At Burmah Castrol our expertise is in blending and marketing lubricants under the Castrol name. We also have a portfolio of speciality chemicals businesses trading under names such as Foseco and Sericol. We now employ staff in 54 countries and the number grows every year. Our products are of course sold in many more through agencies and distributors. We have more than 20,000 staff but these are fairly thinly spread around the world. A typical unit might have only 100-200 employees and we have very few large concentrations of staff.

Across the world we have in excess of 70 retirement benefit arrangements, in which are included pension plans, provident funds, etc. For historic and business reasons many countries have a lubricants plan and a chemicals plan, and some countries operate more than one plan according to level of seniority. Some plans are defined benefit (DB), some defined contribution (DC) and in some countries, particularly where there is generous state provision, we have nothing. Hybrid schemes are quite rare in our group. Our central records show we have about 160 employees who could be considered international".

We are a decentralised group and are becoming more so as we introduce a new structure of 12 strategic business units (SBUs) in our chemicals division and refocus Castrol into four SBUs. As a general rule, those of us who work in group headquarters cannot impose our policies. If they are to succeed it is essential that we consult line management about their requirements and ensure we meet their needs.

Although a lot of our international moves are still from or to the UK, we are increasingly seeing moves between all sorts of countries across the world.

Whichever category an employee falls in, some basic elements need to be established to determine his retirement benefit provision. First, we have to fix his pensionable salary (including of course the currency) and set a mechanism for reviewing this from year to year. Second, we need a formula. This can either be a DB formula, for example, percentage of final pensionable salary accrued for each year of service, or a DC formula, for example, x% of pensionable salary paid into an individual's account. Third, we need to decide how the benefit is going to be financed - for example, using an existing company plan for local employees (either book reserved or separately invested), or another vehicle such as an offshore plan. Lastly, we need to decide who is paying for the employee's benefits. Normally, this will be the business currently employing him, but it may sometimes involve a previous employer or the centre.

It is convenient to class internationally mobile empolyees as secondees, transferees or nomads. By a secondee I mean someone temporarily assigned from country A to country B with an expected stay of two to three years. Secondments sometimes last longer, but after around five years serious questions will start to be asked about whether the employee should continue to enjoy all the perks of an expatriate assignment. There is always an expectation that the secondee will return to his home country, even if it is to retire or be re-assigned to a new project in a third country. Secondees will be distinguished from local employees by their expatriate terms, including a salary which is derived from their home country salary and a variety of benefits which may include housing, education, flights home, etc.

As well asan actual cash salary, secondees are given a notional home base salary. This is the salary they would have been paid at home if they had remained there but on their new job grade. It is primarily used for pension purposes.

A secondee will remain in the home base pension plan, to provide continuity of service on return. It also means that on retirement in country A (which is assumed) he will get a pension which is consistent with his peer group of country A employees and appropriate in amount and currency for retirement there,

If the country A plan is contributory for employees, there are two ways of giving effect to this: we can arrange for part of the employee's salary to be paid in country A. This will be sufficient to cover his pension contributions, but may also be used to meet mortgage repayments, etc, if he so requests. Alternatively, we can get his employing company to remit the total joint contribution to the UK and then deduct the employee's share from his local salary.

A more interesting and difficult category is transferees. These are people who have moved from country A to country B on a permanent basis, at least as far as we know. Sometimes, the employee knows from the start that the move is permanent. Sometimes a period of secondment eventually turns into something more permanent. Typically, acquiring a green card or landed immigrant status would go with being a transferee. Importantly, transferees will not normally be offered, or will have been weaned off, the (often generous) expatriates' perks and be employed on local terms.

In each of our 54 countries we will have either a DC or DB plan, or no plan at all. If we didn't have any pension plans in the group or if they were all DC, life would really be a lot simpler. In the real world we have to consider lots of DB plans.

The matrix illustrates the nine possible changes in pension arrangement which might apply to someone who transfers from country A to country B. The green sections are straightforward, for example, if neither country has a pension plan or the employee moves from country A having no plan to country B with a DC plan. Even a move from one DC plan to another is pretty simple as the employee just gets two lumps of benefit.

The amber combinations are less straightforward, and the red square, is the most difficult: moving from one DB plan to another.

An employee who has either had no pension plan or a DC arrangement in country A and then moves to country B which has a DB plan isn't really a major problem. We would normally put him into the new DB plan for future service and he would eventually draw from it a benefit based on his local final salary. For pre-transfer service he would either get nothing or a DC benefit as appropriate. We would usually resist granting back service in country B's DB plan, although there have been occasional exceptions when country A's DC plan has been broadly comparable in value.

In the first column are people who have been in either a DC or a DB plan in country A and move to country B where there is no company provision.

Very occasionally somebody moves on a permanent basis to a country like Spain, where we have hitherto had no company plan because of the high level of state benefits. Such a person could be retained in his previous plan in country A but this would be the only feature that distinguishes him from other local employees. We are not keen on using offshore plans for secret payments and we wouldn't normally go through the process of working out a notional home base salary for somebody who transfers. So although it is not really very satisfactory, we might just compensate the individual by giving him either a higher salary or a lump sum.

Next we have the employee who leaves country A which has a DB plan to go to country B which operates a DC plan. For future service, he will join country B's plan and contributions will be made at the appropriate percentage of his local salary, so that's very straightforward. For past service, he'll get a deferred pension from country A's DB plan. If country A happens to be the UK, that vested benefit will automatically be revalued during deferment. In other countries we may need to convince local management that their DB plan has made a "withdrawal profit" if the benefit is based on salary at exit and persuade them to apply some sort of revaluation until the employee retires.

Now we get onto the difficult moves where the employee leaves a DB plan in country A to go to a new DB plan in B. This used to apply to the great majority of international moves, but the growth of DC plans is reducing the numbers.

Our philosophy is essentially that when an employee moves permanently to a new country he should get benefits appropriate to that country, where it is now assumed he will retire. His pension should be consistent with those of other local employees, it should be based on total service in the group in all countries, and it should be based on his final actual salary.

The employee will join country B's DB plan and his pension for future service will be based on his local final salary. As far as his past service in country A is concerned he will get a vested benefit based on his last salary in country A. That's certainly the starting point. Next we could consider re-valuing the plan A benefit during deferment as above. The other possibility, which is what we normally seek to do, is to credit all the employee's past group service in country B's DB plan, and then to offset the vested benefit earned in country A's plan.

If we go down the route of crediting past service in country B's plan and offsetting country A's vested benefit, we then have an issue about who pays for the net amount of past service benefit in plan B. Company A no longer employs the individual so is not interested in contributing but, on the other hand, by only paying a vested benefit, company A is spared the cost of future salary increases, which it would otherwise have borne. Company B is employing the person, wants to motivate him by giving him a fair pension benefit, but is reluctant to pay for previous service. Sometimes, therefore, we reach a deadlock and despite having continuity of group employment, the employee ends up with a vested benefit from plan A and future service only in plan B.

Finally, the people we call "nomads". Nomads are difficult to define, but are essentially people who have lost their "roots" and can no longer easily be identified with any particular country for retirement benefit purposes. They are still a small minority; you could certainly count them on two hands, but we do expect their numbers to grow. We have tended to look at them on an individual basis. However, we are starting to feel that it is time for a consistent policy that treats the employees equitably, is not too difficult to operate and can accommodate growing numbers and unknown complications.

We have looked at various possible models for providing pensions to nomads. Here are our views about three of them: using the head office plan, using the host country plan, and putting them into an offshore arrangement. The fourth option, putting them into a home country plan, isn't really feasible because, by definition, they no longer have a home country. The old chestnut of DB or DC needs to be addressed for nomads and there is the requirement to specify a benefit or contribution formula, a salary for pension purposes and a financing vehicle, as well as the need for clear agreement on who pays.

Taking the head office model, which in our case means the UK, we would use the Burmah Castrol Pension Fund. This has a good level of benefits; it already covers both our UK employees and a number of expatriates; we have good administrative staff and systems able to cope with nomads; and there is a good chance that many of our nomads will be high flyers who end up working in headquarters. However, using our UK fund, which is a DB plan, may cause problems if our nomads don't stay to normal retirement age; they will be pensioned by reference to a UK notional home base salary and this may not produce an appropriate level of pension if they retire somewhere else; adopting the head office plan is arbitrary and inflexible, and to join it the employee needs some sort of UK connection; he's also limited to a maximum of 10 years in the plan.

The local model entails putting the nomad into the plan of whichever country he's working in at any point in time. This is administratively convenient; it reflects whatever salary he is earning; and it's unlikely to be difficult to sell to the management of the company employing him. However, if he keeps moving around the globe, he may find himself leaving plans before he's completed enough service to qualify for a benefit; he'll lose out in DB plans by only getting a benefit based on salary at the time of leaving; and by the time he retires he will have a collection of different benefits in different forms and different currencies in different countries. In aggregate these are likely to fall short of what he would have got by staying in one country. There is then a key issue about whether anyone will identify and make good this shortfall. If so, should the cost be spread across the whole group, should it be borne by head office or should it be re-charged to the companies for which the individual has worked? If there is not going to be any top-up, mobility may be seriously inhibited.

The third model is the offshore one which entails using a special plan in an offshore location, such as Guernsey. We would define a formula for all nomads. In practice, this is much more likely to be DC than DB. We would pick a contribution rate to be used in all cases. It is difficult to say what that rate should be, but it might be the typical cost of providing benefits in a good DB plan, say 15% or 20%. Contributions to the plan would be this percentage of the employee's local salary and it might be that those contributions and investments are expressed in say sterling or US dollars. An arrangement like this is going to provide a benefit which is arguably reasonable for most people and it will ensure that all nomads are treated fairly; it's simple to operate and should be capable of application in most cases. One drawback may be an inability to deduct contributions for tax purposes, although this may be mitigated by tax free benefits.

The offshore defined sontribution approach appears to offer some advantages over the other models. However our thinking is still at an early stage and there is a lot more work to be done until we finaly settle how we are going to pension our nomads.

Michael Thomas is head of the pensions policy and development department at Burmah Castrol. This article is based on a speech given at a recent IBC conference in London