ASSESSING THE EQUITY OF AUSTRALIA S RETIREMENT INCOME SYSTEM

Transcription

1 ASSESSING THE EQUITY OF AUSTRALIA S RETIREMENT INCOME SYSTEM George Rothman 6 July 2009 Retirement & Intergenerational Modelling & Analysis Unit Department of the Treasury Contact Details: Dr George Rothman ( ), The Retirement & Intergenerational Modelling & Analysis Unit Department of the Treasury, Langton Crescent, Canberra, ACT, 2600 Australia George.Rothman@treasury.gov.au, Website: The views in this Paper are those of the author and do not necessarily reflect those of the RIMA Unit, the Treasury or the Government. This paper has benefited from comments and suggestions provided by various members of the RIMA Unit of Treasury. 1

2 Abstract ASSESSING THE EQUITY OF AUSTRALIA S RETIREMENT INCOME SYSTEM George Rothman Most analyses of Australia s retirement income system address only the accumulation phase; they draw the clear conclusion that the system strongly favours higher income earners as these make most of the concessional contributions and receive a higher tax saving per dollar contributed. This paper places greater emphasis on a whole of life or lifecycle perspective where age pension benefits in the retirement phase are included. The cost to government of its retirement income policies as a whole is modelled using Treasury s relatively comprehensive RIMGROUP model. Variations by income and gender are presented and discussed. There is also some discussion of the appropriateness of the framework and of benchmarks against which concessionality can be assessed. As well as the base findings, examples are given of how the equity impacts of actual or possible policy changes can be measured using this approach. The paper finds that two key retirement income measures in the 2009 Commonwealth Budget have added considerably to the equity of Australia s retirement income system, both by gender and income. The first such measure is the increase in age pension payments and accompanying increase in the income taper rate. The second is the concessional contribution cap reduction measure which is assessed as adding considerably to vertical equity.. This paper complements a sister paper at this Colloquium (by David Tellis) which analyses tax expenditures in the accumulation phase in much more detail than usual. 2

3 INTRODUCTION This paper proposes a framework for assessing the vertical equity of Australia s retirement income system (by lifetime income deciles) and also some aspects of horizontal equity, particularly in respect of gender. The most common approach to such an assessment is through the tax expenditures on superannuation which are generated almost entirely in the accumulation phase. Tax expenditures are defined as tax concessions that provides a benefit to a specified activity or class of taxpayer. Australian superannuation tax expenditures are indeed large aggregating to an estimated $25 billion in (TES 2008). Any distributional analysis of this aggregate will draw the clear conclusion that the superannuation arrangements strongly favour higher income earners as these make most of the concessional contributions, receive the bulk of the tax concessions on earnings and also receive a higher tax saving per dollar contributed. Such an analysis is provided by David Tellis in considerable detail in a paper at this Colloquium. Some limited analysis is in the Retirement Income System interim report of the review of Australia s future tax system (AFTS 2009). This paper does not stop at the accumulation stage but uses a whole of life or lifecycle perspective where age pension benefits in the retirement phase are included. It seems obvious that the age and service pension payments currently aggregating around $34 billion annually should be included in an assessment of system equity. Many analysts lack the data or models to undertake a comprehensive analysis. Treasury s RIMGROUP model is relatively comprehensive and is a suitable vehicle for modelling the cost to government of its retirement income policies as a whole. Variations by income and gender are presented and discussed. As well as the base findings, examples are given of how the impact on equity of important policies or possible policy changes can be measured using this approach. FRAMEWORK FOR THE ANALYSIS A number of issues need to be considered in setting up a suitable but manageable framework for an analysis of the equity of the arrangements for generating the retirement income of Australians. The first is what aspects of a very complex and extensive web of government involvement should be included. As noted above, it is considered essential to include both tax expenditures on superannuation and government spending on age and service pensions. Additionally the government co-contribution towards the superannuation of low and middle income earners is included in this analysis. While not technically a tax expenditure it is a direct contribution towards retirement incomes and is a targeted payment with clear equity implications. It is also topical as there were some temporary changes made in the 2009 Commonwealth budget, is already included in the RIMGROUP framework and easy to include. On the other hand there are numerous exclusions from the analysis, partly for simplicity and mainly because the author does not have the data and modelling capacity to include them. Firstly, the concessional taxation of owner occupied housing is excluded. In an ideal analysis this would be included as the quality of housing clearly rises with income decile. Housing is clearly relevant to standard of living in retirement and can directly provide income through home equity conversions. Secondly the (concessional) taxation of some other forms of saving for retirement outside of superannuation has not been included; there is some interaction as the net level of saving outside of superannuation in counted in assessing the level of age pension paid but no direct allowance has been made. 3

4 Thirdly, government transfers during the accumulation stage such as family benefits Newstart and disability pensions have not been included. With the emphasis being on the retirement income system this seems a reasonable stance to take. Consumption taxes during both accumulation and retirement phases have also been excluded. Finally, there is a grey area which has the capacity to swamp the analysis. The personal income tax system is progressive and applies during both accumulation and retirement phases. It is implicitly in the analysis in the calculation of tax expenditures. It is problematic as to whether to consider personal income tax directly in the analysis. Personal income tax clearly contributes to retirement incomes through at a minimum the substantial government spending on age and service pensions (and aged care). But there is no earmarking of taxes for social security purposes as there is in many countries, and personal income tax contributes to a host of other government expenses such as defence and education as well as retirement incomes. In this paper the primary analyses will not include personal income tax directly but rather a limited number of analyses will be shown which demonstrate the dramatic and overwhelming impact of including personal income tax. As well as coverage of the analysis there are further choices to be made in setting up the analysis framework. The estimation of tax expenditures is not unique. This paper starts with the key elements of its analysis of tax expenditures being the same as the Tax Expenditures Statement published annually by the Treasury (Tax Expenditures Statement 2008). The key commonalities are: Use of an income tax benchmark refined to a superannuation benchmark as in the Treasury Statement. Specifically the income tax benchmark treatment of superannuation is that contributions are taxed like any other income in the hands of the fund member, earnings are taxed like any other investments in the hands of the investor and benefits from superannuation are untaxed. Any costs associated with superannuation investments are deductible under the benchmark. Using the same general framework as in Treasury s TES in calculating the revenue forgone because of concessional treatment while assuming no behaviour change. In practice if the concessional treatment were withdrawn in whole or part, the individual could be expected to seek another tax preferred way of saving for his or her retirement such as negatively gearing shares or rental property this is not allowed for or estimated. The key differences for this part of the analysis are: The emphasis on individual cohorts and inclusion in the analysis of all relevant years of a member of a cohort experience within and outside the workforce; Emphasis on distributional aspects by gender and income; Ignoring some minor aspects of the system such as spouse contributions; By way of comparison the standard published TES covers tax expenditures in a given year not disaggregated by distributional characteristics and does not track through consequences of action taken in earlier years. David Tellis s paper to this Colloquium provides a disaggregation of key parts of the standard TES in given years but does not track through consequences of action taken in earlier years. 4

5 METHODOLOGY The analysis extracts selected cohorts from runs of Treasury s RIMGROUP model tracking their working life and retirement. Treasury s RIMGROUP model is relatively comprehensive and is a suitable vehicle for modelling the cost to government of its retirement income policies as a whole within the limitations discussed above in the Framework Section. RIMGROUP covers the Australian population of working age and older. It starts with population and labour force models, tracks the accumulation of superannuation in a specified set of account types, estimates non-superannuation savings, and calculates tax liabilities, social security payments including pensions and the generation of other retirement incomes. These projections are done for each year of the projection period separately for each birthyear gender decile cohort. The model projections begin in July RIMGROUP is a very large model incorporating 99,600 records, with many variables calculated for each record and with subgroups formed for those with different superannuation accounts and different retirement ages. Nonetheless, it is not an individually based microsimulation and there is some necessary pooling of work experiences, account balances and income levels. For example, unemployment is viewed as a temporary phenomenon and superannuation accumulation is shared by those working and (temporarily) not working 1. Similarly migrants are pooled with others in the model and may dilute the assets of the group they join. Aggregate modelling based on RIMGROUP has been of considerable policy significance, see for example Gallagher (1995), Rothman (1997), Rothman (2007). It has been used in preparing both the First and Second Intergenerational reports (Intergenerational Report, 2007) and in the analysis of implications of the 2009 Budget changes to age and service pensions. More details of the RIMGROUP model and the current set of economic parameters used are in Attachment A. Specifically, cohorts of specified gender and lifetime income decile are tracked over their working life, including particularly the accumulation of superannuation both compulsory and voluntary, recording the payment of superannuation and personal income taxes and allowing for variation in working life, including unemployment and early retirement. At the time of retirement any end benefit taxes are paid and savings outside superannuation are allowed for. Some moderate spending (dissipation) at the time of retirement is allowed for and a variety of retirement income investments are allowed; retirement incomes are then estimated for each year of retirement including importantly any age and service pension eligibility. The key calculations for each cohort are the net present value (NPV) of government payments and net tax expenditures; for a few limited analyses NPV of personal income tax paid is also taken into account. To maximise time in work and allow for a sufficient retirement period within the model cohorts born in 1960 form the basis for most of this analysis. With the new pension age specified in the Budget, these cohorts retire over a range of ages but first receive age pension in RIMGROUP follows the cohorts in detail to 2060 and an allowance is made for pension payments to those still surviving. It is worth noting that in RIMGROUP life expectancy does of course vary by gender and time but does not vary by income decile. 1 But those permanently unable to work through disability are distinguished and treated separately. 5

6 RESULTS Base Case The base case for this analysis is the retirement income framework following the 2009 budget. As is well known, the 2009 budget included significant changes to this framework. Specifically the relevant changes are: A significant increase in age pension payments of $32.49 a week for single pensioners and $10.14 a week combined for couple pensioners; Together with this increase were changes to the income test whereby the pension taper rate for new pensioners will be 50% rather than 40%, as well as a reduction in the effective taper rate for income from employment for both continuing and new pensioners; Gradually increasing the age for eligibility for an age pension beginning in 2017 so that this age reaches 67 in 2023; Reducing the annual cap on concessional superannuation contributions from $50,000 to $25,000 for those aged under 50 and the transitional cap for over 50 s from $100,000 to $50,000: and A temporary reduction in the matching rate of the superannuation co-contribution to 100% in to and then 125% til , reverting to the usual 150% thereafter. Please refer to the Budget papers 2009 for more detail. These changes are all incorporated in the base case assessment of equity using the methodology described above. Table 1 summarises the Net Present Values (NPV) of government spending in this base case by gender and decile. The cohorts considered here are all born in 1960 and are therefore become eligible for age pension in The net present values are in year 2000 dollars as this is the start year for the analysis. To give an idea of scale, the average NPV of spending per person in 2000 is about $115,000 per person. Brought forward to 2027 the average NPV here is around $490,000 in 2027 dollars.. Table 1. Net present value of cost to government of retirement income system for specified cohorts. NPV1-women NPV1-men NPV1-both decile $m $m $m 1 $1,750 $1,600 $3,350 2 $1,700 $1,600 $3,350 3 $1,650 $1,600 $3,300 4 $1,650 $1,600 $3,300 5 $1,650 $1,600 $3,200 6 $1,650 $1,750 $3,400 7 $1,900 $1,800 $3,700 8 $1,850 $1,950 $3,800 9 $1,850 $2,050 $3, $1,800 $2,450 $4,250 all $17,500 $18,050 $35,550 6

7 Chart 1 below is a graphical representation of Table 1. Chart 1 $3,000 Base Case - Cost to Government of 1960 birth cohort: Women relative to men Lifetime Cost to government $m $2,500 $2,000 $1,500 $1,000 $500 NPV1-women NPV1-men $ decile Table 2 expresses Table 1 in proportionate terms and adds additional information on tax expenditures. Table 2. Distribution of NPV of cost to government of retirement income system for specified cohorts and proportion for each decile and gender that is tax expenditure rather than pension. decile NPV1-women NPV1-men NPV1-both TE-women TE-men 1 9.9% 8.9% 9.4% 0.3% 0.4% 2 9.8% 8.9% 9.4% 0.5% 2.9% 3 9.6% 9.0% 9.3% 0.9% 6.8% 4 9.5% 9.0% 9.2% 5.7% 9.1% 5 9.4% 8.7% 9.1% 8.4% 14.9% 6 9.5% 9.7% 9.6% 13.2% 25.7% % 10.0% 10.4% 15.0% 33.3% % 10.7% 10.7% 22.2% 43.3% % 11.5% 11.1% 34.3% 55.7% % 13.6% 11.9% 71.3% 90.4% all 100.0% 100.0% 100.0% 17.7% 32.2% For example Decile 6 women have 9.5% of the total NPV for women (would be 10% if totally evenly distributed) and of this 13.2% is tax expenditure (about 86.8% is pension). Decile 6 men have 9.7% of the total NPV for men and of this NPV a much higher proportion, 25.7%, is tax expenditure (74.3% is pension). 7

8 Table 3 below sets out some summary equity measures that will be useful when later assessing the impact of policy variations from this base case. Table 3. Summary measures of the equity of NPV and tax expenditures NPV1-women NPV1-men NPV1-both TE-women TE-men share of total 49.2% 50.8% 100% 34.8% 65.2% ratio top 5deciles /bottom five deciles ratio top 3 deciles to bottom ratio decile 10 to decile Table 3 uses summary measures to demonstrate the dramatic difference between the more inclusive measure of equity advocated in this paper and a tax expenditure measure. It is important to note that the tax expenditure measure used here includes the government co-contribution; in the absence of this definition the tax expenditure for some lower deciles would be negative rather than small (compare David Tellis s paper this Colloquium). So Table 3 shows that on the recommended NPV measure there is fairly even split of cost between men and women and by income the differences are not large, particularly for women. For example taking the NPV for the top 3 deciles as a ratio of the NPV for deciles 1 to 3, we get 1.08 for women and 1.34 for men. This compares dramatically with the corresponding ratios for tax expenditure at 81.9 and 25.8 for women and men respectively. As many people will be partnered/married, the NPV averaged over men and women is also relevant. Finally in respect of the base case, as foreshadowed in the framework section Charts 2 and 3 show the completely different picture presented when the net present value of personal income tax paid by the cohorts is added to the NPV1 previously calculated to form the NPV2 measure.. Chart 2. Lifetime Cost to government of cohort NPV1 and NPV2, women NPV cost $m decile decile NPV1 NPV2 8

9 Chart 3 Lifetime Cost to government of cohort: NPV1 and NPV2, men NPV cost $m decile NPV1 NPV decile As noted in the framework section, the strongly progressive nature of the income tax arrangements overwhelm the rest of the analysis, Fortunately there are deciles that are a negative cost to government in this framework as there are host of other government expenses to be funded, such as defence and education, as well as retirement incomes. There are of course other tax bases as well. Charts 2 and 3 should be regarded as providing some perspective, but not as part of the mainstream analysis of this paper. Variations from the base case A primary objective of this paper is to provide an analysis framework for assessing the equity implications of changes in retirement income policy. In illustrating this 4 policy cases are considered. Starting with full post budget base above we successively make various changes. The first variation is to remove the budget pension rate increases, together with the income taper change and earned income change Theses changes together create a situation we call v1. To this new policy situation we reverse the changes in the 2009 budget to concessional contribution caps, calling the policy thus arrived at v2. As a further policy experiment we remove from v2 all government co-contributions creating v3. Thus all these cases, the base and v1 to v3, have pension age eligibility at 67 (from 2023 onwards). The summary measures developed in Table 3 are used to assess the impact on equity of these changes. Finally we also compare these situations with the full pre budget case with an age 65 pension age; this case is simply termed pre-bud. Tables 4 and 5 show the impact of the specified variations. Moving to v1 (old lower pension) reduces women s share of the NPV1 measure and moving to v2 (old higher concessional contribution caps) further reduces women s share while moving to v3 has no impact. The pre budget base has a lower share for women than the base but a higher share than v2. In terms of vertical equity moving from the base to v1 ( old lower pension) reduces vertical equity measure and moving to v2 (old higher concessional contribution caps) further reduces vertical equity (a bit more than the move to v1); then the further move to v3 has minimal impact. The pre budget base has lower vertical equity than the base but higher equity than v2. Vertical equity changes are also summarised in Charts 4 and 5 below and the relative impact on vertical equity is a perhaps easier to see in the Charts. 9

10 Table 4a Variations from Budget base of NPV by decile, Women Women all PV1 base lower pension remove lower conts caps remove coconts prebudget base decile base v1 v2 v3 pre-bud 1 9.9% 9.9% 9.7% 9.7% 9.8% 2 9.8% 9.7% 9.6% 9.5% 9.7% 3 9.6% 9.5% 9.3% 9.3% 9.4% 4 9.5% 9.4% 9.3% 9.2% 9.4% 5 9.4% 9.3% 9.2% 9.1% 9.2% 6 9.5% 9.5% 9.4% 9.4% 9.4% % 10.7% 10.6% 10.6% 10.7% % 10.6% 10.5% 10.5% 10.5% % 10.7% 10.6% 10.7% 10.5% % 10.7% 11.8% 11.9% 11.4% all 100.0% 100.0% 100.0% 100.0% 100.0% Table 4b Summary Measures Of Variations From Budget Base Of NPV, Women Women base v1 v2 v3 pre-bud share of total NPV 49.2% 48.3% 48.0% 48.0% 48.4% ratio top 5 deciles /bottom five ratio top 3 deciles to bottom ratio decile 10 to decile Table 5a Variations from Budget base of NPV by decile, Men Men all PV1 base lower pension remove lower conts caps remove coconts prebudget base decile base v1 v2 v3 pre-bud 1 8.9% 8.6% 8.4% 8.3% 8.8% 2 8.9% 8.6% 8.4% 8.3% 8.8% 3 9.0% 8.8% 8.5% 8.4% 8.9% 4 9.0% 8.8% 8.6% 8.5% 8.8% 5 8.7% 8.6% 8.4% 8.4% 8.5% 6 9.7% 9.6% 9.4% 9.4% 9.5% % 10.0% 9.8% 9.9% 9.7% % 10.9% 10.7% 10.8% 10.5% % 11.8% 11.9% 12.0% 11.4% % 14.3% 15.9% 16.0% 15.1% all 100.0% 100.0% 100.0% 100.0% 100.0% Table 5b Variations from Budget base of NPV, Men Men base v1 v2 v3 pre-bud share of total NPV 50.8% 51.7% 52.0% 52.0% 51.6% ratio top 5 deciles /bottom five ratio top 3 deciles to bottom ratio decile 10 to decile

11 Chart 4 Impact of policies on summary vertical equity measures - Women Ratio bud v1 v2 v3 prebud policy ratio top 5 deciles /bottom five ratio top 3 deciles to bottom 3 ratio decile 10 to decile 1 Chart 5 Impact of policies on summary vertical equity measures Men 2.50 ratio ratio top 5 deciles /bottom five ratio top 3 deciles to bottom 3 ratio decile 10 to decile bud v1 v2 v3 pre-bud policy SENSITIVITY ANALYSIS In the projection process which generates the cohort s experiences many judgements need to be made, including on future population, immigration levels and age distributions, participation rates, retirement ages, future expected returns of superannuation funds, and future levels of voluntary contributions. There is also sensitivity to the changes over time of personal income tax rates and the discount rate used in the NPV calculation. 11

12 The demographic and labour force projections are in line with the approach used in the two intergenerational reports as updated for the 2009 budget and have not been varied. Sensitivity to the following variations from the base case has been tested: The discount rate used for the NPV calculations an increase of half a percent in the discount rate is termed the s1 case; Variation in personal income tax rates a one percentage point reduction in the tax rate for each decile is termed s2 ; The choice of cohort an analysis of cohorts born 3 years later is termed s3; The earnings rate for superannuation funds an increase of 1 percent this rate is termed s4. The results of this sensitivity analysis are shown in Tables 6 and 7 below using the summary measures developed earlier. Table 6 Sensitivity of Variations from the Budget base of NPV1 by decile, Women Women base higher discount lower tax rates later cohort higher super returns s1 s2 s3 s4 share of total NPV 49.2% 48.6% 49.5% 49.2% 48.9% ratio top 5 deciles /bottom five ratio top 3 deciles to bottom ratio decile 10 to decile Table 7 Sensitivity of Variations from Budget base of NPV1 by decile, Men Men base higher discount lower tax rates later cohort higher super returns s1 s2 s3 s4 share of total NPV 50.8% 51.4% 50.5% 50.8% 51.1% ratio top 5 deciles /bottom five ratio top 3 deciles to bottom ratio decile 10 to decile Please note that unlike earlier in the paper these are individual variations from the budget base, not sequential changes. The patterns are clear and generally fit a priori expectations of direction. A higher discount rate reduces gender and vertical equity (as it reduces the impact of long delayed payments of pensions which favour lower cohorts). Lower tax rates improve equity (as they reduce tax expenditures going mostly to higher deciles). Selecting a later cohort has minimal impact on gender and vertical equity, apparently marginally improving vertical equity for women and marginally making it worse for men. Consistently increasing the return on super fund investments reduces both gender and vertical equity (as tax expenditures are increased and these go mostly to higher deciles and additionally pensions paid are reduced). These small variations are symmetrical; thus a decrease in the return on super fund investments would improve equity. Importantly the impacts of the variations in Tables 6 and 7 are not large; this suggests that the framework, within its noted limitations, is fairly robust and the findings in the results section do not depend in a major way on which plausible assumptions are made. 12

13 CONCLUSIONS The methodology presented in this paper clearly demonstrates the dramatic difference between considering the equity of superannuation on its own and the more comprehensive approach demonstrated in the paper. Using this more comprehensive approach the retirement income system after the 2009 Budget is found to be more equitable, both vertically and by gender, than many would have expected. If personal income tax is added, this overwhelms the analysis and thus is not generally useful. Though the approach used in this paper is considered much better than considering superannuation alone, it is far from fully comprehensive. Importantly saving through one s own home and other saving for retirement outside of superannuation lie outside the analysis; this tends to understate the cost to government of higher income deciles. On the other hand transfer payments other than age and service pensions are also excluded, and these tend to favour lower income deciles. Consumption taxes during both accumulation and retirement phases have also been excluded. Another important limitation is that in this modelling life expectancy varies by gender and time but does not vary by income decile. Because of such exclusions and limitations, the approximations that need to be made and the range of assumptions that can be plausibly varied, the findings should be regarded as strongly indicative but neither definitive nor final. Sensitivity analyses suggest that the framework, within its noted limitations, is fairly robust and key findings do not depend in a major way on the plausible assumptions made. The paper finds that two key measures in the budget have added considerably to the equity of Australia s retirement income system, both by gender and income. The first is the increase in age pension payments which is analysed together with the accompanying increase in the income taper rate. This measure adds considerably to equity (but with, of course, a considerable increase in government outlays). The concessional contribution caps budget measure appears to add even more to vertical equity with the saving to government impacting mostly on the top two deciles for both men and women. The government co-contribution changes in the Budget are temporary and have minimal impact on equity as measured in this paper. Comparison with the full pre budget arrangements shows a considerable improvement in equity overall coming from the 2009 Budget changes. This is despite the scheduled increase in age pension age having some negative impact as it reduces the amount of pension paid (impacting more on lower deciles). The impact of the increase in age pension age cannot be satisfactorily measured by the methodology in this paper; this is because some of those no longer eligible for age pensions because of the new age limits will receive other transfer payments, such as disability pensions, instead and these payments are outside the analysis framework. 13

14 REFERENCES Australia s Future Tax System, 2009, The retirement income system: Report on strategic Issues, May 2009, Canberra Budget Papers, 2009, circulated by Wayne Swan, Treasurer, and Lindsay Tanner Minister for Finance and Deregulation, May Available at Gallagher, P., 1995, The Policy Use of the Products of the Retirement Income Modelling Task Force, Third Annual Colloquium of Superannuation Researchers, University of Melbourne. Rothman, G., 1997, Aggregate Analyses Of Policies For Accessing Superannuation Accumulations, Fifth Annual Colloquium of Superannuation Researchers, University of Melbourne Rothman, G., 2007, The Adequacy Of Australian Retirement Incomes New Estimates Incorporating The Better Super Reforms, Fifteenth Colloquium of Superannuation Researchers, University of New South Wales Intergenerational Report 2007, circulated by Peter Costello, Treasurer, April Available at Tax Expenditures statement, 2008, The Treasury, Canberra, January

15 ATTACHMENT A: THE RIMGROUP MODEL RIMGROUP is a comprehensive cohort projection model of the Australian population which starts with a population and labour force model, tracks the accumulation of superannuation in a specified set of account types, estimates non-superannuation savings, and calculates tax payments and expenditures, social security payments including pensions and the generation of other retirement incomes. These projections are done for each year of the projection period separately for each birthyear gender decile cohort. The model projections begin in July 2000 and incorporate government policies including the Better Super reforms of July 2007 and changes in the May 2009 Commonwealth budget. Aggregate modelling based on earlier versions of RIMGROUP has been of policy significance, for instance, in Gallagher (1995) and Rothman (1997). Some more details of the RIMGROUP model are given in Rothman (1997) and Gallagher (1995). Strengths and Limitations The strengths of RIMGROUP lie in: The major new parameter research underlying the model in relation to many distributional aspects of superannuation, non-superannuation savings, labour force dynamics and retirement documented in earlier papers (including Bacon (1995)). Research has been carried out on superannuation sectors not previously extensively researched, such as the public sector, self employed and rollover funds. An extensive set of decrements have also been researched to account for losses on job change, disability, hardship and death as well as retirement. A number of significant new data sets have been created as part of this research. For the current projections RIMGROUP has been benchmarked to the latest available ABS distributional data. The comprehensiveness of the model. This includes the integration into RIMGROUP of a full population model, labour force projection model, the endogenous calculation of GDP, an extensive study of retirement,coverage of saving other than superannuation and wide coverage of government payments to beneficiaries and pensioners, together with modelling of taxation, tax expenditures, and national savings. The detail incorporated into the model, particularly the strong distributional framework which distinguishes by superannuation account, age, income and gender. Taxation and government payments are also coded in considerable detail. A wide range of distributional results are available as well as key aggregates. The very long time frame, to June 2060 if required and appropriate. The facility to make changes in all underlying parameters and assumptions including the ability to make direct changes through a user friendly interface to the most frequently changed policy and economic parameter settings. 15

16 The principal limitations of RIMGROUP lie in: in the essential nature of a group model. The model is a very large one incorporating 99,600 records, with a large number of variables calculated for each record and with subgroups formed for those with different superannuation accounts, different ages of retirement and so on. Nonetheless, it is not an individually based microsimulation and there is some necessary pooling of work experiences, account balances, income levels and so on. For example, unemployment is viewed as a temporary phenomenon and superannuation accumulation is shared by those working and (temporarily) not working 2. Similarly migrants are pooled with others in the model and may dilute the assets of the group they join; in macroeconomic linkages being externally imposed rather than endogenous to the model. For example unemployment is exogenously supplied and does not respond automatically to the build up of superannuation or changing retirement rates or other aspects of the economy; and some data which continue to be unavailable in the detail needed. The extensive and demanding data base continues to need maintenance and fine tuning. DEMOGRAPHY AND LABOUR FORCE The base demographic scenario is very similar to the middle scenario as published by the ABS. The labour force scenarios have been generated specifically by RIMAU. Retirement Retirement can be a complicated process whereby full-time workers may pass through a period of part-time work or become a discouraged job seeker before leaving the work force permanently. Operationally RIMGROUP is based on the concept of full retirement, defined as a person leaving the workforce and not re entering it. Despite some considerable data difficulties, retirement has been researched in detail by the RIMA Unit, and a sub-model called RETMOD constructed which provides annual projections of full retirement by gender, age and income decile. Based on these retirement rates, RIMGROUP calculates the number of people retiring each year from each account type and the aggregate value and components of their retirement benefits categorised by the type of retirement (disability or age). Additional to the basic grouping by gender age and income, 12 retirement subgroups are created depending on type of superannuation coverage and age range at retirement, as there are usually significant differences in retirement income and taxation for such subgroups. Retirement benefits are then allocated for each sub-group of retirees to six destinations. These are: Eligible Termination payments (ETPs) dissipated with no impact on retirement income; ETPs invested in interest bearing accounts; ETPs invested in rollover accounts for those under 65; ETPs invested in shares or other assets with likely long term capital gains; 2 But those permanently unable to work through disability are distinguished and treated separately. 16

17 Monies rolled over into allocated pension accounts; and Benefits taken as superannuation pensions or monies rolled over to a complying lifetime annuity. The allocation can be specified by the user. Numbers of Social security recipients and payments to them are projected by the model both in relation to unemployment and sickness benefits during working life and age and disability pensions upon retirement. Thresholds and withdrawal levels associated with Social Security income and asset tests are modelled in detail, with the user being able to specify the type of indexation to be applied to the tests and to base levels of payment. PARAMETER STRUCTURE Parameters which vary by many of the attributes of gender, age, decile and account type are generated as files in a standard format and input through a parameter integration program (which also sets up the basic 99,600 records referred to above). It is expected that these parameters will be varied only infrequently by expert users. Many other parameters of an economic or policy significant nature can be varied readily through a user friendly interface which handles variables which vary by time and/or account type. Examples of variables that can be input through the interface include the returns of various superannuation accounts and retirement investments, rates of compulsory superannuation contributions, inflation, rates of increase in average weekly earnings, various social security and taxation rates and the mode of indexation to apply to them. BASE PARAMETER SETTINGS These parameters are adjusted to historical rates, with a gradual transition over the forward estimates period to the following long term settings: 2.5 per cent per annum for inflation; 4.14 per cent per annum for growth of average full-time wages for a person of given age and gender 3 ; this reflects long term productivity growth of 1.6% pa; 5.5 per cent per annum for the long term bond rate; 6.5 per cent per annum for the average pre-tax return of superannuation funds (after expenses of managing funds but before tax and administrative expenses are deducted separately on a per capita basis); and effective tax rates on the earnings of superannuation funds of 3 per cent for defined benefit funds, 4 per cent for established defined contribution funds, 5 per cent for SG funds and 10 per cent for rollover funds. In RIMGROUP we differentiate between the annual returns for defined benefit funds, defined contribution funds, industry funds and rollover funds. Currently these differences are set at percentage points, with the defined benefit schemes having the highest rates and rollovers the lowest. 3 The actual wage outcome is impacted by demographic and structural change such as the increasing proportion of work which is part time. 17