Taxation of Life Insurance in South Africa Revisited PDF

TAXATION OF LIFE INSURANCE IN SOUTH AFRICA REVISITED By RJ Clover ABSTRACT This paper describes the theoretical framework underlying the four funds basis for taxing life insurance in South Africa and records the historical steps in the development of the ba...

TAXATION OF LIFE INSURANCE IN SOUTH AFRICA REVISITED By RJ Clover ABSTRACT This paper describes the theoretical framework underlying the four funds basis for taxing life insurance in South Africa and records the historical steps in the development of the basis from inception to its current form. It goes on to evaluate the basis in its current form and to propose possible changes that should be considered. KEYWORDS Life insurance taxation; four funds basis; trustee principle; South Africa CONTACT DETAILS Rob Clover, Department of Statistics and Actuarial Science, University of Stellenbosch, Private Bag X1, Matieland, 7602. Tel: (021) 808 3878. e-mail: [email protected] 1 INTRODUCTION The origins of this paper lie in the author’s personal quest for a document that could explain to him the rationale behind the expense relief ratio formulas that appear in the current four funds basis of taxation. But for a few minor differences in the formula, this search would have ended with the discovery of Taxation of life insurance by Theo Hartwig (1994), which is the seminal work on the four funds basis. The topic is revisited here in part to explain the differences between the basis as described by Hartwig and the basis as it currently stands. The four funds basis appears to be unique in the world of taxation not only in the approach that it adopts to tackling the problem of taxing the life insurance industry but also in the remarkably small amount of public discussion and debate that it has generated over an extended period of time. This may be due in part to the fact that much of the debate, if it occurred, would not have occurred in the public arena; but it may also be evidence that the four funds basis cannot be substantially improved. 1 The purposes of this paper are twofold. Firstly, the paper describes the theoretical framework underlying the four funds basis (section 2) and records the historical steps in the development of the basis from inception to its current form (section 3). It is hoped that this will be of use to others who, like the author, have not been personally involved in the development of the basis but would like to understand how it came to be as it is. Secondly, the paper evaluates the basis in its current form (section 4) and proposes possible changes that should be considered (section 5). It is hoped that this will stimulate fresh debate on the problem of using an average tax rate to tax individual policyholders who would otherwise pay tax at differing marginal rates, a problem that is not new but is still relevant. 2 THEORETICAL FRAMEWORK 2.1 Difficulties of taxing life insurance 2.1.1 Applying normal income tax rules There are a number of difficulties in applying normal income tax rules to life insurance, as evidenced by the fact that many countries, including South Africa, have developed special tax rules that apply specifically to the life insurance industry. If normal income tax rules were applied to the life insurance industry, the life office as a corporate entity would be taxed on income (including premium income) less allowable expenses (including benefit payments). The policyholder, as a purchaser of goods and services, would not be subject to any tax on the benefits received. It is interesting to note that the problems with this approach do not arise directly as a result of the risk-pooling mechanism of insurance. Hartwig (1994) argues that normal income tax rules are appropriate for a life insurance company that sells only “pure insurance” business, where pure insurance is defined as “the pooling of risks where there is a zero sum game for the group as a whole, after allowing for expenses, profit for the underwriter and interim reserving for fluctuations”. An example would be a company that sells only single-year term assurance policies. Premiums received less benefits and expenses paid would be a reasonable basis on which to tax the company. Assuming that mortality experience is in line with the pricing assumption, the policyholders of the company would, as a group, make a net loss equal to the expense and profit margins included in the premiums and should not be subject to tax. 2 Difficulties arise as a result of the long-term, contractual nature of life insurance policies, which enables life offices to structure the premium income stream in such a way that the premium received in a particular period need not match the cost of benefits and expenses in the same period. The first implication is that premiums received less benefits and expenses paid in a particular year may no longer be an appropriate basis on which to tax the company. An extreme example would be a block of single premium whole life assurance policies. If normal income tax rules were to be applied, the company would have a large taxable gain on this block of business in the first year and taxable losses in each subsequent year until the business went off the books. A second implication of the mismatch between income and outgo is the existence of a “savings element” in those products where the cumulative income exceeds the cumulative outgo at some point during the term of the policy. Because the investment return earned on any cumulative excess of income over outgo can be returned to the policyholders in the form of benefits, insurance need not be a zero sum game and the policyholders as a group can make a net gain equal to the investment return earned (less margins). In the example of the single premium whole life policies, the single premium paid would be expected to be smaller than the death benefit received. It could reasonably be argued that this savings element of the benefits received by policyholders should be liable for tax. The long-term nature of life policies therefore leads to “the mix of savings return, savings and risk intermediation, and risk pooling (all of which can give rise to income) inherent in a life policy” (Oliver, 2004). Many of the difficulties in taxing life insurance relate to the problem of separately identifying each of these components and taxing them in an appropriate manner. Following from the discussion above, it would seem reasonable for the risk pooling element of the benefits to be free of tax, the savings return element of the benefits to be taxed as policyholder income and the profits generated by carrying out the business of a life office to be taxed as corporate income. 2.1.2 Taxing policyholder savings Having identified the portion of the investment return that belongs to policyholders, there is the additional practical problem of deciding when and how it should be taxed. Taxing the full savings element at the time of the benefit payment would be a deferral of tax which is 3 unlikely to be acceptable to the revenue collectors. It would also necessitate dividing the net gain in the policy into the risk pooling element and the savings element, and dividing the savings element into the different forms of investment return, if they are taxed differently. In some cases this could be a “difficult or impossible task” (Hartwig). An alternative to taxing benefits is to tax the savings income as it accrues. If, however, the tax liability for income is passed on to the policyholder as the income accrues, there is the problem that tax becomes payable before the individual has access to the income that generated the tax liability. One solution is for the life office to pay tax on behalf of policyholders, but taxing the life office as a proxy for policyholders introduces the problem of deciding on an appropriate tax rate, if the policyholders are subject to different marginal rates of personal income tax. This is perhaps the single greatest weakness of the four funds basis that is currently used in South Africa and will be discussed in more detail in subsequent sections. 2.2 Principles of taxing life insurance Hartwig proposed the following as being fundamental principles governing the design of a basis for taxing life insurance in the South African context: “(a) The insurer must pay tax in respect of policyholders’ interests during the roll-up stage, so that there is no tax deferral. The full payout to the policyholder may then be regarded as having been taxed already. (b) If there are to be reasonably level tax playing fields in the tussle between life insurers and other savings media, the insurer should be acknowledged as being taxed as a proxy for the policyholder. This has implications for the tax rate (which should be in line with that applicable to the body of policyholders) and the tax basis (which should be consistent with principles of individual taxation). (c) The profits of the corporate entity must be separately determined and taxed in accordance with company tax principles. Although one can debate it in theory, in practice one probably has to extend this to mutuals as well.” (Hartwig ¶ 4.1, emphases as in the original) 4 2.3 The Hartwig model The four funds basis originally proposed by the Jacobs Committee1 was consistent with the theoretical model developed by Hartwig. The following extracts outline the development of the model. 2.3.1 Basis for the taxation of policyholders’ interests “Looking at the whole body of policyholders together, the net increase in their interests over a year may be expressed as the benefits they have received, less premiums paid, plus the increase in reserves set aside for them – i.e. (^R + B – P). However it is also obvious that the reserve build-up may be expressed as follows: ^R = Ip + P – B – Ep – T i.e. ^R + B – P = Ip – Ep – T For this purpose ^R = increase in policyholder reserves B = benefits paid to policyholders P = premiums paid by policyholders Ip = total investment fruits earned on policyholders’ funds Ep = expenses incurred on behalf of policyholders T = profits withheld from policyholders This leads to the general conclusion that (Ip – Ep – T) is a reasonable representation of the increase in policyholders’ interests and hence a starting point for calculating the taxable income on their behalf.” (Hartwig, ¶ 5.1) “This can also be intuitively reasoned: Considering the policyholders and their insurance interests as a single entity, the only gain to the entity comes from investment accruals and the only reductions come from expenses incurred and profits withdrawn from their pool.” (Hartwig, ¶ 5.2) 1 Discussed in section 3.1. 5 “Note that in developing this argument it is reasonable to exclude profits/losses from underwriting experience and discontinuances provided these are retained within the policyholder group. The “gains” of some individuals (e.g. early deaths) are losses to others, with the net gain being zero, as long as nothing other than expenses passes out of the pool. The only exception is profit withdrawn from the policyholders’ pool for the benefit of the underwriter as a corporate entity – this should be taxed as trading profit to the underwriter, and as a deductible loss to the policyholder body, as dealt with above.” (Hartwig, ¶ 5.3) “The proxy principle suggests that for tax purposes, Ip should comprise only those elements that an individual would be taxed on.” (Hartwig, ¶ 5.5) “Considering an individual policy in isolation, it is reasonable to concede that the expenses attributable to a policy should only be allowed as a deduction when that policy is contributing enough income against which the expenses may be set off – i.e. the expenses should be carried forward for tax purposes, until the policy generates enough taxable income. This is difficult to deal with when hundreds of thousands of policies are aggregated, but may be dealt with pragmatically by spreading initial expenses over a period of years.” (Hartwig, ¶ 5.6 (a)) “If the whole of Ip is not taxed, there is a strong argument that the whole of Ep should not be deducted – i.e. you should reduce Ep by the ratio of taxed investment fruits over total investment fruits”. (Hartwig, ¶ 5.6 (b)) “Summing up, therefore, it is reasonable to calculate the tax payable on policyholders’ interests by the formula (Ip – Ep.r – T) at tax rate t. Ip includes those elements that would normally be taxed in individuals’ hands Ep includes all expenses attributable to policyholders, with initial expenses being spread over a period of 5 to 8 years r represents a ratio of taxed investment income to total investment income t represents an average policyholder tax rate T represents profits/surplus earned from policyholder funds but not distributed to policyholders or held in reserve for them” (Hartwig, ¶ 5.9) 6 2.3.2 Taxation of the corporate entity “Looking at the insurer as an ordinary company trading for profit, one would easily arrive at the conclusion that its taxable profit is total income less total outgo, with amounts put to reserve for policyholders being allowed as an expense. This would be consistent with the formulae for banks and general insurers. Hence the taxable income for the corporate entity = (P + I) – (B + E + ^R) = I – E – (B – P + ^R) (rearranging) = I – E – (Ip – Ep – T) (referring to 5.1) = (I – Ip) – (E – Ep) + T (rearranging) = Ic – Ec + T where the subscript c refers to income and expenses not allocated to policyholders – i.e. allocated to the corporate entity.” (Hartwig, ¶ 6.2) 2.4 An adapted model The author is of the opinion that there is a small but significant oversight in the development of the model described in 2.3.1. The derivation of the formula for the total increase in policyholders’ interests, (Ip – Ep –T), is sound, but the adaptation of this formula to allow for the fact that only part of the income is taxable, in order to arrive at the formula for the taxable portion of the increase in policyholders’ interests, (Ip – Ep.r – T), is flawed. The flaw in the formula is more easily seen by replacing the Ip in the second formula, which is defined differently to the Ip in the first formula, with the equivalent quantity in terms of the original Ip definition. The second formula then becomes, in terms of the original definitions of Ip and Ep, (Ip.r – Ep.r – T). A more correct formula would be (Ip.r – Ep.r – T.r). The fact that (Ip.r – Ep.r – T) is flawed can be illustrated by considering a company that writes a new block of profitable business that does not generate any taxable income. If this business is written in a policyholder fund that happens to be taxed at the same rate as the corporate entity and which has sufficient taxable income against which the profits can be offset, the additional tax payable in the corporate entity will be offset by the additional deduction in the policyholder fund and the profit on the block of business will essentially be 7 tax free (although there is a slight distortion if the ratio of income generated to expenses incurred is not the same for the new block of business as for the rest of the business). This effect was illustrated in the model office scenarios run by Hartwig, but was accepted as a legitimate consequence of the model. Commenting on the effect of transferring profits from policyholder funds to the corporate entity, it was noted that: “the increase in tax is not as great as might have been expected. Once again the reason is that as long as the IPF2 has taxable income against which to offset the transfer, the overall effect is to tax the transfer at the net differential between the IPF and CF tax rates, which is fairly small at only 10%3. If one were to introduce a UPF into the model, so that the transfers cannot be deducted in the remitting fund, the effect of increasing transfers would undoubtedly be much greater.” (Hartwig, ¶ 10.4) The conclusion that T.r and not T should be deducted from the policyholder taxable income can also be reached using a slightly different approach. As shown by Cole (1995), it can be useful to consider premiums as comprising different components. For this argument it is assumed that each premium consists of a pure risk cover component that contributes to the pooled risk benefits, an expense fee component that is used to cover the expenses of the life office and an investment capital component that will ultimately be returned to the policyholder, with interest, in the form of a benefit. The interest earned on the investment capital component represents the savings element of the policyholder’s gain. The component of the premium representing the risk profit margin is ignored for this illustration. Looking at the group of policyholders together, the net increase in their interests over a year would exclude the risk cover component of the premiums (on the grounds that it is a zero sum game) and the investment capital component of the premiums (which represents a capital investment). The net increase in their interests would be the total income earned on the policyholder funds for the benefit of policyholders (which will be called Ipp) less the expense fee component of the premium (which will be called F). The income referred to here is not the full Ip as defined for the Hartwig model, but Ip less that portion of Ip which will 2 IPF is the individual policyholder fund, UPF is the untaxed policyholder fund and CF is the corporate fund under the four funds tax basis. Refer to section 3 for details. 3 This was the differential at the time. It is currently 1%. 8 ultimately be transferred to the corporate entity as part of the profits earned on policyholder funds. The net increase in the policyholders’ interests is then Ipp – F. Setting this equal to the previous expression for the same quantity, we have: Ipp – F = Ip – Ep – T and, equivalently, T = (Ip – Ipp) + (F – Ep) i.e. the profit generated from the policyholder funds is that portion of the investment return not allocated to policyholders plus the expense profit. This full profit is transferred to the corporate entity and taxed there. The income that should be taxed on behalf of the policyholders is (Ipp.r – F.r), where r again represents the ratio of taxed investment income to total investment income. Now, if the tax calculation was based on Ip and Ep and no adjustment was made for the profit transfer, tax would actually be applied to (Ip.r – Ep.r). In that case, policyholders would be overtaxed by an amount equal to (Ip – Ipp).r + (F – Ep).r = T.r. It is therefore sensible to deduct T.r, and not T, from (Ip.r – Ep.r), before applying tax. 3 A BRIEF HISTORY OF THE FOUR FUNDS BASIS 3.1 The Jacobs Committee Report (1992) The four funds basis for the taxation of life insurance in South Africa was first formally proposed in the Report of the Committee of Investigation into the Promotion of Equal Competition for Funds in Financial Markets in South Africa (September 1992), commonly referred to as the Jacobs Committee Report. This report did not, however, present a starting point for discussion on the matter, but rather it presented a system that had already been agreed upon after “much debate and consultation between the life-insurance industry, the Financial Services Board and the Commissioner for Inland Revenue” (Jacobs, 1992). It is not surprising, therefore, that the four funds basis as described by the Jacobs Committee Report was enacted largely unchanged. As shown by Hartwig, it was also consistent with the theoretical framework that he had developed. 9 3.1.1 Underlying Principles The principles underlying the proposed basis were similar to the principles identified by Hartwig, quoted in section 2.2. “The following principles were considered and should govern the suggested tax system: - The “trustee principle” should be adhered to in respect of all income representative of the insurer’s constituent body of policyholders and should reflect all relevant aspects of their taxation, including the effective tax rate. - All income that an insurer receives and that is not representative of the policyholders (and hence not subject to the “trustee principle”) should be subject to normal corporate tax. - Tax neutrality and competitive neutrality between life insurers inter se and between the life-insurance industry and other financial industries must result, as far as possible, from the new system. - Tax neutrality must prevail, as far as possible, between different classes of policyholders. In particular, there should be no tax advantages for corporate policyholders. (It is accepted that this principle cannot be fully served as to allow for the various individual tax rates of the individual constituent policyholders of an insurer, and that an average rate must be used in this case.)” (Jacobs, p89) It is interesting to note that the Achilles’ Heel of the proposed system, the use of an average rate for taxing the insurer as a proxy for policyholders, was conceded at the same time as setting down the very principle that it violates. The “trustee principle” is a term commonly used to describe the proxy basis of taxation central to the four funds basis. It is described in the report as follows: “This principle, in short, entails that life insurers are deemed to be “holding” and investing funds on behalf of their policyholders, and they should pay income tax on the income derived therefrom on a similar basis.” (Jacobs, p88) 10 3.1.2 The four funds The proposed mechanism by which the above principles would be put into practice was “the maintenance of four funds for tax purposes to which, save as may be qualified, the general principles of taxation will apply.” (Jacobs, p90) The proposed four funds were: “The Individual Policyholders’ Fund (IPF): A fund for taxed policies owned by individuals, which is to be taxed, on a representative basis, at the average tax rate of individual policyholders. The Corporate Policyholders’ Fund (CPF): A fund for policies belonging to companies and other corporate bodies that are subject to company tax, which is to be taxed, on a representative basis, at the corporate tax rate. The Untaxed Policyholders’ Fund (UPF): A fund for approved fund business, annuity business and policies belonging to those bodies that are not subject to tax, which will not be taxed. The Fund for Non-insurance Business, Shareholders and Corporate Reserves – the Corporate Fund (CF): A fund representing the corporate reserve of the life insurer, which is to be taxed on normal company-tax principles.” (Jacobs, p90) Hartwig notes that initially only three funds were planned, but it was then felt that policies owned by companies should not be pooled with those owned by individuals as that would create a tax advantage for corporate policyholders4. Jacobs (p92) proposed that 30% would be an appropriate average individual policyholder tax rate. Hartwig points out that the use of 30% for the IPFs of all insurers is a pragmatic alternative to the theoretically more correct approach of calculating the actual average rate applying to each company-specific policyholder group. He adds, however, that “Having agreed the principle, 4 The corporate tax rate of 40% at the time exceeded the proposed IPF rate of 30%. Currently the corporate tax rate is 29% whereas the IPF rate is 30%, which means that corporate policyholders would be disadvantaged if the distinction between the two groups was not made. 11 the average rate needs to be tested from time to time , and be revised if the tax environment has changed.” (Hartwig, emphasis added) 3.1.3 Calculation of the taxable amount For each of the three policyholder funds, the calculation of the taxable amount proposed by Jacobs was consistent with the Hartwig model and of the form: Taxable income less Adjusted expenses less Profits transferred to the corporate fund. The first adjustment to be made to expenses was that selling expenses were to be spread over a period of five years. This does not seem unreasonable, since “if these are allowed in full when incurred, it dramatically reduces the taxable income, especially during times of high inflation or expansion.” (Hartwig) The second adjustment to be made to expenses was that they were to be reduced for the purposes of tax deductibility by applying the ratio of taxed income to total income in the fund in question. The ratio proposed by Jacobs was I / (I + D + C) where I represents the taxable income (interest and rentals5), D represents dividend income and C represents all other gains and losses, realised or unrealised. It was noted in the report, however, that “there is still debate and uncertainty concerning the appropriateness of the inclusion of capital gains and losses (“C” in the formulae) in the calculation of the deductible expenses.” (Jacobs, p92) Hartwig notes that the inclusion of capital gains in the formula was “strongly resisted” and the conclusion of the debate was that initially they would be excluded. The transfer of profits to the corporate fund is the mechanism by which the principle of taxing corporate profits at the corporate tax rate is achieved. As in the Hartwig model, Jacobs proposed that such profits (assuming that profits are positive) be fully deductible in the policyholder fund from which they are being transferred and fully taxable in the corporate fund. The appropriateness of the transfer being fully deductible in the policyholder fund has been questioned in section 2.4. 5 Capital Gains Tax had not been introduced at the time. 12 In the event of losses in a policyholder fund, the transfer in respect of profit would be a transfer from the corporate fund to the policyholder fund. Jacobs proposed, as does Hartwig, that transfers in this direction also be fully deductible in the fund from which the transfer is being made (now the corporate fund) and fully taxable in the fund to which the transfer is being made (the policyholder fund). The net effect is to tax positive corporate profits, which could be those generated in one policyholder fund (and deducted in that fund), at the tax rate applicable to another policyholder fund (the fund in which the loss was incurred), instead of at the corporate rate. This is not sensible, and the net effect could even be that a negative effective tax rate is applied to the profits. For the corporate fund, the calculation of the taxable amount proposed by Jacobs was consistent with the Hartwig model and of the form: Taxable income less Expenses plus Profits transferred from policyholder funds. Again the taxable income is the taxable income arising from the assets held by the fund in question (i.e. the corporate fund), which in this case includes non-insurance taxable income earned by the fund. The expenses are “all deductible expenses and tax allowances that would be available to a company in normal circumstances” (Jacobs, p92). 3.1.4 Calculation of the transfer amount The calculation of the amount to be transferred from each policyholder fund to the corporate fund proposed by Jacobs was consistent with the Hartwig model and of the form: Market value of assets less (actuarial liabilities plus defined margin) The basis proposed by Jacobs for the calculation of the actuarial liabilities was the prescribed minimum basis6 with certain modifications, the key modifications being the addition of accrued interim bonuses and accrued non-vested bonuses (except for those relating to single premium and paid-up conventional business) to the prescribed minimum reserves. Hartwig explains the rationale behind the proposed basis: 6 The statutory reserving basis that preceded the Financial Soundness Valuation. 13  “The actuarial valuation basis is a very important element of the four-fund basis, because it determines the quantum of surplus and hence the quantum of inter-fund transfers. The following were seen as necessary qualities of the basis: (a) It has to be prescribed; a discretionary basis leaves too much scope for tax management, especially since the inter-fund transfers are a major aspect. (b) It has to be reasonably stable; changes are difficult to deal with because of their impact on the inter-fund transfers. (c) It has to be realistic and be appropriate for use in conjunction with assets being valued at market value.” (Hartwig, ¶ 8.4) “In the end, the prescribed minimum basis was chosen, but with certain modifications to attempt to satisfy criterion (c) above.” (Hartwig, ¶ 8.5) The use of market values for assets, together with the proposed basis for valuing liabilities, “has the inevitable consequence that for all business where the liabilities are not market- related, rises and falls in the asset values have a major effect on surplus in the funds.” (Hartwig, ¶ 9.2.1) It is for this reason that a defined margin that could be added to the actuarial liabilities was proposed. The proposed margin calculation would effectively enable the insurer, at their discretion, to smooth the tax on undistributed surplus over a five year period in order to “avoid large inter-fund transfers, which would otherwise be necessary purely because of fluctuations in the market value of assets.” (Jacobs, p96) Hartwig says of the defined margin: “This was considered essential for business where bonus rates are smoothed; otherwise asset value fluctuations would cause wild fluctuations in inter- fund transfers and hence in tax. In effect it represents surplus that has been earmarked for policyholders but not yet distributed, and may justifiably be included with policyholder liabilities.” (Hartwig, ¶ 8.2) At the time of the Hartwig paper the scrapping of the Act Minimum basis in favour of the Financial Soundness basis for supervisory purposes was already a possibility. In the 14 discussion on the Hartwig paper, A McNulty commented: “If the FSV basis is used, then I would assume that the basis applies to the valuation of both the assets and the liabilities i.e. consistency is applied in determining the assets and liabilities. This would most likely help to reduce the fluctuations in the transfers to the corporate fund and “smooth” the revenue to the Receiver.” The use of market values for assets also implies that “undistributed surplus arising from capital appreciation (realised or unrealised) will be taxed on transfer to the CF. This feature of taxing profit arising from capital appreciation has been criticised as unfair to life insurers. Even the revenue authorities seemed uncomfortable, but there was no way round the problem.” (Hartwig, ¶ 9.2.1) Hartwig added: “In the end, I have peace with the concept that tax should be levied on profits permanently withheld from policyholders regardless of the source of those profits”. 3.1.5 Special transfers Jacobs proposed that in two specific circumstances transfers between funds be allowed to take place without the transferred amount being taxable in the fund into which the transfer was made (or deductible in the fund from which the transfer was made). The specific circumstances, and the rationale for the proposal, are described in the Explanatory Memorandum on the Income Tax Bill, 1993: “Subsections (10) and (11) provide for the calculation of certain special transfers which, in terms of subsection (14)(d), are disregarded for tax purposes. The first of these, under subsection (10), deals with the situation where an insurer has to transfer assets from his CF to make good a shortfall in his IPF or CPF, but has insufficient taxable income in his CF to offset the transfer. The transfer in these circumstances is essentially one of capital, and while it would under the ordinary rules be subject to tax in the IPF or CPF, there would be no corresponding relief in the CF. The subsequent reversal of such a transfer in a later year will be similarly disregarded for tax purposes. The second special transfer, under subsection (11), is a transitional measure. Because of the high initial costs incurred by an insurer in the issue of a policy, the premiums paid by the policyholder in the first years do not cover the insurer’s liabilities under the policy. 15  Consequently, when establishing the separate policyholder funds, the insurer will have to employ some of his own capital in providing the assets required to meet the liabilities in the fund. This subsection permits the insurer to identify these capital amounts, which may then be transferred back to his corporate fund over a period of eight years without being subjected to tax (and without being allowed as a deduction in the policyholder fund).” (RSA, 1993) Jacobs referred to the capital transfer giving rise to the second special transfer described above as “‘locked-in’ new business (sic)” and set out guidelines for the calculation thereof. Hartwig notes that although there is a clear theoretical justification for this concession, particularly in the case of a closed fund, “for an expanding office the strain of putting new business on books depresses tax payable, and a further deduction for locked-in strain may result in tax levels that are seen by the authorities as being unacceptably low during the first 8 years.” (Hartwig, ¶ 9.1.2) 3.1.6 Demarcation of life insurers and banks Although not directly related to the four funds basis, it is worth noting that the Jacobs Committee also made recommendations on the issue of the demarcation of the business of life insurers and the business of deposit-taking institutions. It was appropriate that this should happen at the same time as introducing a new tax system, as the demarcation between the two had previously been governed by the infamous Sixth Schedule to the Income Tax Act, which formed part of the tax legislation dealing with life insurance. “The demarcation was not achieved through a legal definition of either the markets or the products, but by the imposition of a punitive tax system on the proceeds of such policies, thereby discouraging the investor from using the investment medium concerned. This method of demarcation was not very successful and has progressively been circumvented by an astute life-insurance industry, which increasingly applied tax-based product development and marketing strategies. In turn, this caused progressive amendments to the 6th Schedule to curtail these activities, finally resulting in a most complex and difficult piece of legislation developing on the statute-book.” (Jacobs, p86) 16 Jacobs recommended that the demarcation “should take place in topical legislation, not in fiscal legislation”, which practically meant moving the legislation from the Income Tax Act (deleting the Sixth Schedule) to the Insurance Act (s59D) and later to the Long-Term Insurance Act (s54). The principles proposed to govern the resolution of the demarcation question were that: “- Life insurers should be prohibited from paying any investment-based benefit under any class of policy they may issue, within five years from the inception of the policy. In this regard, withdrawal benefits should be limited to capital plus five per cent compound interest within five years and the balance, if any, after five years. - Life insurers should not be limited to (sic) paying any risk-based benefits arising from death, sickness, disability, insolvency or incapacity within this five-year period or thereafter. - Other than the above, life insurers should not be restricted in terms of markets or products, provided the products conform to those statutorily defined for life insurers as a condition of registration.” (Jacobs, p88) 3.2 The Income Tax Act, 1993 The Income Tax Act, 1993 enacted the majority of the changes proposed by the Jacobs Committee by introducing a new section (s29) to the Income Tax Act, No. 58 of 1962. The taxation of life insurers had previously been dealt with in s28. The following features of the four funds basis as described in section 3.1 were included in s29 of the Act: a) The establishment of the four funds themselves (s29(4)). b) The spreading of selling expenses over a period of five years (s29(14)(a)). c) The annual transfer of the amount of assets in excess of liabilities from a policyholder fund to the CF (s29(6)(a)) (or vice-versa in the event of liabilities exceeding assets (s29(6)(b)), with such amount being deducted from the income of the fund from which it is transferred and included in the income of the fund to which it is transferred (s29(14)(d)). 17 d) The use of a prescribed basis for the valuation of liabilities. The adjustments to the prescribed basis, proposed by Jacobs, were not included in the Act, although allowance was made in the calculation of the prescribed value for “modifications as may from time to time be determined for the purposes of this section by the Chief Actuary of the Financial Services Board.” (s29(1)). e) The addition of a defined margin to the liability value for the purposes of calculating the transfer amount (s29(8),(9)). f) The use of a special transfer in the event of transferring capital from the CF (s29(10)). g) The use of a special transfer in the event of locked-in new business strain (s29(11)). The Act also specified that premiums and claims (including reinsurance premiums and claims) shall be disregarded for purposes of determining taxable income (s29(14)(f)). This point was not specified by Jacobs, but was clear in the development of the Hartwig model. The one significant difference between the four funds basis as proposed by Jacobs and Hartwig and the basis as described in the Act was that the Act placed no restriction on the proportion of expenses that could be deducted when calculating the taxable amount in policyholder funds. Given that a significant portion of the total income in the policyholder funds would be expected to be in non-taxable form, this omission would significantly reduce the amount of the taxable income under the new basis. 3.3 The Revenue Laws Amendment Act, 1999 The original s29 of the Income Tax Act was replaced by s29A in 1999. The four funds basis was retained, but a number of significant changes were made. The Explanatory Memorandum on the Revenue Laws Amendment Bill, 1999 sets out the reasons for the tax regulators revisiting the subject of life insurance. “It … became apparent after the phasing-in period of the four-fund approach that the amount of tax payable by the long-term insurance industry was decreasing, despite the fact that substantial profits were reflected in the annual financial statements of the insurers. The method of the taxation of long-term insurers has, therefore, been investigated. Certain 18  deficiencies in the current method of calculation of tax were identified, which appear to be the reason for the relatively small amount of tax collected from the industry.” (RSA, 1999) 3.3.1 Deductibility of expenses The first of the “deficiencies” identified in s29 was described as follows: “In terms of section 29, the insurer’s expenditure is required to be allocated to the fund in which the business, to which the expenditure exclusively relates, is conducted. Where the expenditure does not so exclusively relate to business in any one fund, the expenditure is allocated to the different funds in the proportion that business is conducted in the respective funds. The bulk of these expenses are allowed as a deduction against the investment income of policyholder funds. As the base of the taxable income which mainly consists of investment income in each respective policyholder fund is relatively small compared to the amount of the deductible expenses, especially the selling expenses, the taxable income in the policyholder funds is drastically reduced.” (RSA, 1999, emphasis added) It is not surprising that the deductible expenses were unacceptably high in the view of the tax authorities, as the original proposal to restrict the proportion of the total expenses that could be deducted to the ratio of taxable income to total income was, for reasons unknown to the author, rejected at some point between the Jacobs Committee Report and the Income Tax Act, 1993. The solution adopted, as set out in section 29A(10) of the Act, was to allow the full “amount of expenses and allowances directly attributable to the income of the particular fund” to be deducted, but only allow a proportion of all other expenses allocated to that fund to be deducted. This proportion is sometimes referred to as the expense relief ratio and was specified as (I + R) / (I + 3R + 6D), where I is interest income, R is rental income and D is dividend income. Although no details on the derivation of the formula are given, the Explanatory Memorandum does shed some light on the thinking behind the formula: “The underlying principle of apportionment is to exclude that portion of the expenses attributable to non-taxable income, such as dividend income and capital gains. An appropriate method of apportioning these expenses would have been an asset-based approach. It does, however, appear that such a basis is susceptible to manipulation. For 19  that reason an income-based formula is proposed on a basis that would more of less produce the same result as an asset-based approach. Such formula will be closely monitored to ensure that it produces the appropriate apportionment ratio.” (RSA, 1999) It would appear that the intention of the tax authorities was to use a formula along the lines of (I + R) / (I + R + D + G) where I, R and D are as above and G is capital appreciation (Hartwig, ¶ 5.6), but due to concerns about the calculation of G it was decided to use an adjusted formula based only on I, R and D as a proxy. The proposed formula would then be sensible if the total return on property was expected to be around 3 times the rental yield and the total return on equities was expected to be around 6 times the dividend yield. Given that the deduction of the high level of initial expenses was identified as a particular problem with the original s29, it is interesting to note that with s29A the restriction requiring that initial expenses be spread over 5 years was lifted. At the same time, though, s29A did not specifically allow for special transfers relating to locked-in new business and limited the extent to which the unused special transfers arising from s29 could be utilised going forward. 3.3.2 Deductibility of transfers The second of the “deficiencies” identified in s29 was described as follows: “An insurer charges administrative and management fees in respect of the income or assets of the policyholders. These charges are at present not taxed appropriately as transfers from one fund to another. Although taxable in the transferee fund, these charges are fully deductible in the transferor fund.” (RSA, 1999) The issue referred to here is the problem identified in section 2.4. It is not appropriate to deduct the full transfer amount in the policyholder fund because only the taxable portion of the income, and therefore a similar proportion of the profits, is subject to tax in the policyholder fund. A more correct approach would be to deduct only a portion of the transfer amount in the policyholder fund, that portion being determined using the ratio of taxable income to total income in the fund (i.e. the expense relief ratio). The approach adopted in the legislation (s29A(11)(a)(iii)) is explained as follows: “The deductible portion of the transfer will be calculated by applying the ratio (used to determine the allowable selling and administration expenses in respect of policies in the 20  relevant policyholder fund), to 50 per cent of the transfer. It is, furthermore, proposed that the deductible portion of the transfer may not result in an assessed loss in the relevant policyholder fund. No deduction will of course be allowed in the untaxed policyholder fund. This deduction will also be monitored on a regular basis to determine whether it appropriately compensates for any possibility of double taxation.” (RSA, 1999) No justification was given for applying the expense relief ratio to only half the transfer amount when calculating the deductible amount, and the reference to double taxation may suggest that the tax authorities were aware that the other half of the transfer (multiplied by the expense relief ratio) was theoretically subject to tax firstly in the policyholder fund and then again in the corporate fund. In fact, the problem of double taxation may have been overlooked altogether when it was first decided to reduce the deductibility of transfers in the policyholder fund, as the 50% allowance was itself only granted “after intense lobbying by life offices against what would have clearly amounted to double taxation” (Carroll, 1999). Another weakness relating to transfers in the original Section 29, discussed in section 3.1.3, was the fact that transfers from the CF to policyholder funds were tax deductible in the CF. This was also corrected in Section 29A (s29A(11)(e)) and was explained as follows: “Transfers to the policyholder funds will … not be taxable in such funds. Transfers from the corporate fund will also not be deductible in the corporate fund, but any subsequent return of such transfer to the corporate fund will not be taxable in the corporate fund” (RSA, 1999). This change also meant that it was no longer necessary to make allowance for special transfers relating to capital transfers from the CF. 3.3.3 Calculation of liabilities The Explanatory Memorandum (1999) states that: “It is proposed that the existing valuation basis be changed from the prescribed valuation basis to the financial soundness valuation method.” Although the FSV basis is not specifically mentioned in s29A, the reference to the basis “specified in the Insurance Act” has been removed and liabilities are to be calculated “on the basis as shall be determined by the Chief Actuary of the Financial Services Board in consultation with the Commissioner.” (Income Tax Act, No. 58 of 1962, s29A(1)) Appropriately, at the same time as moving from the prescribed minimum basis to the FSV basis for valuing liabilities, the defined margin that could be added to liabilities under s29 was 21 removed from the legislation. The defined margin was presumably the issue being referred to in the third of the “deficiencies” identified in s29, namely: “The current system of transfers provides for opportunities to defer taxation.” (RSA, 1999) It is worth noting that the tax regulators did not presume that the amendments to the four funds basis in 1999 would necessarily be the final word on the subject: “The effectiveness of the proposed measures will, however, once implemented be monitored closely and, if necessary, adjusted appropriately. Such adjustments may also include a re-evaluation of the appropriateness of the four-fund approach and the trustee principle.” (RSA 1999, emphasis added) 3.4 The Taxation Laws Amendment Act, 2001 The Taxation Laws Amendment Act, 2001 added capital gains tax to the South African tax landscape. It was decided that realised capital gains would be added to the taxable income of life insurers’ funds, using an inclusion rate of 25% in the case of the IPF and 50% in the case of the CPF and CF. It was also decided that capital gains made on second-hand life insurance policies would be liable for income tax under the capital gains rules. The inclusion of gains (as well as foreign dividends, co-incidentally at the same time) in the taxable income of the policyholder funds resulted in a need to change the expense relief ratio formula that had been introduced in 1999, as explained in the Explanatory Memorandum on the Taxation Laws Amendment Bill, 2001: “This formula limited the expenses on the basis that dividends and capital gains were not taxable. As certain dividend income (foreign dividends) and capital gains will now become taxable, it is proposed that the formula be adjusted to eliminate the possibility of a double taxation of amounts transferred from policyholder funds to the corporate funds and to allow a portion of selling and administrative expenses in respect of capital gains taxed in policyholder funds which would not otherwise be allowed as a deduction. As the inclusion rates for capital gains of the individual policyholder fund and the company policyholder fund are different, it is proposed that separate formulae be introduced for the two funds which would take into account the inclusion rate of 25 per cent for the individual 22  policyholder fund and the inclusion rate of 50 per cent in the case of the company policyholder fund. Only capital gains accruing from 1 October 2001 will be subject to tax, therefore, the full impact of tax on capital gains on the value of assets of the individual policyholder fund and the company policyholder fund of an insurer will only be felt a number of years after the introduction of CGT. For this reason the amended formula will be phased in over a period of five years from years of assessment commencing on or after 1 January 2002.” (RSA, 2001) Although the memorandum clearly explains why it is appropriate to increase both the expense relief ratio applicable to the IPF and to a greater extent the expense relief ratio applicable to the CPF, it again does not give any detail on the methodology used to arrive at the new formulas. The “old” formula, adjusted for foreign dividends, becomes: (I + R + F) / {I + 3R + 6(L + F)} while the new IPF formula is: (I + R + F) / {I + 2,5R + 4,75(L + F)} and the new CPF formula is: (I + R + F) / {I + 2R + 3,5(L + F)} where I is interest income R is rental income F is foreign dividend income L is local dividend income. 4 EVALUATION OF THE CURRENT FOUR FUNDS BASIS 4.1 Evaluation against the theoretical framework The current four funds basis is largely consistent with the original theoretical framework developed by Hartwig, but differs in the following areas: 23 a) The valuation basis used has changed to the FSV basis, which is not as strictly prescribed as the basis originally proposed. This is arguably a more appropriate basis for the calculation of the transfers between funds, as evidenced by the fact that the defined margin originally proposed to smooth transfer values is no longer required. b) Transfers from the CF to policyholder funds are no longer deductible in the CF and taxable in the policyholder fund. This appears to be an improvement on the original theoretical framework. c) Transfers of profits to the CF are no longer fully tax deductible in the policyholder fund. This also appears to be an improvement on the original theoretical framework. d) Initial expenses are no longer spread over five years. This is not inconsistent with the theoretical framework, as the spreading of expenses was essentially a concession for the benefit of the tax authorities. e) The formula for calculating the expense relief ratio differs from that originally proposed. Arguably this is not inconsistent with the theoretical framework, but rather a pragmatic approximation aiming to achieve the same result. It would also be expected to differ to the extent that it has been affected by the introduction of CGT. From a theoretical point of view, the weaknesses of the current basis are as follows: a) Only 50% of the transfer amount multiplied by the expense relief ratio is deductible in the policyholder fund, whereas theoretically this full amount should be deductible. There is therefore an element of double-taxation. b) The formulas used for the expense relief ratios have not been satisfactorily justified (either theoretically or empirically). It could be argued that they represent an arbitrary component in the current basis. c) The average tax rate applied to the IPF could also be regarded as an arbitrary component of the basis. Not only was no concrete evidence given to support the initial choice of this 24  rate, but it is highly unlikely that, as is the case for the proxy rate, the actual average individual tax rate has remained unchanged since 1993. 4.2 Evaluation against general tax policy objectives Although each country is free to choose its own tax policy objectives, it is useful to briefly evaluate the current four funds basis against some of the policy objectives that Oliver (2004) proposes tax rules for the insurance sector should aim to achieve. a) Neutrality between different forms of financial intermediation. This was one of the key principles identified by the Jacobs Committee, so at least at a principle level the four funds basis is consistent with this objective. In practice, however, the fact that a proxy rate is used for taxing IPF policyholders means that for many policyholders the marginal rate of tax paid on the savings element of a life policy is different to the marginal rate that would be paid on income generated through other forms of savings. The fact that policyholders cannot use their individual interest and capital gains allowances to reduce the tax payable on income earned within a life policy can exacerbate this problem. b) Appropriate treatment of both the profits attributable to shareholders and the savings income attributable to policyholders. Again this was one of the principles underlying the development of the four funds basis. Except for the weaknesses identified in section 2.4, this objective is achieved by the current basis. c) Administrative compatibility with existing tax procedures to minimise enforcement and compliance costs. Although this was not a central principle underlying the development of the four funds basis, the high level at which the tax calculations can be done and the fact that the life insurer pays tax as a proxy for policyholders has significant administrative advantages. This issue would be a key consideration when contemplating changes to the current basis. 4.3 Evaluation against alternative tax systems Oliver (2004) identified the following four broad approaches that have been adopted by countries around the world in order to tax life insurance. The comments under each of the approaches are also taken directly from Oliver. 25 a) U.K. model – life office taxed on net investment income as proxy for policyholder tax on savings income. The life office is taxed on net investment income (I – E) as if it were just a savings intermediary paying tax on behalf of policyholders. The obvious disadvantage of this approach is that it leaves the pure insurance component preferentially taxed, and there can be complexities over issues such as the allocation of overhead expenses between investment and underwriting activities. b) U.S. model – underwriting income taxed and policyholder benefits taxed on deferral. Underwriting income is taxed, and, thus, insurance activity is within the tax net, but the taxation of the savings component is deferred until savings income is attributed to policyholders. This gives life insurance a tax advantage in the savings industry. c) Continental Europe model – concession applied to savings. Underwriting income is taxed, but either the savings component is exempt or policyholder benefits are taxable but after being offset by premiums. d) New Zealand model – underwriting income taxed and policyholder benefits taxed on accrual. This is arguably the purest model, in that both underwriting and savings income is taxable on an accrual basis. However, as with the U.K. model, savings income is taxed at the life office level without regard for the varying rates applicable to individual policyholders. The four funds basis would fall under the New Zealand model in this categorisation. The strength of this approach is that it is theoretically sound, the weakness is the practical problem of taxing the life office as a proxy for policyholders. 4.4 Concerns and criticisms Publicly voiced concerns and criticisms have centred around the use of a proxy rate of 30% for taxing the IPF. These can be divided into concerns around the principle of taxing a group of policyholders at an average rate and concerns around the level of the average rate itself. It is interesting to note that the concerns raised at the time of introducing the basis are still the major concerns today. 26 4.4.1 The principle of using an average tax rate In a minority view in the Jacobs Committee Report, Mr I Meiklejohn (Commissioner for Inland Revenue at the time) expressed his view that the trustee principle was a compromise necessitated by factors unique to the long-term insurance industry and should not be extended to areas where the normal principle, that the income of an investor be taxed in his hands, could be applied. (Jacobs, p70) Hartwig, in describing potential problems associated with using an average rate, notes that “The use of an average creates the opportunity for exploitation by high marginal taxpayers whereas the low income policyholders pay too much.” (Hartwig, ¶5.8) In opening the discussion on the Hartwig paper, B Gouws commented: “If there is an area of vulnerability in the four fund basis, I believe it to be the treatment of individual policyholders’ tax. The use of an average tax rate, particularly under South African circumstances with such wide divergence in incomes between poor and affluent policyholders, is unlikely to endure.” (Hartwig, Discussion) In the “unified framework for life insurance” described by Hudson (unpublished), one of the tax principles is that “The proceeds of investments by private individuals in collective vehicles are taxed as if the individuals held the underlying assets.” Hudson argues that this principle “can only be achieved by transferring the responsibility for paying tax on income and deducting tax on expenses to the policyholder, as is the case with mutual fund investors”. 4.4.2 The level of the IPF rate In its submission on the proposed introduction of CGT, the Life Offices’ Association of South Africa proposed that the rate on the IPF be reduced from 30% to 25% in order to compensate policyholders for the fact that they cannot use their individual capital gains exclusions or interest exemptions to reduce the tax payable on their savings income in the IPF (LOA, unpublished a). In its representation on the Taxation Laws Amendment Bill, 2005, the Banking Association of South Africa requested that the IPF rate be reviewed in the light of the fact that it exceeded 27 the CPF rate and was “substantially higher than the average rate applicable to individual taxpayers.” (SARS, unpublished a) 4.4.3 The position of the South African Revenue Services As mentioned in section 3.3, SARS indicated in 1999 that the appropriateness of the trustee principle and the four funds basis may need to be re-evaluated at some point. The author is not aware of any public document setting out in detail the current thinking of SARS on this issue, but previous comments made by SARS help to paint a general picture of its views. In responding to representations made to it on the proposed introduction of capital gains tax, SARS commented on the appropriateness of the rate applied to the IPF as follows: “The view is held that the average rate of 30% remains appropriate. A further reduction will lead to a greater differential between the 42% maximum marginal rate and the 30% rate for the IPF, which will lead to greater distortions in the taxation of investment products.” (SARS, unpublished b). In the same document, SARS explained why it was reluctant to adjust the IPF rate in order to compensate individuals for not being able to use their interest exemptions and capital gains exclusions to reduce the tax payable in the IPF, as had been proposed by the life industry. It also implied that it was not willing to re-open the debate on the IPF rate: “The proposal will be difficult to entertain for the following reasons: The policyholder may have other gains as well during a year against which he / she may offset the annual exclusion. Allowing the annual exclusion for the policyholder in his / her personal capacity, plus a form of benefit in the insurer will grant a double benefit to policyholders.” “If we allow a complete look-through approach to accommodate all the circumstances of the individual (policyholder) the debate of the rate at which the IPF is taxed will also have to be re-opened.” In the Explanatory Memorandum on the Taxation Laws Amendment Bill, 2001, SARS gave the following reasons for imposing CGT on second hand policies: 28  “The preferential tax treatment afforded to insurance policies encourages long term savings. Second hand policies do not necessarily comply with this objective as the longer term investment objective is broken. These policies contain a speculative element that would otherwise escape taxation… The large majority of people who invest in these policies are high income earners paying tax at 42 per cent. By investing in second hand policies on a short term basis they enjoy the benefit of the low preferential tax rate of 30 per cent. By levying CGT on these policies the gap is closed to a large extent.” More recently, SARS made the following comments as part of its response to written representations on the Taxation Laws Amendment Bill, 2005: “It is important to note that the 30% policyholder fund rate was informed by the average individual tax rate. Given the dynamic nature of the taxpayer distribution, further quantitative work is needed to establish the weighted average individual income tax rate. The long run policy goal is to create more neutrality in the tax treatment on the returns of capital invested by individuals. Clearly, from a tax policy perspective one would like to strive for equity and neutrality in order to create an environment conducive to retirement savings with the maximum utilization of competitive market forces that tend to reduce transaction costs in the long-term savings market, with commensurate higher investment returns for individual policyholders. In conclusion, this matter has been noted but no rate adjustment can be contemplated at this stage, as this would pre-empt the recommendations of the retirement fund tax review.” (SARS, unpublished a) 4.4.4 The author’s views The author is of the opinion that the four funds basis is an elegant solution to the difficult problem of taxing life insurance, being both sound in theory and simple in application. The one aspect of the basis that is troubling, however, is the inequitable tax treatment at an individual policyholder level. While it could be argued that the use of a proxy tax rate is an 29 unfortunate but tolerable side-effect of the provision of risk benefits which cannot be accessed other than through the products of life insurers, there is no such defence in the case of life- wrapped pure investment products. It is almost unthinkable that there should exist a savings medium that over-charges low- income earners for tax and under-charges high income earners for tax, particularly if similar underlying assets can be accessed through other savings media that have no such built-in tax distortions. The argument that investors in a life-wrapped pure investment product need be rewarded, in the form of a preferential tax rate, for the lack of liquidity in the first five years of a life policy does not hold water, for the following reasons: a) The minimum-term restriction applying to life insurance policies was introduced to demarcate the areas of life insurance business and banking business primarily for purposes of regulation, not taxation. The intention was not to differentiate between short-term savings and long-term savings so that a tax incentive could be offered to long-term savers. One of the principles underlying the development of the four funds basis was tax neutrality; the preferential tax treatment of high marginal tax payers was an undesired consequence rather than a design objective. b) If it was considered appropriate to reward investors for committing to a minimum five-year investment term, then it would not be sensible to reward only those investors that are high income earners. 4.5 Comparison of the taxation of life-wrapped investment policies to the taxation of unit trusts 4.5.1 Proxy versus individual tax The key tax difference between investing through unit trusts (collective investment schemes) and through life insurance policies is that the investment return arising from a unit trust investment is taxable in the hands of the unit holder, whereas the tax due on the investment return arising from a life insurance policy is paid by the insurer on the behalf of the policyholder. Consequently: 30 a) Unit trusts are required to distribute income to unit holders at regular intervals, whereas income rolls-up inside life policies and is only distributed in the form of benefits. b) Unit holders are taxed at their marginal rates on taxable investment income, whereas the life office applies an average proxy rate to the taxable income of all individual policyholders. c) Unit holders are able to use their individual interest exemptions and capital gains exclusions7 to reduce the tax payable on the returns from the unit trust, whereas policyholders cannot use these allowances to reduce the tax payable within a life policy. This means that high income earners that pay tax at a marginal rate of 40% can, by investing through a life-wrapped product instead of a unit trust, benefit from a relative tax saving of 10% of the taxable income arising from their investments (assuming that they have already used their interest exemption and capital gains exclusion). As there is a 25% inclusion rate applicable to capital gains, this equates to a 2.5% tax saving on realised capital gains. Investors that have not used their interest exemptions and capital gains exclusions can, by investing through a unit trust instead of a life-wrapped product, benefit from a relative tax saving of 30% of the taxable income (equivalent to 7.5% of the capital gains) arising from their investments, assuming that the allowances would not otherwise be utilised and subject to the specified allowance limits. It is interesting to note that even an investor with a marginal rate of 40%, but who would not otherwise use any of his interest exemption, would under the current tax rules need to earn more than R72 000 of interest income, per annum, in order for the life-wrapper to be the more tax-efficient option. This threshold increases to R104 000 for investors over 65. 4.5.2 Capital Gains Capital gains realised on the trading of assets within a unit trust are not subject to capital gains tax, but unit holders are subject to tax on the gains realised when selling units. In the 7 Currently the first R15 000 of capital gain and the first R18 000 (persons younger than 65) or R26 000 (persons 65 or older) of interest income, per annum, is tax free. 31 case of life-wrapped products, capital gains realised on the trading of assets within the fund do give rise to a tax liability, which the insurer must pay on behalf of the policyholders. Unit holders may benefit to the extent that there is a deferral of tax on realised capital gains. The significance of this benefit would depend on how actively the underlying shares are traded. On the other hand, policyholders may benefit if a significant portion of the gain at the time of a benefit payment is on assets that the insurer only expects to realise at some future date. The capital gains tax provision at the time of the benefit payment need not cover the full gain to that date, but rather the discounted value of the gain, taking into consideration the future date at which the gain is expected to be realised. In the case of a unit trust, however, the full gain is realised, for tax purposes, at the point of selling the units. 4.5.3 Expenses The taxable income arising from both savings media is net of a portion of the expenses incurred by the service provider. In the case of a life policy, the portion of expenses that can be deducted is determined by the expense relief ratio, which is theoretically intended to be the ratio of taxable income to total income (including realised and unrealised appreciation). In the case of a unit trust, permitted deductions are made before distributing income to the unit holder. The deductions should be allocated to interest income and dividend income in the same ratio that each component forms of the total income. This would avoid the situation where all the expenses are used to reduce the interest income distribution, which is taxable in the hands of the unit holder, and the non-taxable dividend distribution is gross of expenses. Note, however, that the equivalent of the expense relief ratio here is the ratio of interest income to total income excluding appreciation, which would suggest that a higher portion of the expenses are offset against taxable income in the case of a unit trust than for a life policy. The preceding paragraph applies to expenses that are allowed for in the unit price and are implicitly netted off the distribution amount. Expenses that are paid in the form of reduced allocation to units or as explicit unit deductions are not used to reduce the income distribution and neither could they be treated as deductible expenses in the unit holders’ personal tax returns. There may, therefore, be a tax advantage to life policies to the extent that all selling 32 and administration expenses are, after applying the ratio, available to be offset against taxable income, whereas some of those expenses may not be offset against distributions in the unit trust context. For the same reason, there is a tax disadvantage associated with structuring the initial payment to a sales intermediary in the life insurance context as a direct fee payable by the policyholder to the intermediary instead of a commission payable by the life office to the intermediary. 4.5.4 Non-tax differences The main non-tax difference between life-wrapped investments and unit trusts is that unit trusts are not subject to the restriction period that applies to life policies, which limits the amount that can be withdrawn during the first five years of a policy (except in the case of risk benefits). It is the existence of the restriction period that has led to life-wrapped investments being perceived as a trade-off between preferential tax treatment and reduced flexibility. A further difference worth noting is that the range of investments offered through life- wrapped investments is broader than that available through unit trusts, because there are greater restrictions imposed on unit trust providers. In particular, life offices are able to offer products that require capital backing, such as smoothed bonus funds. 5 ALTERNATIVES TO THE CURRENT FOUR FUNDS BASIS 5.1 Options for achieving neutral taxation of policyholders In order for individual policyholders to be taxed at their marginal rates and have access to individual interest and capital gains allowances, it is necessary to allocate to each policyholder his or her portion of the total taxable income earned by the IPF. Assuming that this can be done, Stobo (unpublished) describes the following four possible approaches to the problem of ensuring that each policyholder pays the correct amount of tax. 5.1.1 Tax liability with insurer, single proxy rate, tax credits In this scenario, the responsibility for paying tax sits with the insurer and tax is paid at a single proxy rate (as is the case with the four funds basis). The insurer then issues tax credits to policyholders, indicating the amount of taxable income that has been taxed on their behalf and the amount of tax paid on their behalf. The policyholders then include the income and tax 33 credits in their personal income tax returns, as a result of which the income earned within the life policy is ultimately taxed at the policyholders marginal rates. The difference between the actual tax paid on behalf of the policyholder and the correct tax payment at his or her marginal rate only comes out in the wash when the policyholder completes his or her personal tax return, so there is a slight deferral advantage for high rate tax payers and a deferral disadvantage for low rate tax payers. There is also the possibility that low rate taxpayers will not have sufficient taxable income to make full use of the tax credits, in which case they should theoretically be refunded by the tax authorities. Such a system of tax credits was one of the possible solutions suggested by Gouws in the discussion on the Hartwig paper. 5.1.2 Tax liability with insurer, final tax at variable rates In this scenario, the life insurer would again be liable for paying the tax. The tax rate used, however, would vary according to the marginal rate of the policyholder in question and there would be no need for a system of tax credits. The tax paid by the insurer would be a final tax, as is the case with the current four funds basis, so the policyholder would not have access to individual allowances. Two disadvantages with this approach, identified by Stobo, are that: a) It would require different after-tax pricing streams from one asset. b) It is not clear how the insurer would monitor the policyholders’ marginal tax rates. 5.1.3 Tax liability with policyholder, proxy withholding tax rate In this scenario the tax liability flows through to the policyholder. The life office calculates distributions on a regular basis that become taxable as part of the policyholder’s personal income, as is the case for unit trusts in South Africa. The difference is that the distributions are deemed distributions, from which a withholding tax is deducted before they are re- invested with the insurer. At the end of the tax year the policyholder needs to settle the difference between the correct tax payment and the withholding taxes already deducted. The effect is therefore similar to the tax credit system, but in this scenario the life insurer’s unit pricing, for example, would be done on a gross of tax basis. 34 5.1.4 Tax liability with policyholder, variable withholding tax rates In this scenario the insurer would again make deemed distributions that are taxable in the policyholders’ hands, but would deduct withholding tax at a variable rate according to the marginal tax rate of each policyholder. 5.1.5 Practical difficulties Implementing any of the above options is likely to be difficult in practice. Identifying the portion of the total taxable income that should be allocated to each of the policyholders may be relatively easy to do in the case of unitised, market-linked products, but in the case of non- unitised products and smoothed portfolios this could be a difficult task and at best would require some degree of pragmatism and imprecision. For products other than unitised, market-linked products, the regular allocation of taxable income (either as deemed distributions or in the form of tax credits) may also create confusion amongst policyholders. Changes in taxable amounts from one period to the next will not necessarily be directly related to changes in policy values, and in the case of risk products the policyholders may be unconvinced that they should be paying any tax at all. Another practical consideration is that in order for either a tax credit or a “flow through” system to function correctly, the policyholders need to complete personal tax returns. It may be that low earning policyholders are not otherwise required to register for tax purposes, in which case the additional inconvenience to the policyholder and additional administrative cost to SARS would make such a solution less attractive. As discussed in section 4.2, ease of administration and compliance is a desirable tax policy objective. LOA (unpublished b) indicates that the option of attributing income to policyholders, to be taxable in their hands, has previously been suggested by the Financial Services Board, but that “SARS opposed the Attribution system because it would put a large number of SITE taxpayers back on assessment. The LOA was concerned about the issuing of tax certificates”. The same LOA circular reported that after a meeting of the FSB’s Demarcation Committee in August 1998, the proposed addition of a fifth “Max-tax” fund to the four funds basis “will be 35 supported by the FSB, SARS and the various industries affected by Section 59D8 demarcation”. This fund was to be taxed at the highest marginal tax rate but be exempt from the restriction period applying to other life insurance policies, so that life offices could compete directly with unit trusts for short-term investments. The author does not know the reasons for the “Max-tax” fund failing to come into existence, but notes that it would have entrenched the thinking that the reason for the restriction period on life insurance policies is to define long-term savings in order to reward such savings with a tax incentive. The flaws in this way of thinking are discussed in section 4.4.4. 5.2 Practical options for consideration 5.2.1 Separate treatment of investment-only life insurance products Rather than changing the four funds basis as it applies to all products, an argument could be made for just changing the tax treatment of investment-only products. This could be argued on the grounds that: a) Investment-only products are more easily substituted with non-insurance alternatives. In discussing the policy objective of neutrality, Oliver (2004) notes that “It is especially important to have consistent tax rules for easily substitutable activities. Otherwise, financial activities will be encouraged to move from industries with high taxes to those with lower taxes, irrespective of the relative efficiency of the industries.” b) The task of allocating taxable income to individual policyholders may be accomplished more easily and more precisely in the case of investment-only products. There are, however, likely to be practical and theoretical difficulties associated with treating investment-only products separately, including the following: a) Rules would be required for defining what constitutes investment-only business. This could, for example, necessitate imposing lower limits on the level of risk cover included in “non-investment” products. 8 The section dealing with the five year restriction period applicable to life insurance policies, now section 54 of the Long-term Insurance Act. 36 b) The existence of risk cover does not necessarily have a significant impact on the level of the savings element of the return earned by a policyholder. Even with a non-trivial sum assured, the characteristics of whole life products are in many ways similar to those of long-term pure savings products. c) To the extent that life insurers are able to offer investment options that are not available through other savings media (such as those requiring capital support), the substitutability argument is weakened. d) The absence of risk cover does not necessarily make the allocation of investment income at an individual policyholder level easier. If the ease of income attribution is to be used as the basis for differentiating between products, then it may be more appropriate, as suggested in recent proposals for life insurance tax reform in New Zealand (New Zealand, 2007), to distinguish between “unit-linked products” and “participating policies and other” for the purposes of tax treatment. If investment-only products are to be treated differently to other products, one alternative would be to scrap pure-investment life insurance products altogether. This approach would essentially be one of changing the basis of demarcation of life insurance business. Although the Jacobs Committee recommended that the term of the contract be used as the basis for distinguishing between banking and life insurance business, it was noted that “Other considerations that may be valid in distinguishing between the industries include the nature of their liabilities and the fact that insurers also underwrite risks” (Jacobs, p87). It is difficult to argue that, in essence, a five-year market-lined pure investment contract is a form of life insurance but a three-year term assurance contract is not. Following the demarcation route could also lead to a distinction between certain types of investment portfolio that are defined to be life insurance business (particularly those that require capital as a result of offering underlying guarantees) and other types of investment portfolio that are defined not to be life insurance business (and are available through alternative savings media). A second alternative would be to leave the demarcation of life insurance business unchanged, but change the basis of taxation for investment-only products. Investment income could be allocated at an individual policy level and taxed using one of the four approaches described in 37 section 5.1. This approach would avoid the disadvantages associated with using a proxy rate for the IPF, but only for the investments-only business. 5.2.2 Increasing the number of funds An alternative solution to the problem of taxing all individual policyholders at an average proxy rate would be to increase the number of policyholder funds. So instead of just having an IPF and a CPF, there could, for example, be three different funds for individual policyholders: one for policyholders who pay tax at the highest marginal rate, one for policyholders who pay tax at the lowest marginal rate and a third for all other individual policyholders. This is essentially the option discussed in section 5.1.2, but without catering for every one of the different marginal rates9. The advantages of this approach are as follows: a) It does not require that the total taxable income of each fund be allocated at an individual policyholder level. b) The tax liability remains with the life office and is paid according to the trustee principal, which avoids the administrative disadvantages associated with attributing income to policyholders, to be taxed in their hands. c) The principle has already been applied in the four funds basis as it currently stands, in that corporate policyholders and individual policyholders are separated for tax purposes. d) The tax rates applying to the individual policyholder funds could be more easily updated to reflect changes in prevailing tax rates, at least for the funds relating to the maximum and minimum marginal rates. This is illustrated by the fact that the CPF rate has changed several times since the introduction of the four funds basis but the IPF rate, which has no direct link to prevailing tax rates, has remained unchanged. The disadvantages of this approach are as follows: 9 South Africa’s progressive tax system currently comprises 6 different marginal rates. 38 a) There may be practical problems associated with allocating policyholders to the correct fund. The life office would presumably require evidence to support a policyholder’s claim that he or she should be allocated to a fund other than the fund taxed at the highest marginal rate (as is currently required for policyholders in the UPF). There is also the problem of policyholders moving between tax brackets, particularly into or out of the lowest or highest bracket, which would theoretically require that they move between policyholder funds. Although movements between the IPF and CPF are allowed for in the current system, the number of such movements would be expected to be much larger. A pragmatic approach may be to allow policyholders to provide evidence supporting any claim to be taxed at a lower than maximum marginal rate at inception and thereafter only at intervals of, say, three years. If at any of these dates no such evidence were to be provided, the policy would default to the maximum rate fund. b) As the liability to pay tax remains with the life insurer, individual policyholders would still not be able to access their individual interest and capital gains allowances. Individuals below the tax threshold would also not be able to access the tax rebate to offset income earned within a life policy, unless there was an untaxed fund catering for such individuals. c) There would still be an average proxy rate applied to the middle income group, with the associated over- and under-charging of tax at an individual level and the problem of calculating an appropriate rate. The inequities would, however, be significantly less than under the current basis and the rate for this fund could be defined in such a way as to change automatically with changes in the prevailing rates (e.g. the median of the marginal rates). The above options, while they may not be the only or the best solutions to the problem of over-taxing low income earners and under-taxing high income earners within life insurance products, appear to the author to have sufficient merit to warrant further consideration by the industry. 6 CONCLUSIONS The four funds basis in its current form is largely consistent with the underlying theoretical framework developed by Hartwig, and in some areas is an improvement on the original framework. Theoretically, the basis represents a more sound approach to taxing life insurance 39 than some other models in widespread use, in that it separates the income of the policyholder and the income of the corporate entity and taxes each appropriately. Other than on the issue of tax neutrality, the basis fairs well in terms of achieving generally desirable tax policy objectives. The most important weakness of the basis is that it applies a single tax rate to the income earned by individual policyholders, irrespective of their marginal tax rates. Life insurance policies therefore have tax advantages for high income investors, and tax disadvantages for low income investors, relative to other forms of saving. This lack of tax neutrality is a particular concern for investment-only life insurance products. One proposed solution to this problem is to treat pure investment policies separately, either by changing the demarcation of life insurance business so that they cease to exist or by taxing them in such a way that they are tax neutral relative to other forms of savings. An alternative proposed solution to this problem is to expand the number of policyholder funds so that the tax paid within each fund more closely represents the marginal tax rates of the policyholders in the fund. 40 REFERENCES Carroll, T (1999). The Revenue Laws Amendment Act 1999. Insurance and Tax, 14(4), 24–30 Cole, TC (1995). The puzzle of life office tax. BAJ 1(I), 79–105 Hartwig, TE (1994). Taxation of life insurance. TASSA X(I), 196–239 Hudson, NB (unpublished). A unified framework for life insurance. Convention, Actuarial Society of South Africa, 1998 Jacobs Committee (1992). Report of the committee of investigation into the promotion of equal competition for funds in financial markets in South Africa. RSA LOA (unpublished a). LOA submission: proposed introduction of capital gains tax. LOA Circular no. 2/2001, http://www.loa.co.za/ (login required) LOA (unpublished b). s59D demarcation meeting held at the FSB offices on 7 August 1998. LOA Circular no. 65/98, http://www.loa.co.za/ (login required) New Zealand (2007). Life insurance tax reform: Officials’ paper No. 2 – suggestions for reform. Oliver, R (2004). Insurance companies. In Zee, HH (2004: 16–31) RSA (1993). Explanatory Memorandum on the Income Tax Bill, 1993. W.P.2 –‘93 RSA (1999). Explanatory Memorandum on the Revenue Laws Amendment Bill, 1999. W.P. – ‘99 RSA (2001). Explanatory Memorandum on the Taxation Laws Amendment Bill, 2001. W.P.1 –‘01 RSA. Income Tax Act, No. 58 of 1962. SARS (unpublished a). Responses to written representations by organisations to the Portfolio Committee on Finance and Select Committee on Finance on the Taxation Laws Amendment Bill, 2005. SARS, http://www.sars.co.za/ SARS (unpublished b). Preliminary response to the representations made to the PCOF and the SCOF on the proposed introduction of capital gains tax. SARS, http://www.sars.co.za/ Stobo, C (unpublished). Towards consensus on the taxation of investment income. Internal Revenue Department, New Zealand, http://taxpolicy.ird.govt.nz/ Zee, HH (ed) (2004). Taxing the Financial Sector: Concepts, Issues and Practices. International Monetary Fund, Washington, D.C. 41