andspills are adjusted with respect to vessel size. Finally, unlike previous studies,we investigate whether vessel operations, weather/visibility characteristics,type of waterway, and the vessel’s safety record are determinants of vessel oiltransfer spillage.

WAYNE K. TALLEY ET AL.94

2. OPA-90

OPA-90, the US Congress response to the Exxon Valdez accident,strengthened the regime of accountability for vessel oil spills in US waters.Under OPA-90, the strict liability of the vessel owner is limited to $1200 pervessel gross ton or $10 million for vessels over 3000 gross tons, whichever isgreater. If the vessel owner, or other liable party, is found guilty of grossnegligence or in violation of laws, liability is unlimited and liable parties aresubject to both criminal and civil sanctions. Further, claims can be filedagainst all parties who have an interest in the ownership or operation of thevessel and fit OPA-90’s definition of a ��responsible party’’. OPA-90 alsorequires vessels utilizing US waters to carry certificates of financial respon-sibility (COFRs), proving that the owners have funds or insurance to coverthe maximum liability limits under the law.4

OPA-90 mandates double hulls for tank ships and tank barges traveling inUS waters by the year 2015 – the assumption being that double hulls willreduce the likelihood or severity of damage to a vessel’s tanks and therebyreduce the amount of oil spilled.5 OPA-90 also requires interim structuraland operational measures to reduce the outflow of oil in the event of an oilspill (until 2015) for tank ships and barges of 5000 gross tons or more withoutdouble hulls. The Coast Guard’s rule requiring operational measures, how-ever, did not take effect until November 1996 (Kim 2002).

The total number and volume of oil spills in US waters from tank shipsand barges have declined considerably since enactment of OPA-90 and theimplementation of the financial responsibility regulations (US Coast Guard2002). The number of tank ship (barge) oil spills declined from 2383 (5048)for the 1981–1990 period to 1375 (3051) for the 1991–2000 period. Thevolume of tank ship (barge) oil spills declined from 27,651,000 (21,380,091)gallons for the 1981–1990 period to 1,390,138 (5,015,244) gallons for the1991–2000 period. For the same time periods, the number of oil spills by non-oil-cargo vessels increased from 13,805 to 46,728, while the volume of spillsdecreased from 5,307,440 to 3,384,730 gallons.

The double-hull regulation of OPA-90 was subsequently adopted by theinternational community, when the International Maritime Organization(IMO) adopted regulations banning single-hull tank ships by the year 2015.6

However, the sinking of the single-hull tank ship, the Prestige, in Novemberof 2002 off the northwest coast of Spain has accelerated this time table forEurope. The Prestige was carrying 20.5 million gallons of heavy fuel oil whenit broke in half and sank. In March 2003 the transport ministers of the 15European Union (EU) member states agreed to the prohibition of all single-hulled tank ships by 2010, with the ban for larger ships starting in 2005. Theministers also unveiled plans for heavy fines and jail sentences for shipownersand other transport operators held responsible for polluting EU waters.

EFFECTIVENESS OF COAST GUARD ENFORCEMENT 95

These liability regulations are similar to those of OPA-90 and are expected tolead to a similar reduction in tanker spills in EU waters.

3. The Model

We consider vessels of different types ( j) (e.g., tanker and fishing boat). Avessel (i.e., firm) of type j randomly causes an externality (i.e., oil spill). Thefirm can make certain prevention efforts (e) to reduce the resulting level ofpollution. In a given period (e.g., year), the vessel operator’s problem is tomaximize the expected net benefit (U )

max EðUÞ ¼ bjðqjÞ � cjðqjÞ � E½vjðqj; z; ejÞ� � ej ð1Þwhere bj is the private benefit, qj is the activity level,7 cj is the cost associatedwith vessel operations, vj is the total spill-related payment, z is a vector ofexogenous variables affecting the occurrence and size of spills, and ej is thelevel of care (i.e., pollution prevention effort).8

The total spill-related payment (vj) includes all spill-related costs such aspenalty payment, environmental damage, cleanup cost, lost cargo, and otherprivate losses. vj is a stochastic variable affected by the activity level (qj), otherfactors (z), and the pollution control effort (ej).

Since the total pollution (i.e., gallons of oil spilled) associated with qj isdetermined by the number of spills in the time period (nj) and correspondingspill size (xj), the expected value of vj in Equation (1) can be expressed as

E½vjðqj; z; ejÞ� ¼Z Z

n;x

vjðnj; xj; z; ejÞujðnj;xj; z; ejÞdxdn ð2Þ

where ujð�Þ is the probability density function. The level of care (ej) affectsthe level of pollution by shifting ujð�Þ. In fact, ej can reduce both the size (xj)and number (nj) of spills. The optimal level of care (ej*) by the vessel operatorcan be determined by substituting Equation (2) into (1) and then differenti-ating Equation (1) with respect to ej.

However, the resulting ej* is conditioned on the firm’s total spill-relatedpayment (vj), which is affected by government regulations (e.g., penaltystructure and liability rules).9. Ideally, regulations and enforcement are de-signed optimally so that vj covers the social cost of pollution and ej* reflectsthe socially optimal level of care.10 Specifically, the social planner (i.e.,regulator) is to maximize the expected net social benefit (W ) from all vesselactivities.

max EðWÞ ¼Xj

Xk

E½Wð j; kÞ� ð3Þ

The social benefit is the sum of all individual vessels (k) in different vesseltype ( j) groups. For a vessel of type j


E½Wð jÞ� ¼ BjðqjÞ � cjðqjÞ � E½Djðqj; z; ejÞ� – ej – mj ð4Þ

where Bj is the social benefit, Dj is the spill-related social damage, and mj is avector of Coast Guard monitoring and enforcement activities. Theoretically,mj > 0 may be necessary, depending on other factors that are not included inour simple model, such as the level of spill detection by the regulator, riskpreference of the firm, and risk-sharing arrangements between the regulatorand the firm (see Cohen 1987).

To ensure the maximization of Equation (3), the regulator has to choosethe optimal design of regulation and enforcement (mj) so that the socialdamage (Dj) and social benefit (Bj) are fully captured in the private paymentfunction (vj) (see Equations (1) and (2), also Cohen (1987)).

In this analysis, we focus on the effect of Coast Guard enforcement (mj) onthe size of spills given occurrence (xj), using Coast Guard pollution incidentdata. As in other empirical studies (Epple and Visscher 1984; Grau andGroves 1997), we assume that xj and nj are independent, so thatujðnj; xj; z; ejÞ ¼ hjðnj; z; ejÞgjðxj; z; ejÞ. We combine spills from different vesseltypes in a joint density function (g), where vessel type ( j) is one element in z.

x � gðe�ðmÞ; zÞ ¼ gðm; zÞ ð5ÞRecall that the firm chooses the optimal level of care (e*) for a given penaltystructure and enforcement level (m) using Equations (1) and (2). Thus, e* is afunction of m. Equation (5) allows us to investigate determinants (z) of spillsize (x) and effectiveness of Coast Guard enforcement activities in reducing x.

In our analysis, the vector z includes a number of variables: the type ofvessel, vessel characteristics, nature of vessel operations, weather/visibilityconditions, type of waterway, and the price of oil. The vector of vessel safety/environmental regulation enforcement activities (m) includes the time (hours)spent by the US Coast Guard on inspection of hulls and machinery, boatpatrols, and aircraft patrols11 (see Table I).

4. Data

Vessel out-of-water (OWATSP) and in-water (IWATSP) oil transfer spillfunctions are estimated utilizing detailed data on individual vessel oil transferspills investigated by the Coast Guard as well as data on Coast Guard safety/environmental enforcement activities at the port level during the five-year(post OPA-90) time period 1991–1995. Variables used in the estimations andtheir specific measurements appear in Table I.

Data for all variables except PRICEOIL were extracted from the USCoast Guard Marine Safety Management System (MSMS) database.12

Monthly oil price data were obtained from the US Department of Energy.


Table I. Variable definition and descriptive statistics

Measurement Mean (SD)

Dependent variables

Out-of-water oil-spillage,

OWATSP

Gallons 14.5 (253)

In-water oil-spillage,

IWATSP

Gallons 77.5 (1600)

Explanatory variables

Type of vessel a

TANKBARGE 1 if a tank barge, 0 otherwise 0.057 (0.231)

TANKSHIP 1 if a tank ship, 0 otherwise 0.031 (0.172)

FISHBOAT 1 if a fishing boat, 0 otherwise 0.110 (0.313)

PASSBOAT 1 if a passenger boat, 0 otherwise 0.024 (0.153)

FRTSHIP 1 if freight ship, 0 otherwise 0.048 (0.214)

RECREBOAT 1 if a recreation boat, 0 otherwise 0.093 (0.291)

Vessel characteristics

Age, VAGE Years 22.1 (14.4)

Size, VSIZE Gross tons 5570 (15271)

US Flag, USFLAG 1 if a US Flag, 0 otherwise 0.917 (0.276)

POORSAFb 1 if vessel has a poor safety record,

0 otherwise

0.156 (0.362)

Vessel operation phase c

UNDERWAY 1 if vessel is underway, 0 otherwise 0.042 (0.200)

MOORDOCK 1 if vessel is moored or docked, 0 otherwise 0.298 (0.457)

TOWED 1 if vessel is towed, 0 otherwise 0.002 (0.041)

ANCHORED 1 if vessel is anchored, 0 otherwise 0.020 (0.139)

Weather/visibility characteristics

COLDd 1 if cold temperature, 0 otherwise 0.994 (0.080)

HIGHWINDSe 1 if high winds exist, 0 otherwise 0.002 (0.048)

PRECIPTN 1 if precipitation weather, 0 otherwise 0.003 (0.058)

NIGHT 1 if nighttime, 0 otherwise 0.002 (0.045)

Type of waterwayf

COAST 1 a coastal waterway, 0 otherwise 0.264 (0.441)

OCEAN 1 an ocean, 0 otherwise 0.082 (0.275)

RIVER 1 a river, 0 otherwise 0.271 (0.444)

HARBOR 1 a harbor, 0 otherwise 0.108 (0.310)

Vessel safety/environmental regulation enforcement activities

HULLINSP Coast Guard vessel boarding hours

(average daily hours per month per port)

in hull inspection

8.96 (8.96)


Price index data published by the US Department of Commerce were used toconvert current into real (i.e., in December 1995 dollars) oil prices.

Descriptive statistics (mean and standard deviation) for our variables alsoappear in Table I. The mean statistics reveal that 5.7%, 3.1%, 11.0%, 2.4%,4.8%, and 9.3% of the vessels investigated for an oil transfer spill (notinvolving a vessel accident) in our data were tank barges, tank ships, fishingboats, passenger boats, freight ships, and recreational boats; 4.2%, 29.8%,0.2%, and 2.0% were underway, moored or docked, towed, and anchoredwhen a pollution incident occurred; and 26.4%, 8.2%, 27.1%, and 10.8% werein coastal, ocean, river, and harbor waterways.

The weather was cold, winds were high, and precipitation was present in99.4%, 0.2%, and 0.3% of the potential vessel oil spills; 0.2% occurred atnight. The mean statistics also reveal that the average age and size of thevessels are 22.1 years and 5570 gross tons; 15.6% have a poor safety record;and 91.7% are US flag. Coast Guard average daily vessel boarding hours permonth per port for hull and machinery inspection were 8.96 and 2.42. CoastGuard average daily safety/environmental patrol hours per month per portby boat and aircraft were 3.17 and 0.06.

Table I. Continued

Measurement Mean (SD)

MACHINSP Coast Guard vessel boarding hours


in machinery inspection

2.42 (2.97)

BOATPATR Cost Guard safety/environmental patrols

(average daily hours per month per port) by boat

3.17 (69.2)

AIRPATR Coast Guard safety/environmental patrols


by aircraft

0.06 (0.216)

Price variable

Price of oil, PRICEOIL Real monthly price ($/barrel) 15.24 (2.03)

aIn addition to the six vessel types listed below, there are a number of other vessel types in ourdata set, such as industrial vessel, offshore supply vessel, research vessel, and tugboat.bThe variable is from the MSMS Vessel Identification Table (vidt). POORSAF = 1 for vesselsthat have been boarded by the Coast Guard for safety concerns.cThe operation phase was unknown for the rest of the vessels.dIncluding observations with temperature below 31� Fahrenheit or description of icycondition.eIncluding observations with wind speed above 20 miles per hour or description of high wind

condition.fAlso included in the data are lake and other waterways.


One problem with the data is the large number of missing values for vesseloperations (see Table I). We handle the missing values problem by definingan indicator variable MISSING, where MISSING equals 1 if data aremissing for the nature of vessel operations and zero otherwise.13

5. Empirical Results

Given that a vessel investigated for an oil spillage may not have spilledoil, some of the values of the dependent variables OWATSP and IWATSPmay be zero. If so, their distributions will be left-censored. Since ordinaryleast squares ignores censoring, its parameter estimates may be biased. Weavoid such bias by utilizing tobit regression analysis, which explicitly ac-counts for censored dependent variables. Also, possible estimation biasfrom omission of relevant explanatory variables is addressed by consid-ering yearly binary variables and DIST binary variables that distinguishamong Coast Guard Districts.14 Since the sum of UNDERWAY,MOORDOCK, TOWED, ANCHORED, and MISSING equals 1 and inorder to avoid the problem of perfect multicollinearity, we arbitrarily dropUNDERWAY in the estimations.

Table II reports tobit regression estimates of vessel out-of-water and in-water oil spill functions. Focusing initially on the OWATSP results (2ndcolumn in Table II), we see that the function fits the data well. The likelihoodratio statistic is 2741, well above the critical value for significance at the 0.01level. As expected, the results suggest that vessel out-of-water oil transferspillage for the post OPA-90 period is greater for oil-cargo vessels, tankbarges and tank ships, than for non-oil-cargo vessels. Among the lattervessels, spillage is greater for freight ships.

Spillage is less for larger and US flag vessels, but greater for an oldervessel, when the vessel is moored or docked15 and when the spill incidentoccurs in a river and a harbor. Most of these results are as expected. In thestudy, we distinguish between vessels flying the US flag and vessels operatingunder non-US flags. Since the US is a nation that has some of the highestmaritime safety/environmental standards in the world (Gracey 1985), anegative relationship is expected between US FLAG and vessel oil transferspills. Since vessel structural failure is expected to increase with age, the apriori sign of the relationship between vessel age (VAGE) and oil transferspills is positive. Since out-of-water oil transfer spillage (OWATSP) isexpected to be greater when the vessel is moored or docked, a positiverelationship is expected between the former and the latter. A positive rela-tionship is expected between out-of-water oil spills and river and harborwaterways, given their proximity to land. However, the result with respect tovessel size (VSIZE) is not a priori expected to be negative. On-board oil


Table

II.Vesseloiltransfer

spillage:

tobitregressionestimates

Variable

OWATSP

coeffi

cient(t-statistic)

OWATSPT

coeffi

cient(t-statistic)

IWATSP

coeffi

cient(t-statistic)

IWATSPT

coeffi

cient(t-statistic)

Typeofvessel

TANKBARGE

212.9*(7.40)

0.366*(4.18)

195.1*(2.80)

0.118(0.65)

TANKSHIP

245.8*(5.79)

0.059(0.43)

58.91(0.58)

)0.278()1.00)

FISHBOAT

)16.38()0.71)

)0.026()0.35)

)111.7**()2.35)

0.630*(4.46)

PASSBOAT

)137.7*()3.25)

)0.414*()3.37)

)109.7

()1.30)

0.382***(1.76)

FRTSHIP

178.3*(5.02)

0.187***(1.65)

13.91(0.17)

0.143(0.63)

RECREBOAT

)71.97*()2.60)

)0.402*()3.58)

)220.6*()4.72)

0.954*(5.32)

Vesselcharacteristics

VAGE

0.379*(19.28)

)0.003*()27.71)

)0.107*()2.86)

0.009*(50.57)

VSIZ

E)0.002**()2.46)

0.003·10)4(0.13)

0.005*(2.85)

)0.002·10)3()0.48)

USFLAG

)142.3*()5.04)

)0.244*()2.64)

)152.9**()2.41)

)0.218()1.19)

POORSAF

31.80(1.42)

0.076(1.12)

)23.52()0.45)

)0.065()0.47)

Vesseloperationphase

MISSIN

G)44.25()1.32)

0.064(0.59)

)586.3*()9.21)

)1.69*()8.85)

MOORDOCK

120.0*(3.69)

0.357*(3.48)

)384.3*()6.12)

)0.834*()4.62)

TOWED

)268.1

()1.47)

)1.16***()1.77)

1163*(3.97)

0.659(0.78)

ANCHORED

33.88(0.65)

0.189(1.16)

)551.0*()5.31)

)0.756*()2.59)

Weather/visibilitycharacteristics

COLD

)19.02()0.19)

)0.272()0.83)

1539*(7.94)

)1.38**()2.44)

HIG

HWIN

DS

)238.7

()1.36)

)1.02***()1.78)

)2107*()7.08)

3.44*(4.23)

PRECIPTN

44.72(0.34)

0.290(0.70)

2927*(11.39)

1.16(1.64)

NIG

HT

)3.58()0.02)

)0.114()0.20)

4839*(15.46)

4.87*(5.67)


Table

II.Continued

Variable

OWATSP

coeffi

cient(t-statistic)

OWATSPT

coeffi

cient(t-statistic)

IWATSP

coeffi

cient(t-statistic)

IWATSPT

coeffi

cient(t-statistic)

Typeofwaterw

ay

COAST

)7.73()0.38)

)0.122***()1.66)

)28.95()0.82)

0.211(1.51)

OCEAN

)156.9*()4.59)

)0.444*()3.75)

130.6**(2.52)

0.091(0.44)

RIV

ER

34.49***(1.76)

0.112(1.64)

3.07(0.09)

0.102(0.77)

HARBOR

45.42***(1.66)

)0.076()0.74)

35.47(0.74)

)0.026()0.13)

Vesselsafety/environmentalregulationenforcem

entactivities

HULLIN

SP

)3.80*()2.88)

)0.009**()2.09)

)3.98()1.56)

)0.029*()3.32)

MACHIN

SP

3.82(0.96)

0.002(0.12)

7.40(1.01)

0.030(0.11)

BOATPATR

)0.134()1.07)

)0.001·10)1()0.46)

0.003(0.02)

)0.005·10)2()0.10)

AIR

PATR

)70.46***()1.86)

)0.304**()2.26)

)23.63()0.39)

)0.137()0.58)

Price

variables

PRIC

EOIL

9.87**(2.21)

0.036**(2.32)

15.68**(1.96)

0.020(0.66)

Districts

DIST2

95.18**(2.28)

0.191(1.40)

)39.64()0.44)

)0.041()0.15)

DIST5

9.65(0.29)

)0.174()1.40)

22.51(0.38)

0.004·10)1

(0.002)


DIST7

)103.3*()3.55)

)0.109()1.04)

67.49(1.37)

)0.386***()1.94)

DIST8

30.76(1.17)

0.142(1.54)

72.93(1.47)

0.360**(1.98)

DIST9

)49.07()1.04)

)0.436**()2.45)

)99.07()1.19)

)0.562***()1.76)

DIST11

)32.30()0.99)

0.075(0.61)

20.05(0.36)

)0.241()1.03)

DIST13

84.39*(2.58)

0.368*(3.12)

47.80(0.81)

)0.124()0.53)

DIST14

150.2*(3.75)

0.606*(4.16)

58.64(0.76)

1.15*(4.06)

DIST17

114.4*(3.38)

0.349*(3.03)

62.85(0.94)

)0.302()1.33)

Years

1992

)24.09()0.27)

)0.041()0.15)

)153.9

()0.89)

)0.589()1.07)

1993

62.40(0.70)

0.172(0.60)

)148.5

()0.86)

)0.702()1.27)

1994

17.60(0.20)

0.150(0.52)

)135.3

()0.77)

)0.574()1.02)

1995

19.73(0.22)

0.259(0.89)

)136.0

()0.78)

)0.571()1.01)

Constant

)537.8*()3.27)

)1.61*()3.00)

)1064*()3.40)

1.12(1.13)

#Observations

18708

18708

18708

18708

Likelihoodratiotest

statistic

2741

4217

679.0

9863

*(**,***)significantatthe1(5,10)percentlevel.


volume is expected to increase with vessel size, and therefore larger spillsmight be associated with larger vessels.

A negative relationship is expected between vessel inspection/patrol hoursand oil transfer spills. In our results, the negative coefficients of HULLINSPand AIRPATR suggest that an increase in Coast Guard on-board hullinspections and aircraft patrols will reduce out-of-water oil transfer spills.According to the literature (Epple and Visscher 1984; Cohen 1987), the priceof oil (PRICEOIL) is negatively related to transfer spills. For oil-cargovessels, PRICEOIL represents the unit value of oil potentially lost in a spill;in addition, when oil prices are high, carriers may be better able to afford theinvestment necessary to prevent spills. However, the inverse might be true fornon-oil-cargo vessels. The explanation for the positive coefficient ofPRICEOIL in our analysis may be the complex relationships among oil price,oil production, demand for oil tanker shipments, and number of transfers.For example, rising oil prices may result in an increase in production incertain oil exporting counties, which in turn leads to rising transport volumeand an increase in the number of transfers. Large spills could be associatedwith this increase in workload at transfer facilities. Relative to other CoastGuard Districts, this spillage is greater in Coast Guard Districts 2 (theMidwest), 13 (Washington and Oregon), 14 (Hawaii) and 17 (Alaska), butless in District 7 (South Carolina, Georgia and Florida).

Like the OWATSP model, the IWATSP model (4th column in Table II)fits the data well – the likelihood ratio statistic is large and statistically sig-nificant at the 0.01 level. The estimation results suggest that in-water oiltransfer spillage for the post OPA-90 period is greater for tank barges, butless for non-oil-cargo vessels (e.g., fishing and recreational boats). Like out-of-water spillage, in-water oil spillage is less for US flagged vessels. Unlikethe OWATSP results, in-water oil spillage is less the older the vessel, whenmoored or docked and when anchored, but greater when towed and when thespill incident occurs in the ocean. The in-water spillage is greater during coldweather, during precipitation, at nighttime, and during higher oil prices, butless during high winds. Many of these results are as expected. For example, apositive relationship is expected between in-water oil transfer spills (IW-ATSP) and the vessel operation phases UNDERWAY and TOWED. Also,adverse weather and visibility conditions are likely to increase the risk of oilspills, thus their impact on the amount of oil transfer spillage is positive. Animportant result from this analysis is that no Coast Guard spill preventionenforcement activity was found to be statistically significant in reducing in-water vessel oil transfer spills.

Further insight into whether Coast Guard enforcement activities reduceoil transfer spills may be gained by normalizing this spillage for vessel sizeand re-estimating. Since spill size is expected to be positively related to theamount of oil being transferred, our regression analysis should control for


transfer quantity at the time of spill. Ideally, this can be done by usingthe percentage of spill (i.e., quantity spilled/quantity transferred) as thedependent variable in our analysis. Since the information on quantities ofindividual transfers is not available, we use vessel tonnage as a proxy.Generally, the amount of oil transferred is proportional to vessel tonnage(i.e., size). Hence, vessel tonnage is a proxy for the total oil carrying capacityfor oil-cargo vessels.

Tobit regression estimates for the dependent variables, out-of-water andin-water oil transfer spillage per vessel gross ton (OWATSPT andIWATSPT), also appear in Table II. The results for OWATSPT (3rd columnin Table II) are generally similar to those of OWATSP. However, tank shipsdo not spill more oil than other non-oil-cargo vessels when spillage is ad-justed for vessel size. As for OWATSP, the coefficients of HULLINSP andAIRPATR are negative and statistically significant, suggesting that an in-crease in Coast Guard on-board hull inspections and aircraft patrols willreduce out-of-water vessel oil transfer spillage per vessel gross ton.

Although the coefficient signs for the IWATSPT (5th column in Table II)and IWATSP estimates are the same for a number of statistically significantvariables, sign differences do exist. Unlike the IWATSP results, the coefficientsigns of the IWATSPT results are positive for FISHBOAT, PASSBOAT,RECREBOAT, VAGE, and HIGHWINDS and negative for COLD. Theseresults suggest that the in-water transfer oil spillage per vessel gross ton isgreater for a fishing boat, a passenger boat, a recreation boat, for oldervessels, and during high winds, but less during cold temperatures. Thecoefficient of HULLINSP is negative, but unlike the IWATSP estimate, it isalso statistically significant, suggesting that an increase in Coast Guard on-board hull inspections reduces in-water oil transfer spillage per vessel grosston. Relative to other Coast Guard Districts, the in-water spillage per vesselgross ton is less in Coast Guard Districts 7 (South Carolina, Georgia andFlorida) and 9 (the upper Midwest), but greater in District 8 (the Gulf coast)and District 14 (Hawaii).

Overall, our results for the enforcement variables are somewhat mixed.They suggest that Coast Guard hull but not machinery inspections areeffective in reducing both out-of-water and in-water spills. Coast Guardpatrols by air, but not by boat, are effective in reducing out-of-water spills,but neither is effective in reducing in-water spills.

A recent study by Sharit et al. (2000) provides a plausible explanation forour mixed results for Coast Guard enforcement. Their study grouped thecauses of tank barge transfer spills into two main categories: human error(e.g., tank monitor (overflow) or hose connection) and equipment failure(e.g., hull defect or valve failure). They found that there were no statisticallysignificant differences between the two causes; spill causation was equallyattributable to human error and equipment failure.


Sharit et al. (2000) also state that Coast Guard regulations have empha-sized engineering solutions (e.g., specification of hull plating thickness) andhave been successful in reducing equipment failures. However, Coast Guardregulations have failed to address the problem of human error. By the late1990s, the Coast Guard had begun to address the human error problemthrough such programs as Prevention Through People. However, our dataand analysis do not capture these more recent developments.

The different structural relationships across the two vessel spill typessuggest different policies for reducing these spillages. Specifically, policies forreducing the spillages of tank ships should focus on reducing out-of-water asopposed to in-water spills. Spill-reduction policies for older vessels shouldfocus on reducing out-of-water spills, but for larger vessels and in coldweather, the focus should be on reducing in-water spills. Policies for reducingvessel spills at night and during precipitation should focus on reducing in-water spills, whereas the focus of policies for reducing spills in rivers andharbors should be on out-of-water spills. Coast Guard Districts 2, 13, 14 and17 should focus on reducing out-of-water spills.

6. Adjusted Tobit Coefficients

Unlike ordinary least squares, tobit regression coefficients do not measure thecorrect change in a dependent variable from a change in an explanatoryvariable for non-zero observations of the dependent variable. However, tobitcoefficients can be adjusted to obtain such measures. McDonald and Moffitt(1980) show that the change in the dependent variable (for its observationsabove a limit such as zero) from a change in an explanatory variable can bemeasured as the product of the explanatory variable’s tobit coefficient andthe adjustment factor A:

A ¼ f1� ½sfðsÞ=FðsÞ� � ½ fðsÞ2=FðsÞ2�g ð6Þwhere s represents an evaluation (at the means of the explanatory variables)of the tobit equation divided by the equation’s standard error; f(s) is the unitnormal density; and F(s) is the cumulative normal distribution function. Werefer to the product of A and a given tobit coefficient as the latter’s ��adjustedtobit coefficient’’.

The adjusted tobit coefficients that correspond to the tobit coefficients inTable II are presented in Table III. The OWATSP coefficients (2nd columnin Table III) indicate that a tank barge, a tank ship and a freight ship spill18.3, 21.1 and 15.3 more out-of-water gallons of oil, respectively, and apassenger boat and a recreational boat spill 11.8 and 6.18 fewer out-of-watergallons of oil, respectively, than other vessels involved in an oil transfer spill.Out-of-water transfer spillage is 0.032 gallons more per year of vessel age,


Table III. Vessel oil transfer spillage: adjusted tobit coefficients

Variable OWATSP

coefficient

OWATSPT

coefficient

IWATSP

coefficient

IWATSPT

coefficient

Type of vessel

TANKBARGE 18.3 0.008 98.4 0.008

TANKSHIP 21.1 0.001 29.7 )0.019FISHBOAT )1.41 )0.005 · 10)1 )56.3 0.043

PASSBOAT )11.8 )0.009 )55.3 0.026

FRTSHIP 15.3 0.004 7.01 0.010

RECREBOAT )6.18 )0.008 )111.2 0.065

Vessel characteristics

VAGE 0.032 0.006 · 10)2 )0.054 0.006 · 10)1

VSIZE )0.002 · 10)1 0.006 · 10)6 0.002 )0.001 · 10)4

USFLAG )12.2 )0.005 )77.1 )0.015POORSAF 2.73 0.002 )11.86 )0.004

Vessel operation phase

MISSING )3.80 0.001 )295.6 )0.116MOORDOCK 10.3 0.007 )193.8 )0.057TOWED )23.0 )0.024 586.4 0.045

ANCHORED 2.91 0.004 )277.8 )0.052

Weather/visibility characteristics

COLD )1.63 )0.006 776.1 )0.095HIGHWINDS )20.5 )0.021 )1063 0.236

PRECIPTN 3.84 0.006 1476 0.080

NIGHT )0.307 )0.002 2440 0.334

Type of waterway

COAST )0.663 )0.003 )14.6 0.014

OCEAN )13.5 )0.009 65.9 0.006

RIVER 2.96 0.002 1.55 0.007

HARBOR 3.90 )0.002 17.9 )0.002

Vessel safety/environmental regulation enforcement activities

HULLINSP )0.326 )0.002 · 10)1 )2.01 )0.002MACHINSP 0.328 0.003 · 10)2 3.73 0.002

BOATPATR )0.011 )0.003 · 10)3 0.001 )0.003 · 10)3

AIRPATR )6.05 )0.006 )11.9 )0.009

Price variables

PRICEOIL 0.847 0.007 · 10)1 7.91 0.001


0.0002 gallons less per vessel gross ton, 12.2 gallons less for a US flaggedvessel and 10.3 gallons more when a vessel is moored or docked. Spillages ina river and harbor are 2.96 and 3.90 gallons more, but 13.5 gallons less in theocean, than in other types of waterways, and 0.326 and 6.05 gallons less perport hour of Coast Guard hull inspections and aircraft patrols, respectively.Among Coast Guard Districts, spillages in Districts 2 (the Midwest), 13(Washington and Oregon), 14 (Hawaii) and 17 (Alaska) are 8.17, 7.24, 12.9and 9.82 gallons more, respectively, but 8.86 gallons less in District 7 (SouthCarolina, Georgia and Florida).

The IWATSP adjusted tobit coefficients (4th column in Table 3) indicatethat a vessel spills 56.3 and 111.2 fewer in-water gallons of oil if it is a fishingand a recreational boat, respectively, but 98.4 gallons more if it is a tankbarge. The spillage is 0.054 gallons less per year of vessel age, but 0.002gallons more per vessel gross ton. The spillages are 77.1, 193.8, 277.8 and1063 gallons less for a US flagged vessel, when moored or docked, whenanchored and during high winds, respectively, but 586.4, 776.1, 1,476, 2,440and 65.9 gallons more when towed, during cold weather, during precipita-tion, at nighttime, and when the spill occurs in the ocean, respectively.

A comparison of OWATSPT type-of-vessel adjusted tobit coefficients (3rdcolumn in Table III) indicates that a tank barge and freight ship spill 0.008and 0.004 more out-of-water gallons of oil per vessel gross ton, respectively,

Table III. Continued

Variable OWATSP

coefficient

OWATSPT

coefficient

IWATSP

coefficient

IWATSPT

coefficient

Districts

DIST2 8.17 0.004 )19.99 )0.003DIST5 0.83 )0.004 11.35 0.003 · 10)2

DIST7 )8.86 )0.002 34.03 )0.026DIST8 2.64 0.003 36.77 0.025

DIST9 )4.21 )0.009 )49.95 )0.039DIST11 )2.77 0.002 10.11 )0.017DIST13 7.24 0.008 24.10 )0.009DIST 14 12.9 0.012 29.57 0.079

DIST17 9.82 0.007 31.69 )0.021

Years

1992 )2.07 )0.008 · 10)1 )77.6 )0.0401993 5.36 0.004 )74.9 )0.0481994 1.51 0.003 )68.2 )0.0391995 1.69 0.005 )68.6 )0.039


but passenger and recreation boats spill 0.009 and 0.008 fewer gallons pervessel gross ton, respectively, than other vessels involved in oil transfer spills.The spillage per vessel gross ton is 0.00006 gallons less per year of vessel ageand 0.005 gallons less if the vessel is US flagged. Also, the spillage per vesselgross ton is 0.024, 0.021, 0.003 and 0.009 gallons less when the vessel istowed, during high winds, when the spill occurs along the coast and in theocean, respectively, but 0.007 gallons more when the vessel is moored ordocked. OWATSPT is 0.0002 and 0.006 gallons less per port hour of CoastGuard hull inspections and aircraft patrols, respectively. Among CoastGuard Districts, the out-of-water spillage per vessel gross ton is 0.008, 0.012and 0.007 gallons more in Districts 13, 14 and 17, but 0.009 gallons less inDistrict 9.

IWATSPT adjusted tobit coefficients (5th column in Table III) indicatethat a fishing boat, a passenger boat and a recreational boat spill 0.043, 0.026and 0.065 more in-water gallons of oil per vessel gross ton than other vesselsinvolved in an oil transfer spill. Also, the spillage per vessel gross ton is0.0006 gallons more per year of vessel age. The gallons per vessel gross tonspilled are 0.057, 0.052 and 0.095 less when the vessel is moored or docked,when anchored, and during cold weather, respectively, but 0.236 and 0.334more in high winds and at nighttime, respectively. IWATSPT is 0.002 gallonsless per port hour of Coast Guard hull inspections. Among Coast GuardDistricts, the in-water spillage per vessel gross ton is 0.026 and 0.039 gallonsless in Districts 7 and 9, respectively, but 0.025 and 0.079 gallons more inDistricts 8 and 14, respectively.

7. A Comparison

The vessel oil transfer spill model and empirical results found in this study arein sharp contrast to those found in the literature. Epple and Visscher (1984),Cohen (1987) and Grau and Groves (1997) restrict vessel oil transfer spills tothose of oil-cargo vessels, tank barges and ships. Epple and Visscher (1984)utilize oil spill-size data for the 1973–1975 period and express the log of spillsize as a function of the logs of price of oil, vessel size and Coast Guard spill-reduction enforcement activity. Cohen (1987) utilizes spill-size data for the1973–1977 period and estimates a similar function. Grau and Groves (1997)also utilize pre-OPA-90 data, but for the 1984–1987 period, and express thelog of spill size as a function of the log of Coast Guard enforcement activityand Coast Guard District binary variables.

Epple and Visscher (1984) utilize an overall measure of Coast Guardenforcement activity, i.e., man-hours per transfer of oil; Cohen (1987) sep-arates Coast Guard enforcement activity into vessel compliance inspections,monitoring oil transfer operations, and patrolling ports; and Grau and


Groves’ (1997) measure Coast Guard enforcement activity by the probabilityof oil transfer operations being monitored. These studies, however, do notmake a distinction between in-water and out-of-water spills. Further, spillsize is not adjusted for the oil carrying capacity of the vessel and vessels thatare not involved in a detected spill are not considered.

The results of this study suggest that the previous studies are subject tomodel misspecification. If relevant (i.e., statistically significant) explanatoryvariables are omitted from the model, a specification error occurs, since thisomission will bias the empirical results. Unlike previous studies, this studyconsidered and found explanatory variables such as vessel characteristics,vessel operating conditions, weather/visibility conditions, waterway type, andtype of non-oil-cargo vessel to be statistically significant determinants of vesseloil transfer spills.

Epple and Visscher (1984) and Cohen (1987) found negative and positiverelationships between transfer oil spill size and the price of oil and vessel size,respectively. Conversely, in this study, a positive relationship was foundbetween spill size and the price of oil for both in-water and out-of-waterspillages. Further, while a positive relationship was found between spill andvessel sizes for in-water spillage, the relationship for out-of-water spillage wasnegative. Unlike Cohen (1987), this study found that vessel complianceinspections are effective in reducing vessel transfer oil spills. Specifically, hullbut not machinery inspections were found to be effective in reducing both in-water and out-of-water spillages. Cohen (1987) found that Coast Guardpatrols of ports reduce transfer spills. This study found that Coast Guardpatrols by air but not by boat are effective in reducing out-of-water spills, butneither is effective in reducing in-water spills. Grau and Groves (1997) found,as for this study, that the size of transfer spills varies among Coast GuardDistricts.

8. Conclusions

This study has investigated determinants of the size of vessel oil transfer spillsas well as the effectiveness of Coast Guard enforcement activities in reducingspillage during the post OPA-90 period. Out-of-water and in-water vessel oiltransfer spill functions were estimated, utilizing tobit regression analysis anddetailed data on vessel oil transfer spill investigations by the Coast Guardduring the five-year time period 1991–1995.

The study differs in several ways from previous studies of vessel oiltransfer spills: (1) we examine spills for the post OPA-90 as opposed to thosefor the pre OPA-90 period; (2) in addition to spills from oil-cargo vessels, wealso consider spills from non-oil-cargo vessels; (3) we make a distinctionbetween in-water and out-of-water spills; (4) we also consider vessels for


which a spill was not detected; (5) we adjust spills for vessel size; (6) we makea distinction between hull and machinery inspections and between air andboat patrols by the Coast Guard; and (7) we consider vessel characteristics,the phase of vessel operation, weather/visibility conditions, type of waterway,and the vessel’s safety record as determinants of vessel oil transfer spills.

Like the pre OPA-90 studies, our investigation found negative relation-ships between vessel oil transfer spills and Coast Guard enforcement activi-ties. However, unlike Cohen (1987), this study found that vessel complianceinspections are effective in reducing vessel transfer oil spills. Specifically, hullbut not machinery inspections were found to be effective in reducing bothout-of-water and in-water spills. A vessel’s out-of-water spillage per vesselgross ton is 0.0002 gallons less and in-water spillage per vessel gross ton is0.002 gallons less per port hour of Coast Guard hull inspections. Further,Coast Guard patrols by air, but not by boat, are effective in reducing out-of-water spills, but neither is effective in reducing in-water spills.

The estimation results suggest that the type of vessel, vessel characteristics,vessel operations, weather/visibility conditions, and type of waterway aredeterminants of post OPA-90 vessel oil transfer spills. Among Coast GuardDistricts, out-of-water transfer spills per vessel gross ton are greater in Dis-tricts 13 (Washington and Oregon), 14 (Hawaii) and 17 (Alaska), but less inDistrict 9 (the upper Midwest). In-water spills per vessel gross ton are greaterin District 8 (the Gulf coast) and District 14 (Hawaii) but less in Districts 7(South Carolina, Georgia and Florida) and 9 (the upper Midwest).

Acknowledgements

The authors benefited from discussions with Andy Solow and Porter Hoag-land. An earlier draft of this paper was presented at the 2002 AnnualMeeting of the Eastern Economic Association in Boston. We thank JeffCohen and other participants for their constructive comments at the meeting.This is WHOI Contribution number 11183.

Notes

1. After discharging its oil cargo, a vessel takes ballast water into its cargo tanks to ensurestability on the return trip, but then dumps the dirty ballast, an oil-in-water mixture, on orbefore arrival at the loading port.

2. A number of studies have investigated vessel accident oil spills from oil-cargo vessels. Oilspillage from a tanker accident is greater when the tanker is adrift and when the accidentoccurs in a coastal waterway, but less for larger and US flag tankers (Anderson and Talley1995). Tanker accident oil spillage per vessel gross ton is less for a US flag tanker, but

increases with the vessel damage severity of the accident (Talley 1999). For tank bargeaccidents, oil spillage is greater for collision and material/equipment failure accidents(Anderson and Talley 1995) and greater if the accident occurs in a river, at nighttime and


for older barges, but less if there is precipitation at the time of the accident (Talley 2000).Whether the oil spillage of oil-cargo vessel accidents is greater than that of non-oil-cargovessel accidents has been investigated by Talley, Jin and Kite-Powell (2001). The empirical

results suggest that for the post OPA-90 (Oil Pollution Act of 1990) period, tank barges, butnot tanker vessels, have incurred greater vessel accident oil spillage than non-oil-cargo vessels.

3. Data from the Coast Guard’s Pollution Incident Reporting System (PIRS) were used byEpple and Visscher (1984), Cohen(1987), and Grau and Groves (1997) in their investi-

gations of vessel oil transfer spill sizes. This database includes all vessel oil spills inves-tigated by the Coast Guard that were not vessel-accident related.

4. When a vessel accident spill occurs, the US Coast Guard designates and notifies the party

at fault. If the vessel has a COFR, the guarantor or insurer is also notified. The gov-ernment then takes care of the clean-up and subsequently sends the bill for expensesincurred to the appropriate party.

5. The cost effectiveness of double hulls in reducing vessel accident oil spills has beenquestioned. Hopkins (1992, p. 59) concludes that ‘‘costs appear substantial relative tobenefits, and lawmakers’ emphasis on design standards deflects attention from alternativerisk reduction strategies, e.g., operation and maintenance measures that warrant equal

attention.’’ Brown and Savage (1996) found that the expected benefits of reduced tankerspillage from the double-hull requirement are only 20% of the increased construction andoperation costs of double-hulled tankers. In addition, Jin, Kite-Powell and Broadus (1994)

found that electronic charts may be a far more cost-effective approach than double hullsfor marine pollution control.

6. The IMO is a UN agency responsible for improving the safety of international shipping

and preventing pollution from ships. In 1948 the UN convened in Geneva an internationalconference to consider establishing an international organization devoted exclusively tomaritime matters. It was believed that such matters as maritime safety would be addressed

more effectively at the international level than by individual countries acting unilaterally.The conference lead to the establishment of a UN agency known as the Inter-Govern-mental Maritime Consultative Organization (IMCO). In 1982 its name changed toInternational Maritime Organization. The IMO has 158 member and two associate

member states. In the last 30 years, the IMO has promoted the adoption of 30 conventionsand protocols and adopted over 700 recommendations and codes for maritime safety, theprevention of pollution, and related matters.

7. The specification covers different vessel types. For example, for tankers, bj is revenue fromshipping service, and qj is the volume of oil shipped; for fishing vessels, bj is revenue fromfish landings, and qj is the number of fishing trips; and for recreational boats, bj is the

benefit from boating activities, and qj is the number of boating days.8. For a general discussion of this type of analysis, see Shavell (1987).9. In U.S. waters, key regulations for oil-cargo vessels include the liability rule regarding

natural resource damages specified in OPA 90 (P.L. 101–380). For other types of vessels

(e.g., fishing and recreational vessels), the oil discharge regulations are under the FederalWater Pollution Control Act (33 U.S.C. 1321). Specific penalties for vessel spills can befound in 33 CFR 151.04.

10. In addition to the level of care, the choice of activity level is important. For analysis of thesocially optimal activity level in the case of tanker transport, see Jin andKite-Powell (1999).

11. In our analysis, data for the enforcement variables are ex ante monitoring activities,

checking for compliance with safety and environmental regulations. They are at the portlevel and not vessel specific. Thus, the enforcement is exogenous to vessel operators.

12. 10 MSMS data tables were merged to obtain the data set for this study. The 10 data tables

include: the Marine Casualty and Pollution Master Table (cirt), the Marine Casualty


Vessel Supplement Table (civt), the Vessel Identification Table (vidt), the Marine CasualtyWeather Supplement Record (cwxt), the Marine Pollution Substance Table (cpdt), theMarine Inspection Activities tables (crst and irit), the Marine Investigation Module Data

Table (mcrt), and the Port Safety Activities tables (brst and brit).13. This procedure is found in Cropper et al. (1992).14. The Coast Guard 1st District covers the New England and New York Atlantic coast, 2nd

District the Midwest, 5th District the mid Atlantic coast (southern New Jersey to North

Carolina), 7th District the southern Atlantic coast (South Carolina to Florida), 8thDistrict the Gulf coast, 9th District the Great Lakes, 11th District the California coast,13th District the Pacific northwest coast, 14th District Hawaii, and 17th District Alaska.

15. Since UNDERWAY is dropped in the estimations, the discussions of these vessel oper-ation phase variables are relative to UNDERWAY.

References

Anderson, E. E. and W. K. Talley (1995), �The Oil Spill Size of Tanker and Barge Accidents:Determinants and Policy Implications’, Land Economics 71, 216–228.

Brown, R. S. and I. Savage (1996), �The Economics of Double-Hulled Tankers’, Maritime

Policy and Management 23, 167–175.Cohen, M. J. (1995), �Technological Disasters and Natural Resource Damage Assessment: An

Evaluation of the Exxon Valdez Oil Spill’, Land Economics 71, 5–82.

Cohen, M. A. (1987), �Optimal Enforcement Strategy to Prevent Oil Spills: An Application ofa Principal- Agent Model with Moral Hazard’, Journal of Law and Economics 30, 23–51.

Cropper, M. L., W. N. Evans, S. J. Berardi, M. M. Ducla-Soares and P. R. Portney (1992),

�The Determinants of Pesticide Regulation: A Statistical Analysis of EPA DecisionMaking’, Journal of Political Economy 100(1), 175–197.

Epple, D. and M. Visscher (1984), �Environmental Pollution: Modeling Occurrence, Detec-tion, and Deterrence’, Journal of Law and Economics 27, 29–60.

Gracey, J. S. (1985), �Many Question Value of Drive for Vessel Safety Standards’, Journal ofCommerce, May, p. 10C.

Grau, M. V. and T. Groves (1997), �The Oil Spill Process: The Effect of Coast Guard Mon-

itoring on Oil Spills’, Environmental and Resource Economics 10: 315–339.Hopkins, T. D. (1992), �Oil Spill Reduction and Costs of Ship Design Regulation’, Contem-

porary Policy Issues 10, 59–70.

International Tanker Owners Pollution Federation (ITOPF), 2002. Oil Tanker Spill Statistics(http://www.itopf.com).

Jin, D., H. L. Kite-Powell and J. M. Broadus (1994), �Dynamic Economic Analysis of Marine

Pollution Prevention Technologies: An Application to Double Hulls and ElectronicCharts’, Environmental and Resource Economics 4, 555–580.

Jin, D. and H. L. Kite-Powell (1999), �On the Optimal Environmental Liability Limit forMarine Oil Transport’, Transportation Research Part E 35(2), 77–100.

Kim, I. (2002), �Ten Years after the Enactment of the Oil Pollution Act of 1990: A Success or aFailure’, Marine Policy 26: 197–207.

McDonald, J. F. and R. A. Moffitt (1980), �The Uses of Tobit Analysis’, Review of Economics

and Statistics 62, 318–321.Sharit, J., S. J. Czaja, E. lakovou and J. Moses (2000), �A Human Factors Systems Approach

to Examining Oil Spills in Tankbarge Operations: Analysis and Reappraisal of Manage-

ment Policies’, Maritime Policy and Management 27(1), 41–52.Shavell, S. (1987), Economic Analysis of Accident Law. Cambridge, Massachusetts: Harvard

University Press.


Talley, W. K. (1999), �Determinants of the Property Damage Costs of Tanker Accidents’,

Transportation Research Part D 4, 413–426.Talley, W. K. (2000), �Oil Spillage and Damage Costs: U.S. Inland Waterway Tank Barge

Accidents’, International Journal of Maritime Economics 2, 217–234.

Talley, W. K., D. Jin, H. Kite-Powell (2001), �Vessel Accident Oil-Spillage: Post US OPA-90’,Transportation Research Part D 6, 405–415.

US Coast Guard (2002), Pollution Incident Compendium (http://www.uscg.mil/hq/gm/nmc/response/stats/summary.html).

White, I. C. and J. M. Baker (1998), �The Sea Empress Oil Spill in Context’, A Paper Presentedat the International Conference on the Sea Empress Oil Spill, Cardiff, Wales.


Documents

Post OPA-90 Vessel Oil Transfer Spill Prevention: The