Socioeconomic development and the risk of maritime piracy

A broadly similar pattern is found in the classification of airline hijackings, and Holden (1986) illustrates that the motivation for, and the probability of an attack, is influenced by the specific nature of attacks.

There is well established literature on the economic and security implications of airline hijackings, for example, Dugan et al. (2005). And although shipping piracy in the P-HV form is another type of hijacking, surprisingly similar literature is not well established. If ship piracy attacks are linked to terrorism objectives, they may corroborate the findings of Dugan et al. (2005) that increased security measures did not influence terrorist motivated hijacking attempts in the same way as non-terrorist motivated attempts. This may lead to enquiries to attempt to separate some forms of P-HV from terrorism.

A test of this discrepancy would be complex. Yet subject to data this could be assessed. Such an analysis would require inter-alia, marine insurance premiums for various markets/routes, a comparison of premium revenue and cost to marine insurance companies over time, total revenue derived from premiums relative to total costs of payments to shipping companies, aggregate marine insurance premium revenues over time/ annual pay-out costs made to shipping companies, and further data on premium revenue, disaggregated into premium revenue for piracy routes and premium revenue for non–piracy routes would be required. There would also be a need to consider cost to shipping companies. This would include data on aggregate insurance costs over time - annual premium costs paid by shipping companies over time – for piracy and non-piracy routes, and premium costs for piracy routes and non-piracy routes. Further granularity would be obtained by analysing these costs by ship size, type of cargo, value of cargo, flag of vessel, nationality of captain/crew etc and a comparison of Horn of Africa insurance and South-East Asia insurance premiums-where piracy is prevalent and insurance premiums where incidences of piracy are less. Also it is worth noting that in the P-HV version, the amount of ransom demanded and the amount actually paid are often misstated. Likewise in P-AR, the value of goods which the victims claim they are robbed of is often overstated for insurance claims or understated to avoid possible prosecution by authorities for fiscal and other violations. In sum, while it is an important issue to consider and is worthy of further enquiry, measuring the exact impact, costs, benefits and incentives of marine insurance is difficult without the data. The one industry which does not benefit from such a subsidy is marine insurance. Marine insurance is a key aspect of shipping operations and is affected by piracy. As with any commercial underwriter, the ship insurance underwriter makes an actuarial assessment of the maximum cost of a catastrophe and relates this to the probability of such an event taking place. The underwriter places as many safeguards and compliance requirements on the insured party as is possible, relative to the premium cost that the ship owner pays. Such requirements includes, routes which the ship may operate in ports it may call at, types of cargo it may carry, the minimum quality of the officers and crew and the quality of the charter party and ship manager. In the main, given the potential costs of catastrophe, a subsequent increase in the premium that would be due as a result of non-compliance with underwriting requirements is unlikely to be enough of an incentive for the underwriter to seek to benefit from piracy activity.

Such a situation enables a number of negative activities to take place which might be checked if democratic institutions and processes existed. Even where they are elected, the party in power stays in office for so long that the civil service and other arms of the constitutional system become co-opted. Individuals become substitutes for institutions at great cost to society at large. A key feature of such a system is that there is a large network of informers loyal to the key individual- this makes it highly unlikely that the head of state would be unaware (if only, for self-interest) of anything that has economic or security implications.

It is also possible that the frequency of attacks is not strongly correlated with the amount of revenue derived and such revenues derived from piracy activities are not expended in locations in close proximity to the attacks and so the level of imported goods and port activities, growth in capital stock and real estate demand and prices will receive no noticeable multiplier effects.

The forms of violence vary from threats of physical harm, gun-fire to deaths of a single crew member or murder of entire crew. The database highlights whether a particular incident is reported to a nearby coastal authority. Further, the database describes the nature of action taken by the particular coastal authority(ies) that receives the report. There are cases where private ships and warships in the vicinity respond to the attack; where these occur, such incidents are also captured.

Removing vessels without IMO numbers from the data would have reduced valuable observations, so instead they were allocated unique numbers with the assumption that they had only one attack.

The empirical analysis also considered univariate and multivariate logistic regression models and the results of these estimates are available from the authors.

Correlation plots and their fit statistics for economic and governance indicators are available upon request from the authors.

Other goodness of fit procedures for the logistic regression have been suggested by Tang (2001) amongst others.

We apply the exchangeable correlation structure that assumes the same correlation between vessels within the same cluster of attacks. See Fig. 1 with a homogeneous variance and constant correlation. This does not bias the parameter estimates since GEE parameter estimates are consistent as long as (6) is correctly modelled. \( \left[\begin{array}{ccccc}\hfill {\rho}^2\hfill & \hfill \rho \hfill & \hfill .\hfill & \hfill .\hfill & \hfill \rho \hfill \\ {}\hfill \rho \hfill & \hfill {\rho}^2\hfill & \hfill .\hfill & \hfill .\hfill & \hfill .\hfill \\ {}\hfill .\hfill & \hfill .\hfill & \hfill .\hfill & \hfill .\hfill & \hfill .\hfill \\ {}\hfill .\hfill & \hfill .\hfill & \hfill .\hfill & \hfill .\hfill & \hfill .\hfill \\ {}\hfill \rho \hfill & \hfill .\hfill & \hfill .\hfill & \hfill .\hfill & \hfill {\rho}^2\hfill \end{array}\right] \)

A logistic regression with random intercepts, an extension of the mixed model approach, popularly used to model non-normal data was also considered. However the specification of the conditional model and the distribution of random effects presented challenges given that the dependent variable is categorical. These challenges were overcome by specification of a GEE model.

The functional forms of the 5 multivariate GEE models vary. A GEE model is fitted in each of the regions and for the global data. Therefore there are 6 models per table multiplied by 5 tables which gives 30 models. Details on differences in the specification of the functional forms are available from the authors at www.​baragh.​com.

Tables of the results of models 1 and 4 are available and can be provided by the authors.

We had tried to consider the impact of military intervention more directly via the specification of a variable to measure naval capacity to prevent or apprehend pirates. The intervening naval capacity (INC) is naval capacity that is capable of providing a direct military response to a vessel attack. The INC variable is calibrated as the location of total military tonnage as a percentage of total available tonnage. The majority of INC is located in North America, Russia, China and Japan. Non-intervening Naval Capacity - NINC is naval capacity that is unable to respond directly to attacks outside of a port area. There might be a high correlation between NINC, low economic development and low institutional quality. In particular, it may be useful to consider whether the probability of piracy attack increases the further distance from the operational base of intervening naval capacity (INC), and the alternative hypothesis. So with appropriate data, we could test whether or not the probability of a piracy attack is influenced by INC. However the absence and sensitivity of the data made these tests unachievable.

The Kenyan economy is the largest in East Africa. However on a per capita basis the GDP of Djibouti is almost twice as high as that of Kenya, and almost three times higher than Somalia and Tanzania. The data indicates that Djibouti also has highest quality of governance. In the governance measures of control of corruption, rule of law, government effectiveness and political stability, Djibouti performs better that the rest of the countries with Somalia providing the worst outcomes. Net ODA is also lower for Djibouti compared to the rest of the countries.

An attempt was also made to assess whether unofficial flows from piracy can be discerned. The key focus was on Kenya – which is the regional financial centre. The avenues of enquiry were of two types. A direct approach involved attempting to obtain data on money transfers from Somalia and Djibouti. An indirect approach involved estimating the variance between official sources of mortgage and housing finance and the extent and level of development of housing and other capital stock. However the results proved spurious due to absence of detailed data.