1Faced by largely the same demographic challenges as the other OECD countries, Sweden opted in 1992/94 for a radical reform of its national old-age pension system. Most of the legislation on the new system was passed in 1998. Parliament adopted the “final” legislation,  providing for the automatic balance mechanism, in May 2001.
2As with most other countries in the process of pension reform, the principal driving force behind the Swedish reform was the current and projected financial deficits that would require a sharp increase in contribution rates unless the pension system was changed. Thanks largely to a substantial buffer fund, the financial stability of the pension system in the early 1990s was probably somewhat better in Sweden than in most other European countries. However, projections showed that under existing rules, together with projected increases in longevity and decreases in size of the labour force, the buffer fund would be exhausted by 2015-2020. The contribution rate required to sustain the system was heavily dependent on the assumed rate of economic growth. It was estimated that with slow growth it would be necessary to increase the combined contribution rate from 18 to 25-30 per cent of the total earnings. Such increases were considered both politically and economically unfeasible.
3Similar financial problems in the pension systems of OECD countries have elicited different political responses. In general, the financial problems are met by changing certain parameters of the pension plan. The most widespread of these parametric changes is to increase the contribution rate. To reduce costs, a frequent parametric change is to increase the number of years required to earn a “full” pension. Another measure taken relatively often is to change the indexation of benefits. In Sweden, the response to the financial challenges noted above was a systemic reform rather than changing the parameters of the old system. In fact, the pension system has been so radically changed that it is somewhat misleading to use the word “reform” for this process. Today, it would be more correct to speak of an entirely new system. 
This article is intended to provide a concise presentation of the principal elements of the new system while at the same time indicating some of the “results” of the changes. A further ambition, which may be rather premature, is to present some of the initial experiences with the new system and how it has been received by the public. While the reasons for the reform strategy chosen are not the primary focus of discussion, some of the considerations in the design of the new system occasionally emerge.
4The principles guiding the reform and the design of the new system are quite simple and clear-cut. The following principles have been fundamental in the decade of research and decision-making on the reform:
- In the new system there should be a one-to-one relationship between contributions paid to the system by or for an individual, and the pension credit of that individual – i.e. no pension credit without a corresponding contribution, and no contributions without corresponding pension credit. The relationship is set in nominal terms, with one krona in contribution equal to one krona in pension credit.
- The financing of pension payments should be guaranteed by a fixed contribution rate.
- The average pension in relation to the average income (here referred to as the pension level) in the new system should be the same as in the old system under the following assumptions: an average of 40 years worked, life expectancy as measured in 1994, and annual growth in average income of 2 per cent. The pension level in the old system is about 50 per cent, while the average replacement rate at retirement is about 60 per cent. 
- There should be a guaranteed minimum pension.
5There is, of course, a social dimension of financial stability in that the financial properties of the system are intimately related to the issue of intergenerational equity.  Financial stability is a necessary, but insufficient, prerequisite, for this difficult goal of the new Swedish pay-as-you-go system (with the exception of the fully funded component of the new scheme, which does not have this objective).
The inter-related concerns of financial stability and inter-generational equity are partly justified by the fundamental change in the nature of pension insurance resulting from the increase in life expectancy in the previous century. As mortality rates have declined, the risk component in pension insurance has diminished significantly, whereas the savings component has come to predominate. With a smaller risk component and a larger savings component, it is more natural, economically efficient and fair to base the size of an individual’s pension benefit on lifetime earnings and/or contributions rather than on late-career income.
The ex post timing of the reform
6The reformation of the Swedish pension system has been a discouragingly lengthy process. The first draft report on a new system was published in August 1992 by a committee that had designated itself the Pension Working Group.  This report was issued as a booklet of 89 pages in the traditional orange cover for publications of government ministries. It had been drafted during a brief period starting at the end of 1991. At that time, following the September general election, a centre-right coalition had succeeded the Social Democratic government. Topping the new government’s agenda was the aim of reaching a broad consensus on securing the financial stability of the public pension system. As it would turn out, all the principal participants from the political parties were already included in this group. One reason why the committee was able to present a solution so quickly was that the pension issue had been thoroughly analysed during the 1980s by a parliamentary committee that had not agreed on any proposals to change the old-age pension benefit provided by the so-called ATP system.  Another reason was probably that the group had an uncharacteristic composition for Sweden; neither unions nor employers were invited to participate in the negotiations.
7After the political agreement and the adoption by parliament of the principles for a new system in 1994, the Pension Reform Working Group transformed itself into the “Implementing Group”. This group was assigned to “nursing the agreement” and implementing the principles that parliament had decided upon. In the main, this group consisted of the same representatives of the five political parties behind the reform proposal, aided by economic and legal experts. All proposed legislation was scrutinized and agreed on, often after lengthy and difficult negotiations, in this group. Until recently, the chairman of the group was the minister responsible for social insurance, an exception indicating the importance attached by the government to the issue. The existence and status of the “Implementing Group” within the Ministry of Health and Social Affairs provided the energy, power and stability that probably kept the reform from collapsing during the difficult period from 1994-1998.
According to the 1994 Government Bill on the principles for a new pension system, the new scheme should have been in operation from 1 January 1996. Because of the problems that arose, however, the legislation did not pass parliament until 1998. The proposals to parliament have consistently been supported by some 85 per cent of its members. This high percentage in favour of the reform is evidence both of the consensus at the political leadership level and of the firm party discipline that is a tradition in the Swedish parliament. Before the proposals reached parliament they were often quite firmly resisted. The process survived these difficulties, and as from 2003 the reformed scheme and its administration are in full operation. 
The ex post timetable of the Swedish pension reform
The ex post timetable of the Swedish pension reform
8The most critical period was the time after the principles for a new pension system had been adopted by parliament in June 1994 and until the principal legislation had been drafted and finally passed in June 1998. During this period, the principles agreed on in 1994 were transformed into legislation. This process, which entailed renewed negotiations within the government and between the political parties that had supported the principles for a new pension system, proved more demanding, both politically and technically, than had been foreseen. Chiefly the delay was caused by the crossfire that the reform was subject to from the political left (traditionalists) within the Social Democratic Party and the Ministry of Finance. The Social Democratic traditionalists disliked the reform and argued that it would make pensions unacceptably low, whereas the Ministry of Finance considered the reform too expensive.
9It is difficult to say how close the opposition was to succeeding in stopping the reform; nobody knows for sure. My guess is that they were quite close, and that only determination, some skilful manœuvring, the strong loyalty within the Pension Reform/Implementing Group, and an element of luck saved the process.
One surprising aspect of the reform process – perhaps second only to its tardiness – is that the end result is so similar to the first draft. This outcome is most likely due equally to the inherent logic of the concepts advanced by the pension reformers, and to their dedication to implementing these concepts.
What has happened – the direction of the Swedish pension reform
10In essence, the scheme proposed by the Pension Working Group was a cross between private and public pension principles. The design was aimed at combining the best features of both – including the financial stability of a defined-contribution system, and the economic efficiency of pay-as-yougo public insurance. The originality of the ideas presented – at least in Sweden, the design of the system is perceived as an innovation – had an additional and politically decisive advantage: since the design was new, it belonged to no political party; thus, there would be no clear ex-ante political winner or loser if the proposed idea was implemented.
11The reform implies a shift from a traditional income-related defined-benefit (DB) system, complemented by a flat-rate system, to two types of definedcontribution (DC) systems, complemented by a guaranteed pension. In the new system, about 14 per cent of contributions will go into individual financial accounts (fully funded), while the remaining 86 per cent will be channelled into the new pay-as-you-go system. The survivor and disability schemes that were a part of the old pension scheme have been transferred to separate systems. Thus, the new pension system is exclusively an old-age pension scheme; in regard to risks, it covers only those related to survivorship. The previous flat-rate pension (folkpension) has been replaced by a so-called guaranteed pension, which provides an additional pension benefit only to persons with a low public earnings-related pension. The guaranteed pension is financed not by contributions, but by general tax revenue. The revenue needed to finance this benefit is equivalent to about 2.5 per cent of total earnings.
12DC systems have traditionally been associated with fully funded schemes. In Figure 2, DC schemes are represented by quadrants I and II. It may be argued that quadrant I does not really represent a genuine DC system, largely because the pension liability is not backed by funded assets and hence the return on contributions will differ from the market return on capital. To distinguish between DC systems that are fully funded and those that are financed on a pay-as-you-go basis, the latter are often called Notional Defined Contribution (NDC) systems.
National pension schemes have universally been DB and financed almost entirely on a pay-as-you-go basis. Schemes so designed are found in quadrant III of Figure 2. Furthermore, public pension plans have normally covered both survivorship and disability risks.
Four generic types of pension systems and the direction of the Swedish reform
Four generic types of pension systems and the direction of the Swedish reform
13The demographic and economic developments that may lead to changes in the contribution rate or the value of pensions can be treated on a national level as uninsurable risks. In principle, a DB system should accommodate uninsurable risks by altering the contribution rate. In practice, however, DB schemes are known to manage uninsurable risks by also adjusting the value of pensions through changes in the rules of the system. In a DC scheme (whether funded or pay-as-you-go) accommodation must take the form of adjusting the value of pensions. As a response to a deficit, increasing the contribution rate is not viable in the long run. 
14It might be considered a disadvantage of an NDC system that economic and demographic risks must be accommodated by changes in benefit levels. But this very feature might just as well be considered the strong point of this type of pension scheme. The inflexibility of an NDC design fosters discipline and forces legislators to manage existing risks explicitly since the system is able to produce information about those risks. Both the need for and the availability of information on the effects of economic and demographic developments are greater in an NDC scheme than in a traditional pay-asyou-go DB pension plan. For example, with the new pension system, annual economic growth in Sweden has become a matter of individual financial interest for the large and politically important group of retirees. This was not the case with the old DB scheme. Furthermore, the need to postpone retirement as life expectancy increases, or to accept lower replacement rates, is quite explicit in the new scheme.
15Pensionable income consists principally of all annual earned income exceeding the minimum taxable income (?9,000 SEK) but below the maximum taxable income (306,750 SEK in 2003). Also included in pensionable income are all social-insurance benefits in the nature of income replacement, such as sickness, unemployment, disability and maternity/paternity benefits. In addition, pension credits are also granted for “childcare years”, university studies and compulsory national service. Pension credits for childcare years can be credited to men as well as women. This provision was included in the reform since it was judged both fair and politically necessary to minimize or eliminate the negative impact that the change from the DB formula to the life-income principle of an NDC formula otherwise would have had for women.
Pension credit and pension contribution base in 2001
Pension credit and pension contribution base in 2001
16The pension credit for income-replacement benefits from the social-insurance system, e.g. for childcare years, is financed annually by corresponding transfers from the national budget to the buffer fund. At present about 17 per cent of all contributions and pension credits derive from such transfers.
Transition from the old system to the new
17Individuals born in 1937 or earlier will have their pensions calculated according to the old rules. For individuals born in 1938-1953, pensions will be calculated according to the rules of both systems. Persons born in 1938 will receive 16/20 of the pension they would have drawn from the old system; the remaining 4/20 of their pension will be from the new system (4/20 of pension credits in the old system as from 1960 have been converted to pension credit in the NDC system). Those born in 1939 will receive 15/20 from the old system and 5/20 from the new, and so on. Thus, individuals born in 1953, the last transitional year, will receive 1/20 of their pensions from the old system and 19/20 from the new. For persons born in 1954 or later, all accumulated pension points in the old system have been converted to pension credit in the NDC system.
18Since 2002, the consumer-price indexation of pensions in the old system has been replaced for all income-related pensions by the new income indexation, i.e. the change in nominal average income, minus 1.6 per cent.
The retroactive recalculation of pension points in the old ATP scheme to pension credit in the new NDC system, together with the change from CPI indexation of ATP to the new indexation based on growth in average income minus 1.6 per cent, entails a rather rapid transition from the old to the new system. Since 31 December 2002, 72 per cent of the liability to the economically active population has derived from the new system and only 28 per cent from the old.
How does the notional DC system work?
19In the NDC system, the equivalent of 16 per cent of each insured person’s annual pensionable income will be credited yearly to his or her notional account. The corresponding amount will be transferred in monthly instalments to the system’s buffer fund, which finances pension payments. The “interest” earned on the notional account is either the increase in nominal average (pensionable) income as measured by an income index or an approximation of the internal rate of return in the system, as measured by the balance index, explained in the section below, “Guaranteed pension – a remaining DB component”. Each year, the indexation of the notional accounts will be augmented by inheritance gains and reduced by administrative costs.
20There is no formal retirement age in the new system. Pension credits will always be earned and added to the notional (as well as financial) accounts if the individual has pensionable income, regardless of his or her age and of whether the individual has begun to draw a pension. Pension can be drawn at 25, 50, 75 or 100 per cent beginning at any time chosen by the individual after reaching the age of 61, i.e. without any upper age limit.
21Pensions from the pay-as-you-go system are calculated at the time of retirement by dividing the notional-account balance by a so-called annuity divisor. The annuity divisor reflects the unisex life expectancy at retirement. This life expectancy is measured annually as a five-year moving average. Further, the annuity from the notional account is calculated at an interest rate of 1.6 per cent. The pension is subsequently indexed by the growth in the income index, or the balance index, minus this interest rate. The interest rate of 1.6 per cent is imputed in the conversion to an annuity to achieve a more even distribution of real income during the retirement period. As pensions are annually indexed by the increase in nominal average wages minus 1.6 per cent, or more correctly, divided by 1 016, they will be fixed in real terms only if nominal wages increase by exactly 1.6 per cent more than the inflation rate.
Since 2002, this new type of indexation by the increase in the nominal average wage divided by 1 016 has also applied to pension benefits calculated according to the old rules. Under those rules, in effect since the ATP scheme was introduced in 1960, pensions were indexed by the change in the Consumer Price Index (CPI). From figure 4 it is clear that retirees have now benefited for two years from this change in the manner of annually recalculating pensions. In 2002 pensions increased in real terms by 0.6 per cent  and in 2003 by 1.8 percent.
Actual and Forecast Annual Changes in NDC-significant Indices 2000-2006
Actual and Forecast Annual Changes in NDC-significant Indices 2000-2006
22However, the forecast for 2004 indicates that the income index will increase by only 1.0 per cent more than CPI; thus, the real value of pensions will fall by approximately 0.6 per cent. The projection for 2005 and 2006 also show that in each year the indexation of pensions will be 0.1 percentage point less than the CPI. Nevertheless, projected public earnings-related pensions will still be higher in each of the years 2002-2006 than they would have been if the old system had been retained. It is a bit of a paradox that the Swedish system has been made financially stable and that as a consequence pension benefits – so far – have increased relative to their level had the old scheme been retained.
As noted, in a DC as well as an NDC scheme, the pension level is a function of the system’s real or “implicit” rate of return and is thus in a sense unpredictable. This feature has been advanced as an argument in favour of the reform. In fact, it is impossible to say that the reform by necessity will lead to lower pensions; whether pensions will decrease or increase after the reform is a function of demographic and economic developments. Reformers have argued that if the DB scheme had required either higher contribution rates or benefit cuts, the response would have been benefit cuts, since the economy would not sustain increases in the contribution rate. It is preferable to implement those benefit reductions through timely, “transparent” and co-ordinated measures automatically taken within an NDC scheme that – unlike a DB scheme – will also automatically pay higher pensions with a fixed contribution rate when the economy can afford it. To put it more crudely, if a DB scheme is reformed into a new DB scheme, it is rather easy to determine who would gain and who would lose from such a change. With a shift from a DB to an NDC system, it is not so easy to identify winners and losers.
Annuity divisors – the main source of financial stability
23Final annuity divisors, one for every age over 64, are determined for all individuals in the same birth cohort in the year when the cohort reaches 65. If life expectancy increases, the same notional capital will produce a successively lower yearly pension for younger cohorts if conversion to an annuity (pension) is made at the same age. To maintain a fixed pension level when life expectancy increases, retirement must be postponed. Table 1 shows the increase in remaining life expectancy at age 65 currently (2003) projected by Statistics Sweden. In addition, the table shows the projected effect on pensions with an unchanged retirement age, as well as the change in the retirement age required to keep the pension level fixed.
24In the NDC scheme, the annuitization divisor is the principal feature that promotes financial stability. Swedish pension reformers have had relatively few problems in defending this actuarial and pedagogical method of achieving financial stability and, it may be argued, inter-generational fairness. Like all DB public pension plans, the previous Swedish DB pension system tended to subsidize early departure from the labour market. As life expectancy increases, so does the size of the subsidy. The abolition of such automatic increases in subsidies, with their distorting effects on the economy, is one important aspect of the Swedish pension reform.
Effect of projected increase in life expectancy on pension levels or retirement age
Effect of projected increase in life expectancy on pension levels or retirement age
The buffer fund of the NDC system
25The NDC scheme has a rather substantial buffer fund, which actually consists of four different National Pension Funds. The primary purpose of the funds is to serve as a demographic buffer so that mere variations in cohort sizes, in conjunction with the fixed contribution rate, will not cause the value of pensions to differ between different cohorts.
26Each buffer fund has its own board appointed by the government. In investment policy, these funds are obliged by law to consider only what is optimal from the standpoint of the insured. The funds were established as of 1 January 2001, at which time they inherited from the “old” buffer funds a capital of 540 billion SEK (?54 billion euros), apportioned equally among the funds. Although the ratio of total assets to expenditure was almost 400 per cent, in relation to the pension liability the buffer fund was no larger than 10 per cent.
27The pension reform has entailed significant liberalization of the investment policy of the funds. From 1 January, a minimum of only 30 per cent of assets has had be invested in bonds. In other words, up to 70 per cent of assets may be invested in equities. A maximum of 40 per cent of assets may be exposed to currency-exchange risks. There are some additional restrictions on fund investment policy that need not be mentioned in the present context.
28In spite of their independence, the buffer funds have all chosen similar investment strategies. In the spring and summer of 2001, they adjusted their portfolios to their perceived long-term optimal investment mix. Roughly this mix has been judged to consist of 60 per cent equities and 40 per cent bonds. In absolute terms, investment performance so far has been extremely poor, with total losses of 25 billion SEK (minus 5 per cent) in 2001 and 85 billion SEK (minus 15 per cent) in 2002. In relative terms, i.e. relative to the market in general and to their respective reference portfolios, the buffer funds have performed marginally better. The total size of the fund as of 31 December 2001 was 487 billion SEK, and the fund-to-payment ratio had dropped to 320 per cent. The possible effects of these losses and the debate thereon are briefly discussed below.
A transfer of some 250 billion SEK from the fund to the national budget was made in 1998-2000. This transfer was a form of compensation from the pension system to the central-government budget for the increased costs initially incurred as a result of the reform. The national debt was reduced by the assets transferred (government bonds). The size of the initial buffer fund was determined by calculating how much could be transferred without triggering the automatic balance mechanism (see section below on “Automatic balance”) in any year during the period 2001-2050, under the assumptions of a specific scenario. In brief, this scenario, which was developed from the demographic forecast by Statistics Sweden, was based on the assumptions of “recent” age-related income patterns, 2 per cent annual growth in average income, and a real return of 3.25 per cent on the buffer fund.
The system of financial accounts
29The contribution base in the DC system of financial accounts, or the FDC system, is identical to that in the NDC system. Contributions to the two systems are collected together. When  the amount of contributions for the taxable year is definitely established, the Premium Pension Authority (PPM), which is the agency charged with administering the system of financial accounts, distributes the contributions as a lump sum to the mutual funds chosen by the insured.
30In principle, any investment fund that complies with the UCIT directive and accepts the fee structure and contractual demands set by the PPM is accepted in the premium-pension system. As of 31 December 2002, the FDC system included 644 different funds, managed by 87 different fund managers. The insured may choose a maximum of five different funds, and the number of fund changes is unrestricted. The average number of funds for individuals who actively chose funds is 3.7. Contributions of individuals who do not respond when requested by the PPM to choose a fund are invested in the default fund(s), which is (are) always the latest choice of funds. The capital of each individual is increased by annual contributions and the return on investment. In addition, inheritance gains are credited and administrative costs are deducted, both on a yearly basis.
31At retirement, the insured can choose to convert his or her capital either to a fixed-interest annuity or to a variable annuity. In the former case, the PPM assumes/distributes the risk that the actual life span of the cohort will exceed its estimated life expectancy, as well as the investment risk. In the latter case, the monthly pension is re-calculated every year, depending on the return on investment of the funds chosen by the individual and on changes in the life expectancy of the individual’s birth cohort. Thus, in this case the insured cohort assumes the longevity risk, while the insured individual assumes the investment risk.
All information between the insured and the fund is transmitted through the PPM; in fact, the funds generally do not know who their individual customers are. From their perspective, the PPM is the customer. The PPM negotiates fund fees and acts as a clearinghouse within the system. The total cost of operating the system is presently estimated at approximately 0.75 per cent of assets, with the PPM costing 0.3 per cent. It is assumed that the total cost of operating the system will decline as the system matures, perhaps to about 0.25 per cent of assets. Different funds charge different fees, and PPM negotiates rebates from the funds to the insured individuals.
To choose or not to choose
32In the autumn of 2000, over four million Swedes were requested to choose a fund. Some 66 per cent responded to this request. Since that year, such a request is sent every year to all individuals that have made their first contribution (2.5 per cent) to the system, but the percentage of active choosers has dropped rapidly, as shown in Table 2.
Rate of active choice of fund management
Rate of active choice of fund management
33One possible reason for the dramatic fall in the number of active choosers is that the first round of selection in 2000 involved investing the accumulated contributions of four years (1995-1998); thus, the capital concerned was substantially larger in 2000 than in subsequent years. In addition, the “opportunity” to choose a fund manager was much more intensely advertised in 2000. The fact that the average age of those choosing for the “first” time was considerably lower in 2001-2003 than in 2000 may also have contributed to the overall decreased interest in making an active choice, as may the dramatic fall in equity prices since 2000. Except for this first year, women have been slightly more active than men in choosing funds.
The assets of the initial “non-choosers” were transferred to a separate default fund, called the Seventh AP fund, which is an independent government agency. This fund, which has broad discretion in the allocation of investments, has invested about 90 per cent in equities and 10 per cent in bonds. All insured individuals can make an active choice at any time during the saving phase; once an active choice is made, the individual will thereafter be treated as an active chooser, the latest selection of funds always being the default choice. The level of activity by the insured in terms of fund changes is very low, even though there are virtually no restrictions on changing fund(s) or costs of doing so.
Enormous losses in absolute terms – small losses in relation to the total public notional pension capital
34Both the funds chosen by individuals and the default fund have invested some 90 per cent of their assets in equities.  In conjunction with the fall in equity prices since 2001, this investment strategy has led to loss of some 40 per cent of the capital invested in the premium-pension system. It is a challenge to find anything positive to say about this development. One possible source of limited comfort is that the system and the capital invested in it are still very small in relation to the notional asset of the insured in the pay-as-you-go system, which increased rapidly during the same period because of positive growth in average income. Had the return on premium-pension capital been zero instead of minus 40 per cent, this capital would have represented 1.8 per cent of the combined real and notional capital of the insured in the public pension scheme. Thus, the actual loss of 40 per cent represents less than one per cent of total notional capital.
35The losses have, of course, cast a shadow on the reputation of the new system and on the entire reform, especially as a large share of the public perceives the pension reform as synonymous with the small funded component. In general, however, there seems to be substantial broad acceptance of the price volatility in the equity market, and the negative trend has caused little commotion so far. Furthermore, since the capital in the system of financial accounts is still quite limited, the losses have had only marginal effect on projected pensions.
The losses sustained by the buffer funds of the pay-as-you-go system caused the ratio of buffer fund assets and contribution flow to pension liability, or the balance ratio, to decrease from 1.03 to 1.01 between 2001 and 2002. This reduction in solvency has heightened the risk that the so-called balance mechanism (see section on “Automatic balance”) will be triggered. In that case, the indexation of pension balances and pension benefits will be lowered, at least temporarily. As buffer-fund losses have only reduced solvency by some two per cent, their psychological effect is probably greater than their economic effect in terms of a greater risk that pension benefits will be cut. Psychology, or the image the insured have of the new system, is probably as important as economic reality, and the substantial losses of the buffer funds have been an unpleasant experience for pension reformers.
Guaranteed pension – a remaining DB component in the new system
36Persons with low lifetime incomes will have accumulated only limited notional and financial capital. These individuals may be entitled to a supplement guaranteeing them a minimum pension, called the guaranteed pension. The guaranteed pension is financed by general tax revenue and not by contributions to the earnings-related scheme.
37To be eligible for a guaranteed pension, an individual must be at least 65 years of age and a resident of Sweden or another EU/EEA country with which Sweden has signed a convention. For a “full” guaranteed pension, the individual must, in principle, have resided for 40 years in Sweden after reaching age 25. The guaranteed pension is designed as a supplement to the public earnings-related pension.  For single persons with no public earnings-related pension, the guaranteed pension is 2.13 price-related base amounts; for married couples it is 1.90 price-related base amounts per person.  The guaranteed pension is reduced by 100 per cent of any public earnings-related pension received up to 1.26 price-related base amounts (for married couples up to 1.14 per spouse). The guaranteed pension is also reduced by 48 per cent of any earnings-related pension exceeding 1.26/1.14 price-related base amounts. The guaranteed pension thus drops to zero for persons with a public earnings-related pension of 3.07 price-related base amounts (2.72 for married couples) or more. The coverage of the guaranteed pension is illustrated in the diagram below.
Yearly earnings-related and guaranteed pensions, in price-related base amounts. In 2003, one price base amount = 38,600 SEK (?3,860 euros)
Yearly earnings-related and guaranteed pensions, in price-related base amounts. In 2003, one price base amount = 38,600 SEK (?3,860 euros)
38The guaranteed pension is indexed yearly by the change in the Consumer Price Index (CPI). By implication, the total pensions of those also receiving a guaranteed pension will differ from those of persons who receive only an earnings-related pension. If nominal average income grows at a rate that exceeds the CPI by more than 1.6 per cent, the total pensions of those with some guaranteed pension will increase more slowly than the pensions of those with only an earnings-related pension. On the other hand, if nominal average income grows at a rate exceeding the CPI by less than 1.6 per cent (i.e. a situation where the indexation of the earnings-related pension is less than the CPI and pensions lose in real value), the total pension of those with some guaranteed pension will grow faster. Individuals with earnings-related pensions below 1.26 (1.14) price-related base amounts will not be affected at all by slower or faster growth in average income or by the automatic balancing mechanism. As the guarantee pension is only tested against the size of the public earnings-related pension, it is not a means-tested benefit in a traditional sense.
The dilemma of guaranteed pension – work incentives
39Persons born in 1938 are the first to be eligible for a guaranteed pension according to the new rules. Of those who were born that year and have retired between January and April of the current year, some 30 per cent are entitled to a guaranteed pension. Since many of these individuals receive only a small guaranteed pension, the cost is minor as a share of their total earnings-related pension. The percentage is slightly less than what was projected in 1998. The total cost of the guaranteed pension in 2003, both for persons born in 1938 and for those born earlier, is estimated at 16 per cent of the cost of the earnings-related pension. If the price indexation of the guaranteed pension is unaltered, and the average income grows in real terms, both cost of this pension and the number of persons receiving it will decline. Even with moderate rates of real growth, the cost decrease is rapid.
40The size and importance of the guaranteed pension have been much debated. According to some, so many persons will receive a guaranteed pension that the desired reduction in labour-market distortion – one objective of the pension reform – will be neutralized. This argument is given added weight by the relatively low ceiling on pensionable income. Others argue that the guaranteed pension is too low to provide a decent standard of living.
41Unfortunately, there is considerable validity to both arguments. With the guaranteed pension, especially in combination with housing allowances and income-related charges of nursing homes and for other forms of care, it is indeed impossible for many low-income workers to improve their standard of living as retirees by working more or by earning higher pay. For these persons, most or all of the contribution to the earnings-related pension system will effectively constitute a tax. This fact is an unpleasant one for Swedish pension reformers, who have worked very hard – and successfully – to maintain the actuarial character of the earnings-related scheme.
However, the problems of disincentives to work are not new. They were equally serious in the old scheme and have been inherited by the new one. Indeed, there is an unavoidable dilemma: either a significant number of people will be relatively poorer, or a more “generous” system will make it more difficult for relatively poor persons to improve their economic situation by working more. The choice of linking the guaranteed pension to a price index may reflect the view of pension reformers as to the more pressing problem in Sweden today: namely, that social insurance, taxes and other provisions make it difficult for the relatively poor to improve their economic situation.
Annual information – Sweden’s “orange envelope”
42All insured persons over 26 who have not started to receive a full pension receive an annual statement containing joint information on the changes and current status of their notional and financial account. This statement shows how the latest closing balance is derived from the closing balance in the statement of the previous year, new pension credit in SEK earned in the past year (i.e. the contributions paid by the insured, the insured’s employer and in some cases by the government with general tax revenue), the amount of indexation in SEK, the amount of inheritance gains in SEK and the reduction for administrative costs. At retirement the annuity divisor is applied to this notional pension capital to calculate the pension.
The annual statement also presents projections of the insured’s public pension, i.e. NDC plus fully funded pension and any guaranteed pension. The projection shows what the pension would be for three different retirement ages – 61, 65 and 70 – and for two different rates of growth in average income – zero and two per cent. In the zero-per-cent scenario, the real return on the funded pension is assumed to be 3.5 per cent, net of administrative costs. In the two-per-cent growth scenario, this net real return on the funded pension is assumed to be six per cent. The latter assumption especially has been criticized as over-optimistic. Defenders of the assumption argue that the real return of six per cent represents the historical average for a portfolio of 90 per cent equities and 10 per cent bonds – roughly the present investment “strategy” of those participating in the funded scheme. In a sense, this projection is auto-corrective since it always starts with the actual value of the capital in the funds as well as the actual value of the notional capital in the NDC system.
Anyone interested in pensions? Proportion in per cent*
Anyone interested in pensions? Proportion in per cent*
43Since 1998, the National Social Insurance Board (RFV) and the Premium Pension Authority (PPM) have tried to measure some aspects of public opinion and awareness concerning the new pension system. Each year, some 1,000 people are interviewed by telephone. Among other things, they are asked whether they remember having received the annual information (“the orange envelope”), whether they have read it, what they have read, etc. Table 3 presents the findings on some of the survey questions.
Administrative issues and costs
44A new IT system has been established to manage the records and benefits of both the old and new pension systems. It has been a large, costly and apparently successful project. Since 1 January 2003, all pensions are calculated and paid with the new IT system. A substantial share of the costs would probably have been incurred even without the reform, as the old IT system needed updating.
45The funded scheme also required a new IT system. The first attempts to design one failed, leading to a legal dispute with the consultant involved. This problem delayed the launch of the system by over a year and caused considerable debate and criticism. However, the system that eventually was built has been functional since its start, and this issue now seems more or less settled.
46The cost of administering the NDC system is presently some 0.05 per cent of notional capital, or 0.9 per cent of contributions. Half of this cost is for buffer-fund management and half for insurance administration. The insurance administration of the fully funded scheme is estimated to cost some 0.3 of the accumulated capital. Fund-management costs are some 0.4 per cent of accumulated capital. As mentioned, the high ratio of these costs to assets is expected to drop substantially, to about 0.25 per cent, as system assets grow. Unfortunately and rather surprisingly, there has been very little debate about the high costs of administering the FDC system. Hopefully there will be more debate.
Automatic balancing and the annual report of the NDC system 
47One primary objective of the NDC system is to provide for a stable relationship between the average pension and the average income. Consequently, the indexation of both notional capital and pensions is based on growth in average income. However the growth in average income may deviate from the internal rate of return of the NDC system. Indexation by the change in average income will imply a return on “notional” capital and pensions higher than the internal rate of return of the NDC system if, for example, the number of persons working decreases. This will be the case whether the fall is due to a lower rate of labour-force participation or to demographic factors, such as low nativity.
48There are a number of other reasons why the indexation of notional pension capital and pensions by the increase in average income may be higher (or lower) than the internal rate of return in the NDC system. For example, if life expectancy increases, the annuitization divisors will, ex post, prove to have been calculated too generously. Another example, and one of current interest, is the case where the return on buffer-fund capital is less than the increase in average income. The loss sustained by the buffer fund in 2002 was the main reason why the system’s internal rate of return was less than the indexation of the pension liability this year.
49The automatic balance mechanism was developed to make the system fully stable financially, while indexing the pension liability by the averageincome index as frequently as possible. The balance mechanism consists of rules for annual calculation and presentation of the assets and liabilities of the pay-as-you-go system, thus providing that system with a balance sheet. The valuation of the “asset” represented by the contribution flow is determined with the aid of a concept termed expected turnover duration – to my knowledge, this concept is new. Turnover duration is a measure of the expected turnover time of the pension liability. The formula for calculating the assets and liabilities of the system is prescribed by legislation. Aside from the buffer fund, which is valued on the basis of capital-market transactions, the calculation used in the automatic balance mechanism is based exclusively on transactions that are recorded in the pension system. There is thus no element of forecasting in the calculation of the balance sheet. The financial status of the scheme is summarized in the balance ratio, i.e. buffer-fund assets and value of contribution flow over pension liability.
If and when liabilities should exceed assets, the basis for indexation automatically changes to an approximation of the system’s internal rate of return, thus automatically adjusting pension levels as well. Figure 6 illustrates how balancing works in a scenario where it is first activated and later discontinued. 
Income index and balance index
Income index and balance index
50With growth in average income as the “main” index and intervention in the system only if and when necessary to secure its financial balance, the volatility of the pension level relative to other design options will be minimized. The asymmetric design of the indexing of the new Swedish pay-as-you-go system has been made possible by determining the relevant time aspect of the system. The inverse of expected turnover duration can be regarded as the discount rate for valuing the flow of contributions to a pay-as-you-go pension system.
51The balance mechanism ensures the financial stability of the NDC and provides for what might be called actuarial accounting, a form of double-entry bookkeeping for a pay-as-you-go pension system. The method was used in the Annual Report of The Swedish Pension System for 2001 and 2002. These publications show a new way of presenting information about the financial status and development of a public pay-as-you-go pension system. The reporting closely resembles that of a private insurance company on its operations. The Annual Reports of the Swedish Pension System are available on: www.rfv.se/english/publi/index.htm.
52As yet there has been no ambitious attempt to evaluate the results of the Swedish pension reform. Indeed, any serious effort to so at this point would be premature. However, on some technical issues definite conclusions can be drawn. The pension reformers have succeeded in creating a system where in principle every contribution made is reflected by the equivalent in pension credit and where no pension credit is granted without a corresponding contribution. The exceptions to this rule are very minor and attributable to administrative imperfections. In addition, the objective of financial stability has been met – probably to a greater extent than was thought possible in 1994. Also fulfilled is the conditional promise that the average pension in the new system would equal the average pension of the old in a scenario of 2 per cent average growth, 40 years of work and the same life expectancy as in 1994.  The Annual Report of the Swedish Pension System provides evidence that this objective and the one of financial stability have been attained.
53Even if the new pension system, for some reason, were abolished in the near future – a very unlikely event – I consider it fair to say that the reform has been a major political achievement. The five political parties succeeded in abolishing a very popular social insurance programme that had proved unsustainable and economically detrimental, and in replacing it with a logically coherent and financially sustainable structure. Normally, logic and coherence do not characterize products of political negotiations; the Swedish pension reform is an exception in this respect. The reform has also been successful in terms of administrative implementation, which has taken a long time and cost a lot of money, but nevertheless proceeded without serious setbacks. Perhaps some of the innovations in the scheme represent intellectual advancements as well.
54Indeed, the Swedish pension reform exhibits so many exceptional features that its very occurrence is astonishing. Many of the delegations from pension-burdened countries that have visited Sweden to study the new system are justly surprised over what has happened. In their efforts to understand the Swedish process, neither these visitors nor Swedish commentators have realized that the pension reform here was more than just a matter of cutting costs. To an equal or perhaps greater extent, the reform was about improving a flawed public pension system. In this respect, the Swedish pension reform represents an intelligent repositioning of the government’s role in providing old-age pensions. This repositioning was both defensive and offensive, a synchronized contraction and expansion of the role of government. With a complex but coherent reform strategy, the pension reformers prevailed over their many political opponents on both left and right. The financial margin necessary for the reform of the pension system was present partly because the new scheme was designed in the 1990s, when the expansion of pension costs had been slow, and partly because Sweden had accumulated a substantial buffer fund in its pay-as-you-go pension system.
However, the reformed plan may be considered a success only if it survives for a long time. As yet there are no tangible threats to the reform, but future troubles are not difficult to envisage. The principal danger probably lies in the still rather tacit unpopularity of the new scheme with the public. To survive, the new pension plan must gain acceptance by at least a critical mass of the population; its supporters must become more numerous than the handful of politicians and experts that designed the new scheme, a small number of approving editorialists and a few economists. In this regard, one of the best features of the new scheme, its transparency, may in fact work against it. As media tend to focus more on bad news than on good, there is a danger that the information communicated will have a negative bias, even if on average there is a balance between good and bad news. The first two years of the system provide ample proof of this “negative spin”. Even though the dramatic fall in equity prices has reduced the buffer fund of the system by a full quarter, the “solvency” of the NDC system has decreased by only two percentage units. The media focus on the lost quarter has been striking; the much more limited impact on solvency has been a well-kept secret. The surge in average wages over the past year has increased the value of the notional accounts by two and a half times the amount of the losses sustained by the buffer fund, raising the real value of pensions for the first time in Swedish history. This good news has hardly been reported at all and is thus unknown to the public. In the end, the success or failure of the new Swedish pension system may depend on the skill of its administrators at communication – an area in which they have little experience.
56A number that reflects the average remaining life expectancy at retirement, discounted by an “interest” rate of 1.6 %. In the calculation of the annual inkomstpension benefit, which is made at the time when this pension is first withdrawn, the individual’s pension balance is divided by the annuitization divisor. Because of the interest credited at 1.6 %, the annuitization divisor at the time of retirement is always less than the remaining average life span.
58A method of indexing the pension liability to ensure that the disbursements of the inkomstpension system will not exceed its revenue in the long run. If balancing is activated, the pension liability increases at a compounding rate approximately equal to the system’s internal rate of return.
60Here, income as measured by the income index.
62The assets of the pay-as-you-go system – i.e. the contribution asset and the buffer fund, divided by the pension liability of the system. The balance ratio can be considered equivalent to the consolidation ratio of a funded system. Unlike the consolidation ratio, however, the balance ratio provides no information on the amount of funded assets in relation to the pension liability.
64Absorbs interperiod discrepancies between pension contributions and pension expenditure in a pay-as-you-go system. The primary purpose of a buffer fund is to stabilize pension levels and/or pension contributions against economic and demographic fluctuations.
65The buffer fund
66of Sweden’s national pension system comprises the First, Second, Third, Fourth and Sixth National Pension Funds. Legally and administratively, the buffer fund of the pay-as-you-go system thus consists of five different funds. Pension contributions are apportioned equally to the First-Fourth National Pension Funds, which also contribute equally to the payment of pensions. A reason for dividing the pension capital among four initially identically large funds, with an identical mission and identical regulations is to limit the concentration of power that otherwise would have resulted. Another reason is that such a division might increase risk diversification and competitiveness in fund management. The much smaller Sixth National Pension Fund receives no pension contributions and pays no pensions and it can be considered a kind of anomaly within the new pension system inherited from the old system. From the standpoint of the pay-as-you-go system, the five buffer funds may be viewed in some respects as a single fund. Information on the buffer funds is available on: their respective websites at: http://www.ap1.se, http://www.ap2.se, http://www.ap3.se and http://www.ap4.se.
67Defined-benefit pension systems
68Pension systems in which the insurer bears the financial risk of possible variations over time in mortality and the rate of return on the assets of the system. In a public pension system, the insurer is the taxpayer, which means that the contribution to the system may vary.
69Defined-contribution pension systems
70Pension systems in which the insured bears the financial risk of possible variations over time in mortality and the rate of return on the assets of the system. Consequently, the value of pensions may fluctuate. A supplementary definition is that in a defined-contribution pension system pension credit accrues by the same nominal amount as the contributions paid by or for the individual. In principle, pension credit accrues at the time when a contribution is paid into the system.
72Beginning in 1999, the incomes in the income index (U) consist of pension-qualifying income (incl. income in excess of the ceiling on pension-qualifying income and income in the form of sickness and activity allowances), less the national pension contribution, earned by persons aged 16-64. The sum of these incomes is divided by the number of persons who have earned them. If the income index for year t is designated I (t), then I (t)
74Ut?1 = forecast of the average income for year t?1
75Ut?4 = settled average income for year t?4
76CPIt?1, CPIt?2, CPIt?4 = consumer price index for June in years t?1, t?2, and t?4
77k1 = correction factor for subsequent (more exact) forecast of income in year t?2
78k2 = correction factor for difference between actual and forecast income in year t?3.
79To reduce the volatility of the index, a three-year moving average is used. To counter the slow compensation for changes in inflation rate that this smoothing otherwise would cause, the inflation rateduring the three-year period is derived from the change in nominal average wages, and the change in the CPI in the last year is subsequently added.
81Here, the term indexation is used both for the annual “interest” accrued on the notional pension balances (pension accounts) and for the annual revaluation of the pension benefit. The indexation of the pension balances is based on the income index or, if the balance mechanism is triggered, by the balance index. The revaluation of the pension is based on the change in these indices divided by 1 016, a reduction for the interest rate that is credited in the calculation of the annuitization divisor.
83The unused pension balances or premium-pension capital of deceased persons, shared by all insured survivors. Inheritance gains are allocated as a percentage increase in the pension balances or premium-pension capital of all insured survivors in each birth cohort. At age 65, inheritance gains are estimated at about 8 per cent of the pension balances and premium-pension capital.
84Internal rate of return
85Here, the compounding rate of the pension liability so that it increases at the same rate as the assets of the system. The internal rate of return is determined by the change in the contribution revenue of the system and the change in the extent to which these contributions can finance the pension liability – in other words, the change in turnover duration – and the return on the buffer fund, in addition to the cost (gain) due to changes in average life span. If balancing is activated, the pension liability is compounded at a rate approximately equal to the internal rate of return of the pay-as-you-go system.
86Pay-as-you-go pension systems
87Pension systems in which the pension liability need not be backed by a certain amount of funded assets. A pay-as-you-go system is often described as a system where contribution revenue is used directly to finance pension disbursements. This description is inaccurate in the case of a pay-as-you-go system with a buffer fund.
89The sum of the annually determined pension credit, which is successively recalculated in accordance with the income index, – or alternatively the balance index, – inheritance gains and costs of administration.
91An individual’s pension credit is 18.5 per cent of his/her pension base and is equal to the pension contribution. Accrual of pension credit increases the individual’s pension balance and premium-pension capital.
92Pension level Here, the average pension in relation to the average pension-qualifying income.
94In this report, the financial commitment of the pension system at the end of each year. The pension liability to economically active persons is the sum of the pension balances of all these individuals. The liability is calculated by multiplying the amount of the pension of each birth cohort by the average remaining (economic) life expectancy of that cohort. Through 2017, a pension liability will also be calculated for the ATP credit earned by the economically active.
A basis for granting pension credit apart from taxable income or any actual earned income. Pension-qualifying amounts, or imputed pension-qualifying income, can be credited for sickness and activity allowances, childcare years, study, and compulsory national service.
The income used in calculating the pension credit of the insured. In principle, it consists of annual earnings reduced by the general individual pension contribution. Before deduction for this contribution, the maximum pension-qualifying income is 8.07 income-related base amounts; after this deduction, the maximum pension-qualifying income is 7.5 income-related base amounts.
Senior economist at the Swedish National Social Insurance Board (Riksförsäkringsverket: RFV).
Few matters are ever decided definitively, especially in politics. In the reformation of the Swedish pension scheme there officially remain two unresolved issues. The most important is how surpluses should be calculated and distributed. An expert committee will present a proposal on this matter in March 2004. The second outstanding problem is how (and perhaps whether) the political agreement to share the contribution rate equally between employees and employers should be implemented. The present contribution rate is 10.21 per cent for employers and 7 per cent for employees.
However, one written guideline unconditionally followed by the Ministry of Health and Social Affairs during the 1990s was to use the expression “the reformed pension system”. The earnings-related pension system, or the ATP system, introduced in 1960, was regarded by many Social Democrats as the “crown jewel” of the Swedish welfare state. Labelling the system as new might well have aroused even greater opposition to the process.
The average pension level is lower than the replacement rate since pensions have been indexed to the change in consumer prices, while incomes have grown faster than consumer prices.
My preferred definition of inter-generational equity in this context is a zero standard deviation in the ratio of each birth cohort’s benefits to contributions. For reasons beyond the scope of this paper, benefits and contributions should be discounted by the growth in average income in the calculation of the benefit/contribution ratio.
The members of the Pension Reform Work Group that developed and agreed upon the reform were Anna Hedborg and Ingela Thalén (Social Democrats), Margit Gennser (Conservative), Bo Könberg (Liberal), Åke Pettersson (Centre) and Pontus Wiklund (Christian Democratic Party). Bo Könberg, Anna Hedborg and Ingela Tahlén have served as ministers of Social Affairs and chaired the Pension Reform/Implementing Group.
However, the committee did propose that the widower’s pension, which was also integrated in that system, be abolished. This proposal was approved by parliament, with a very long transitional period.
However, pension points in the old scheme (ATP) will be earned to an annually decreasing extent up to the year 2017, after which the old scheme will no longer be economically significant.
In an NDC system, a temporary deficit can be remedied by increasing the contribution rate, but it is risky to do so. If the original cause of the deficit persists, the deficit may become even larger than at the outset since higher contributions in an NDC scheme also imply that more pension credit is earned. In both a fully funded DC system and an NDC system, the long-term pension level can be increased or maintained by increasing contribution rates, but the pension level in the short term cannot. Some analysts have considered the NDC “formula” to be another version of a career-average defined-benefit formula (see, for example, Cichon, 1999). This view fails to recognize that uninsurable risks in a defined-contribution system must in principle be accommodated by the pension level, rather than by the contribution rate, as is in fact the case with the Swedish NDC system.
By special legislation required for phasing in the new system, pensions in 2002 were recalculated by the increase in the income index divided by 0 996 rather than 1 016. This explains why the red curve (change) in income index/1 016 crosses the change in income; as from 2003, the red curve will be at a constant distance of 1.6 per cent below change in the income index.
The average delay is 18 months pending final determination of annual tax. During this time the contributions earn market interest rate on fixed income instruments, PPM is in charge of these investments.
However, there are pure fixed-income funds, even index-linked, fixed-income funds, within the PPM system.
In fact, the guarantee is calculated as a supplement to the pension from the (old) DB scheme, ATP and the NDC pension; the actual size of the premium pension is not considered. In the calculation of the guaranteed pension, the premium pension is assumed to be of the size it would have been if it had earned the same return as the contributions in the NDC system. The reason for this peculiar provision is mainly administrative and the legislative bill indicates that it may be changed so that the actual size of the guaranteed pension would be considered in calculating that pension.
In 2003, one price-related base amount is 38,600 SEK.
A debate has been in progress at least since 1994 on the merits of notional DC systems. A criticism of NDCs has been that they would not be financially stable (Valdés-Prieto, 2000; Disney, 1999), contrary to the more or less explicit claims of their advocates (Palmer, 2000; Fox and Palmer, 1999). This criticism of NDCs is unjustified, at least in the special case of the Swedish system. The general outline of the balance mechanism was described in the legislative history of the income index and automatic balancing mechanism (1997, in Swedish). For an introduction to the balance mechanism in English, see Settergren, 2001.
For details on the automatic balance mechanism, see the legislative history of the reform (in Swedish) or the non-technical description in Settergren (2001), also The Swedish Pension System Annual Report 2002 provides information on the balance mechanism.
Three technical changes that explicitly cut costs relative to the principles decided in 1994 were proposed in 1995 and legislated in 1998. The amendments lowered the pension level (replacement rate) by some two per cent relative to ATP replacement rates in the specified scenario.