6. determinants of vessel flag

47
6. DETERMINANTS OF VESSEL FLAG Jan Hoffmann, Ricardo J. Sanchez and Wayne K. Talley 1. INTRODUCTION For a vessel operator, the choice of the flag is one of the main business decisions. For the countries that register vessels under their flag, this operation implies responsibilities as well as income. For international organizations such as the International Maritime Organization (IMO), the United Nations Conference on Trade and Development (UNCTAD), the Organization for Economic Cooperation and Development (OECD), the International Labour Organization (ILO), the Food and Agricultural Organization (FAO), and also national governments who have an interest in cleaner oceans and safer shipping, the compliance with international environmental, safety and labour regulations is of high priority – and this compliance is assumed to be closely related to a vessel’s flag. Most recently, for security reasons, interest in the linkages between flags and those who actually control the ships has re-emerged, although OECD (2003a) finds that in practice it is not so much the open registries themselves that enable reclusive owners to remain anonymous, but the corporate instruments and structures that are freely available internationally. In order to increase transparency about the linkage between the flag states and those who actually control the ships, UNCTAD (1986) developed a “Convention on the Registration of Ships,” which, however, never entered into force. Ample literature exists about comparisons between different flags concerning their compliance with international standards. OECD (2003b, p. 8), for example, highlights that the “principal responsibility for complying with the IMO’s Shipping Economics Research in Transportation Economics, Volume 12, 173–219 Copyright © 2005 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0739-8859/doi:10.1016/S0739-8859(04)12006-4 173

Upload: unam

Post on 14-May-2023

1 views

Category:

Documents


0 download

TRANSCRIPT

6. DETERMINANTS OF VESSEL FLAG

Jan Hoffmann, Ricardo J. Sanchez and Wayne K. Talley

1. INTRODUCTION

For a vessel operator, the choice of the flag is one of the main businessdecisions. For the countries that register vessels under their flag, this operationimplies responsibilities as well as income. For international organizationssuch as the International Maritime Organization (IMO), the United NationsConference on Trade and Development (UNCTAD), the Organization forEconomic Cooperation and Development (OECD), the International LabourOrganization (ILO), the Food and Agricultural Organization (FAO), and alsonational governments who have an interest in cleaner oceans and safer shipping,the compliance with international environmental, safety and labour regulationsis of high priority – and this compliance is assumed to be closely related to avessel’s flag.

Most recently, for security reasons, interest in the linkages between flags andthose who actually control the ships has re-emerged, although OECD (2003a) findsthat in practice it is not so much the open registries themselves that enable reclusiveowners to remain anonymous, but the corporate instruments and structures that arefreely available internationally. In order to increase transparency about the linkagebetween the flag states and those who actually control the ships, UNCTAD (1986)developed a “Convention on the Registration of Ships,” which, however, neverentered into force.

Ample literature exists about comparisons between different flags concerningtheir compliance with international standards. OECD (2003b, p. 8), for example,highlights that the “principal responsibility for complying with the IMO’s

Shipping EconomicsResearch in Transportation Economics, Volume 12, 173–219Copyright © 2005 by Elsevier Ltd.All rights of reproduction in any form reservedISSN: 0739-8859/doi:10.1016/S0739-8859(04)12006-4

173

174 JAN HOFFMANN ET AL.

regulatory framework has always remained with Flag States. These statestraditionally exercise direct control over national fleets and their crews that tendedto be nationals of those states. However, the development of ‘open’ registries –where non-national shipowners could register their ships in national registries witha sometimes-tenuous link to the flag – saw the direct ship-Flag State-national crewlink weakened. The development of open registries and the international sourcingof crews has offered cost savings to owners and new employment opportunitiesfor seafarers around the world. However, this shift of registries has rendered thecontrol of the quality of world fleets and their crews more problematic. MostFlag States carry out their regulatory responsibilities either directly or throughintermediary Class Societies. However, a certain number of states have sought toreduce their expenditures related to the administration of their fleet and/or havesought to develop their registry solely as an income-generating venture. These andother smaller states simply do not have the budgets and/or administration necessaryto ensure that their fleets continue to meet IMO requirements. Class Societies haveplayed an increasingly more important role in ensuring the safety, seaworthinessand quality of these national registries. Yet, it is commonly recognised that stiffcompetition in the classification/certification market has led to the emergenceof certain Class Societies willing to cut corners in order to gain or retainclients.”

The OECD thus assumes a link between the flag of a vessel, its classificationssociety, and the likelihood of being non-seaworthy. For the specific case of bulkcarrier accidents investigated by the U.S. Coast Guard, Talley (2002) analysesthe non-seaworthy risks, and does not find them to be related to the ship’s flag.Roberts and Marlow (2002) on the other hand found that the risk of foundering wasrelated to the ship’s flag of registration. Li and Wonham (1999) state that open-registry country ships tend to be substandard, although the safety records of someopen-registry countries are quite acceptable. Similarly, Alderton and Winchester(2002) conclude that, on average, vessels registered in open registries do havehigher ratings, but big differences persist, and the countries with the worst safetyrecords are small national, but not open, registries. UNCTAD (1994) finds thatopen registry ships were involved in a higher proportion of General Average casesthan would correspond to their share of the world fleet. With regard to classificationsocieties, Talley (1999) concludes that there is evidence of variance in the safetyperformance of different classification societies.

For those countries that provide registration services, many of which arerelatively small developing countries, the income generated from this businessis important for their national economies (Thanopoulou, 1995). For Panama, forexample, just the direct income from the registry is around 60 million USD peryear. If, as is assumed, lower registration fees imply a saving for the shipping

Determinants of Vessel Flag 175

company, this has a bearing on freight rates, which in turn benefit global trade,although – as some claim – possibly at the cost of more accidents and pollution(see for example Thanopoulou, 1998).

This leads us to two related questions: (a) why are some flags safer than others?;and (b) what are the determinants of a vessel’s flag? Although the paper onlydirectly investigates the latter question, the two questions are related when thedeterminants of a vessel’s flag are also determinants of its safety record. Forexample, if a given flag attracts more dry bulk than liquid bulk vessels, then itmay have a worse average safety record than an alternative flag with more liquidbulk vessels in its registry, because the latter vessel type is historically less likelyto sink than dry bulk ships. Hence, the first flag may appear worse in casualtystatistics, although it may have the same quality controls as the second.

Of interest in this context is also the impact of IMO conventions, the safetyrecord of the operator country’s national register, and possibly the income percapita. Will the ratification of many IMO conventions scare away vessel operatorsinto foreign registers? May a relatively bad safety record be an indicator of a laxnational safety regime and encourage operators to stay in the national register?Are high wages, which are related to the national GDP per capita, a measurableburden which may increase the likelihood to choose foreign flags?

The remainder of this paper will look at quantifiable determinants of a vessel’sflag. It practically covers the world fleet of commercial vessels. What is to beexplained is the probability that a given vessel – or rather its operator – chooses ornot a foreign flag. To “choose a foreign flag” is meant to stand for a situation wherethe vessel’s operator’s country of domicile is different from the vessel’s flag stateor where a second register is used. In a topically related paper, Bergantino andMarlow’s (1999) have analysed the out-flagging of U.K. vessels. In our paper, wewill later further distinguish between specific major Latin American and Caribbeanopen registers.

We will use data about the known commercial fleet of vessels of 300 GrossTons (GT) and above, based on information provided by LRFairplay1 and analyseit with regard to possible determinants of the choice of flag. The available dataallows us to consider binary or quantifiable information concerning the age, size,country of build, vessel types, the classification society, and various characteristicsof the operator’s country of domicile.

2. THE WORLD’S COMMERCIAL FLEET

Figures 1 and 2 depict the size of fleets according to the country of domicile of thevessel operator.

176 JAN HOFFMANN ET AL.

Fig. 1. Top 30 Operator Countries, Number of Vessels, January 2003.

Lloyds Register Fairplay provided data for 47,740 commercial vessels, includingships on order. Of the delivered vessels, 45.9% use a foreign flag. Larger andcargo vessels are more likely to use a foreign flag than smaller and miscellaneousor passenger vessels. As a consequence, the proportion of foreign flaggedtonnage is higher than is the case for the number of vessels: Only 34.6% of theworld’s GT use the national flag, i.e. almost two out of three gross tonnes areregistered under a foreign flag. Big differences exist between operator countries.In terms of vessel numbers, Japan is the most important operator country.Regarding GT, the Greek operators control the biggest proportion of the world’stonnage. Of the major 30 operator countries, Iran (96%) and India (88%) havethe highest proportion of nationally flagged GT, whereas operators based inMonaco (1%), Switzerland (7%) and Belgium (9%) are least likely to use theirnational flag.

Table A1, in the Annex, contains information about open and national registriesfor vessels above 300 GT. The total fleet grand total of delivered vessels is 584million GT, 43,878 vessels, 8.1 million TEUs and 847 million DWT. With regard toGT, national registries reach a 37.8% of world total and 57.2% of number of vessels.61.8% of TEU carrying capacity is flagged under open registries. The largest FlagStates in terms of GT are Panama (22%), Liberia (9%), Bahamas (6%) and Greece(4.7%). Japan is the second largest Flag State in terms of number of vessels, whichreflects its large fleet of inter-island cargo and passenger transport and also fishingvessels. One third of the world’s tonnage is registered in Latin America and the

Determinants of Vessel Flag 177

Fig. 2. Top 30 Operator Countries, GT, January 2003. Source:Authors, based on dataprovided by LRFairplay. Notes: “National” and “Foreign” refers to the vessel’s flag.

“Foreign” includes second registries.

Caribbean. As sample cases, this paper will look at eight major registries of thisregion in more detail (marked “∗ ∗” in Table A1). Of these Latin American andCaribbean registries, Bahamas has the largest vessels (average 26,082 GT per unit),whereas Honduras specializes in the smallest vessels (1,719 GT).

3. MODEL

In the basic Model (Eq. (1)), the choice of a foreign flag for the ith vessel byits operator is posited to be a function of vessel age, size and carrying capacity;country of build; characteristics of the country where the operator is domiciled (i.e.the “operator country”); vessel type; and its classification society. Specifically:

FOREIGNi = F(VAGEi , VSIZEi , VCAP−DWTi , VCAP−TEUi ,

BUILTOPERi , OPERDEVi , CARGOi , CONTi , CIACSi ) (1)

. . . where FOREIGN is a binary variable equal to 1 for a foreign flag and 0for a national flag for the ith vessel; VAGE is vessel age; VSIZE is vessel

178 JAN HOFFMANN ET AL.

size in GT; VCAP-DWT is the vessel’s carrying capacity in deadweight tons(DWT); VCAP-TEU is the vessel’s container carrying capacity in twenty-foot-equivalent units (TEUs); OPERDEV and BUILTOPER are binary variablesrepresenting a vessel operator country that is a developed country and thecountry where the vessel was built, respectively; CARGO and CONT arevessel-type binary variables that represent vessels that can transport cargo andcontainers, respectively; and CIACS is a binary variable denoting that the vesselis classed by a member of the International Association of Classification Societies(IACS).

In a more detailed model (Eq. (2)), by substituting alternative measuresfor the vessel type, classification and operator country variables, CARGO,CONT, VCAP-TEU, CIACS and OPERDEV, the FOREIGN function can berewritten as:

FOREIGNi = G(VAGEi , VSIZEi , VCAP−DWTi , BUILTOPERi ,

OPCASUALi , OPOPENREi , OPIMONUMi , OPGDPCAPi ,

OPPOPULAi , OPLITERAi , OPLIFEEXi , VGCARGOi ,

VCONTi , VLBLKi , BDBLKi , VPASSi , VROROi , VREEFi ,

VOROILi, VORSHOREi , VFSTUGi , CABSi , CBUVi , CCCSi ,

CDNVi , CGELi , CKORi , CLLRi , CNIKi , CRINi , CRUSi ) (2)

The variables describing the operator’s country of domicile OPCASUAL,OPOPENRE, OPIMONUM, OPGDPCAP, OPPOPULA, OPLITERA andOPLIFEEX represent the operator country’s past national flag casualty rate(log); its characterization as open registry (binary); the number of ratified IMOconventions; the GDP per capita (log); the population (log); the literacy rate; andthe life expectancy, respectively.

The binary vessel-type variables, VGCARGO, VCONT, VLBLK, VDBLK,VPASS, VRORO, VREEF, VOROIL, VORSHORE and VFSTUG representgeneral-cargo; container; liquid-bulk; dry-bulk; passenger; roll-on/roll-off; reefer;ore/oil combined; off-shore; and fishing, tug or diverse miscellaneous servicevessels, respectively.

The binary vessel-classification variables, CABS, CBUV, CCCS, CDNV,CGEL, CKOR, CLLR, CNIK, CRIN and CRUS, represent the American Bureauof Shipping, the Bureau Veritas, the China Classification Society, the Det NorskeVeritas, the Germanischer Lloyd, the Korean Register of Shipping, the LloydsRegistry, the Nippon Kaiji Kyokai, the Registro Italian Navale and the RussianMaritime Register of Shipping classification societies, respectively.

Determinants of Vessel Flag 179

The choice of a specific foreign flag for a vessel by its operator is alsoposited as a function of the same explanatory variables as found in Eqs (1)and (2). These specific foreign flag choices are analyzed for eight Latin Americanand Caribbean open registries. The binary foreign-flag variables, PANAMA,BAHAMAS, ST. VINCENT AND THE GRENADINES, BERMUDA, ANTIGUAAND BARBUDA, BELIZE, HONDURAS and BOLIVIA, represent the Panama,Bahamas, St. Vincent and the Grenadines, Bermuda, Antigua and Barbuda, Belize,Honduras and Bolivia open registries, respectively.

4. DATA

Variables used in the estimations of Eqs (1) and (2) and their specific measurementsappear in Table 1. Descriptive statistics (mean and standard deviation) of thedata for our variables also appear in this Table. The mean statistics reveal that45.7%, 12.9%, 3.1%, 2.4%, 0.3%, 1.9%, 1.2%, 0.7% and 0.1% of the vessels wereforeign, Panama foreign, Bahamas foreign, St. Vincent and the Grenadines foreign,Bermuda foreign, Antigua and Barbuda foreign, Belize foreign, Honduras foreignand Bolivia foreign flagged.

The average age and size of the vessels are 18.1 years and 13,428 GT; theaverage vessel carrying capacity of vessels in DWT and TEUs are 19,483 and 186,respectively. The observations for GT range between 300 and 261,453 and forDWT between 0 and 564,650. The mean statistics also reveal that 56.3% of thevessel operators were from developed countries, 33.0% of the vessels were builtin a shipyard of the operator country, 92.9% of the vessels have cargo carryingcapacity, and 20.7% can transport containers.

Thus, 76.4% of the vessels were classed by an IACS member and 10.9%, 9.4%,2.7%, 10.4%, 8.8%, 2.1%, 10.3%, 14.0%, 1.9% and 6.0% were classed by theAmerican Bureau of Shipping, the Bureau Veritas, the China Classification Society,the Det Norske Veritas, the Germanischer Lloyd, the Korean Register of Shipping,the Lloyds Registry, the Nippon Kaiji Kyokai, the Registro Italian Navale and theRussian Maritime Register of Shipping, respectively.

Regarding the information about the operators’ countries of domicile, theinformation in Table 1 is complemented by country level data (Table 2).Whereas the statistical measures in Table 1 are in fact weighted averages basedon the number of vessels operated in each country, Table 2 provides simpleaverages and Std. Devs. based on the 114 available country values for eachvariable.

The average number of ratified IMO conventions by country is 24.8 (Table 2),yet the number of IMO conventions that applies to the average vessel is 32

180 JAN HOFFMANN ET AL.

Table 1. Variable Definitions and Descriptive Statistics.

Measurement Mean (Std. Dev.)

Dependent variablesFOREIGN 1 if vessel flag country is foreign, i.e. different from

the operator’s country of domicile, 0 otherwise0.457 (0.498)

PANAMA 1 if vessel flag foreign country is Panama,0 otherwise

0.129 (0.335)

BAHAMAS 1 if vessel flag foreign country is Bahamas,0 otherwise

0.031 (0.173)

ST. VINCENT ANDTHE GRENADINES

1 if vessel flag foreign country is St. Vincent andthe Grenadines, 0 otherwise

0.024 (0.153)

BERMUDA 1 if vessel flag foreign country is Bermuda,0 otherwise

0.003 (0.051)

ANTIGUA ANDBARBUDA

1 if vessel flag foreign country is Antigua andBarbuda, 0 otherwise

0.019 (0.138)

BELIZE 1 if vessel flag foreign country is Belize,0 otherwise

0.012 (0.108)

HONDURAS 1 if vessel flag foreign country is Honduras,0 otherwise

0.007 (0.084)

BOLIVIA 1 if vessel flag foreign country is Bolivia,0 otherwise

0.001 (0.033)

Independent variablesVessel characteristics

VAGE vessel age, years 18.1 (11.4)VSIZE vessel size, GT 13428 (23049)VCAP-DWT vessel carrying capacity in DWT 19483 (40732)VCAP-TEU vessel carrying capacity in TEU 186 (669)

Vessel operator country characteristicsOPERDEV 1 if a vessel operator’s country is a developed

country, 0 otherwise0.563 (0.496)

OPCASUAL Log of the operator country’s 1997–99 vesselcausality rate

0.150 (0.619)

OPOPENRE 1 if the country where the vessel’s operator isdomiciled is on open registry, 0 otherwise

0.029 (0.167)

OPIMONUM Number of conventions ratified by the countrywhere the vessel’s operator is domiciled.

32.0 (8.18)

OPGDPCAP Log of the GDP per capita (purchase power parity,ppp) of the country where the vessel’s operator isdomiciled.

9.57 (0.813)

OPPOPULA Log of the population size of the country where thevessel’s operator is domiciled.

17.5 (1.76)

OPLITERA Literacy rate of the country where the vessel’soperator is domiciled.

0.935 (0.102)

OPLIFEEX Life expectency of the country where the vessel’soperator is domiciled.

73.0 (8.65)

Determinants of Vessel Flag 181

Table 1. (Continued)

Measurement Mean (Std. Dev.)

Vessel construction informationBUILTOPER 1 if the vessel operator’s country is where the

vessel was built, 0 otherwise0.330 (0.470)

Vessel typeCARGO 1 if a vessel has cargo carrying capacity,

0 otherwise0.929 (0.256)

CONT 1 if a vessel has container carrying capacity,0 otherwise

0.207 (0.405)

VGCARGO 1 if a general-cargo vessel, 0 otherwise 0.167 (0.373)VCONT 1 if a container vessel, 0 otherwise 0.069 (0.254)VLBLK 1 if a liquid-bulk vessel, 0 otherwise 0.193 (0.395)VDBLK 1 if a dry-bulk vessel, 0 otherwise 0.197 (0.398)VPASS 1 if a passenger vessel, 0 otherwise 0.068 (0.251)VRORO 1 if a roll-on/roll-off vessel, 0 otherwise 0.046 (0.209)VREEF 1 if a reefer vessel, 0 otherwise 0.041 (0.197)VOROIL 1 if an ore/oil combined vessel, 0 otherwise 0.004 (0.064)VOFSHORE 1 if an off-shore vessel, 0 otherwise 0.069 (0.254)VFSTUG 1 if a fishing, tug or miscellaneous vessel,

0 otherwise0.146 (0.353)

Vessel classificationCIACS 1 if vessel is classed by a member of the IACS,

0 otherwise0.764 (0.424)

CABS 1 if vessel is classed by the American Bureau ofShipping, 0 otherwise

0.109 (0.312)

CBUV 1 if vessel is classed by the Bureau Veritas,0 otherwise

0.094 (0.292)

CCCS 1 if vessel is classed by the China ClassificationSociety, 0 otherwise

0.027 (0.161)

CDNV 1 if vessel is classed by the Det Norske Veritas,0 otherwise

0.104 (0.306)

CGEL 1 if vessel is classed by the Germanischer Lloyd,0 otherwise

0.088 (0.284)

CKOR 1 if vessel is classed by the Korean Register ofShipping, 0 otherwise

0.021 (0.143)

CLLR 1 if vessel is classed by the Lloyds Registry,0 otherwise

0.103 (0.304)

CNIK 1 if vessel is classed by the Nippon Kaiji Kyokai,0 otherwise

0.140 (0.347)

CRIN 1 if vessel is classed by the Registro Italian Navale,0 otherwise

0.019 (0.135)

CRUS 1 if vessel is classed by the Russian MaritimeRegister of Shipping, 0 otherwsie

0.060 (0.237)

182 JAN HOFFMANN ET AL.

Table 2. Descriptive Statistics for Vessel Operator Country Characteristics.

Variable Max Min Mean Std. Dev.

Casualty rate 1997–1999 23.33 0.20 2.26 2.80Variable: OPCASUALa

OPOPENRE 1 0 0.22 0.42OPIMONUM 48 0 24.80 11.03GDP percapita, USD ppp, 2000 35894 498 10719 9432.69Variable: OPGDPCAPa

Population 1273111290 27649 49175678 156029150Variable: OPPOPULAa

OPERLITERA 1.00 0.31 0.85 0.17OPERLIFE 80.80 36.40 70.00 9.17

Sources:OPCASUAL: Alderton and Winchester, 2002; OPENRE: Authors; OPIMONUM:www.imo.org; OPGDPCAP, OPPOPULA, OPERLITERA and OPERLIFE: www.mrdowling.com/800gdppercapita.htm.

a In the regression, the natural log of the ∗marked variable values is used.

(Table 1), indicating that countries with a larger operated fleet tend to ratifya higher number of IMO conventions. Equivalently, the average casualty rateby country (Table 2) is far above the weighted average by fleet size (Table 1),implying that larger registries have lower casualty rates. Twenty-two percent ofcountries can be considered open registries (Table 2), but only 2.9% of vesselsare operated by operators that are domiciled in these countries (Table 1).

Among the types of vessels, 16.7%, 6.9%, 19.3%, 19.7%, 6.8%, 4.6%, 4.1%,0.4%, 6.9% and 14.6% were general-cargo, container, liquid-bulk, dry-bulk,passenger, roll-on/roll-off, reefer, ore/oil combined, off-shore and fishing, tugor diverse miscellaneous service vessels, respectively. Within these broad vesseltypes, the composition of specific sub-types is given in Table A2 in the Annex.

5. EMPIRICAL RESULTS

5.1. Foreign Flag

Equations (1) and (2) are estimated via binominal probit analysis rather thanordinary least squares (OLS). Unlike OLS, probit analysis restricts the predictionsof FOREIGN to lie in the interval between zero and one. The probability ofobserving FOREIGN = 1 in the probit model may be expressed as:

Prob(FOREIGN = 1) = �(�′x) (3)

Determinants of Vessel Flag 183

Table 3. Vessel Foreign Flag Choice: Eq. (1) Probit Estimates.

Variable Coefficient (t Statistic) Marginal Probability

Vessel age, size and capacityVAGE −0.005 (−8.36) −0.002VSIZE 0.001 × 10−2 (11.8) 0.005 × 10−3

VCAP-DWT −0.008 × 10−4 (−1.38) −0.003 × 10−4

VCAP-TEU −0.003 × 10−2 (−2.52) −0.001 × 10−2

Vessel operator country characteristicsOPERDEV 0.730 (51.9) 0.289

Vessel construction informationBUILTOPER −0.786 (−51.0) −0.311

Vessel typeCARGO 0.439 (15.5) 0.174CONT 0.373 (19.3) 0.148

Vessel classificationCIACS 0.221 (13.7) 0.088Constant −0.970 (−28.1) −0.384

# Observations: 43140

�2 statistic: 8999

. . . where� is the standard normal distribution function; � is a vector of parameters;and x is a vector of explanatory variables as found in Eqs (1) and (2). The marginalprobabilities in the probit model for the x explanatory variables are:

∂E[FOREIGN]x

∂x= �(�′x)� (4)

. . . where � is the standard normal density function.

5.1.1. Eq. (1) EstimatesTable 3 reports the probit estimation results (coefficients and marginalprobabilities) for Eq. (1). The probit estimate fits the data well. Its chi-squaredstatistic is 8,999, well above the 21.7 critical value necessary for significance atthe 0.01 level for 9 degrees of freedom.

The estimation results suggest that the likelihood that an operator will choosea foreign flag for his vessel decreases with vessel age. This result is somewhatsurprising in that open registries tend to be associated with substandard ships,which in turn are more likely to be older. Possible explanations include pro-activeattempts by major registries to attract younger tonnage, giving them discountsand other benefits. Also, older vessels may have difficulties in complying with

184 JAN HOFFMANN ET AL.

international standards and thus do not trade internationally. Further, cabotageand inter-island services in many countries tend to be undertaken by older vesselsthat have to be nationally flagged as a result of cargo reservation regimes. In theUnited States (U.S.), the Jones (cabotage) Act requires domestic water traffic to betransported by U.S. flagged, built and crewed vessels, making the renewal of thefleet more expensive, thereby leading to an above-average aged nationally flaggedfleet. The probit marginal probability estimate for VAGE indicates that an increasein vessel age by one year decreases the probability by 0.002 that it will be foreignflagged.

The likelihood that a foreign flag will be chosen versus a national flag increaseswith the size of the vessel, but decreases with the carrying capacity of the vessel,all else held constant. Having adjusted for vessel carrying capacity, the positiverelationship with respect to vessel size may simply reflect the fact that larger vesselsare more likely to trade internationally than smaller ones. For fishing vessels, forexample, the coastal fleet of smaller vessels will almost always use the nationalflag, whereas larger vessels that catch fish in international waters may find it moreconvenient to use a foreign flag. The same may apply to other non-cargo vesselssuch as tug boats, dredging vessels, or passenger ferries. An increase in vessel sizeby one GT increases the probability by 0.000005 that the vessel will be foreignflagged.

The negative coefficient for VCAP-TEU suggests that the larger the carryingcapacities for container vessels that trade internationally, the lower is the likelihoodthat they will be foreign flagged. One possible explanation is that crewing costsare a smaller proportion of vessel operating costs for larger than smaller containervessels. An increase in container vessel carrying capacity by one TEU decreasesthe probability of a vessel being foreign flagged by 0.00001. The insignificance ofVCAP-DWT may be attributed to a multicollinearity problem involving VCAP-TEU and VSIZE.

The likelihood that the flag choice is foreign increases if the vessel operator’scountry is a developed country, but decreases if the vessel was built in the vesseloperator’s country. One possible explanation for the former is that wages are higherin developed countries, motivating vessel operators to reduce crewing costs byflagging out. Bergantino and Marlow (1999, p. 30) found that for United Kingdomoperators, labour and crewing factors account for 39% of the “factors affecting theuse of a foreign flag.” Also, environmental and safety controls tend to be less strictin developing countries, which would reduce the motivation for substandard shipsto move to possibly a laxer open registry. Further, some of the largest nationallyflagged fleets are from developing countries such as China, India or Indonesia,where markets are less liberalized and state-owned shipping companies still exist,making it simply impossible to use a foreign flag. When the vessel operator’s

Determinants of Vessel Flag 185

country is a developed country, the probability of the vessel being foreign flaggedincreases by 0.289.

One explanation for the negative relationship between FOREIGN andBUILTOPER is that an operator who buys or charters a nationally built vessel islikely to have other national linkages, thereby increasing the likelihood of a nationalflag. Also, it may be that the national shipyard was chosen because of governmentalfinancial incentives, which are linked to the obligation to use the national flag,at least during the initial years after construction. Further, government-operatedvessels are more likely to be built in national shipyards, and thus are almost certainto use the national flag. Among categorical variables, BUILTOPER has the largestmarginal probability effect on a vessel being foreign flagged. When the vesseloperator’s country is where the vessel was built, the probability that the vessel willbe foreign flagged decreases by 0.311.

The positive coefficients for CARGO and CONT suggest that cargo vessels, andin particular those with container carrying capacity, are more likely to be foreignflagged. This follows from the clear positive correlation between a vessel tradinginternationally –typically cargo vessels – and being foreign flagged. Passengervessels tend to have shorter journeys, thereby trading in just one country, which isusually the operator’s domicile, and flying its flag. Operating cargo and containervessels increases the probability that a vessel will be foreign flagged by 0.174 and0.148, respectively.

The positive coefficient for CIACS suggests that foreign flagged vessels aremore likely to be classed by an IACS member than nationally flagged vessels. Fora vessel to trade internationally, it is obliged to be “classed,” i.e. inspected andcertified regularly by a classification society, 10 of which are a member of theIACS. Many non-IACS-member classification societies are more likely to workat the national level for nationally flagged vessels. The IACS is a self-regulatedbody whose members claim that their inspections tend to be of high quality, whichwould suggest that a vessel classed by one of the ten IACS members is more likelyto comply with international safety and environmental standards. A vessel classedby an IACS member increases the probability of vessel foreign flagging by 0.088.

Based on the available data for this paper, our “forecast” concerning thechoice of a foreign flag was correct in 72% of the 43,140 observations whenestimating Eq. (1) and in 76% of the 41,470 observations when estimating Eq. (2),below.

5.1.2. Eq. (2) Estimates for all VesselsProbit estimation results for Eq. (2) are found in Table 4. Unlike Table 3, theregressions include only significant explanatory variables. The probit estimatefits the data well. The �2 statistic of 14,383 is well above the 41.6 critical

186 JAN HOFFMANN ET AL.

Table 4. Vessel Foreign Flag Choice: Eq. (2) Probit Coefficient Estimates.

Variable Coefficient (t statistic) Marginal Probability

Vessel characteristicsVSIZE 0.002 × 10−2 (15.0) 0.006 × 10−3

VCAP-DWT −0.005 × 10−3 (−8.72) −0.002 × 10−3

Vessel operator country characteristicsOPCASUAL −0.132 (−10.1) −0.052OPOPENRE 0.163 (3.51) 0.065OPIMONUM 0.020 (20.5) 0.008OPGDPCAP 0.402 (33.3) 0.160OPPOPULA −0.029 (−5.76) −0.012OPLIFEEX 0.009 (10.8) 0.004

Vessel construction informationBUILTOPER −0.902 (−49.9) −0.358

Vessel typeVGCARGO 0.212 (9.92) 0.084VCONT 0.273 (8.72) 0.108VDBLK 0.299 (14.8) 0.119VPASS −0.732 (−22.6) −0.291VREEF 0.362 (9.89) 0.144VFSTUG −0.810 (−31.2) −0.322

Vessel classificationCBUV 0.267 (10.8) 0.106CCCS −0.138 (−2.85) −0.055CDNV 0.207 (8.46) 0.082CGEL 0.319 (12.0) 0.127CKOR 2.34 (24.6) 0.929CNIK 0.525 (24.0) 0.208CRIN −0.294 (−5.56) −0.117CRUS −0.517 (−15.9) −0.205

Constant −4.68 (−28.8) −1.86

# Observations 41470

�2 statistic 14383

Note: Although the sum of the binary vessel type variables equals 1 for all observations, which wouldnot allow to include a constant term, as in this regression the non-significant variables wereexcluded a constant could be included.

value necessary for significance at the 0.01 level for 23 degrees of freedom.The coefficient and marginal probability estimates for VSIZE, VCAP-DWT andBUILTOPER are similar to those found in Table 3. VAGE is not estimated to besignificant any more.

Determinants of Vessel Flag 187

The coefficients of the vessel type variables are statistically significant, exceptthose (not reported) for VLBLK, VRORO, VOROIL and VORSHORE. Asexpected, the coefficients for the cargo vessels are positive and the coefficient forpassenger vessels (VPASS) and Tugs (VFSTUG) are negative. Among the vesseltype variables, a reefer vessel has the largest positive marginal probability effect ona vessel being foreign flagged, followed by dry-bulk and container vessels, respec-tively. Specifically, container, reefer and dry-bulk vessels increase the probabilitythat a vessel will be foreign flagged by 0.108, 0.144 and 0.119, respectively.

The coefficients of the vessel classification variables are statistically significant,except those (not reported) for CABS and CLLR. Among the classificationvariables, the Korean Register of Shipping (CKOR) has the largest positivemarginal effect on a vessel being foreign flagged, followed by Nippon KaijiKyokai (CNIK) and Germanischer Lloyd (CGEL). Specifically, vessels classed byNippon Kaiji Kyokai, the Korean Register of Shipping and Germanischer Lloydincrease the probability that a vessel will be foreign flagged by 0.208, 0.929 and0.127, respectively.

All vessel operator country variables except the Literacy rate are significant.Population and Flag casualty rate have negative signs, while Open registry, Numberof ratified IMO conventions, GDP per capita and Life expectancy have positivesign. The marginal probability of OPCASUAL, OPOPENRE, OPIMONUM,OPGDPCAP, OPPOPULA and OPLIFEEX are estimated as −0.052, 0.065, 0.008,0.160, −0.012 and 0.004 respectively.

A higher past casualty rate OPCASUAL may be an indicator of a lax nationalsafety regime, which appears to encourage ship operators to choose the nationalflag. If port state control inspections or insurance premiums reflected past casualtyrates, then one might have expected the contrary. In that case, a high past casualtyrate should encourage operators to choose other, foreign flags in order not to beaffected by the stigma of a bad national flag. In reality, the opposite is true; astrong national maritime safety regime appears to scare away national operatorsinto foreign flags. Each increase of the past casualty rate of 1% decreases theprobability of choosing a foreign flag by 0.00052. Accordingly, a doubling ofthe past casualty rate decreases this probability by 0.036. Note that taking logsof the original variable leads to a change in the interpretation of the estimatedparameters. A constant change in the explanatory logged variable is equivalentto a constant percentagechange in the original variable. It is then necessary toreduce the parameter values by two digits in relation to the percentage change ofthe independent variable. For example, if the estimated parameter “�” is −0.052, achange of 1% of the dependent variable leads to change in the probability of usinga foreign flag by −0.00052. In order to calculate specific cases, the followingapproach needs to be taken: Multiplying the explanatory variable by a factor Z

188 JAN HOFFMANN ET AL.

leads to a change of the probability to flag out by b (−LN (Z)). Again, by way ofexample, doubling the casualty rate leads to a reduction of the probability to flagout by (−0.052(–LN(2))) = 0.036044.

Similar to the previous point, a high number of ratified IMO conventionsOPIMONUM also appears to drive national operators towards the use of foreignflags. The estimated parameter can be interpreted as each additional ratified IMOconvention increases the probability of choosing a foreign flag by 0.008, i.e. almost1%. Having ratified 30 instead of 10 conventions, by way of example, thus increasesthe probability to flag out by 0.16.

If the operator is domiciled in an open-registry country OPOPENRE, thenthe country per definition becomes his “national flag.” In fact, there exist, forexample, Panamanian nationals who are domiciled in their country and own andoperate vessels. In other cases, internationally active ship operators may movetheir “domicile” to a country which also hosts an open registry. Such operators,however, are now more likely to choose a different (sic) foreign flag than the“national” flag of their country of domicile. If an operator is domiciled in any of thecountries identified as “open registries” in Table A1, he is more likely to choose adifferent flag than if he is domiciled in a national registry country, i.e. a country notdenominated as “open registry.” Although this may come as a surprise if we expectthat operators move to have their domicile in countries where they also flag theirvessel, in reality it appears that open registry countries tend to attract operatorswho are not inclined to use any particular flag. If an operator is domiciled in anopen registry country, the likelihood of using a foreign flag increases by 0.065.

To maintain a registry – be it “national” or open – implies high fixed costs. Therethus exist economies of scale, and it may be advantageous to be a “large” countryto maintain a national fleet. Taking the population OPPOPULA as an indicatorfor “size,” the estimated parameters confirm that operators from large countriesare more likely to maintain their national flag as compared to operators fromsmaller countries. A 1% increase in the operator country population decreases theprobability for an operator to choose a foreign flag by 0.00012. Being operatorin a country of 100 million inhabitants instead of one from a country of justone million leads to a decrease of the probability to flag out by 0.055, i.e.around 5%.

GDP per capita OPGDPCAP and life-expectancy OPLIFEEX are closelycorrelated and are main indicators for the level of a country’s general development.Just as OPERDEV in the regressions on Eq. (1), they have a positive impact onthe likelihood to flag out. A high level of development tends to coincide with highwage levels, strict security standards and corresponding labour regimes. All thesefactors appear to encourage operators to use foreign flags. An increase of the GDPper capita of 1% increases the probability to flag out by 0.0016. Doubling the GDP

Determinants of Vessel Flag 189

per capita leads to an increase of this probability by 0.11, i.e. more than one outof ten, which is a very strong impact given the large GDP differences that existbetween countries.

5.1.3. Eq. (2) Estimates by Vessel TypeTable 5 summarizes the results of regressions with Eq. (2), by vessel type, andTable 6 includes the marginal probability results. Only statistically significantvariables are included.General cargo vessels VGCARGO:Estimated parameter signs coincide with

those for the general estimation of Eq. (2) as presented in Table 4. The choice ofa foreign flag for a general cargo vessel is particularly strongly influenced by itscountry of built.Container vessels VCONT:Parameter signs are the same as in the general

case, except for OPOPENRE, where a negative sign is estimated. The DNVclassifications society is particularly strong in foreign flagged container vessels.Liquid bulk vessels VLBLK:Parameter signs coincide with the general

FOREIGN case, except for the vessel size component measured in GT. Here,larger liquid bulk vessel are less likely to be foreign flagged than bigger ones.Liquid bulk vessel operators appear to be particularly strongly reacting to a badpast safety record of its national registry, being encouraged to stay in the nationalregistry if its casualty rate has been high, with the estimated parameter being 2.5times higher in its value than in the general FOREIGN case. Older liquid bulkvessels are more likely to choose the national flag.Dry bulk vessels VDBLK:All parameter signs coincide with the general case.

Vessel size is particularly relevant as larger dry bulk vessels are more likely to beforeign flagged than smaller ones.Passenger vessels VPASS:In the case of passenger vessels, several parameter

signs are different from the general FOREIGN case. In particular, this vessel typeis the only one where a larger number of ratified IMO conventions appears todecrease the probability to choose a foreign flag. Passenger vessels are more likelyto be classed by the CCS than other vessel types.RoRo vessels VRORO:All parameter signs coincide with the general FOREIGN

case, except for CRUS, which is positive for RoRo vessels. The GL and BUVclassification societies are particularly strong in foreign flagged RoRo vessels.Reefer vessels VREEF:Estimated parameter signs for VSIZE and VCAP-DWT

have opposite signs from the general FOREIGN case. The Japanese classificationssociety is particularly strong in foreign flagged reefer vessels.Miscellaneous other vessels VOROIL, VOFSHORE, and VFSTUG:All

parameter signs are as expected, except for OPPOPULA, which has a positivesign for VOFSHORE, and CDNV, which has a negative sign for VOROIL.

190JA

NH

OFFM

AN

NE

TA

L.

Table 5. Vessel Foreign Flag Choices by Vessel Type: Eq. (2) Probit Coefficient Estimates.VARIABLE VGCARGO VCONT VLBLK VDBLK VPASS VRORO VREEF VOROIL VOFSHORE VFSTUG

Vessel characteristicsVAGE – – −0.010 −0.005 0.011 0.019 – – 0.017 0.006

(−6.21) (−2.81) (5.16) (5.08) (6.46) (3.46)VSIZE – – −0.005 × 10−3 0.005 × 10−2 0.002 × 10−2 0.003 × 10−2 −0.001 × 10−1 0.004 × 10−2 0.004 × 10−2 0.002 × 10−2

(9.62) (6.31) (9.63) (8.49) (−3.67) (2.94) (6.40) (3.97)VCAP-DWT 0.283 × 10−2 – – −0.002 × 10−2 – −0.003 × 10−2 0.002 × 10−1 – −0.001 × 10−2 –

(−5.76) (−3.99) (7.55) (−4.96)

Vessel construction informationBUILTOPER −1.05 −0.726 −1.03 −0.580 −1.13 −0.559 −0.387 – −0.886 −0.964

(−24.4) (−11.0) (−23.4) (−13.2) (−14.9) (−6.78) (−3.60) (−14.4) (−18.2)

Vessel operator country characteristicsOPCASUAL – −0.269 −0.328 −0.098 – −0.285 – – −0.234 −0.084

(−4.48) (−11.2) (−3.06) (−4.71) (−4.77) (−2.41)OPOPENRE – −0.417 0.252 – – – – – – –

(−1.96) (2.20)OPIMONUM – 0.022 0.032 0.042 −0.010 0.020 – 0.114 0.007 –

(4.71) (13.4) (16.5) (−2.34) (3.52) (2.92) (1.80)OPGDPCAP 0.193 0.134 0.266 0.400 0.619 0.381 0.595 2.02 0.272 0.263

(16.7) (1.85) (8.31) (14.4) (10.0) (4.95) (8.33) (3.26) (5.56) (7.61)OPPOPULA −0.016 −0.053 −0.067 −0.025 0.079 – −0.056 – 0.621 −0.109

(−4.08) (−2.41) (−5.47) (−2.09) (3.81) (−2.65) (3.98) (−7.78)OPLITERA 0.172 1.34 0.780 – −0.818 2.42 −1.58 – 1.51 –

(1.93) (2.05) (3.31) (−1.98) (3.79) (−2.90) (4.38)OPLIFEEX 0.006 0.037 0.006 0.016 −0.019 – 0.016 – – –

(5.58) (10.7) (3.83) (7.14) (−4.96) (2.77)

DeterminantsofVesse

lFlag

191

Vessel classificationCABS −0.724 −0.354 0.278 −0.305 −0.0.383 −0.907 −0.418 – 0.363 −0.199

(−7.82) (−4.02) (4.51) (−5.80) (−2.35) (−5.57) (−1.80) (4.84) (−2.35)CBUV – – 0.349 – 0.477 0.783 – – 0.638 0.327

(5.23) (4.98) (6.55) (5.82) (5.04)CCCS – −0.755 −0.263 −0.483 0.459 – – – −0.992 –

(−4.93) (−1.92) (−5.79) (1.74) (−3.35)CDNV – 0.675 0.444 – 0.395 0.195 – −1.92 0.489 –

(2.76) (7.55) (4.64) (1.91) (3.01) (5.94)CGEL 0.274 – 0.354 – 0.259 0.820 – – 0.454 –

(5.83) (4.19) (2.04) (6.43) (3.33)CKOR 2.18 1.64 2.07 1.46 2.60 2.03 0.954 – – 2.96

(5.52) (3.96) (9.00) (8.46) (5.35) (4.43) (2.50) (14.3)CLLR −0.245 −0.340 0.111 −0.192 – – – – 0.182 –

(−4.00) (−3.89) (1.86) (−3.64) (1.73)CNIK 0.418 0.454 0.495 0.306 – 0.517 0.620 – – 0.270

(6.83) (4.69) (10.0) (6.98) (4.93) (5.64) (1.70)CRIN – −0.841 −0.238 −0.880 – −0.500 – – – –

(−3.15) (−2.28) (−5.59) (−2.93)CRUS −0.237 −0.980 −0.692 −0.633 – 0.412 −0.539 – −0.558 −0.423

(−3.47) (−5.25) (−7.75) (−9.41) (2.22) (−5.00) (−2.86) (−4.54)

Constant −5.23 −4.24 −3.40 −5.38 −5.84 −7.18 −4.49 −25.5 −5.87 −1.48(−16.3) (−5.60) (−8.67) (−13.9) (−8.98) (−10.9) (−5.59) (−3.97) (−11.8) (−3.49)

# Observations 6995 2738 8149 8171 2833 1917 1658 169 2945 5895�2 statistic 2073 579.1 2500 2318 826.1 572.6 766.6 193.7 697.5 1582

Note: t-statistics are in parentheses.

192JA

NH

OFFM

AN

NE

TA

L.

Table 6. Vessel Foreign Flag Choices by Vessel Type: Eq. (2) Probit Marginal Probability Estimates.

VARIABLE VGCARGO VCONT VLBLK VDBLK VPASS VRORO VREEF VOROIL VOFSHORE VFSTUG

Vessel characteristicsVAGE – – −0.004 −0.002 0.003 0.007 – – 0.007 0.001VSIZE – – −0.002 × 10−3 0.002 × 10−2 0.006 × 10−3 0.001 × 10−2 −0.004 × 10−2 0.001 × 10−2 0.001 × 10−2 0.003 × 10−3

VCAP-DWT 0.001 × 10−2 – – −0.008 × 10−3 – −0.001 × 10−2 0.001 × 10−2 – −0.006 × 10−3 –

Vessel construction informationBUILTOPER −0.421 −0.243 −0.411 −0.220 −0.286 −0.221 −0.147 – −0.352 −0.205

Vessel operator country characteristicsOPCASUAL – −0.090 −0.131 −0.037 – −0.113 – – −0.093 −0.018OPOPENRE – −0.140 0.100 – – – – – – –OPIMONUM – 0.007 0.013 0.157 −0.003 0.008 – 0.033 0.003 –OPGDPCAP 0.193 0.045 0.106 0.151 0.156 0.150 0.225 0.582 0.108 0.056OPPOPULA −0.016 −0.018 −0.027 −0.009 0.020 – −0.021 – 0.025 −0.023OPLITERA 0.172 0.448 0.311 – −0.206 0.957 −0.597 – 0.599 –OPLIFEEX 0.006 0.012 0.003 0.006 −0.005 – 0.006 – – –

Vessel classificationCABS −0.289 −0.118 0.111 −0.115 −0.097 −0.358 −0.158 – 0.145 −0.042CBUV – 0.019 0.139 – 0.120 0.309 – – 0.254 0.070CCCS – −0.252 −0.105 −0.183 0.116 – – – −0.395 –CDNV – 0.226 0.177 – 0.100 0.077 – 0.553 0.195 –CGEL 0.109 – 0.141 – 0.065 0.324 – – 0.181 –CKOR 0.870 0.550 0.824 0.552 0.655 0.802 0.362 – – 0.630CLLR −0.098 −0.114 0.044 −0.073 – – – – 0.072 –CNIK 0.167 0.152 0.197 0.116 – 0.204 0.235 – – 0.057CRIN – −0.281 −0.095 −0.333 – −0.198 – – – –CRUS −0.095 −0.328 −0.276 −0.240 – 0.163 −0.204 – −0.222 −0.090Constant −2.09 −1.42 −1.36 −2.04 −1.47 −2.84 −1.70 −7.34 −2.34 −0.316

Determinants of Vessel Flag 193

Although differences in the magnitude of estimated parameters persist, ingeneral, the main conclusions concerning the direction of impact of differentvariables on the probability to flag out are not modified if we look at individualvessel types; the one exception being passenger ships, where several parametersare estimated with a different sign from the general FOREIGN case.

5.2. Latin American and Caribbean Foreign Flags

In this section, we investigate determinants of the foreign flag choice for eightspecific Latin American and Caribbean open registries. Probit estimation resultsfor Eqs (1) and (2) for these registries are presented in Tables 7–10.

5.2.1. Eq. (1) EstimatesIn Table 7, probit coefficient estimation results for Eq. (1) are reported and probitmarginal-probability estimates are reported in Table 8. For all eight probit estimatesof Eq. (1) found in Table 7, the estimates fit the data well. The �2 statistics are wellabove the critical value necessary for significance at the 0.01 level for 9 degreesof freedom.

ThePanamaregistry, the world’s largest, has similar characteristics to that of theoverall foreign flagged fleet (see Table 3), albeit with a special emphasis on Asianoperators. In Table 7 the Panama probit estimated coefficients of the variablesVAGE, VSIZE, VCAP-DWT, CARGO and CIACS have the same signs as thosefound in Table 3. Younger vessels and those with greater gross tonnage are morelikely to be registered in Panama, but this likelihood decreases the greater the DWT-carrying capacity (adjusting for gross tonnage). Cargo as opposed to non-cargo andIACS as opposed to non-IACS classed vessels are more likely to be registered inPanama. Coefficient signs differ in Tables 3 and 7 for the container vessel variables,VCAP-TEU and CONT, and country where vessel was built (BUILTOPER). Thelikely reason lies in the fact that Panama specializes in flagging Asian operators,in particular vessels classed by the Japanese and Korean Asian societies, NipponKaiji Kyokai and Korean Register of Shipping (more information about individualclassification societies will be incorporated in Table 10). Since Asian operators areparticularly strong in global container shipping and since Panama has a relativelylarge share of larger container vessels, a priori signs for VCAP-TEU and CONTfor the Panama registry are positive. In Table 7 VCAP-TEU has the expectedpositive sign, but CONT has a negative sign. The Asian countries, Korea, Japan,China and Hong Kong, are the world’s largest ship builders, which increases thelikelihood that a ship constructed in one of these countries is also registered inPanama. In Table 7 BUILTOPER has the expected positive sign. The coefficients

194JA

NH

OFFM

AN

NE

TA

L.

Table 7. Vessel Open Registry Choices: Eq. (1) Probit Coefficient Estimates.*

Variable Panama Bahamas St.Vinc. and the Gren. Bermuda Antigua and Barbuda Belize Honduras Bolivia

Vessel age, size and capacityVAGE −0.007 0.005 × 10−1 0.016 0.003 −0.016 0.014 0.024 0.014

(−9.44) (0.41) (11.9) (0.87) (−7.68) (8.10) (12.4) (3.72)VSIZE 0.001 × 10−2 0.002 × 10−2 −0.003 × 10−2 0.001 × 10−2 −0.002 × 10−1 −0.009 × 10−2 −0.002 × 10−1 −0.007 × 10−2

(12.6) (13.8) (−5.45) (6.55) (−12.9) (−5.94) (−6.75) (−1.77)VCAP-DWT −0.003 × 10−3 −0.006 × 10−3 0.001 × 10−2 −0.004 × 10−3 0.007 × 10−2 0.003 × 10−2 0.007 × 10−2 0.003 × 10−2

(−5.41) (−9.49) (3.86) (−3.36) (7.27) (3.74) (4.76) (1.27)VCAP-TEU 0.003 × 10−2 −0.001 × 10−1 −0.004 × 10−1 −0.008 × 10−2 0.009 × 10−1 −0.001 × 10−1 −0.008 × 10−1 −0.002

(2.06) (−6.70) (−4.48) (−1.68) (12.41) (−0.52) (−1.00) (−1.09)

Vessel operator country characteristicsOPERDEV 0.012 0.559 0.131 0.397 1.27 −0.196 −0.244 −0.297

(0.67) (18.4) (4.47) (4.81) (19.4) (−4.65) (−4.41) (−2.56)

Vessel construction informationBUILTOPER 0.201 −0.598 −0.717 −0.875 −0.622 −0.350 −0.266 −0.347

(11.2) (−16.8) (−15.4) (−5.21) (−13.1) (−6.47) (−3.87) (−2.21)

Vessel typeCARGO 0.272 0.163 0.233 0.124 1.01 0.222 −0.165 −0.134

(7.35) (2.31) (3.37) (0.59) (3.34) (2.64) (−2.08) (−0.81)CONT −0.145 0.287 0.331 0.266 1.08 −0.160 0.062 0.266)

(5.90) (8.18) (7.63) (2.84) (23.3) (−1.88) (0.44) (1.04)

Vessel classificationCIACS 0.249 0.351 −0.100 0.119 0.296 −0.112 −0.190 −0.334

(11.8) (8.09) (−3.06) (1.05) (4.73) (−2.74) (−3.76) (−3.35)Constant −1.65 −2.76 −2.30 −3.51 −3.98 −2.28 −2.30 −2.72

(−36.9) (−31.6) (−27.4) (−12.9) (−12.6) (−22.1) (−21.4) (−12.7)

# Observations 43140 43140 43140 43140 43140 43140 43140 43140�2 statistic 1532 1291 895.5 188.8 3156 587.7 597.6 94.1

∗ t statistics are in parentheses.

DeterminantsofVesse

lFlag

195

Table 8. Vessel Open Registry Choices: Eq. (1) Probit Marginal Probability Estimates.

Variable Panama Bahamas St.Vinc. and the Gren. Bermuda Antigua and Barbuda Belize Honduras Bolivia

Vessel age, size and capacityVAGE −0.001 0.002 × 10−2 0.005 × 10−1 0.001 × 10−2 −0.001 × 10−2 0.001 × 10−1 0.001 × 10−2 0.005 × 10−3

VSIZE 0.002 × 10−3 0.008 × 10−4 −0.001 × 10−3 0.004 × 10−5 −0.001 × 10−4 −0.008 × 10−4 −0.001 × 10−4 −0.002 × 10−5

VCAP-DWT −0.006 × 10−4 −0.003 × 10−4 0.004 × 10−4 −0.001 × 10−5 0.005 × 10−5 0.003 × 10−4 0.004 × 10−5 0.009 × 10−6

VCAP-TEU 0.005 × 10−3 −0.007 × 10−3 −0.001 × 10−2 −0.002 × 10−4 0.007 × 10−4 −0.001 × 10−3 −0.004 × 10−4 −0.005 × 10−4

Vessel operator country characteristicsOPERDEV −0.002 0.026 0.004 0.001 0.009 × 10−1 −0.002 −0.001 × 10−1 −0.010 × 10−2

Vessel construction informationBUILTOPER 0.040 −0.028 −0.024 −0.002 −0.004 × 10−1 −0.003 −0.002 × 10−1 −0.001 × 10−1

Vessel typeCARGO 0.053 0.008 0.008 0.003 × 10−1 0.007 × 10−1 0.002 −0.010 × 10−2 −0.004 × 10−2

CONT −0.029 0.013 0.011 0.008 × 10−1 0.008 × 10−1 −0.001 0.004 × 10−2 0.009 × 10−2

Vessel classificationCIACS 0.049 0.016 −0.003 0.003 × 10−1 0.002 × 10−1 −0.001 −0.001 × 10−1 −0.001 × 10−1

Constant −0.324 −0.128 −0.078 −0.010 −0.003 −0.019 −0.001 −0.001

196JA

NH

OFFM

AN

NE

TA

L.

Table 9. Vessel Open Registry Choices: Eq. (2) Probit Coefficient Estimates.*

Variable Panama Bahamas St.Vincent/ Bermuda Antigua/ Belize Honduras BoliviaGrenadines Barbuda

Vessel characteristicsVAGE – – 0.015 – −0.015 0.016 0.028 0.029

(10.1) (−6.37) (9.17) (12.8) (7.44)VSIZE 0.001 × 10−2 0.001 × 10−2 −0.002 × 10−2 0.009 × 10−3 −0.009 × 10−2 −0.001 × 10−1 −0.002 × 10−1 –

(9.97) (9.87) (−9.20) (8.62) (−7.02) (−6.32) (−4.69) –VCAP-DWT −0.002 × 10−3 −0.005 × 10−3 – 0.004 × 10−2 0.004 × 10−2 0.004 × 10−2 0.006 × 10−2 –

(−4.15) (−6.75) (4..11) (4.11) (4.19) (2.32)

Vessel construction informationBUILTOPER −0.152 −0.526 −0.660 −0.778 −0.596 −0.390 −0.467 −0.634

(−6.41) (−13.7) (−13.3) (−4.31) (−10.7) (−6.65) (−5.45) (−2.27)

Vessel operator country characteristicsOPCASUAL −0.098 – −0.041 – 0.348 −0.129 – −0.189

(−6.00) (−1.69) (6.23) (−3.99) (−2.99)OPOPENRE −0.208 0.518 −0.473 – 1.06 0.412 0.415 –

(−2.76) (7.32) (−5.19) (5.39) (4.63) (4.14)OPIMONUM −0.011 – 0.021 0.029 0.043 −0.010 – –

(−8.38) (9.54) (5.10) (7.52) (−3.90)OPGDPCAP 0.252 0.371 – 0.320 0.382 0.133 0.258 –

(13.4) (10.8) (3.42) (4.05) (4.78) (5.62)OPPOPULA 0.028 – −0.094 – 0.346 – – −0.145

(4.23) (−9.60) (16.1) (−4.98)OPLITERA −0.374 1.58 −1.30 – 2.47 – −0.901 –

(−2.51) (5.08) (−8.66) (2.29) (−2.94)OPLIFEEX 0.013 −0.009 0.006 −0.026 0.031 −0.004 −0.012 –

(11.2) (−6.07) (2.85) (−8.47) (6.71) (−1.86) (−4.10)

DeterminantsofVesse

lFlag

197

Vessel typeVGCARGO – −0.387) 0.424 −0.576 0.826 – 0.672 0.304

(−6.37) (9.99) (−1.99) (14.9) (7.17) (2.33)VCONT 0.203 −0.671 0.184 – 0.958 – 0.564 –

(5.88) (−8.79) (2.03) (12.4) (2.38)VLBLK – −0.665 – −0.582 – −0.147 0.502 0.344

(−11.1) (−4.26) (−2.65) (5.06) (2.66)VDBLK – −0.531 0.523 – – – 0.521 –

(−9.05) (11.9) (4.69)VPASS −0.470 −0.379 – – – −0.582 0.336 –

(−10.0) (−5.47) (−4.23) (2.44)VRORO – −0.651 0.418 0.442 0.831 – – –

(−8.18) (5.73) (3.53) (7.74)VREEF 0.392 – 0.363 0.525 – 0.210 0.723 –

(9.48) (5.14) (3.86) (2.58) (4.96)VOROIL – – – – – – – –VOFSHORE 0.216 −0.459 0.264 – – – – –

(6.11) (−6.87) (4.23)VFSTUG −0.717 −1.01 – – – −0.225 – –

(−17.3) (−12.6) (−3.61)

198JA

NH

OFFM

AN

NE

TA

L.

Table 9. (Continued)

Variable Panama Bahamas St.Vincent/ Bermuda Antigua/ Belize Honduras BoliviaGrenadines Barbuda

Vessel classificationCABS −0.077 0.289 −0.290 0.478 −0.840 −0.259 −0.314 –

(−2.25) (6.01) (−4.09) (4.76) (−2.73) (−3.01 (−2.01)CBUV 0.646 0.388 0.115 – 0.406 −0.189 – –

(1.83) (8.22) (2.43) (5.21) (−2.62)CCCS – – 0.559 – – – – –

(7.60)CDNV −0.199 0.576 −0.246 – – −0.240 −0.392 –

(−5.34) (13.9) (−4.05) (−2.67) (−2.51)CGEL −0.463 −0.188 −0.313 – 1.18 −0.146 −0.271 –

(−10.7) (−2.97) (−5.20) (20.7) (−1.96) (−2.63)CKOR 0.616 −0.946 −0.458 – – −0.246 – –

(11.5) (−3.25) (−3.06) (−1.72)CLLR −0.123 0.364 −0.263 0.353 − −0.277 −0.201 −

(−3.43) (8.21) (−4.52) (3.72) (−3.21) (−1.83)CNIK 0.846 – −0.233 – −0.990 – – –

(31.2) (−3.75) (−3.16)CRIN −0.215 – – – – – – –

(−2.80)CRUS −0.671 −1.08 −0.189 – −0.657 – −0.718 –

(−9.71) (−4.93) (−2.97) (−2.40) (−3.57)Constant −4.48 −6.05 −0.479 −5.43 −18.4 −2.81 −3.90 −1.48

(−20.8) (−19.6) (−1.77) (−5.81) (−20.2) (−11.0) (−10.4) (−3.04)

# Observations 41470 41470 41470 41470 41470 41470 41470 41470�2 statistic 5613 1943 1307 377.4 4115 632.8 702.5 117.1

∗ t statistics are in parentheses.

DeterminantsofVesse

lFlag

199

Table 10. Vessel Open Registry Choices: Eq. (2) Probit Marginal Probability Estimates.

Variable Panama Bahamas St.Vincent/ Bermuda Antigua/ Belize Honduras BoliviaGrenadines Barbuda

Vessel characteristicsVAGE – – 0.004 × 10−1 – −0.008 × 10−3 0.009 × 10−1 0.006 × 10−3 0.001 × 10−2

VSIZE 0.002 × 10−3 0.004 × 10−4 −0.004 × 10−4 0.008 × 10−6 −0.006 × 10−5 −0.007 × 10−4 −0.004 × 10−5 –VCAP-DWT −0.004 × 10−4 −0.002 × 10−4 – – 0.002 × 10−5 0.003 × 10−4 0.001 × 10−5 –

Vessel construction informationBUILTOPER −0.023 0.017 −0.017 −0.006 × 10−1 −0.004 × 10−1 −0.003 −0.001 × 10−1 −0.004 × 10−1

Vessel operator country characteristicsOPCASUAL −0.015 – −0.001 – −0.002 × 10−1 −0.009 × 10−1 – −0.001 × 10−1

OPOPENRE −0.032 0.016 −0.012 – 0.007 × 10−1 0.003 0.009 × 10−2 –OPIMONUM −0.002 – 0.005 × 10−1 0.002 × 10−2 0.003 × 10−2 −0.007 × 10−2 – –OPGDPCAP 0.038 0.012 – 0.003 × 10−1 0.002 × 10−1 0.010 × 10−1 0.006 × 10−2 –OPPOPULA 0.004 – −0.002 – 0.002 × 10−1 – – −0.009 × 10−4

OPLITERA −0.057 0.050 −0.034 – 0.002 – −0.002 × 10−1 –OPLIFEEX 0.002 −0.003 × 10−1 0.002 × 10−1 −0.002 × 10−2 0.002 × 10−2 −0.003 × 10−2 −0.003 × 10−3 –

200JA

NH

OFFM

AN

NE

TA

L.

Table 10. (Continued)

Variable Panama Bahamas St.Vincent/ Bermuda Antigua/ Belize Honduras BoliviaGrenadines Barbuda

Vessel typeVGCARGO – −0.012 0.011 −0.005 × 10−1 0.005 × 10−1 – 0.002 × 10−1 0.002 × 10−1

VCONT 0.031 −0.021 0.005 – 0.006 × 10−1 – 0.001 × 10−1 –VLBLK – −0.021 – −0.005 × 10−1 – −0.001 0.001 × 10−1 0.002 × 10−1

VDBLK – −0.017 0.014 – – – 0.001 × 10−1 –VPASS −0.072 0.012 – – – −0.004 0.008 × 10−2 –VRORO – −0.021 0.011 0.004 × 10−1 0.005 × 10−1 – – –VREEF 0.060 – 0.010 0.004 × 10−1 – 0.002 0.002 × 10−1 –VOROIL – – – – – – – –VOFSHORE 0.033 −0.015 0.007 – – – – –VFSTUG −0.109 −0.032 – – – −0.002 – –

Vessel classificationCABS −0.012 0.009 −0.008 0.004 × 10−1 −0.005 × 10−1 −0.002 −0.007 × 10−2 –CBUV 0.010 0.012 0.003 – 0.003 × 10−1 −0.001 – –CCCS – – 0.015 – – – – –CDNV −0.030 0.018 −0.006 – – −0.002 −0.009 × 10−2 –CGEL −0.071 −0.006 −0.008 – 0.008 × 10−1 −0.001 −0.006 × 10−2 –CKOR 0.094 −0.030 −0.012 – – −0.002 – –CLLR −0.019 0.012 −0.007 0.003 × 10−1 – −0.002 −0.005 × 10−2 –CNIK 0.129 – −0.006 – −0.006 × 10−1 – – –CRIN −0.033 – – – – – – –CRUS −0.102 −0.034 −0.005 – −0.005 × 10−1 – −0.002 × 10−1 –

Constant 0.113 −0.099 −0.013 −0.004 −0.012 −0.020 −0.009 × 10−1 −0.009 × 10−1

Determinants of Vessel Flag 201

of OPERDEV are positive in both tables; the coefficient is highly significant inTable 3 but statistically insignificant in Table 7. The latter may reflect the factthat the Asian countries, Korea, Taiwan and China, do not appear as developedcountries in OPERDEV. Among the categorical explanatory variables, a cargovessel has the largest marginal probability effect on selection of the Panama openregistry, followed by an IACS classed vessel and country where vessel was built.A cargo vessel increases the probability of choosing the Panama flag by 0.053,while an IACS classed vessel and country where vessel was built increases thisprobability by 0.049 and 0.040, respectively.Bahamasis the second largest Latin American and Caribbean register. The

Bahamas flag tends to be used by cruise and other passenger ship operators andtheir vessels are often classed by the American Bureau of Shipping. The Bahamasprobit coefficient estimates for Eq. (1) found in Table 7 have the same signs as thosefor the overall foreign flagged fleet found in Table 3, except for vessel age. TheBahamas coefficient for the latter is positive, but statistically insignificant. Amongthe categorical explanatory variables, country where vessel is built has the largestmarginal probability effect on selection of the Bahamas open registry, followed bya developed country and an IACS classed vessel. If the vessel operator’s country iswhere the vessel was built, the probability of choosing the Bahamas flag decreasesby 0.028, while if a vessel operator’s country is a developed country and if it is anIACS classed vessel, this probability increases by 0.026 and 0.016, respectively.

The third largest Latin American and Caribbean register is St. Vincent and theGrenadines. This registry has a bad reputation and features high on Port StateControl black lists (Winchester, 2003). A comparison of the St. Vincent and theGrenadines probit coefficient estimates in Table 7 with those for the overall foreignflagged fleet in Table 3 reveals that the former has vessel characteristics unlikethat of the latter. Specifically, the St. Vincent and the Grenadines registry attractsolder vessels, smaller gross-ton vessels, and non-IACS classed vessels. Among thecategorical explanatory variables, BUILTOPER decreases the probability of a St.Vincent and the Grenadines flag by 0.024, whereas a container and cargo vesselincreases this probability by 0.011 and 0.008, respectively.

TheBermudaregistry is closely linked to British operators. The Bermuda probitcoefficient estimates for Eq. (1) have the same signs as for the Bahamas coefficientestimates. However, unlike the Bahamas results, the coefficients of CARGO andCIACS are statistically insignificant. If the operator country is also the countryof build, this decreases the probability of a Bermuda flag by 0.002, whereas adeveloped operator country and a container vessel increases this probability by0.001 and 0.0008, respectively.

The Antigua and Barbudaregistry specializes in German container and generalcargo operators. Like the overall foreign flagged fleet, vessels under this flag tend

202 JAN HOFFMANN ET AL.

to be younger, but unlike the former, vessels tend to be smaller. Further, unlikethe former, vessels tend to have larger carrying capacities (adjusting for grosstonnage). Among the categorical variables, OPERDEV has the largest marginalprobability effect (0.0009) on Antigua and Barbuda flagging, likely reflecting thehigh incidence of German operators, followed by container (0.0008) and cargo(0.0007) vessels.

The registries,Belize,HondurasandBolivia, are relatively small open registriesand appear on the black lists of many Port State Control organizations. The signs oftheir probit estimated coefficients are similar, except those for CARGO and CONT.The Belize registry attracts cargo vessels, whereas the Honduras and Boliviaregistries attract non-cargo vessels. The Belize registry attracts non-containervessels, whereas the Honduras and Bolivia registries tend to attract containervessels. Unlike that for the overall foreign flagged fleet (see Table 3), vessels inthese registries tend to be older, smaller in size, come from a developing rather thana developed country, and to be classed by non-IACS societies. Among categoricalvariables, country where vessel was built has the largest marginal probabilityeffect on the Belize and Honduras registry, decreasing the probability Belize andHonduras flagging by 0.003 and 0.0002 respectively.

A comparison of the probit marginal probability estimates (Table 8) for theeight Latin American and Caribbean open registries provide further insight intothe dissimilarities among these registries. An increase in a vessel’s size by one grosston increases the probability of the vessel being flagged in Panama, Bahamas andBermuda by 0.000002, 0.0000008 and 0.00000004, respectively, but decreases theprobability of the vessel being flagged in St. Vincent and the Grenadines, Antiguaand Barbuda, Belize, Honduras and Bolivia by 0.000001, 0.0000001, 0.0000008,0.0000001 and 0.00000002, respectively. A vessel classed by the IACS increasesthe probability of the vessel being flagged in Panama, Bahamas, Bermuda and An-tigua and Barbuda by 0.049, 0.016, 0.0003 and 0.0002, respectively, but decreasesthe probability of the vessel being flagged in St. Vincent and the Grenadines,Belize, Honduras and Bolivia by 0.003, 0.001, 0.0001 and 0.0001, respectively.

5.2.2. Eq. (2) EstimatesIn Table 9 probit coefficient estimation results for Eq. (2) are reported and probitmarginal-probability estimates are reported in Table 10. Unlike Table 7, the resultsinclude only statistically significant explanatory variables. For all eight probitestimates of Eq. (2), the estimates fit the data well. The �2 statistics are wellabove the critical value necessary for significance at the 0.01 level. The statisticallysignificant coefficient estimates for VAGE, VSIZE, VCAP-DWT and BUILTOPERand their marginal probability estimates are in general similar to those found inTables 7 and 8.

Determinants of Vessel Flag 203

5.2.2.1. Vessel types.For the Panamaregistry, the signs of the probit coefficientestimates of the vessel type variables in Table 9 coincide with those of the generalFOREIGN case. As expected, the coefficients for the cargo vessels are positive andthat for passenger vessels (VPASS) is negative. Among the vessel type variablesfound in Table 10, a reefer vessel has the highest positive probability of flyinga foreign Panama flag, i.e. a reefer vessel increases the probability of a foreignPanama flag by 0.06.

For the Bahamasregistry, all vessel type variables except VREEF and VOROILare statistically significant. Tugs and similar vessels are those least likely to flythe Bahamas flag. Among the vessel type variables for the St. Vincent and theGrenadinesregistry, VDBLK has the strongest positive impact on the likelihoodto choose this flag.

Vessel-type estimation results for Bermudasuggest that this registry attractsmostly reefer and roll-on/roll-off vessels and is unlikely to attract liquid-bulkvessels. Also – different from the general FOREIGN case – a general cargo vesselreduces the likelihood of being Bermuda foreign flagged.

The Antigua and Barbudaforeign flag attracts mainly general-cargo, containerand roll-on/roll-off vessels, with all other vessel type variables not estimated to besignificant. A reefer vessel increases the likelihood of choosing the Belizeregister,and a passenger vessel makes this choice particularly unlikely. The Hondurasopen registry is more likely to be chosen if the vessel is a general cargo or a reefervessel.

The Bolivia open registry attracts relatively few different types of vessels, i.e.only the vessel types VGCARGO and VLBLK are statistically significant.

5.2.2.2. Classification societies.For the Panamaregistry, the probit coefficientestimates of the vessel classification variables are statistically significant, exceptthose (not reported) for CCCS. Among the significant classification variables, onlythe coefficients of CBUV, CKOR and CNIK are positive, highlighting the fact thatmany Panama flagged vessels tend to be classed by the Asian classification societiesof Korea and Japan.

Vessels flying theBahamasforeign flag are likely to be classed by the AmericanBureau of Shipping (CABS), Bureau Veritas (CBUV), Det Norske Veritas (CDNV)and Lloyds Register (CLLR). Among these societies CDNV has the highestestimated impact on the likelihood to choose the Bahamas foreign flag.

All vessel classification variables for St. Vincent and the Grenadinesareestimated as negative, except the coefficients for CBUV and CCCS. Hence, vesselsclassed by the French and Chinese classification societies are attracted to theSt. Vincent and the Grenadines foreign flag, and vessels classed by other IACSmembers are less likely to choose the St. Vincent and the Grenadines flag.

204 JAN HOFFMANN ET AL.

For the Bermudaregistry, only the classification societies ABS and LloydsRegister have significant marginal probability effects on this registry, both with apositive sign.

The probit estimate for the Antigua and Barbudaflag reveals that only theclassification societies Bureau Veritas and Germanischer Lloyd have positiveeffects on this registry. The latter is by far the strongest, reflecting the highproportion of German operators in this registry.

For the BelizeandHondurasopen registries, none of the statistically significantvessel classification variables in the probit estimates are positive. This can beinterpreted as negative indicator for the quality of these registries as far as theircontrol and certification of vessel safety is concerned. No vessel classificationvariables were statistically significant in the Bolivia probit estimate.

5.2.2.3. Operator country characteristics.Vessel operators that choose thePanamaflag tend to come from countries with similar characteristics as thegeneral FOREIGN case. Exceptions are the variables OPOPENRE, where thesign is negative for Panama, OPIMONUM, where the sign is negative, andOPPOPULA, where the sign is positive, i.e. different from the general FOREIGNcase. This indicates that vessel operators that choose the Panama flag are lesslikely to be domiciled in other open registry countries, but rather in relativelylarge countries without open registry. Also, Panama does not appear to attractvessels whose operators are trying to avoid a high number of ratified IMOconventions.Bahamasattracts vessels from countries with similar characteristics as the

general FOREIGN case. The impact of a high GDP per capita is particularlystrong.

For St. Vincent and the Grenadines, all operator country variables exceptOPGDPCAP are significant, and all except OPOPENRE have the expected sign.

For Bermuda, most country characteristics variables are not significant, exceptOPIMONUM and OPGDPCAP, which are estimated with the expected positivesign, and BUILTOPER and OPFILEEX, which are estimated with a negative sign.

In the case of Antigua and Barbuda, again the strong participation of Germanoperators appears to influence the estimated parameter values. Unlike in thegeneral FOREIGN case, vessels in this registry do not come from countries witha bad national flag safety record and they are from relatively larger countries.They come from countries that have ratified a relatively high number of IMOconventions.Belizeand Bolivia seem to attract vessels from countries with good national

flag safety records, i.e. a relatively strong estimated negative parameter forOPCASUAL. Different from the general FOREIGN case, Belize and Honduras

Determinants of Vessel Flag 205

both appear to attract vessels from countries with a relatively low life expectancy.Could this be interpreted as implying that those who choose these flags attach alower priority to safety of life at sea?

6. CONCLUSIONS

6.1. Determinants of a Vessel’s Flag

The empirical results of our research suggest that older vessels and vessels thatare not classed by an IACS member are more likely to be nationally flagged thanforeign flagged. This may come as a surprise to some if the expectation was thatforeign flagged ships would be “worse,” i.e. older and less stringently controlled,than nationally flagged ones. A determining factor for a vessel to choose a foreignflag appears to be the likelihood that it trades internationally, as do most cargo andlarger vessels as compared to passenger or smaller units. Further, if a vessel is builtin the operator country this increases the likelihood of remaining in the nationalflag registry.

Another interesting conclusion is certainly the impact of socio economicvariables. The operators from developed countries are more likely to choose aforeign flag than those from countries with a lower GDP per capita or with lowindicators concerning human development such as the literacy rate or the lifeexpectancy. Our empirical results suggest that higher wages and labour standardsmay scare operators away from national registries.

If the operator’s country of domicile has ratified a high number of IMOconventions, and also if its national flag register has a positive past safety record,i.e. a low casualty rate, this increases the likelihood to use a foreign flag. Thisshould of course not be interpreted in a way that ratification of IMO conventionsshould not be recommended – neither would we recommend to reduce the GDPper capita or the life expectancy in order to attract more nationally flagged vessels.However, international organizations and national registries should be aware ofthis situation where high national safety standards appear to encourage vesseloperators to choose a foreign flag.

There exist of course big differences between Flag States. We have shown howdifferent Latin American and Caribbean registries have specialized in certainoperator countries and vessel types and sizes. From the vessel types, their ages,and the chosen classification societies, one would expect that Panama and Antiguawill be less likely to have their vessels detained, whereas St. Vincent and theGrenadines, Belize, Honduras and Bolivia should find their vessels with a higherprobability of being inspected and detained by Port State Control authorities.

206 JAN HOFFMANN ET AL.

For Panama, the world’s largest registry by far, our data suggests that it has afleet of relatively larger and younger vessels, more likely to be classed by an IACSmember. This IACS member is most likely to come from Asia, and the same appliesto the operator’s country of domicile and the country of build. Panama does notappear to be particularly attractive to operators from countries with a high numberof ratified IMO conventions, indicating that Panama is not chosen as a safe havento avoid having to comply with IMO’s rules and regulations.

6.2. Developed vs. Developing Countries – Who Benefitsfrom the Open Registries?

Operators from developed countries are most likely to choose a foreign flag. Thisallows them to remain competitive in a business environment where developingcountries might otherwise have a competitive advantage due to lower wages and,perhaps, less stringent safety and environmental standards. The question then arisesif the whole system of open registries is in general unfavourable for developingcountries. Should international organizations such as UNCTAD continue to try toreduce the use of open registries, in the belief that the system is detrimental todeveloping countries’ interests? Against this background, there are three reasonswhy developing countries, including the eight countries which were particularlylooked at in this paper, might actually benefit from the present system:(1) First, most open registries are based in developing countries (Table A1). For

some of them, this is a relevant business, i.e. an export of a service.(2) As indicated also by our research, one of the main reasons why operators

choose a foreign flag is related to crewing costs – and without open registriesfar fewer seafarers from the Philippines, Indonesia, Honduras etc. would findemployment aboard. These countries would certainly be very much against adismantling of the open registry system.

(3) If the use of open registries really reduces costs, as we assume, this eventuallyleads to lower freight rates. Lower freight rates are particularly to the benefitof developing countries, who pay more on average for the transport of theirimports and exports, and who depend more on maritime transport than do thedeveloped countries.

6.3. Future Research

Our research suggests that there exists a close relationship between the socio-economic development of a country and the likelihood to attract nationally flagged

Determinants of Vessel Flag 207

vessels. The challenges and opportunities that arise for developing countries couldbe a topic for future policy-oriented research.

Concerning maritime safety, more detailed research should attempt to shed lighton the relationships between the determinants of flag choice and flag registries’safety records. As has been shown, younger and IACS classed vessels are morelikely to choose a foreign flag, which should be a positive indicator in relation tomaritime safety. At the same time, however, a bad past safety record and a lownumber of ratified IMO conventions encourages vessel operators to choose thenational register. Put differently, strict national controls appear to scare operatorsaway into foreign flags. In this context, future research might look at the impactof specific IMO and other conventions to identify which conventions exactlyare empirically most likely to attract or scare away vessels into foreign flagregistration.

Flag States could use this type of research to target specific markets toincrease their fleet – or to defend their national flagged fleet. They might,for example, modify their pricing structure based on different variables thatare estimated to have an impact on the likelihood of a vessel to choose aforeign flag. As a second step, Flag States could also take into account whichvariables are most likely to be linked to substandard vessels, and consequentlyapply statistically supported criteria when accepting or rejecting vessels to theirregistries.

NOTE

1. The authors gratefully acknowledge the supply of that information from LloydsRegister Fairplay. For further information visit http://www.ships-register.com/

REFERENCES

Alderton, T., & Winchester, N. (2002). Flag states and safety: 1997–1999. Maritime Policy andManagement, 29(2), 151–162.

Bergantino, A., & Marlow, P. B. (1999). An econometric analysis of the decision to flag out. Cardiff:SIRC.

Li, K. X., & Wonham, J. (1999). Who is safe and who is at risk: A study of 20-year-record on accidenttotal loss in different flags. Maritime Policy and Management, 26(2), 137–144.

OECD (2003a). Ownership and control of ships. Paris, March.OECD (2003b). Cost savings stemming from non-compliance with international environmental

regulations in the maritime sector. Paris, February.Roberts, S., & Marlow, P. B. (2002). Casualties in dry bulk shipping (1963–1996). Marine Policy, 26,

437–450.

208 JAN HOFFMANN ET AL.

Talley, W. (1999). Determinants of ship accident seaworthiness. International Journal of MaritimeEconomics, 1(2), 1–14.

Talley, W. (2002). Non-seaworthy risks of bulk ship accidents. International Journal of TransportEconomics, XXIX(1), 3–15.

Thanopoulou, H. A. (1995). The growth of fleets registered in the newly-emerging maritime countriesand maritime crises. Maritime Policy and Management, 22(1), 51–62.

Thanopoulou, H. A. (1998). What price the flag? The terms of competitiveness in shipping. MarinePolicy, 22, 359–374.

UNCTAD (1986). United Nations convention on the registration of ships. Geneva.UNCTAD (1994). The place of general average in marine insurance today. UNCTAD/SDD/Leg/1.

Geneva, March.Winchester, N. (2003). Flag state audit 2003. Cardiff: SIRC.

APPENDIX A: OPEN AND NATIONAL REGISTRIES,JANUARY 2003

Registry (Ranked by GT) GT Units TEU DWT

Panama** 128847021 5626 1480409 193080967Liberia* 52633013 1645 930017 79423906Bahamas** 34898752 1338 349066 48590858Greece 27719307 1051 169562 46131822Malta* 26808191 1406 212841 43601920Cyprus* 23296834 1351 373280 36876291Singapore* 21512511 1172 350521 33631765Norwegian International

Register*19230527 744 127979 29241576

United States 17223973 1630 297566 19443571China 15558825 1645 187373 23230280Hong Kong 14981444 569 210680 25187110Marshall Islands* 13273846 370 145859 21918659Japan 13202619 2575 49845 16235867Italy 9894662 807 127285 10674426Russia 8077939 2140 69910 7978577United Kingdom 7744380 929 200149 7371573Germany 7275160 584 575675 8664070St. Vincent and the

Grenadines**6634756 1071 78432 9572546

Determinants of Vessel Flag 209

Registry (Ranked by GT) GT Units TEU DWT

Korea (South) 6390606 871 66747 9854163India 6285300 487 21230 10216716Isle of Man* 6155831 300 60621 9772486Netherlands 5760478 913 237710 6443630Turkey 5487251 747 64180 8502229Malaysia 5483118 434 74843 7796908Danish International

Register*5307031 302 273298 6851899

Bermuda** 4912086 113 36074 7042220Philippines 4907163 507 24384 7095152Antigua and Barbuda** 4831335 840 382680 6316247Taiwan 4531387 195 111880 7090064Norway 4129964 901 2835 4842823Iran 4104917 208 34633 7022548Brazil 3461569 242 28126 5529012Kerguelen Islands* 3341217 90 55577 5366928Sweden 3122840 291 29889 1917543Indonesia 2962851 718 33050 4019601Canada 2749531 486 4828 3239376Denmark 2334417 247 61368 2695496Cayman Islands* 2323992 150 7292 3660906Kuwait 2261556 54 15024 3623585Australia 2020625 198 2970 2586501Cambodia* 1922554 486 16957 2727326Thailand 1680996 354 31223 2616066Saudi Arabia 1610243 139 24607 2488681Canary Islands 1600745 183 18964 1807459Finland 1589632 166 18617 1313656Luxembourg 1539974 74 12702 2216857Belize** 1507086 566 10382 2028624France 1502177 185 3100 1452865Egypt 1430899 209 16882 2063956Netherlands Antilles* 1398140 165 47747 1822668Vanuatu* 1355979 238 5480 1516700Vietnam 1196432 184 10887 1843977Ukraine 1012633 302 9554 987816Algeria 944205 80 3580 1053718

210 JAN HOFFMANN ET AL.

Registry (Ranked by GT) GT Units TEU DWT

Madeira 907809 134 9415 1343034Bulgaria 872407 81 5639 1290625Gibraltar* 857971 98 25449 1234394Chile 794177 111 7216 1062048Croatia 784816 91 10513 1175904United Arab Emirates 756495 179 19974 864185Mexico 753470 117 283 1042418Venezuela 709570 85 735 1127547Qatar 656481 56 14653 986921Israel 606510 24 44595 704614Poland 591542 85 541 806939Korea (North) 576428 97 3820 854755Switzerland 570146 27 5798 1002749Honduras** 565427 329 2814 768659Romania 538075 120 2425 698109Nigeria 516487 66 752 881776Spain 478945 233 2627 493744Brunei Darussalam 477229 17 – 442201Azerbaijan 452488 110 376 383646Syria 426312 151 5472 638967Lithuania 399363 76 8064 392774Myanmar* 399105 47 3988 599836Barbados* 394547 54 4402 535320Unknown 392071 89 6743 508345Morocco 371240 81 7844 314367Estonia 343025 75 5653 218538Argentina 324532 84 1326 437498Tonga 319749 130 3908 433887Bangladesh 283136 50 10528 391772Bahrain 278120 32 8472 385058Lebanon* 271394 70 4974 391743Libya 262676 25 1477 359301Yemen 260001 10 – 529475Georgia 255112 82 2298 361024Ireland 246740 75 3050 176760Ecuador 240833 45 – 404224Pakistan 225601 20 7402 339249

Determinants of Vessel Flag 211

Registry (Ranked by GT) GT Units TEU DWT

Iraq 224116 41 772 323083Belgium 200719 92 400 256345Portugal 198703 68 5212 294510Jordan 192980 9 1853 370125Wallis & Futuna Islands 182423 7 – 92719Tunisia 179431 21 696 160767New Zealand 172071 62 587 161472Faroes* 158937 60 94 174802South Africa 141281 51 2423 117560Bolivia** 138699 55 862 187998Iceland 125755 122 724 64412Sao Tome et Principe* 124264 35 660 175104Comoros 85660 16 408 141551Latvia 85285 25 681 60596Peru 84995 54 316 117220Cuba 84760 27 392 110205Ethiopia 81782 9 2032 101375Puerto Rico 77074 4 2868 53024Cameroon Republic 75227 5 – 139593Sri Lanka 72654 27 1729 97215Mauritius* 62780 17 244 75189Papua New Guinea 55649 38 1579 69091Jamaica 54114 7 188 70926Uruguay 52400 19 438 32666Greenland 51699 35 – 18470Seychelles 50532 10 364 63000Tuvalu 45979 7 705 58709Maldives 45525 18 1234 62368Angola 40856 14 226 62330Colombia 35796 24 – 50341Paraguay 35364 25 1107 44132Equatorial Guinea* 33263 41 67 28785Austria 33182 8 2098 42223Reunion 31562 5 – 45885Sudan 29854 4 922 39084Tanzania 29805 10 247 32573Ghana 27351 21 – 33943

212 JAN HOFFMANN ET AL.

Registry (Ranked by GT) GT Units TEU DWT

Falkland Islands 27319 14 12 18107Turkmenistan 26970 11 150 24669French Polynesia 26840 21 298 15304Albania 23599 20 79 33744Eritrea 20010 8 527 25859Fiji* 20004 16 55 19101Belarus 18871 1 – 14947Namibia 17906 15 – 12648Madagascar 15401 11 24 19095Oman 14932 7 74 10055Trinidad & Tobago 14594 11 32 9859Kenya 10397 12 – 10844Samoa 8951 3 650 8492Mozambique 8368 7 77 11295Congo (Democratic Republic) 7469 4 – 10579Slovakia 5994 2 – 8900Cape Verde Islands 5895 8 123 8267Yugoslavia 5714 4 – 2132Micronesia 5362 6 – 4325Mauritania 5012 3 – 3363Senegal 4976 7 – 1832Gabon 4234 9 – 3041Hungary 3784 1 274 5500New Caledonia 3579 4 – 3840Guadeloupe 3370 6 – 214Virgin Islands (British) 3012 2 – 5095Guyana 2946 2 68 4507Kiribati 2840 3 33 2609Costa Rica 2509 2 – 1108Laos 2370 1 – 3110Solomon Islands 2303 3 – 1004Guatemala 2165 1 – 1727El Salvador 2109 1 – 1650Macau 2007 1 – 2184Somalia 1916 3 260 4539Kazakhstan 1897 2 820 1446Dominican Republic 1880 2 – 2581

Determinants of Vessel Flag 213

Registry (Ranked by GT) GT Units TEU DWT

American Samoa 1092 1 – 1200Suriname 981 1 – 1599Niger 942 1 – 1002Togo 836 1 – 1187St Helena Island 789 1 – 478Cook Islands 764 2 – –Haiti 613 1 – 793Ivory Coast 489 1 – 1020Dominica 445 1 – 630Congo 413 1 – 329Slovenia 369 1 – 36Gambia 350 1 – 600Sierra Leone 349 1 – 411St Pierre and Miquelon 324 1 – 280Guinea 321 1 – –Mayotte 300 1 – 30Total national registries 220662866 25083 3087159 299985727Percent of world 37.8% 57.2% 38.2% 35.4%Total open registries* 363217093 18795 4988121 547235423Percent of world* 62.2% 42.8% 61.8% 64.6%

Grand Total 583879959 43878 8075280 847221150

Source:Authors, based on data provided by LRFairplay. All vessels of 300 GT and above deliveredprior to January 2003.

∗Open registries.∗∗Latin American and Caribbean registries included in this paper’s research.

The Grand Total is slightly above the total included in the regressions as somevessels had to be excluded due to missing information such as unknown flag. Thedistinction between open and national registries is of course not always clear cut.Singapore and the Canary Islands, for example, may be border line cases, and wehad to make a subjective choice when assigning them to one or the other group ofregistries. Vessels with an “unknown” flag were included in the “national registries”total. Comments and suggestions are welcome to [email protected].

214 JAN HOFFMANN ET AL.

APPENDIX B: VESSEL SUBTYPES

%

VGCARGOGeneral Cargo Ship 64.97Multi-Purpose Ship 31.83Livestock Carrier 1.25Heavy Lift Ship 1.17Semi-Sub Heavy Lift Vessel 0.41General Cargo/Part Refrigerated Ship 0.33Log Tipping Vessel 0.03Deck Cargo Ship 0.01

VCONTContainer Ship 99.83Container Ship/all Reefer 0.17

VLBLKTanker 24.54Products Tanker 20.77Crude Oil Tanker 16.78Chemical Tanker 13.75LPG Carrier 10.24Chemical/Oil Tanker 6.89LNG Carrier 1.59Bunker Tanker 1.37Asphalt Tanker 1.15Replenishment Tanker 0.91Parcels Tanker 0.64Bitumen Tanker 0.37Wine Tanker 0.31Ethylene Tanker 0.23Storage Tanker 0.15Water Tanker 0.13Fruit Juice Tanker 0.10Sulphur Tanker 0.07

Determinants of Vessel Flag 215

%

VDBLKBulker 58.11Bulk Carrier Ore StrenGThened 28.86Bulk/Container Carrier 5.57Bulk Cement Carrier 3.90Bulk Wood Chip Carrier 1.38Bulker - Great Lakes Only 1.37Ore Carrier 0.43Bulk Vehicle Carrier 0.25Bulker Great Lakes only/Dumb 0.13

VPASSPassenger/Vehicle Ferry 43.55Passenger/Cargo Ship 13.38Passenger Vessel 12.34Multi-Hull Passenger Ferry 11.53Cruise Ship 10.13Passenger/Train/Vehicle Ferry 3.18Multi-Hull Passenger/Vehicle Ferry 2.86River Cruise Ship 0.91Passenger Hydrofoil 0.81Passenger Excursion Vessel 0.42Passenger Hovercraft 0.42Casino Ship 0.23Surface Effect Passenger Ferry 0.13Passenger/Vehicle Hovercraft 0.06Surface Effect Passenger/Vehicle Ferry 0.03

VFSTUGFishing Vessel 37.51Tug 15.92Barge 5.13Hopper Suction Dredger 4.17Oil Barge 4.10Hopper Barge 2.20Salvage Tug 2.16Research Vessel 1.98Suction Dredger 1.93

216 JAN HOFFMANN ET AL.

%

Buoy Tender 1.41Crane/Derrick Barge 1.34Cable Ship 1.28Training Vessel 1.22Oceanographic Vessel 1.18Dredger 1.09Fisheries Research 1.04Icebreaker 0.98Logistics Vessel 0.93Anti-Pollution Vessel 0.88Cutter Suction Dredger 0.84Survey Ship 0.79Bucket Dredger 0.75Fisheries Protection Vessel 0.73Pusher Tug 0.71Survey/Research Vessel 0.59Grab Dredger 0.50Live Fish Carrier 0.49Salvage Vessel 0.49Geophysical Research Vessel 0.46Pontoon 0.46Sail Training Vessel 0.46Sludge Carrier 0.44Museum Ship 0.35Container Barge 0.34Search And Rescue Vessel 0.32Maintenance/Utility Vessel 0.28Pleasure Craft 0.26Sand Loading Dredger 0.26Cable Repair Ship 0.24Icebreaker/Buoy Tender 0.24Oil Storage Barge 0.24Pilot Vessel 0.24Cement Storage Barge 0.21Accommodation Vessel 0.19Diamond Mining Vessel 0.19Weather Ship 0.18

Determinants of Vessel Flag 217

%

Fisheries Training 0.16Restaurant Ship 0.16Ro-Ro Barge 0.16Floating Hotel 0.15Hospital Ship 0.15Spent Nuclear Fuel Carrier 0.15Waste Disposal Vessel(Liq) 0.15Icebreaker/Research Vessel 0.12Floating Dock 0.10Patrol Vessel 0.10Polar Research Vessel 0.10Whaling Vessel 0.10Cutter Dredger 0.07Fire Fighting Vessel 0.07Exhibition Vessel 0.06Floating Power Station 0.06Support Ship 0.06Drilling Barge 0.04Incinerator & Waste Disposal Vessel 0.04Naval Vessel 0.04Tank Cleaning Vessel 0.04Dipper Dredger 0.03Floating Car Park 0.03Radio Station 0.03Transshipment Vessel 0.03Asphalt Barge 0.01Dragger Dredger 0.01Floating Wave Powered Power Station 0.01Pile Driving Vessel 0.01Radioactive Waste Carrier 0.01Rocket Launch Vessel 0.01Sealing Vessel 0.01

VRORORo-Ro 48.67Vehicle Carrier 28.86Pallet Vessel 6.40

218 JAN HOFFMANN ET AL.

%

Ro-Ro/General Cargo 6.15Ro-Lo 5.90Ro-Ro/Heavylift 1.55Barge Carrier 1.20Ro-Ro/Cellular 1.20Multi Hull Ro-Ro Freight 0.05

VREEFReefer 56.47Refrigerated Fish Carrier 43.53

VOROILOre/Bulk/Oil Carrier 51.96Ore/Oil Carrier 35.20Product/Ore/Bulk/Oil 12.85

VOFSHOREAnchor Handling/Tug/Supply 26.76Supply Vessel 25.30Self-Elevating Mobile Offshore Drilling 8.09Semi Submers Mobile Offshore Drilling 5.99Safety Standby Vessel 5.23Anchor Handling/Tug 4.32Seismic Survey Vessel 3.28FPSO 2.68Diving Support Vessel 2.25Crewboat 2.01Offshore Maintenance/Utility Vessel 1.79Offshore Cargo Barge 1.37Drillship 1.34Offshore Construction Vessel 1.16Offshore Support Vessel 1.09Floating Storage Offtake 1.00Survey Ship Rov Support 0.94Offshore Drilling Barge 0.85Pipe Laybarge 0.67Pipe Carrier/Platform Supply 0.61Oil Well Stimulation Vessel 0.58Offshore Accommodation Vessel 0.49

Determinants of Vessel Flag 219

%

Aht/Salvage 0.46Floating Production Unit 0.46Rock Laying Ship 0.43Multi Function Service Vessel 0.24Pipelay Vessel 0.21Self-Elevating Production Unit 0.18Oil Well Production Test Vessel 0.09Trenching Vessel 0.09Semi-Sub Pipe Laybarge 0.06

Source:Authors, based on data provided by LRFairplay.