Factors Influencing Ship Registration

Seediscussions,stats,andauthorprofilesforthispublicationat:http://www.researchgate.net/publication/230877126

Factorsinfluencingthechoiceofflag:Empiricalevidence

ARTICLEinMARITIMEPOLICY&MANAGEMENT·JANUARY1998

ImpactFactor:1.45·DOI:10.1080/03088839800000026

CITATIONS

22

DOWNLOADS

237

VIEWS

128

2AUTHORS,INCLUDING:

AngelaStefaniaBergantino

UniversitàdegliStudidiBariAldoMoro

28PUBLICATIONS116CITATIONS

SEEPROFILE

Availablefrom:AngelaStefaniaBergantino

Retrievedon:14August2015

http://www.researchgate.net/publication/230877126_Factors_influencing_the_choice_of_flag_Empirical_evidence?enrichId=rgreq-6549d1ec-922f-4bf4-b662-10df1b46cc9b&enrichSource=Y292ZXJQYWdlOzIzMDg3NzEyNjtBUzo5OTc2ODA4NDcyOTg1OUAxNDAwNzk3OTczNDY4&el=1_x_2

http://www.researchgate.net/publication/230877126_Factors_influencing_the_choice_of_flag_Empirical_evidence?enrichId=rgreq-6549d1ec-922f-4bf4-b662-10df1b46cc9b&enrichSource=Y292ZXJQYWdlOzIzMDg3NzEyNjtBUzo5OTc2ODA4NDcyOTg1OUAxNDAwNzk3OTczNDY4&el=1_x_3

http://www.researchgate.net/?enrichId=rgreq-6549d1ec-922f-4bf4-b662-10df1b46cc9b&enrichSource=Y292ZXJQYWdlOzIzMDg3NzEyNjtBUzo5OTc2ODA4NDcyOTg1OUAxNDAwNzk3OTczNDY4&el=1_x_1

http://www.researchgate.net/profile/Angela_Bergantino?enrichId=rgreq-6549d1ec-922f-4bf4-b662-10df1b46cc9b&enrichSource=Y292ZXJQYWdlOzIzMDg3NzEyNjtBUzo5OTc2ODA4NDcyOTg1OUAxNDAwNzk3OTczNDY4&el=1_x_4


http://www.researchgate.net/institution/Universita_degli_Studi_di_Bari_Aldo_Moro?enrichId=rgreq-6549d1ec-922f-4bf4-b662-10df1b46cc9b&enrichSource=Y292ZXJQYWdlOzIzMDg3NzEyNjtBUzo5OTc2ODA4NDcyOTg1OUAxNDAwNzk3OTczNDY4&el=1_x_6


PLEASE SCROLL DOWN FOR ARTICLE

This article was downloaded by: [informa internal users]On: 12 May 2010Access details: Access Details: [subscription number 755239602]Publisher RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Maritime Policy & ManagementPublication details, including instructions for authors and subscription information:http://www.informaworld.com/smpp/title~content=t713694970

Factors influencing the choice of flag: empirical evidenceAngela Bergantino a;Peter Marlow a

a Department of Maritime Studies and International Transport, University of Wales, Cardiff, UK

To cite this Article Bergantino, Angela andMarlow, Peter(1998) 'Factors influencing the choice of flag: empirical evidence',Maritime Policy & Management, 25: 2, 157 — 174To link to this Article: DOI: 10.1080/03088839800000026URL: http://dx.doi.org/10.1080/03088839800000026

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial orsystematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply ordistribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contentswill be complete or accurate or up to date. The accuracy of any instructions, formulae and drug dosesshould be independently verified with primary sources. The publisher shall not be liable for any loss,actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directlyor indirectly in connection with or arising out of the use of this material.

http://www.informaworld.com/smpp/title~content=t713694970

http://dx.doi.org/10.1080/03088839800000026

http://www.informaworld.com/terms-and-conditions-of-access.pdf

Factors influencing the choice of flag: empirical evidence

ANGELA BERGANTINO and PETER MARLOW

Department of Maritime Studies and International Transport, University of Wales, Cardiff CF1 3YP, UK

The aim of this paper is to analyse the decision-making process of shipowners when adopting flags of registration. More specifically, it is interested in examining the relative importance of the factors related to an individual company's decision to flag out. The decision to change flag is viewed as similar to any other strategic decision by a profit maximizing firm (shipping company) and therefore those variables which influence the attractiveness, for a given firm, of taking the flagging out decision are analysed. The approach is two-fold in that it employs both qualitative and quantitative analysis. The research is innovative in the sense that it uses an econometric approach and the analysis is based on original data which has been obtained by the authors via a questionnaire and personal interviews with members of the UK shipping industry. The results deal with two particular sectors namely the tanker and general cargo markets and provide an insight into the magnitude and significance of various factors which affect the choice of flag. It is able to provide indications of the likelihood of a particular vessel being flagged out under different circumstances and, further, to consider how changes in these circumstances might affect the probability of the event occurring. The paper is structured in six parts comprising an introduction, background, methodology, qualitative analysis, econometric analysis, and conclusions.

1. Introduction Since its first significant appearance in the nineteen fifties, flagging out-the change of a vessel's registry from a national flag to a n open register-has been the focus of a strong debate in international shipping quarters [I]. As is shown in figure 1, the Open Registry fleet, had their sharpest increase in the 1970s rising from representing 21.6O/0 of the world fleet to represent 3 1.1 O h in 1980. Their increase continued through the 1980s and, in 1988, the fleet of the major Open Registry countries surpassed that of the traditional maritime countries. The difference between the two does not appear to be diminishing. This now widespread phenomenon, afflicting the merchant fleet of the vast majority of the traditional maritime nations, has attracted a great deal of attention for a variety of reasons. First, open registry fleets have expanded a t a faster rate than any other fleet in the world. Second, it has been felt that the expansion of the open registry has limited the growth of the fleets of other countries and has caused the decline of the fleets of the traditional maritime countries with all the related consequences for the defence of the countries, the balance of payments and the disappearance of national trained crews. Finally, the occurrence in recent years of several alarming incidents involving environmental disasters has (rightly or wrongly) increased public awareness of this problem.

From the late 1960s a number of studies have been conducted on the flagging out phenomenon. However, with a few exceptions, most existing works, whilst providing useful insights and interesting information on the issue, are out of date. An in depth

0308 X819,98 $12 00 ( 1998 Taylor & Franc~s Ltd

Downloaded By: [informa internal users] At: 15:23 12 May 2010

A. Bergantino and P. Marlow

- A = World total

8 -MajaFoC countries

Source: Review of Maritime Transport, Unctad (various years)

Figure 1. Composition of the world fleet by type of registry.

review of the literature has pointed to a general lack of theoretical background and empirical analysis in a great many of the previous studies [2] due mostly to the difficulties of obtaining the necessary data. Hence arises the necessity to give a quantitative contribution to the flagging out issue in the most objective way possible.

In the light of this, this paper seeks to explain and evaluate flagging out, inserting the problem into a more general theoretical framework and using recent primary source data. The aim of this paper is not to take part in the debate of the advantages and disadvantages of open registries [3], but to study the determinants of the mobility between the diflerent types of registry, in order to establish the factors which may have influenced the relative decline of the merchant fleets of traditional maritime regimes. Subsequently, the consequences of flagging out may be assessed and policy responses to the problems created by their growth may be reviewed. The empirical part of this analysis will focus on the reasons for the choice of flag for UK owned vessels greater than 500grt in size. The authors are grateful for the assistance in terms of time and information given to them by UK companies, the Chamber of Shipping, and shipowners, since without the involvement of the industry such studies have little or no validity. International comparisons would add considerable value to this work and it is hoped that these will be possible in future.

2. The background The aim of this paper is to model the determinants of the choice o f j a g decision by placing the decision within the basic economic framework of the theory of the firm and, hence, to provide economic explanations for the practice of flagging out adopted by shipping companies facing a competitive international shipping environment. The motivation for transferring a ship from one registry to another is no different in principle from the motivation behind any other strategic decision on the part of a profit maximizing company. The basic principles of the theory of the firm can be applied to the economics of this behaviour.


Factors in$uencing the choice o f j u g 159

For the purposes of this study it is assumed that shipping companies are profit maximisers [4] which strive to reach their objective by seeking the production input combination which allows them to minimize costs. However, their choice of factors of production is constrained by their operating environment. Institutional factors and the characteristics of the market in which they operate condition their ability to make independent decisions. The selection of factors, their quantities, their costs and quality appear to be regulated in most of the so called developed countries.

However, the existence of open registers creates a sort of dualism in the international maritime transport sector splitting the industry into two segments distin- guished by operating characteristics peculiar to the two different scenarios and by lower break-even points. The shipowner, like any other entrepreneur, must choose the optimum amount of inputs to obtain the desired service output and strives to have the freedom to do so.

Flagging out is primarily caused by the desire to minimize costs by placing the vessel under a relatively lower cost regime and it is estimated that crew cost differences between selected EU flags and lower-cost open registry vessels range from +22% to +333% [5] . However, as we shall see, a number of other factors in addition to crew costs can be considered as influencing the shipowners' choice: labour quality, management costs, tax liabilities, and degree of control are just a few of the items which characterize the different regimes and which influence the decisions.

From a detailed literature review, combined with the analysis of the outcome of a number of interviews carried out with shipowners, a list of variables which are considered influential in the decision making process of the shipowner has been produced. With the evolving of the practice of flagging out the reasons, for register- ing a ship under the flag of a country other than that of the owner, have varied. Their ranking has often changed and today, even though the practice still retains most of the same motivations, the primary reason for flagging out is generally accepted as being the need to reduce overall costs.

It seems peculiar that while, until a few years ago, the need to enjoy fiscal advantages was considered by the literature to be a substantial factor in the choice of flag decision, this is now no longer the case. One interpretation is that the importance of this motivation has been diminished since the Traditional Maritime countries have had a chance to respond to the competition set up by the FoC/Open Register countries. The governments of most of the traditional maritime countries have mod- ified their policies to move them closer to the situation created by the legislation of the Open Registry countries. The FoC legislation is used as a benchmark against which to measure the effects of Traditional maritime nations' policies.

A group of factors that might influence the shipowner's decision, but which have been partly ignored by the existing literature, are the characteristics of the shipping companies and of the ships. It is observed that only some companies of the same nationality decide to flag out, and that the decision to flag out might concern either the entire or only part of the fleet of the same shipping company. In particular, with regard to the characteristics of the ship, factors such as: age, size, type of trade, type of vessel, or geographical area of operation, might influence the flagging out decision. From the preceding analysis it has been seen that while, until a few years ago, flagging out seemed to be relegated to sectors with low freight rates (i.e. bulk carriers) and low quality standards [6], this seems to be no longer the case. Containers are beginning to form a significant and growing proportion of the flagged out fleets, augmenting the need to understand the motivations at the heart of the


160 A. Bergantino and P. Marlow

decision making process. Furthermore, while recognizing that a ship is a very mobile asset, it is possible that its area of trade, when identifiable, may have a bearing on the flag chosen [7]. The following mixture of qualitative and quantitative factors helped to inform the formulation of the questionnaire used to obtain data from the industry: crew costs; other costs; labour productivity, manning rules; profitability; fiscal factors; control; economic/political considerations; the attitude of financial institutions.

3. Methodology The methodology used in this paper has been dictated by the fact that we are dealing with a binary choice: either the vessel does or does n o t j y its traditional nationalJag. The choice is clearly binary since the primary decision is whether to uselkeep the national flag or to leave it and choose another flag. The secondary decision is then which other flag to choose. This paper analyses the first stage choice between national and non-national flag and requires a yes/no or binary solution. Using a binary choice model the probability of the outcome of the shipowner's flag decision can be determined depending on a set of variables. The model postulates that the probability that a ship will be flagged out is a function of observable factors both internal and external to the firm and a random element resulting from unobservable or unmeasurable characteristics. In recent years logistic (logit) regression has been suggested as an appropriate analytical technique to use for the multivariate modelling of categorical dependent variables and we intend to use the binary logit model.

Logit analysis will allow the sign and marginal effect on the conditional probability of the event occurring due to a change in any of the explanatory variables to be determined and further it will allow comparisons to be made of the relative importance of the explanatory variables in determining the likelihood of flagging out. Furthermore by evaluating the logit probability function for a ship using its measured attributes, and comparing the output to similar calculations for other ships, a ranking can be found as to the relative probability of being flagged out under different circumstances.

3.1. Model As for any discrete choice model the specification of the model itself relies on the latent variable approach, where it is assumed that there is some underlying (and unobserved) response variable y*, where y* E (-oo, foo). Whilst we do not observe y* directly, we do observe a binary outcome y such that:

Defining the latent variable equation in linear form:

y* = x ' p + u

where u represents an (unobservable) stochastic component. This now gives:

E ( ~ l x ) = P r ( y = llx) = P ~ ( Y * > OIx) = Pr(u > -xp) = 1 - F ( - x p )

where F( . ) represents the cumulative distribution function of u. By specifying the appropriate distribution function for u, (i.e. logistic distribution), the Logit model can be derived. In this specific case the components are an individual i, a shipowner,


Factors influencing the choice oj,flug 161

who is faced with a choice problem and who is hypothesized to behave so as to maximize his utility in choosing between two alternatives. In this case:

Alternative 1: having the vessel fly the national flag. Alternative 2: having the vessel fly a foreign flag.

In this context the interest lies in the set of factors affecting the occurrence or non- occurrence of the event. Information about the attributes of each alternative, characteristics specific to the individual economic agent exercising the choice, and the outcome of their choices is available.

Letting y represent the choice of flag gives:

J.' = 1 if the vessel is flagged out, 1% = 0 otherwise.

The two outcomes, vessel being flagged out o r not flagged out, can be described by the following state-specific utilities (i.e. the utility that the shipowner perceives from flagging out or staying with the national flag in terms of, e.g. cost savings):

where .u is a common set of control variables, and / j o and are vectors of unknown parameters and u,, and u , represent unobservable (state-specific) tatelpreference components. Under this characterization, an individual will choose to flag out if UT > U;, such that:

The underlying function has been assumed to be a utility function measuring the utility of the shipowner deriving from the cost savings caused by his decision. Assuming that the shipowner is a rational economic agent he would be expected to maximize revenues, R , while minimizing his costs, C. Therefore, it can be assumed that his utility, U , can be described as:

In this case the underlying function will be:

U* = U I - Uo

Where: U1 = R1 - C 1 ,

and

Uo = Ro - Co

Therefore:

U* = ( R 1 - C I ) - (RO - Co)

Since in the shipping market the rates are determined a t international level and this study is analysing only a small proportion of the world fleet, it can be assumed that the freight rates are given. Therefore, Rl and Ro can be considered equal, i.e. R I = Ro, from which:



The only relevant information is that relative to the cost function. If Co and C1 represent the costs for the national flagged ships and the costs for the foreign flagged ships respectively:

Co = f (Crew related costs (per category), Other running costs (per category), Capital costs (partial)).

C1 = f (Crew related costs (per category), Other running costs (per category), Capital costs (partial)).

Data on the above listed factors have been derived from the information collected via the questionnaire. The data concern both types of fleet, flagged out and non flagged out and therefore creates the following variables:

Crew cost related information: C, =Manning related costs NAT (per category) - Manning related costs FOC (per category). Other running costs information: 0, =Other running costs NAT (per category) - Other running costs FOC (per category).

Therefore a first statement of the relevant function for the ith shipowner is:

This function can be interpreted as follows: the utility of taking the decision to flag out is a function of the cost saving obtained by taking the decision.

However, there are additional factors that, independent of cost saving considerations, can be assumed to influence the shipowner's decision to flag out and should be included in the model, e.g. characteristics of the company, characteristics of the trade, characteristics of the vessel, etc. These factors might indirectly influence the ability or possibility of the shipowner to change flag. Therefore, the utility of the ith shipowner is also a function of the above variables.

Uf = aCmi + POci + -y Other characteristicsi

Furthermore, since unobservable or unmeasurable characteristics might be part of the decision process connected to flagging out, the complete function to be estimated should be:

Ul* = k + aCmi + POci + 7 Other characteristicsi + uj

where ui = (u,, - uio) and k is a constant. This model postulates that the probability that a ship will be flagged out is a

function of observable factors both internal or external to the firm and a random element resulting from either qualitative (unmeasurable) or non observable factors. In order econometrically to infer the parameters from the available data set the Maximum Likelihood Estimation Procedure was used under which the Maximum Likelihood estimates are obtained by iterative techniques [8].

The estimated coefficients do not provide explicit insights as to the magnitude of the effect of the specific variable on the probability of the event occurring but on the sign, or direction, of the effect [9]. In order to calculate the marginal effect on the conditional probability of the event occurring of a unit change in the jth explanatory variable xi further calculations need to be performed which involve taking the


Fuctors influencing the choice o f f lug 163

derivative of the Logistic function with respect to the variable for which the marginal impact is to be calculated.

For logit models as for standard OLS models statistical inference can be carried out using standard inferential techniques. In particular for testing that a coefficient is significantly different from zero the Wald Statistic, which has an X 2 distribution, can be used [lo]. In order to assess the accuracy with which a binary choice model approximates the observed data, a number of measures are available which follow the principle of the more familiar R~ in standard Least Square regression [I I]. The measure adopted here is based on a ratio of Likelihood from the full model and from a restricted model estimated on the intercept alone-this is the McFadden pseudo R*. Finally to test the joint significance of the slope parameters in the model (known as the F-test in OLS analysis) the X 2 test statistic was used.

3.2. Data collection The previous paragraphs detailed the data requirements necessary to satisfy the Logit formulation. Having chosen the technique and decided on the relevant variables the appropriate data were collected via a questionnaire. The authors conducted numerous interviews with different UK shipping companies to obtain qualitative background information, to discuss the nature of the questions and to clarify any problems of interpretation. Pilot interviews took place to test the validity and com- prehensibility of the questionnaire and then the final version was either left with the companies or mailedlfaxed to them later. Follow-up telephone conversations clar- ified any subsequent difficulties which respondents had and the authors are grateful to all of them for their cooperation.

Data were required on all the factors discussed in Section 2 of this paper and the questionnaire was designed to effect this. The questionnaire was divided into two parts. Part I relating to the shipping company or shipping division while Part I1 related to the individual vessels of the fleet. The response rate within the UK is considered good and stems from the time devoted to personal interviews. The data received currently refers to 186 ships or about 38% of the population. The sample is distributed as shown in table 1.

As can be seen the distribution per type of ship and per group of flag-national (NAT) or foreign (FOR)-is, in some cases, not representative of the true population. In particular, the samples for Bulk cargo, Specialized cargo and Container ships are not representative of the national flag fleet and this will create difficulties

Table 1. Sample covered.

Ship type NAT flag FOR flag

Total Sample Percentage Total Sample Percentage

Bulk cargo 12 0 Container 23 0 Tanker 68 24 General cargo 59 34 Specialized cargo 5 0 Totals 167 58

Total UK owned Response rate 492

0.0 0.0

35.3 57.6 0.0

34.7

Total sampled 186

43 20 46.5 54 29 53.7 87 3 8 43.7

111 3 7 33.3 30 4 13.3

325 128 39.4

Percentage 37.8



for the estimation of the model for these specific categories. Data in the other categories, however, are representative and more appropriate for estimation and the overall response rate is well above the minimum required.

The study is concerned with the reasons for choosing a flaglflagging out and the responses from the industry indicate that both qualitative and quantitative (commercial) factors play their part in decision making. The quantitative data will be analysed in Section 5 using the Logit model described in Section 3.1 but, based on the qualitative responses, some preliminary observations may be made.

4. Qualitative analysis Flag selection is a high level decision usually made, on a vessel by vessel basis, a t the time of vessel acquisition and is generally based on experience. During the interviews it became clear that different companies perceived different factors as being important to their decision on flag. Some companies stated that, having flagged out previously, they would now prefer to flag back in but were prevented from doing so by the high cost of compliance to meet the requirements of the flag. A flag might be chosen for political reasons, to ensure a supply of skilled labour, for public relations reasons, for historical reasons, because of directives from financial institutions, or for reasons related to the trade routes of the vessel or to its characteristics. Hence it is possible for companies to flag out some of their vessels while retaining others under the national flag. Six of the respondents were in this category (i.e. some of their vessels are registered under the national flag while others are flagged out) and this reinforces the view that each vessel is considered individually and is the focus of a separate decision.

When considering the nationality of a company the definition used by Lloyds of 'direct owner' was adopted and, hence, vessels owned by UK based companies which did not fly the national flag were considered as flagged-out vessels no matter what management or company culture they had. Those companies which had chosen not to use the nat ional f ig gave crew costs as the most common reason for their decision. Other factors which had influenced them were: to escape bureaucratic control; high costs of compliance with standards of the national flag; the unavail-

m V ) n ", - R Fr.qu.n<y

Figure 2. Reasons given for choice of flag (flagged out vessels).


Factors influencing the choice of',flag

ID Other

Compliance Costs Fiscal Reasons 4% 1% Cornpllance costs Control

Attitude of F~nanclal lnst

Productlv~ty of Labour Avalllblllty of

Control C] Mannlng Rules 4 7 0 1 1 l 10 Other Cost

2%

Trade :

I Crew Cost

Trade

73';0% Ship Type 2% Avallabll~ty of Labour

r o t h e r costs 5% . Flscal Reasons

Crew Costs 26% Historical

Figure 3. Relative importance of factors affecting the use of foreign flag.

ability of skilled labour (the need to ensure a supply of same); and fiscal reasons. Figures 2 and 3 provide an indication of the influence of individual factors on this decision.

Those companies which had chosen to use the national f i g stated that their decision was affected inter alia by: the type of ship; the trade routes; public relations reasons; marketing considerations; and historical reasons. Figures 4 and 5 depict the situation for the national flag vessels.

From the profiles shown in figures 6 and 7 it is clear that no one reason acts in isolation and that the decision on choice of flag is a subtle amalgam of factors. For companies choosing a foreign flag (figure 6) and for companies choosing the national flag (figure 7).

Most of the companies (65%) which responded were in the No Tax category [I21 with the majority (60%) expecting to remain there for the next 5 years. A company

Figure 4. Reasons given for choice of flag (national flag vessels).



Other

Trade

Ship Type

C l Availabilityof Labour Fiscal Reasons

Legal Requirements

Public Relationsl4% Public Relation

!# Marketing

Legal Requireme Historical Y -10 reasons 5%

Figure 5. Relative importance of factors affecting the use of national flag.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 a ConpknceCorls

company s Oh*,

Figure 6. Companies' responses (foreign flag).

Figure 7. Companies' responses (national flag).

will be in this position if it has accumulated tax allowances from earlier years to the extent that it is now in a position where it is unlikely to be required to pay tax for the indefinite future. Under such circumstances the fiscal factors affecting the decision on flag are mitigated as the company is unable to make immediate use of any favourable changes which might occur. Furthermore, management costs appear not to be affected (or if so only slightly) by the choice of flag.


Factors influencing the choice o f f lug

5. Econometric analysis 5.1. Preliminary dutu unuly.sis The data used to estimate the logit model have been extracted from the responses collected via the questionnaire. The UK sample consists of 186 ships connected with 51 companies. In this estimation the type of flag is considered (divided into two categories, national and foreign) as the dependent variable and attempts are made to estimate the influence and the magnitude of a number of factors listed previously on the choice of the ship's flag.

The total fleet is distributed between the two categories as follows, 35% national and 65% foreign. Since the response to the questionnaire was unbalanced among the different categories of vessels and of flag as shown in table 1 in Section 4, this analysis has been limited to a subsection of the sampled population, namely the data for two categories of ships: tankers and general cargo. The revised total sample therefore consists of 71 general cargo ships and 62 tankers. The flag distribution is as follows: 39% of tankers and 52% of general cargo vessels are registered in the UK. The remaining ships are under foreign flag.

The explanatory variables have been chosen with a view to modelling the decision of the shipowners and to take account of the ship specific characteristics. These have been selected in order to describe the probable effect of a series of factors on the shipowner's decision.

In order to account for the type of ship, a dummy variable has been created using general cargo as the base group. The type of trade has also been considered with the creation of a dummy for shortldeep sea trade and, in this case the base group is the short sea shipping type of trade. Dummies concerning the geographical area of trade were created but proved to be not significant for the specific sample and were therefore excluded from the analysis.

Other characteristics of the ships were included through the following variables: GRT, age, crew number, days of operation in a year (calculated from the information on downtime) and the current valuation of the vessel. A dummy variable was also created to distinguish ships acquired within the last 5 years from those purchased previously.

The variables concerning information on costs were derived from the survey data. In order to account for the effect of size and crew number on the absolute value of the costs connected with running the ships the variables have been divided either by GRT or by crew number depending on which is more appropriate in each case. The criteria followed has been to divide variables representing crew related costs by crew number in order to obtain a 'per crew member' figure and divide the variables representing other operating costs by GRT.

In order to comply with the theoretical requirements of the logit model and its implicit utility function variables representing differences in expenditure have been calculated rather than expenditure in absolute terms. This has been necessary also for the purpose of interpreting the estimated results as will be seen later. In order to obtain these values the cost function was estimated for the two types of ships, flagged out or not flagged out, and the estimated values were used to calculate the differences in costs [13].

To obtain a more detailed insight into the specific factors influencing the overall decision variables have been created representing the separate components of these total costs, where this was possible from the data provided in the questionnaire. The variables concerning crew costs created for this purpose are: difference in basic



salary; difference in national insurance cost; differences in pension cost; difference in crew related other costs (calculated by adding up the expenses related to leave pay, overtime, and other costs); differences in training cost and difference in travelling costs.

Concerning costs related to other running expenses we have: difference in repairs and maintenance costs, difference in docking and surveys costs, difference in stores and provisions costs, and difference in port and fuel costs.

Additionally, at an aggregated level the following variables have been created: total crew costs difference, and total other operating costs difference. However, in the following analysis a partial figure was used for total other operating costs. Missing values for the variable related to fuel and port charges, and its close relationship with the type of trade the ship is employed on, coupled with the fact that some of the data related to docking costs includes specific expenses carried out on a regular basis (and which some respondents have not averaged over the docking cycles as required) have lead us to exclude these two cost components from the calculation of the variable related to the difference in total other operating costs.

A variable representing the difference in the overall running costs of operating the vessel is available as measured by the total figure given by the company itself, in order to consider the costs not explicitly mentioned as specific categories in the questionnaire.

Variables related to the cost of insurance were included to test previous theories and widespread beliefs related to flagging out [14]. Again, the variables were divided by the size of the ship in order to take account of the influence of the size of the vessel on the calculations related to the insurance premia. The variables were calculated for three types of insurance: Hull and Machinery, Protection and Indemnity and for Self insurance. Again, the variable related to total insurance costs was created as the difference between the sum of all the costs related to insurance per GRT according to the two categories of flag. Finally, a variable relating to the difference in registration related costs (initial and annual costs of registration) has been created. A list of all these variables and their nomenclature is given in Appendix A.

For a preliminary analysis of the data we can define the principal average characteristics of the two types of vessels in this survey. From the sample 61% of the tankers are flagged out as opposed to 48% of the general cargo vessels. The average ship size for tankers is 31 228 GRT (and 55 932 DWT) and the average age is 11.95 years; while for general cargo vessels the average size is 2255 GRT (and 3280 DWT) and the average age is 12.69 years. For tankers there is a strong relationship between the flag of the ship and the type of trade since 61% of the sampled ships are employed in deep sea trade (95% of the flagged out fleet) while this is not the case for the general cargo fleet where only 7% of the fleet is employed in deep-sea shipping and all these vessels are flagged out.

Tankers have an average crew on board of 15 members compared to seven for the general cargo ships. Also the value of the vessel varies significantly between the two types of ships. In order to make the figures comparable we have calculated the average value per GRT which is £1001 for tankers and £1062 for general cargo vessels. We also know that 42% of the tankers and 37% of the general cargo ships have been purchased within the last 5 years and these figures represent 31% and 19% respectively of the flagged out tankers and general cargo ships.

On average tankers operate 362.58 days per annum (the flagged out fleet has 359.73 operative days) while general cargo ships have 359.41 operational days on


Fuctors influencing the choice offhrg 169

average (the vessels under foreign flag have 355.89 operative days) a difference of 3 days.

In the conlparison between vessel types we cannot abstract from considerations relating to the differential in costs connected with the vessel's flag. From the preliminary data analysis the costs connected with flagged out vessels can be seen to be generally significantly lower than the costs associated with nationally flagged vessels and that the responses received show interesting insights into some of the precon- ceived ideas related to flagging out and shipowners' behaviour.

In some cases it is difficult to predict the sign of the coefficient of the variables created but, for others, the expected sign is either dictated by the existing literature or by rr priori beliefs based on the experience acquired while dealing with the represen- tatives from the sector. The expected signs of the variables are as follows:

positive signs are expected for the type of trade, size of vessel, age of vessel, national insurance costs, pension payments, training costs, and total crew costs, and negative signs are expected for basic salary costs, registration fees and possibly the ship type variable.

5.2. Enlpirical re,rults Having identified, constructed and explained the variables to include in the model the authors proceed to assess the effects and magnitude of these on the probability of the ship being flagged out. The function needed to estimate therefore will be:

Probability of being flagged out = f'(ship characteristics, ship differences in costs)

The use of a non linear function such as described in Section 3 requires that maximum likelihood estimation of the function be performed using iterative techniques. In this paper the complete data set is used to estimate the parameters differentiating between the two types of vessels via an intercept dummy [15].

In order to obtain the final model (presented below) the general approach has been adopted testing for the significance and relevance of a number of variables. Where a choice of alternative variables was possible the most appropriate measure of the factor has been determined based on econometric. statistical and a priori reason- ing. In such cases our decision has been to choose the variable which could impart the most informatio~l in a single measurement while satisfying these criteria.

For instance, the G R T variable connected with size was initially included in the place of the variable related to the ships' trade (as the two are highly correlated), however, even though it appeared statistically significant at the 5% level it was decided not to leave it in the model since this factor had shown itself to be highly related also to other independent variables which had a stronger case for inclusion. The variable TradeDum was preferred because it implicitly allowed for the vessel size to be taken into account, since in general ships employed in short sea shipping are smaller than ships employed in deep sea navigation, and at the same time provided additional information.

For the purposes of exploring the data available a dummy was included allowing for the number of years since purchase. The estimated value of the coefficient of this variable was negative, which indicated that ships of recent acquisition are more likely to be placed under national flag, however it was not significant and has thus not been included in the final model. The regression was also estimated using variables related



Table 2. Regression results.

Variables Constant TypeDum Age TradeDum d-Basic d-NI d-Train d-Rep

Coefficient -4.79** -2.94* 0.24** 4.48** -0.66** 4.81** 4.40 1.15 Stand. Err. 1.51 1.55 0.10 1.71 0.26 1.74 2.09 0.825

Model Chi-square X2"' * ** *** =1 , , evels of significance, respectively lo%, 5%, 1%.

to the current value of the vessel and the crew size, however the coefficients of the two variables again did not significantly differ from zero and were therefore omitted.

Having considered variables related to generic factors the variables connected with the cost factors were considered. The responses from the questionnaire allowed the authors to divide the total crew and operating costs into several different components and using all of these attempt to estimate the complete model. However, this approach was not successful because of the presence of multico- llinearity between some variables and the matrix could not be inverted. For instance, the difference in pension related costs and other crew related costs such as those connected with leave, overtime, and other allowances had to be excluded. Moreover, in the final equation we have not been able to include the variable related to insurance costs. While our preference would have been to include this factor it may be argued that these costs largely depend on the reputation and record of the company concerned and not on the flag of the vessel, hence the omission is unlikely to pre- judice the results.

As can be seen from table 2 variables related to the characteristics of the ship and variables representing differences in costs have been included. Specifically the variables include: Age, Type of Vessel; Type of Trade; and differences in Basic Salary Costs; National Insurance Payments; Training Costs; and Repair and Maintenance Costs between the two types of registers. The results of the final model shown in table 2.

Table 2 shows that the coefficients of each variable except d-Rep are significantly different from zero. This variable has not been omitted since there are a priori reasons for its inclusion and furthermore its failure to be significant is only marginal. As highlighted previously the coefficients of the variables included in a logit model allow the sign of the effects of the various factors on the probability of the event occurring to be determined directly but do not provide explicit insights as to the magnitude of the effect; further calculations are needed for this and will be presented subsequently.

A consideration of the sign of the estimated coefficients shows that they are all as expected. In particular, it can be deduced that tankers are less likely to be flagged out compared to general cargo vessels and this is not surprising given the specialist nature of the trade. The impact of the variable Age is also as expected since the older vessels are more likely be flagged out. It also agrees with common sense that the vessels employed in deep sea shipping are more likely to be registered in foreign registers. The fact that the dummy related to area of trade is significantly different from zero at the 5% level reinforces the conclusion reached in the previous section concerning the qualitative investigation of the factors influencing the decision to flag out, namely that the type of trade may be an important consideration when choosing the flag.


Factors influencing the choice o f ,flag 171

This model also shows that the difference in basic salary is inversely related to the likelihood of the ship being flagged out. What this actually implies is that an increase in the differences in basic salary decreases the probability that the ship is flagged out. This, even if it initially appears to be counter intuitive, seems on a deeper analysis to be logical since flags of convenience crews generally receive a 'consolidated' wage which includes sums toward pension provision and other allowances. The average basic salary of seafarers employed on the sampled ships is higher than that of seafarers employed on national flagged vessels. This argument is strengthened by the positive relationship which is revealed between the variable d-NI and the likelihood of the ship being flagged out. In this case as the difference in National Insurance related costs increases, the probability of the ship being flagged out also increases. The obvious corollary to this is that a decrease in the difference in these costs would increase the probability of the ship being under national flag. Another interesting result is yielded by the sign of the coefficient for difference in training costs. The positive relationship implies that as the difference in costs increases, the probability of a ship being flagged out also increases.

The final variable d-Rep has a positive sign which indicates that an increase in savings related to repairs would increase the probability of the vessel being flagged out, however, the coefficient of this variable as previously mentioned is not significantly different from zero. From this it can be inferred that the difference in the costs of repair and maintenance between different flags is not salient to the decision and this suggests that there is unlikely to be any difference between the repair and maintenance standards under different flags. Any such differences which may occur may arise from a variety of factors including, for example, differences in company cultures.

In conclusion this model shows that the likelihood of flagging out is positively affected by Age, Type of Trade, National Insurance Payments, Training Costs, and Repairs and Maintenance Costs, but it is negatively affected by the Type of Ship and differences in Basic Wage Costs.

In fact, the calculated probabilities show that for Tankers engaged in Deep Sea Trade the probability of being flagged out is, ceteris puvihu.~, 95% while if it is engaged in Short Sea Trade the chance of it being flagged out is only 18%. This might stem from the deployment of the vessels in coastal trades where the use of local crews is more likely, port calls more frequent and environmental considerations assume greater relevance. For General Cargo vessels engaged in Deep Sea Trades the equivalent probability of being flagged out is 99%, while for Short Sea Trades it is 80%. Again similar arguments apply.

To provide an assessment of the magnitude of the effects of these variables on the decision to flag out we can calculate the impact of changes on the probability of the event occurring. This can be expressed in terms of odds as follows: the odds on a General Cargo vessel being flagged out are nearly 19 times as large as for a Tanker, and the odds on a vessel engaged in Deep Sea Trades to be flagged out are 88 times as large than for one engaged in Short Sea Trades.

6. Conclusion This paper focusses on the factors which influence the choice of flag and embraces both qualitative and quantitative considerations. The analysis is based on a sample of original data obtained by the authors through field survey interviews and ques- tionnaires distributed to members of the UK shipping industry. The results deal with



two particular sectors namely the tanker and general cargo markets and provide an insight into the magnitude and significance of various factors which affect the choice of flag.

These conclusions are based on interpretations of the qualitative aspects of the data and on the outcome of the econometric analysis stemming from the estimation of a Binary Logit Model. The authors have been able to provide indications of the likelihood of a particular vessel being flagged out under different circumstances and, further, to consider how changes in these circumstances might affect the probability of the event occurring.

From the initial literature review there emerged little agreement among the experts on the sign and on the magnitude of the effects of the various factors influencing the choice of flag. This lack of agreement however, had not previously led to a systematic empirically based study designed to determine and verify the effects and this omission provides the rationale for this study. While it is widely recognized that the flagging out phenomenon is primarily fuelled by the desire of the shipowner to minimize both costs and restrictions on his operating freedom, this research has pointed out a multitude of other factors which may have a role to play in the decision making process.

The qualitative analysis has indicated, in the context of remaining under the traditional national flag the importance, inter alia, of factors such as Public Relations Reasons, Historical Factors, Marketing Considerations and the Trade Routes Involved. Within the UK the impression is that many companies operate in a manner consistent with living up to their public expectations and make their decisions accordingly. The list of qualitative factors affecting the use or retention of a foreign flag would include Crew Costs, Availability of Skilled Labour, Search for Less External Control, High Compliance Costs of Flagging Back in, and Fiscal Reasons. This last factor has often been quoted as a major reason for the decision to flag out though it may be the perception of this effect rather than the reality which stimulates the decision. Certainly the questionnaire responses have indicated that most of the companies are in the No Tax position where it is unlikely that fiscal factors could have much impact on the commercial consideration of choice of flag and therefore we remain unconvinced of the effect of fiscal factors in this context.

The quantitative part of this study has confirmed the importance and relevance of factors, such as the Age of the vessel, the Trade on which it is engaged, Basic Wage Costs, National Insurance Payments, and Training costs, to the decision to flag out. It has suggested that Basic Wage costs may have a negative influence on the decision but that it is the total employment costs which matter. Furthermore, the study indicates that differences in the costs of Repairs and Maintenance between different flags is not salient to the decision. It is likely that any difference in maintenance standards is more the results of company culture than of choice of flag. Clearly the decision on which flag to choose is based on a subtle amalgam of both subjective and objective factors.

The major problem encountered during this work has been the confidentiality and availability of data. The authors are grateful to all those who assisted in this regard but objectivity requires a follow-up study to test these results in a wider context. The next stage must be to extend this analysis to other sectors of the UK shipping industry and also to the shipping industries of other countries.


Factors influencing tlze choice o f ' j u g 173

Appendix A. Description of variables

Variable Calculation Contents

DepVbl TypeDum

TradeDum

GeoDum G R T AGE Cr-Nun1 Op-Days YearDum

Dependent variable (1 =flagged out, 0 =otherwise) Type dummy (1 =tanker. 0 =otherwise)

Trade dummy (1 =deep sea trade, 0 =otherwise)

Geographical dummies (not relevant) G RT Age Crew number Days of operation Purchase dummy (1 =within 5 years, 0 =otherwise)

Basic salary for seafarers on national flag ship, basic salary for seafarers on foreign flag ships NI for seafarers on national flag ship, NI for seafarers on foreign flag ships

Pension for seafarers on national flag ship, pension for seafarers on foreign flag ships Other crew costs for seafarers on national flag ship, other crew costs for seafarers on foreign flag ships

Training costs for seafarers on national flag ship, training costs for seafarers on foreign flag ships Travelling costs for seafarers on national flag ship. travelling for seafarers on foreign flag ships Repairs and maintenance costs on national flag ship, repairs and maintenance on foreign flag ships

Docking and survey costs on national flag ship, docking and survey costs on foreign flag ships Stores and provision costs on national flag ship, stores and provision costs on foreign flag ships Fuel and port charges on national flag ship, fuel and port charges on foreign flag ships Total crew costs for seafarers on national flag ship, total crew costs for seafarers on foreign flag ships Total operating costs on national flag ship. total operating costs on foreign flag ships Total running costs on national flag ship, total running cost on foreign flag ships H&M costs on national flag ship, H&M costs on foreign flag ships P&I costs on national flag ship. P&I costs on foreign flag ships Self insurance costs on national flag ship, self insurance costs on foreign flag ships Total insurance cost on national flag ship, total insurance cost on foreign flag ships Registration costs on national flag ship, registration costs on foreign flag ships

Ship flagged outlflagged in Ship type (tankeqgeneral cargo) Type of trade (decpkhort sea) Area of trade Size of the ship Age of the ship Crew size Days of operation Ship purchased within 5 years Difference in basic salary per crew membert Difference in national insurance payments per crew membert Difference in pensions payments per crew membcrt Difference in other crew related costs per crew membcrt Difference training costs pcr crew membert Difference in travelling costs per crew membert Difference in repairs and maintenance costs per GRTf Difference in docking and survey costs per GRT: Difference in stores and provision costs per GRT: Difference in fuel and port charges per GRT: Difference in total crew costs per crew membert Difference in total operating costs per G R T I Difference in total running costs per GRT: Difference in H&M insurance cost per GRT: Difference in P&I insurance cost pcr GRTf Difference in self insurance cost per GRT: Difference in total insurance costs per GRT: Differencc in registration costs per GRTf

?The variable has been divided by 100. f The variable has been divided by 1000


174 Factors influencing the choice of flag

Notes and references 1. For a recent discussion and definition of the term 'flag of convenience' see BERGANTINO,

A. and MARLOW, P., 1996, An analysis of the decision to Jag out: part one. The Seafarers' International Research Centre, University of Wales, Cardiff, UK.

2. The studies by TOLOFARI, S. R. , BUTTON, K. J. and PITFIELD, D. E., 1986, Shipping costs and the controversy over open registry, Journal ofIndustria1 Economics, 19 (4), 409427; TOLOFARI, S. R., 1989, Open Registry Shipping. A Comparative Study of Costs and Freight Rates, (London: Gordon and Breach Science Publishers); METAXAS, B. N. and DOGANIS, R., 1976, The Impact of Flags of Convenience, (London: PCL); METAXAS, B. N., 1985, Flag of Convenience: A Study of Internationalisation (Aldershot: Gower); YANNOPOULOS, G. N., 1988, The economics of flagging-out, Journal of Transport Economics and Policy, 22 (2), 197-207, can be considered exceptions since they offer a detailed analysis of the flagging out phenomenon accompanied by empirical findings.

3. The term 'open registries' embraces not only the familiar and traditional flags of convenience (FoCs) but also the more recent registries that allow free registration (the second or international registers).

4. The authors are well aware that profit maximization may not be the only goal for companies but this assumption is appropriate and realistic for this study.

5. Figures quoted are from Structure and Organisation of Maritime Transport, July, 1995. 6. Recently the factor 'quality' has been assuming increasing importance in the studies

concerning the flagging out phenomena. In particular, the study by CULLINANE, K. and ROBERTSHAW, M., 1996, The influence of qualitative factors in Isle of Man ship registration decisions, Maritime Policy & Management, 23 (4), 321-336, has shown the importance of qualitative factors alongside the widely recognized economic factors.

7. It is widely recognized that some ports (i.e. Australian and Scandinavian in particular) are more likely to place strong emphasis on the control of the living, operating, and safety conditions on board the vessels berthed.

8. For further details see MADDALA, G. S., 1986, Limited-Dependent Variable and Qualitative Variables in Econometrics (Cambridge: Cambridge University Press).

9. This is because the Logit function is monotonically increasing in its arguments and therefore if the parameter P, associated with the jth explanatory variable is positive, then the general conditional probability will increase with an increase in x,. The contrary is valid if the coefficient of x, is negative.

10. The standard t-test can also be performed but the output of the econometric package gave the Wald test.

11. However, it must be remembered that this measure in ML is different from the OLS equivalent.

12. GARDNER, B. M. and MARLOW, P. B., 1983, An international comparison of the fiscal treatment of shipping, Journal of Industrial Economics, X X X I (4), 39741 5 provide a full description of the different tax positions available to shipowners.

13. The OLS technique has been used to estimate the functions, and assumptions about the independence of the error terms have been made see PAGAN, A,, 1989, Two stage and related estimates and their application. Review of Economic Studies, LIII, 517-538. For reasons of brevity details of the estimations have been omitted from this paper but they are available upon request from the authors.

14. For example to test the current validity of the study by Metaxas and Doganis (see [2]). 15. The report by the authors, published by the Seafarers' International Research Centre,

Cardiff in July 1997, contains results obtained by analysing separately the factors affecting the decision of the shipowners for the different types of ships.


Documents

Factors Influencing Ship Registration