Balanced Notional Defined Contribution Schemes: A new “geist” in old bottles?

174 Balanced NDC Schemes: A new “geist” in old bottles? NFT 2/2005 1. Introduction Social policies like all other fields of policy are subject to fashion. Fashion cycles in social policy are much longer than in the garment industry but they are a manifestation of the prevalent zeitgeist. Less than a decade ago a new fashion appeared in pension policy: Notio- nal Defined Contribution (NDC) schemes. In- vented in Sweden and Italy, first applied in Latvia and later introduced in Poland, Italy and Sweden, they have meanwhile been heralded by the World Bank2 as a cornerstone of a possible long-term pan-European pension model. When analysing the NDC method (as it was then known) of calculating pensions the author concluded in 19993 that a)the schemes are not in automatic financial equilibrium without a balancing mechanism, since they can cope with increasing longev- ity but not shrinking workforces resulting from decreasing fertility, Balanced Notional Defined Contribution Schemes: A new “geist” in old bottles? by Michael Cichon1 Michael Cichon cichon@ilo.org Michael Cichon holds a Masters degree in Pure and Applied Mathematics (Technical University, Aachen) a Masters degree in Public Administration (Harvard University) and a Ph.D. in Economics (University of Göttingen). Since 1995 he has been Chief of the Financial, Actuarial and Statistical Services Branch of the ILO’s Social Protection Sector. NDC schemes are not in automatic financial equilibrium without a balancing mechanism, since they can cope with increasing longevity but not shrinking workforces resulting from decreasing fertility. The need for additional mechanisms to keep the schemes in balance is resulting in the adoption of new regulatory mechanisms. This paper tries to trace the principal effects of such balancing mechanisms on a typical European country called Demoland. The analysis heavily draws on the Swedish method of balancing NDCs to analyse the principal effects of a balanced NDC approach in a stylized typical European demographic and economic context. The paper argues that the balanced NDC approach limits the policy space policy makers to find a fair sharing of the financial burden associated with demographic developments between generations. The limitation of the policy space leads to the fact that the financial consolidation of the NDC pension schemes will be done at a high cost to pensioners in a typical European context. It also symbolizes a fundamental shift in the way PAYG pension schemes are functioning, away from a solidarity-based way of coping with emerging new demographic, economic, social and resulting financial burdens to an individualistic approach with uncertain long-term consequences for the future standard of living of pensioners. 175 Balanced NDC Schemes: A new “geist” in old bottles? b)the NDC formula itself was new wine in old bottles as similar financial effects could be obtained by a PAYG Defined Benefit (DB) scheme with a career average pension for- mula and actuarial reductions and incre- ments to compensate for early respectively late retirements. Due to the fact that they are not in automatic financial equilibrium the NDC schemes are now turning into balanced NDC schemes. The need for additional mechanisms to keep the schemes in balance is resulting in the adoption of new regulatory mechanisms. Consequently the principal effects of NDC reforms on pen- sioners and contributors have changed. This paper tries to trace the principal effects of such balancing mechanisms on a typical European country called Demoland. The analysis heav- ily draws on the Swedish method of balancing NDCs – but does not set out to criticize the specific Swedish pension reform. It cannot and does not set out to replicate the numerous and more sophisticated model calculations that were prepared by the Swedish authorities in recent years. It simply uses the defining elements of the Swedish balancing mechanism – which is the first fully developed and fully documented mechanism – to analyse the principal effects of a balanced NDC approach in a stylized typical European demographic and economic context. The paper argues that the balanced NDC approach, which aims at consolidating the fi- nances of PAYG pensions, may do so at high cost to pensioners in a typical European con- text. It also symbolizes a fundamental shift in the way PAYG pension schemes are function- ing, away from a solidarity- based way of coping with emerging new demographic, eco- nomic, social and resulting financial burdens to an individualistic approach. That approach also limits the policy space for politicians to distribute future financial burdens triggered by old age security systems between the active and inactive generations. Balanced NDC schemes reflect a new “zeitgeist”. 2. Conceptual and definitional basics Pension schemes are basically a set of rules that determine the share of total consumption that a society allocates to the elderly. On the surface one can finance that share of national consumption either – as we have tra- ditionally done Europe – from the current income of active workers or – alternatively – by forcing each generation to accumulate fi- nancial or tangible assets and to sell them to the next generation (i.e. saving and dis-saving). By now, however, it should be common knowl- edge that nations cannot – or only to a very limited extent – stockpile (or save) goods for future consumption4. Even if generations save for their retirement, the consumption of the elderly has to be financed from the income generated by the active population. The pro- ceeds that future pensioners need to derive from their savings to finance their day-to-day consumption depends critically on what share of their income the next generation wants to use to buy assets from the pensioner genera- tion, i.e. what share of GDP future generations of actives want to share with the elderly. If the number of actives decreases in an ageing soci- ety, the rate of return on capital stocks are likely to diminish and asset prices are also likely to fall as the demand for assets will most likely decline. Pension levels can be expected to fall likewise. Even the World Bank in its recent pension policy paper adheres to this thinking5 . Nonetheless, a greater reliance on fully funded components in national pension systems is widely recommended by the World Bank an others. However, a complete change- over from a PAYG pension scheme to a fully funded one would create substantial transi- tional financing problems for governments. In this context Notional Defined Contribu- tion (NDC) schemes were invented as a close proxy to “real” fully funded defined contribu- tion (DC) schemes. The basic philosophy of 176 Balanced NDC Schemes: A new “geist” in old bottles? NDC schemes is simple. They mimic (Barr 2004) the principle of fully funded defined contribution schemes without requiring actual resources to finance transition cost. The contri- butions of individuals are credited to a ficti- tious account. That “account” is actually noth- ing more than a record of contributions paid and fictitiously credited interest on these con- tributions. At retirement pension amounts are determined by dividing the fictitious or “no- tional” balance of the “account” by an annuity factor (or “divisor”). That factor or divisor is actuarially calculated –like in any private pen- sion insurance scheme – based on the remain- ing life expectancy and an assumed interest rate as well as the assumed rate of future pension indexation. If the interest rate used for credits to the accounts and the rate used for the calculation of the annuity factor were equal6 then such NDC schemes can be defined as “pure “ NDC schemes. They fully simulate real DC schemes with respect to the pension calculation7. Pensions of different cohorts would thus under ceteris paribus conditions automatically vary in line with their expected average life expectancy at the time of pension award. The contribution rate would be more stable than in a classical PAYG scheme. From the first appearance of the NDC mod- els governments have diverged from the pure emulation of real DC schemes – necessary, in order to “balance the books”. Existing NDC schemes vary according to the interest rates they apply to the fictitious savings and the interest rate used when calculating the annuity factor. If one assumes that the interest rate to calculate the annuity factor is equal to the future rate of pension indexation, then the annuity factor is equal to life expectancy at retirement age. This is the case in Poland, for example. Poland adjusts pensions in payment with the rate of inflation plus 20% of real wage growth. Because of this numerical equivalence between life expectancy and the annuity fac- tor, that rate (inflation plus 20% of real wage growth) is implicitly equal to the assumed interest rate for the calculation of the annuity factor. Savings, on the other hand, are credited with an interest rate that is equal to 75% of the total wage sum.8 The effect is that initial pen- sions are held down and the average replace- ment rate of pensions in payment drops over time. In Sweden, the interest rate applied to savings “in normal times” is equal to the in- crease in average wages. Pensions in payment are indexed with average wage increase minus 1.6%-points. The latter means that the implicit effective interest rate applied to savings is equal to 1.6%9 ,10 . This is generally lower than the rate of change of wages which means that (due to the smaller denominator in the present value calculations) initial pensions are rela- tively high but would then face a declining replacement rate during an individual’s pen- sion life. A crucial difference between real DC and notional DC concepts remains. Real DC schemes are – if all goes according to plan (and according to actuarial calculations) – in auto- matic financial equilibrium since the present value of all pensions to be received by an individual would – at least in theory and on average – match the amount of his/her ficti- tious savings. Collectively this would mean that at any given point in time the present value of all liabilities (i.e. the present value of all pensions in payment and all pension rights earned by still-active insured persons) would be equal to the total value of all balances in the individual pension accounts. This allows for substantial flexibility with respect to retire- ment ages. People would just get out what they put in – regardless of when they retire. Pure NDC schemes on the other hand are not in automatic equilibrium. It is obvious that the actuarial pension formula alone only isolates NDC schemes against the risk of longevity. It does not isolate NDC schemes against the risk of shrinking contribution cohorts due inter alia to decreasing fertility rates. Achieving an auto- 177 Balanced NDC Schemes: A new “geist” in old bottles? matic equilibrium – which is here equated with maintaining a constant contribution rate – sys- tematically requires an additional balancing mechanism –a “crutch” to substitute the ex- penditure and income balancing power of money – of which there is none or relatively little (in form of contingency buffer funds) in the NDC scheme. The need for additional balancing between income and expenditures turns pure NDC schemes into balanced NDC schemes. In Sweden this is achieved through addition- al corrections to the interest rates credited to savings and the adjustment of rates of pension in payment as introduced in 2001. In some cases the indexing of savings or pensions to the rate of change of the wage sum is regarded as a perfect balancing mechanism. This would go some way towards balancing income and ex- penditure but is not always mathematically correct (and counter examples exist11) and does not generally abolish the need for an additional balancing mechanism. 3. The effects of maintaining financial equilibrium in balanced NDC schemes 3.1 Financial equilibrium and policy spaces in PAYG schemes NDC schemes remain PAYG or partially fund- ed pension schemes – which determines the nature of their financial equilibrium. If one abstracts from the possible existence of a con- tingency buffer fund (thus leaving the “pure“ Swedish case) and ignores administrative cost, they have to comply with the basic formula: (1) CRt * AWt * CONSt = APt * PENSt i.e. the product of the average wage (AW), the contribution rate (CR) and the number of con- tributors (CONS) has to be equal with the product of the number of pensioners (PENS) and the average amount of pensions (AP) in any given period t. This can conveniently be written as: (2) CRt = ( APr //AWt )* (PENSt /CONSt ) meaning that the PAYG contribution rate is the product of the financial ratio (the ratio of the average pension to the average wage AP/AW) and the demographic ratio (the ratio of the number of pensioners to the number of contri- butors PENS/CONS). An emerging financial dis-equilibrium would be signalled in this “pure” PAYG world by increasing deviations of necessary contribu- tion rates from actually charged contribution rates. A standard DB PAYG pension scheme as an institution can use at least three policy instruments to react to that situation: i.e. mod- ifying pension levels, pension age and contri- bution rate. The pure NDC scheme gives up one or two of those (i.e. the pension level, and with some limitation the pension age) but leaves the contribution rates – even if this is not always explicitly admitted (see Palmer (2003)) – to accommodate financial pressures that re- sults from factors other than longevity. A bal- anced scheme changes that situation. If – as in the case of a balanced NDC scheme – the contribution rate is fixed and the demo- graphic ratio is outside the direct control of policy makers, the number of contributors is determined by the economy and the size of the cohorts in active age by the demographic envi- ronment, and the number of the pensioners is determined by people’s retirement preferences (with some limitation through the setting of a minimum retirement age), then logically the schemes can only be kept in financial balance if the financial ratio can be modified. With the exception of a ceiling on contributors’ earn- ings, the average insurable wage can also not be influenced by policy decision, thus – in principle – the only policy instrument that can be used in an NDC scheme to maintain its financial equilibrium and to bring a deviating scheme back into equilibrium is to modify the 178 Balanced NDC Schemes: A new “geist” in old bottles? level of pensions. In the prevailing demo- graphic situation in Europe, this will mean in most cases reducing the level of pensions12. The balanced NDC scheme thus deliberately and severely limits the policy space for policy- makers13. 3.2 Maintaining financial equilibrium in pure NDC schemes If a financial imbalance is due to increasing longevity then the pure NDC mechanism copes with it through the reduction of new pensions at each single retirement age – except possibly for some time lag problems. Individuals can counter this by retiring later – if they have the freedom to do so. Alternatively they can choose other means of individual social risk manage- ment. They can choose to retire at the time planned but draw a pension later. They might bridge the time gap by using other transfer payments – if accessible – or “buy” additional periods of leisure out of private savings – if they have the means to do so. There are various ways of individually managing the longevity risk. However, these options generally favour the better off and the better informed. Less well off people might prefer to take pension later and yet might be subject to pressures to retire earlier than planned. What is meant as an incentive for change in retirement behaviour might just turn into a straight reduction of current income for the less fortunate. If, however, the financial imbalance occurs due to a contracting volume of contribution income, then a pure NDC scheme would have to resort to increasing retirement age or in- creasing contribution rates, although the latter measure has it’s own disadvantages. Each in- crease of the contribution rate to balance cur- rent accounts, creates new future pension rights that may very well cause new disequilibria problems in the future (Scherman 2003). The only way to avoid this would be to split the contribution into a share that is credited to the individual accounts and one that is credited to the contingency buffer fund without affecting pension amounts. In any case, raising retire- ment age or increasing the contribution rate are measures that could be applied in any other PAYG scheme – without the special disadvan- tages that are associated with increasing con- tributions in an NDC scheme. 3.3 Maintaining financial equilibrium in balanced NDC schemes and its likely effects If the scheme were to maintain automatic fi- nancial equilibrium with a constant contribu- tion rate, other measures would be needed to cope with the financial imbalance from a con- tracting contribution base, for example, by introducing a balancing mechanism. This sec- tion establishes the possible effects of such a balancing mechanism. Indeed, the politically tenable options for the actual design of such balancing mechanisms are limited. Rather than reducing the value of actual savings and pen- sions in payment, the rate of increase of both would probably be slowed down, i.e. the annu- al adjustments of pensions and the interest rate credited to pension savings would be reduced by applying a certain reduction factor to the “normally” applicable rates of increase and interest. Such is the example of Sweden, and this mechanism is used here as a concrete example to analyse the potential effects of such a balancing mechanism on the long-term re- placement rates of pensions. A brief introduc- tion of the mechanism is therefore in order. Other NDC countries such as Latvia, Poland and Italy have not yet introduced such explicit automatic stabilisers14 although the necessity is acknowledged.15 Interestingly two of the older classical PAYG DB schemes (i.e. the statutory pension scheme in Germany and the earnings related pension component in Japan) have introduced so-called explicit demograph- ic factors16 or sustainability factors17 that aim explicitly at the financial stabilization of the schemes. 179 Balanced NDC Schemes: A new “geist” in old bottles? A prominent example: The mechanics of the Swedish balancing mechanism The Swedish method to determine the balanc- ing factor is new. Its full mathematical descrip- tion can be found in The Social Insurance Office (2004, pp. 71-73). Essentially, the bal- ancing formula is a rule-of-thumb simplifica- tion of an actuarial present value calculation. Instead of calculating the ratio of the expected present value of all pension liabilities (ac- quired pension rights and pensions in pay- ment) and the sum of the present value of all future contribution income plus the value of the initial reserve, the formula used here esti- mates pension liabilities and contribution as- sets by using rules of thumb that do not require any projections.18 The ratio of assets and lia- bilities provides a balancing factor. If that factor is smaller than unity, interest rates cred- ited to the retirement savings in the individual accounts and the rate of adjustment of pension have to be reduced compared to the normal rates of interest and adjustment of pensions by multiplying the normal rate with the balancing factor. The balancing factor (which we assume in a normal stylized European case will be smaller than 1, i.e. the ratio of the contribution assets and the pension liability is smaller than unity due to the above mentioned upward trend of the demographic ratio in Europe during the next decades) will be applied to the normal rate of pensions and savings indexation. In Sweden this would mean that if the balancing factor is, for example, 0.99 (i.e. that the contribution assets – including the value of the buffer fund if any – are 1% smaller than the pension liabilities) and if the normal wage increase shows a value of 3%, then savings are only credited with an interest rate of 1.97% (.99*1.03=1.097) and pensions are adjusted only by 0.4% (0.99 * (1.03/1.016) = 1.004)19 instead of the normal rate of 1.4% (1.03/ 1.016=1.0138). The new rate of 1.97% is called the “internal rate” of return of the pension scheme20. If the balancing ratio recovers, pen- sions and balances are adjusted at a higher rate than the normal until they regain the index level they would have had reached without the temporary reductions due to the activation of the balance level in the first place. Effects of the balancing mechanism on pension levels In the latter case pension levels are restored but annual losses during the years with less than normal adjustment are not compensated. The present value of pensions in payment will thus always be reduced whenever the balancing mechanism is activated. By contrast, and de- pending on when during the contribution life of an insured person reduced interest rates for account balances are triggered through the balancing mechanism and the consequential recovery is activated – he/she might actually benefit from the balancing procedure if the same “recovery rates of adjustment” are ap- plied to the account balances and pensions. This is an obvious effect of the asymmetric adjustment of pensions and balances21. While bringing pensions back onto the normal index- ing track, the value of the accounts might be overcompensated for the loss. The following Box 1 illustrates this effect by an example. The “cost” of short-term shocks in the system is thus most likely entirely borne by pensioner generations. The worrying fact is that the overcompensa- tion of the active generation’s savings balanc- es might trigger another activation of the bal- ancing mechanism which could then hit the loosing pension generation again. If the period of below unity balancing factors is not fol- lowed by a recovery period of positive factors due to a systemic deterioration of the demo- graphic situation or a general contraction of the economy then future generations of pensions will also lose pension income but to a lesser extent than the pensioner generation during whose pension period the necessary down- 180 Balanced NDC Schemes: A new “geist” in old bottles? -6 -4 -2 0 2 4 6 010203040 year per c ent age Dif ference betw een savings balanced and savings normal The following graphs describe a simple example. Cohort II is starting to contrib- ute in year one an amount of 10 cur- rency units (CUs). It contributes for forty years. Contributions are increas- ing by 3% per annum. Cohort I starts to receive a pension in the same year when cohort II starts contributing. It receives a pension of 70 CU. In the base case in a normal situation annual retirement savings are credited with an interest of 3% and pensions by 1.4% (i.e. 3.0 – 1.6% =1.4%) which would simulate the Swed- ish case. In a second scenario the inter- nal rate of return is reduced due to a triggering of the balancing mechanism to 1.97% for a duration of 10 years. This simulates a period of a limited economic shock, which could be triggered by in- creased unemployment, for example. Retirement savings are thus credited with an interest rate of 1.97% while pensions are increasing in nominal value only by 0.4% p.a. The loss in pension level is subsequently recovered through a faster adjustment of pensions (which automatically also benefit the balances on the savings accounts) for seven years. Box Figure 1 shows how the adjustment index recovers over the years. Box Figure 2 shows the parallel picture for recovering pension levels. Box Figure 3 shows that retirement sav- ings under the recovery scenario are overcompensated, if the systematic dif- ference between savings and pension indexation is maintained. The differential effect of the situation on pension levels and retirement savings is obvious. The pensions of cohort I lose about 3.1% of their present value while the retirement savings of cohort II will gain about 2.3%. If no further balancing periods are triggered then even in this relatively unspectacular example the pen- sions of cohort II are about 5.5% higher that those of cohort I. Box 1: The Swedish-type balancing mechanism, pension levels and retirement savings under short-term shock conditions. Box Figure 3: An example of the effect of a balancing mechanism on the level of retirement savings. Box Figure 2: An example of the effect of a balancing mechanism on pension levels. Box Figure 1: An example of the effect of a balancing mechanism on the pension indexation index. 60 70 80 90 100 110 120 130 1 4 7 10131619222528 years P ens ion l ev el s Pension Normal Pensionbalanced 0,8 1 1,2 1,4 1,6 1,8 2 2,2 2,4 2,6 1 4 7 10131619222528 years Index at ion i n dex Index Normal Index balanced 181 Balanced NDC Schemes: A new “geist” in old bottles? ward adjustment of pension levels occurs.22 If there is a long sequence of consecutive below unity balancing ratios without recovery over long periods or even decades then retire- ment savings will suffer (and hence future pension levels) a greater loss than pensions in payment. This, in a European context, is the much more likely scenario. A mental exercise helps to understand the potential dimension of the cumulative effect of successive balancing on pension levels. In 2005 a country – that we may call Demoland – has a contribution rate of 16% and a demo- graphic ratio of 0.33 (i.e. there would be 33 old age pensioners for 100 contributors). This de- mographic ratio of 0.33 could be typical in any ageing European country23 if all people were retiring at age 65 and 90% of the people in active age groups were employed and contrib- uting. According to our formula (2) this would then yield a financial ratio (or an average replacement rate of pensions) of 0.485 (i.e. the average pension would amount to 48.5% of the average insurable wage). If the demographic situation in the model country were to develop as the UN projections forecast for our model country Demoland indicate, then the demo- graphic ratio (without a change of retirement age) would increase to 0.57 in 2050. To keep the contribution rate stable we would need to bring the average replacement rate down to 28.1%.24, 25 The automatic downward adjustment of the level of new pensions in line with increasing life expectancy (due to the annual adjustment of the annuity factor or divisor) would go some way towards achieving that “objective”. But it would fall far short of target. In 2005, all (old age) pensioners have been born before 1940. The pensioners of 2050 will have been borne roughly between 1960 and 1985. According to Settergren (2003, table on page 104) in Swe- den the latter group would experience on aver- age a reduction of their pension by about 10% due to increased longevity compared to the cohorts born before 1960. Due to the identity of the demographic structure and development of Sweden and Demoland, we can use these factors here. The order of magnitude of the reduction is most likely not atypical for other European countries. Meaning that the average replacement rate would decrease to 43.7%. This is the effect of the pure NDC automatism triggered through increases of the NDC divisor (or annuity factor). If Demoland were to follow a strict balancing policy (i.e. maintaining a constant contribution rate), then pension levels would be forced down over time through the balancing mechanism by another 36%. This roughly means that only about 24% of the total consolidation need would come from the lon- gevity effect on the pension levels and 76% through the balancing mechanism. This is roughly equivalent to the permanent use of a balancing ratio of 0.99 for about 45 years. Using the jargon of the World Bank, roughly three quarters of the “implicit pension debt” that the system is incurring at a constant con- tribution rate of 16% would be cancelled by reductions in pension levels while one quarter could be cancelled by the increase of retire- ment age (if retirees prefer later retirement to an equivalent reduction of pensions). If a contingency buffer fund is available (which in Sweden at the end of 2003 stood at 370% of annual expenditure), it could be used to mitigate against the fall in replacement rates over the decades. However, one has to note that – in our example – in the year 2050 alone the income from the buffer fund needed to fully stabilize the replacement rate would amount to 12% of the total wage or about 43% of annual expenditure. Much more exact actuarial calcu- lations and projections are needed to confirm this order of magnitude but there is reason enough to believe that even the existence of a sizeable buffer fund could not prevent a drama- tic drop in replacement rates in balanced NDC schemes operating in a typical European demographic environment. The problem of 182 Balanced NDC Schemes: A new “geist” in old bottles? declining replacement rates would, of course, be much bigger and surface much earlier in countries without such buffer funds that might currently be contemplating an NDC-type reform. But back to our case without a buffer fund. If people were far-sighted enough and were com- pensating prospective reductions of the re- placement rate by higher retirement ages they would have to increase the average rate of retirement age from 65 to about 73 years in 205026. Many more people than today would never experience retirement. In addition, pro- spective later retirement is highly unlikely as there is no way that people would be able to forecast the long-term decline of replacement rates years or even a decade before they plan to retire. The effect on individual cohorts would be rather dramatic. The following graph shows the effect of the continuous application of a balancing factor of 0.99 on the average re- placement rate of a cohort of pensioners in Demoland that starts out with a replacement rate of 41% (earned after 40 years of pension savings at a rate of 16% of an average income which has increased by a nominal rate of 3% throughout the savings period)27,28. The top line describes the “normal” decline of the replacement rate due to the asymmetric adjust- ment of pensions vis-à-vis the interest earned on the fictitious retirement savings. The sec- ond line describes the effects of a continuously applied balancing ratio of 0.99, and the last line describes the development of the replacement rate of a minimum pension which was set at 33% of the average wage in the start year and is consequently only adjusted for inflation.29,30 That amount could be interpreted as a relative poverty line. The figure shows that at a contri- bution rate of 16% and an average nominal wage increase of 3% and a sequence of balanc- ing ratios triggered by a demographic develop- ment, the application of the balancing ratio would bring the pension level of the standard beneficiaries in this cohort down to the poverty level31 . Most of the drop in replacement rates would occur after retirement, so that pension- ers would no longer have the option to com- pensate replacement rate losses through in- creasing retirement age. Incidentally, the replacement rate in the above example – after 30 years of contributions – would only be in the order of 31%32 . Even if that were to be increased by proceeds from the real DC component which the reformed sys- tems have also introduced as second pillar the overall replacement rates would most likely fall short of 40%. This raises the interesting question if – and for how long – some of the European NDC schemes will be able to meet the standards of the ILO convention (No. 102 of 1952) on minimum standards in Social Security or the European code of Social Secu- rity (1964). Actual replacement rates depend, of course, critically on the level of the contri- bution rates. As long as these are locked in at the present levels, some of the present Europe- an NDC schemes might be heading for legal complications. The issue justifies an in-depth actuarial analysis which is far beyond the scope of this short paper. In Sweden, the existence of a buffer fund and liabilities stemming from the old ATP system provide for temporary deviations from the prin- cipal development.33 Without the buffer fund (in 2003 equal to 10.6% of contribution as- sets34 ) the balancing ratio would already be smaller than one and the decline of the replace- ment rates would be accelerated due to the application of the balancing factor. In addition, pension liabilities are still dominated by the old ATP burden which are based on generally higher pension levels that will be reached under the new system, thus the transition to lower replacement rate rates is slowed down. However, the above figures show the princi- pal trends that balanced NDC schemes are most likely to face. The balancing of the books will be at the cost of dramatic reductions in 183 Balanced NDC Schemes: A new “geist” in old bottles? 0,2 0,25 0,3 0,35 0,4 0,45 1 3 5 7 9 11131517192123 years R epl a c ement r a te s RR RR Normal RR RR adjus ted Minimumpension Min. P. pension levels. As it looks, pensioners can only compensate about one quarter of such losses through postponing retirement in line with longevity gains. Three quarters of these losses would most likely occur after they have retired, unless they (i.e. the generation of the 20 to 45 year-olds of today) would be wise and healthy enough – with little advance information on post retirement reductions in pension levels to push retirement far beyond the age of 70 and beyond longevity gains. Possible system side effects Balanced NDC reforms set out to keep the contribution rate to the NDC tier of the overall national pension system constant. With the help of a balancing mechanism that objective can be achieved. However, the NDC scheme generally is only the first tier in the pension system. The second tier in all recent European reforms is a real DC scheme. According to the calculations of the Swedish Social Insurance Office35 the overall replacement rates for an average pensioner at age 65 are expected to drop from roughly 65% for those born in the early 1940s to about 51% in the medium vari- ant and to 47.5% in the pessimistic variant. In order to avoid such drops in the replacement rate present contributors would have to in- crease their savings in the second tier schemes or in a voluntary third tier by 150% to 200%. Similar orders of magnitude would apply to our Demoland case. This means that overall contributions to the pension system as a whole would have to go up in order to maintain present replacement rate levels. Governments might need to legislate hikes in the second pillar if too many people fall under the guaran- teed minimum pension levels (which are an integral part of most pension reforms). This means that while the NDC scheme might be able to maintain a constant contribution rate, the pension system as a whole might not. Possible social budget side effects As Hagemejer (2004) points out, the reduction of pension levels will most likely trigger in turn compensation strategies of future pensioners. They will delay the date of pension application to recoup some of the losses inflicted on them by the NDC pension formula and the balancing mechanism. However, that does not mean that they will delay actual retirement from the la- bour market, they may well try to use other transfer payments as a substitute for pensions to bridge the gap between desired and afford- able retirement age, such as social assistance, unemployment benefits and disability bene- fits. This option could at least defer the age of entry into pension receipt until the age of 65 (after that age, in most countries no alternative transfers are payable). Part of the retirement cost might thus be shifted to alternative trans- fer mechanisms. If the benefits under these schemes are relatively generous and pension contributions are paid by the state during the receipt of these benefits the incentives for behavioural adjustments of this sort are sub- Figure 1: Simulation of the effect of the balancing mechanism on the pension replacement rate of a standard pension recipient during the period of pension receipt in Demoland. 184 Balanced NDC Schemes: A new “geist” in old bottles? stantial. If average levels are still declining then the state may have to “remedy” some of the effects through the financing of an over- proportional share of total pension expendi- ture through the guaranteed minimum pen- sion. The NDC scheme and the balancing mechanism might thus consolidate the financ- es of the old age pension schemes without necessarily achieving a consolidation of the overall level of social transfers. In other words, while the pension scheme might be in financial equilibrium the social budget of the nation as a whole might not. In view of the above principal problems of the “balanced NDC” one might query why policy makers chose a relatively complex and new system to consolidate pension systems. One possible reason is that it was the only way to consolidate these systems. The following section rejects that hypothesis. 4. Are NDC reforms necessary? Let us assume, contrary to the previous exam- ples, that there exists a simplified PAYG pen- sion scheme in Demoland. People retire at age 60 with an average replacement rate of 50% of average earnings. We assume that all people presently making use of de-facto early retire- ment through the use of alternative transfer benefits such as social assistance, unemploy- ment benefits and invalidity pensions are inclu- ded in the old age system. Society ages rapidly. The objective of the consolidation mechanism is to keep the contribution rate constant or in a narrow range around the present starting rate of 23.5% (which is the product of a financial ratio of 0.5 and a demographic ratio of 0.47) in the start year 2000. According to experience a contribution rate of between 20% and 25% seems to be a realistic order of magnitude for a PAYG pension scheme operating in a typical European demographic environment36. There are various ways to keep the contribu- tion rate in check. One is described by a simple modification of formula (2), i.e.: (3) 0.235 = (PENSr /CONSt * 0.5 This means that we would want to keep the contribution rate and the replacement rate con- stant implying that we have decided not to burden the active generation further. We also do not want to reduce the relative standard of living of the pensioner generation (symbolized through keeping the replacement rate constant at 50%). This can only be done by increasing retire- ment age. In Demoland we do this in steps of one year. To roughly maintain therefore the equilibrium of formula (3) we must raise retire- ment age seven times between 2000 and 2035, which means that the effective retirement age will increase by about 7 years. The model triggers an increase of the retirement age by one year each time the demographic ratio (DR) exceeds 0.5. The effect of the measure on the development of the demographic ratio is demonstrated in Figure 2. Figure 2: The develop- ment of the demo- graphic ratio of Demoland 2000-2050 with and without successive increases of retirement ages 0 0,2 0,4 0,6 0,8 1 Y ear 2007 2015 2023 2031 2039 2047 year D e mog rap hi c r a tio Mod. DR Orig. DR 185 Balanced NDC Schemes: A new “geist” in old bottles? 0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4 0,45 Ye ar 2005 201 1 201 7 20 23 20 29 2035 2041 2047 year P A Y G c ont ri but ion r a te PAYG - Status quo PA Y GCR - RA PA Y GCR- MODIN However, the increase of the de facto retire- ment age over 3.5 decades by seven years may not be feasible politically. There is another option. One could simply abolish the automat- ic adjustment of pensions in line with wages (i.e. waving the condition that the replacement rate stays constant). It is assumed here that wages will increase by 3% p.a. and pensions by 1.6% less – simulating an annual indexation of pensions in line with prices. The following figure shows the PAYG contribution rates from the year 2000 onwards a) under status quo conditions without consol- idation (curve PAYG-status quo), b)under consolidation exclusively through the increase of retirement ages (curve PAYG CR –RA) and c) under consolidation by replacing wage in- dexation by price indexation (curve PAY- GCR-MODIN). What the graph shows is that both consolida- tion measures could have “balanced the books”. However, a mono-dimensional approach us- ing just one of these tools would most likely not be acceptable (for example, an exclusive con- solidation through pension adjustments would lead to a dramatic halving of the initial average replacement rate). A pragmatic combination of the two consolidation measures and a mod- erate increase of the contribution rate could help to broker a fairer sharing of the consolida- tion burden between actives and pensioners. So the answer to the above question as to whether a balanced NDC reform (probably with a second-tier DC scheme) would be nec- essary to maintain the relative stability of con- tribution rates from a financial and technical point of view? Clearly not. There is enough reason to believe that classical instruments could have achieved the same effect. A careful balance of the use of the three main policy instruments: reducing pension levels, increasing pension age and increasing the con- tribution rate would have a) balanced “the books”, b) probably created a different inter-genera- tional sharing of the consolidation burden, and c) also probably created positive economic side effects. With respect to the latter point, it should be noted that the financing of pension schemes is only one problem that the ageing of European societies will have to cope with. The more central problem will be the negative or low economic growth rates that could potentially be triggered by a contraction of the labour force. European economies might need a much steeper and/or earlier increase of retirement ages – once some of them will have come out of the present unemployment trough – than can possibly be triggered through the longevity- based decreases in pension levels. A further exploration of the subject is outside the scope Figure 3: Projected PAYG contribution rates in Demoland under status quo, increasing retirement ages and a modified pension adjustment, 2000-2050. 186 Balanced NDC Schemes: A new “geist” in old bottles? of this paper but has been done elsewhere.37 It is obvious that the traditional bundle of policy measures opens a much wider and more flexible policy space for decision-makers than the balanced NDC approach. So the question remains why this approach was chosen in some countries and why is it promoted by institu- tions like the World Bank? The following sections tries to find an explanation. 5. Why then NDC reforms? On the surface of national and international policy debates the prevailing objective of pen- sion reforms these days seems to be the main- tenance of financial equilibrium or – better – guaranteeing long-term financial sustainabili- ty and stability. However, there may also be hidden political agendas which may have to do with the huge amounts of monies that will be passing through financial institutions (banks, pension funds and insurance companies) when public social security schemes are wholly or partially privatized.38 This again – fascinating as the topic may be – is not the subject of this paper, but the observations may help to make the case that there may be non-apparent expla- nations for some pension and social policy reforms. Financial consolidation generally means in the context of an ageing society that expendi- ture or prospective expenditure has to be brought in line with prospective income. In a genuinely fully funded DC scheme this is automatically the case. The scheme simply does not pay out more than what has been saved on an individual cohort basis and if the management of the reserves is functioning properly and the actuarial annuity calculations are sufficiently risk averse then the schemes should be in automatic equilibrium. It is per- ceived to give “people their money back” which in turn is increasingly being seen as fair from an intergenerational and inter-personal point of view. NDC schemes suggest to the general public that they operate in the same fashion as real DC schemes, i.e. that people “will get out what they put in”.39 And if that should not be the case (as it will most likely not, as we have seen) then what they will get out is at least as closely related to their personal inputs as possible. This is an essentially individualistic consol- idation approach – which appears to constitute one part of the paradigmatic foundation of the approach.40 The overall financial consolida- tion of the combined NDC and DC two-tier systems forces individuals to develop individ- ual retirement strategies. If future pensioners want to safeguard their pension levels they have to adjust their individual retirement age upwards or must begin to increase their DC savings from an early age on. The development of the right individual strategy is subject to substantial uncertainty about future demo- graphic and economic developments all com- pounded by information uncertainties (e.g. about the potential size of the future reduction of NDC pension levels). The old PAYG DB approach was based on collective societal re- sponsibility which guaranteed an adequate level of consumption for the elderly and a collective shouldering of risks and uncertainties. These responsibilities are now being delegated from societies to the individual. That reflects a new Zeitgeist. 6. By way of conclusion: New “geist” in new bottles? When analysing the mechanics of a balanced NDC reform some technical findings stand out: 1)The system can – in theory – most likely put a pension system into long-term financial, equilibrium – provided the downward pen- sion adjustments will be tolerated by the population in future. 187 Balanced NDC Schemes: A new “geist” in old bottles? 2)While the pension system might be in finan- cial equilibrium the social budget of the country as a whole might not. The size of the potential shifting of expenditure from the pension system to other social transfer schemes is unknown, but – if present prac- tice of early retirement through other trans- fer schemes in Europe is anything to go by – then the risk is substantial. 3)The burden of the financial consolidation under balanced NDC schemes will be over- whelmingly borne by pensioners during the next three decades. Losses of pension levels through a balancing mechanism in first-tier NDC schemes are not likely to be compen- sated through pension earned in second-tier real DC schemes – without substantial in- creases in their contribution rates. 4)The balanced NDC approach needlessly lim- its the policy space. The balanced NDC reforms are not necessary to consolidate the financial equilibrium of the national pension system. Financial equilibrium can be main- tained by classical means using a combina- tion of the policy instruments: raising retire- ment age, reducing pension levels and in- creasing contribution rates. 5)Policy space can be regained in NDC schemes if a certain increase of the contribution rate were permitted without triggering benefit longer-term increases. This could be done by splitting the contribution rate into an individual component (that would determine the amounts “saved” in individual accounts) and an solidarity component (that would be paid into a general buffer fund to help cope with a part of the increasing demographic burden). The individual component could be kept constant and the solidarity component could be allowed to fluctuate within limits. A new balancing mechanism could try to distribute inevitable consolidation burdens fairly between active contributors and pen- sioners.41 The obvious reason to use the NDC approach or better the combined NDC/DC approach was to achieve a fundamental paradigm change in the method of consolidation. The consolida- tion is perceived as being “fair” in the sense that contributors perceive that they “get out” what they “pay in”. Individual equity reigns over societal responsibility. In that respect I have to revise my findings of 1999 referring to the unbalanced NDC ap- proach. If one includes a balancing mechanism – prescribing constant contribution rates for the active population – NDC reform embodies a fundamental shift in the meaning of solidar- ity. In that sense, there is a new spirit (a new Zeitgeist) in the old PAYG bottle. In German the word for ghost and spirit is identical (i.e. geist). It appears likely, that once uncorked, the new “zeitgeist” of the brave new balanced pension world will haunt us all – during our retirement. Notes 1 The author is grateful for the detailed review of the text by Karuna Pal, Karl Gustaf Scherman and Diane Vergnaud and constructive comments received from Warren McGillivray, Florian Léger and Robert L. Brown. Factual errors and errors of judgment, however, remain the respon- sibility of the author. Views expressed in this paper are private and those of the author and do not commit the International Labour Office. 2 Holzmann (2003), p. 15. 3 See Cichon (1999). 4 See inter alia Barr (2000) and Brown (2002). 5 See Holzmann and Hinz (World bank, 2005), p. 70. 6 Except for annual deviations of interest rates (used to credit interest to the accounts of actives) from assumed long-term average interest rates (needed to calculate annuities). 7 This definition is independent of the annual indexation of pensions as long as the indexation follows an established rule. 8 See ILO(Fultz, 2002), pp. 124 and 125. 9 See Scherman (1999), p. 21. 188 Balanced NDC Schemes: A new “geist” in old bottles? 10 Also The Social Insurance Office, p.36. 11 Palmer (2003), p.13 claims that “the NDC scheme is in principle stable, if the figure for life expect- ancy used in computing the NDC annuities is on average correctly estimated, and if the rate of return in the account scheme follows the rate of growth of the contribution base. In addition, reserves in the demographic buffer fund would need to earn a rate of return also equivalent to the rate of growth of the contribution base. These conditions are both necessary and sufficient ...”. Indeed they are not, as the following example shows – the example refers to a case with buffer fund zero, but could easily be generalized. Take, for example, a cohort that experiences an atyp- ical increase in their savings, say, 10 years before their retirement due to an atypical in- crease in employment (causing the wage sum to increase). At that time pensions in payment are increasing in line with the wage sum keeping the contribution rate constant. When the cohort with the high employment phase retires total expend- iture will increase faster than the sum of wages due to the higher pension level of the entering cohort, causing at least a temporary increase in the contribution rate even though savings and pensions in payment continue to increase in parallel with the rate of change of the wage sum. 12 This can only be avoided if people postpone retirement fast enough to counteract the emerg- ing imbalance. However, that can be regarded as rather unlikely as long as pensions do not de- cline. According to the NDC formula, initial pensions at time of award are immune to shrink- ing active populations as long as retirement savings are not indexed by wage sums. Even in times of shrinking workforces pensioners would thus not have any incentive to retire later than the individually preferred time. Retirement behav- iour could actually be pro-cyclical. In times when employment shrinks for economic rea- sons, older workers may be forced to retire earlier rather than later, to contribute to the clearance of the labour market. 13 Brooks and Weaver (2005) describe this state of affairs as being “lashed to the Mast” (i.e. a stable contribution rate) to avoid following the siren’s call (i.e. political calls for more leniency when combating old age poverty or a different distri- bution of future financial burdens between con- tributors and pensioners). 14 Lequiller (2004), p.11. 15 Franco and Sartor (2003) state for Italy: “ Stabil- ity of the equilibrium contribution rate therefore requires either the presence of built-in stabilis- ers, such as those incorporated into the Swedish system … or periodic ad hoc adjustments to the changed scenario.” (p. 9) 16 In the case of Japan there are two explicit demo- graphic factors. One reduces pension levels to take account shrinking active contributor co- horts, the second corrects pension levels for increased life expectancy. Both factors take the form of constant average long term reduction factors applied till 2023/2025 (Takayama 2004). 17 In Germany a so-called Nachhaltigkeitsfaktor was introduced and is to be applied as of 1 July 2005. It corrects annual pension indexation by a factor that reflects the change in the relationship between “full” pensioners and “full” contribu- tors thus incorporating the effect of shrinking contributor cohorts and am increasing longevi- ty. The factor also incorporates a parameter that allows for consolidation burdens to be shared between pensioners and contributors (von Broekel, 2005). 18 This means that the procedure is applied without the longer-term view into the future. The neces- sity to apply the factor annually embodies a further limitation of policy space for decisions- makers. If a classical actuarial procedure for the determination of the balancing factor were used then one would calculate the ratio between the present value of all future pension expenditure and the present value of all future contribution income. If – in case of a temporary contraction of the contribution base – the long-term equilib- rium is expected to return to normal or one could stretch policy measures over a certain time peri- od this might make adjustments more acceptable to the pensioners and contributors. Of corse, the actuarial approach would require a set of as- sumptions on future demographic and economic developments which might make the system vulnerable to political interference. However, the actual number and nature of assumptions that enter implicitly into the asset and liability approach that is used by the Swedish system is actually similar to those which are explicitly employed by the actuarial approach. The implic- it approach, for example, assumes stable demo- graphic development. A no less stringent as- sumption than any other actuarial assumption. 19 See also The National Social Insurance Board (2004), p.35. 189 Balanced NDC Schemes: A new “geist” in old bottles? 20 For an interesting analysis of the nature of the internal rate of return one might wish to consult Settergren and Mikula (2005) 21 The asymmetry stems from two effects. First, the rates of indexing of savings and pension are different by definition; secondly these different rates are applied to mathematically different aggregates, i.e. a flow variable (the pensions) on the one hand and a stock variable (savings) on the other hand. 22 Aware of this situation, the designers of the balancing mechanism hesitate to remedy it for fear – understandably – of overcomplicating the mechanism (O. Settergren, in personal commu- nication, 1 February 2005). 23 For the purpose of these calculations, the demo- graphic structure and development as given and forecasted by the UN population projections (median variant) for Sweden were used. 24 According to formula (2): 0.16/0.57 =0.2807. 25 This may seem to be exaggerated, but in the Swedish case (in the pessimistic scenario) PAYG pension replacement rate for new pensioners at age 65 would fall roughly from 65% to 40% from today until 2055, i.e. a drop of 38% in relative terms, whereas the rough calculations here envisage a fall of 42%. In the base scenario of the Swedish calculations the drop in the replacement rates would only be in the order of 35% (figures were estimated from graphs and have thus some margin of uncertainty), see The Social Insurance Office (2004), pp.47 and 48. The fact that initial replacement rates are higher stems from the levels inherited from the old system. The fact that the drop in replacement rates is slightly less than the ones predicted here is probably due to a more optimistic demograph- ic scenario but is also certainly due to the fact that the Swedish rate applies to new pension awards (rather than all pensions in payment) whose replacement rates tend to fall throughout the individual periods of pension receipt. 26 This is probably a conservative estimate as it is based on a simple extension of Settergren’s table (2003, p 104). The extension ignores the effect of increased mortality between age 65 and 73. 27 The assumptions describing the example are identical with those assumed for the example in box 1. 28 The replacement rate may appear low but that is as shown by the actuarial calculations. In Swe- den, a standard member of the cohort may earn another 5 to 7% replacement rate from the fund- ed tier. 29 This is the case in Sweden (see Scherman 2004, p. 309) 30 The 33% roughly reflects the present level of the minimum pension guarantee in Sweden. 31 Even at – in relative terms – a declining poverty line. 32 At a value of 15.7 for the annuity factor, i.e. the 2005 rate. 33 Again, the existence of the buffer fund will delay the violation of the 40% level but it will – most likely – given demographic developments not postpone it forever. 34 See The Social Insurance Office (2004), p.8. 35 See The Social Insurance Office (2004), p.47/ 48, average replacement rate calculations for the base scenario and pessimistic scenario displayed in graphs. 36 To maintain a replacement rate of 50% and a retirement age of 60 the Swedish pension system would also require an overall contribution rate of over 20%. 37 For a more detailed analysis of the potential effects see Cichon et al. (2003) 38 All of these institutions will take a “cut” for handling the savings of individuals and turning them into annuities. These “cuts” can be sub- stantial and reach easily up to more than 25% of contributions and hence savings (Thompson 1998, pp 106,107). 39 See Takayama (2005, p.10). 40 Another part may be that dropping replacement rates under the NDC tier might force up the voluntary levels of savings in real DC pillars. 41 The new German Nachhaltigkeitsfaktor envis- ages a sharing of the burden between contribu- tors and pensioners. The exact numerical shar- ing of that burden can be corrected in future. The effects of the factor is discussed in some detail by Borsch-Supan et al. (2003), pp. 15 -18. References Barr, N: 2000. Reforming pensions: Myths, truths and policy choices, Working Paper of the Inter- national Monetary Fund WP/00/139 (Washing- ton, DC, IMF). 190 Balanced NDC Schemes: A new “geist” in old bottles? Barr, N. 2004. Notional defined contribution pen- sions: What they can do, and what they can’t do? In NFT 3/2004, pp 203- 213. Börsch.Supan, A. et al.2003. Der Nachhaltigkeits- faktor und andere Formelmodifikationen zur langfristigen Stabilisierung des Beitragssatzes in der GRV, MEA working paper, Mannheim, Germany. von Broeckel, J.2005. Der Nachhaltigkeitsfaktor, Http://www.janvonbroekel.de. Brooks, S.M. and Weaver, R.K. 2005. Lashed to the mast?: The politics of Notional Defined Contribution Pension Reforms, Center for Re- tirement Research Working Papers, Boston College. Brown, R.L. 2002. Security for social security: Is pre-funding the answer?, mimeo (University of Waterloo, 2002). Cichon, M. 1999. Notional defined contribution schemes: Old wine in new bottles? In ISSA review 4/99, pp. 87 - 109. Cichon, M.; Knop, R.; Léger, F. 2003. White or Prosperous: How much migration does the age- ing European Union need to maintain its stand- ard of living in the twenty-first century? Paper presented at the 4th International Research Con- ference on Social Security, “Social security in a long-life society”, Antwerp, Belgium, May 2003. Franco, D. and Sartor, N. 2003. Notional Defined Contribution in Italy: Unsatisfactory Present, Uncertain Future, Paper presented in Stock- holm, September 2003. Fultz, E. (ed.) 2002. Pension Reform in Central and Eastern Europe- Volume 1: Restructuring with privatisation: Case studies of Hungary and Poland, Budapest (ILO). Hagemejer, K.: Accounting for the new Swedish Old-age pension scheme: A comment, Interna- tional Workshop on the Balance Sheet of Social Security Pensions, Tokyo, 1-2 November 2004. Holzmann, R.2003. Toward a reformed and coor- dinated pension system in Europe: rational and potential structure, mimeo, World Bank, Wash- ington . Lequiller, F. 2004: Recording liabilities of Pen- sion Schemes: The very interesting case of the Swedish “Inkomstpension”, mimeo, OECD, Paris. Palmer, E. 2003: The new Swedish Pension Sys- tem, mimeo, Stockholm Settergren , O. : Financial and Inter-generational Balance? In NFT 2/2003, pp 99 - 114. Settergren, O. and Mikula, B.: The rate of return of PAY-AS-YOU-GO pension systems: A more ex- act consumption-loan model of interest, mimeo, to be published 2005. Scherman, K.G.: The Swedish pension Reform: A good model for other countries? In NFT 4/2003, pp. 304 - 318. Scherman, K.G. The pension reform debate. In NFT 2/2004, pp. 97. Scherman, K.G. 1999. The Swedish pension Re- form, Issues in Social Protection - Discussion paper No 7, ILO. Takayama, N. 2005. The Balance Sheet of Social Security Pensions in Japan, in: The Balance Sheet of Social Security Pensions, pp.1-11, In- stitute of Economic Research, Hitotsubashi University, February 2005. Thompson, L. 1998. Older and Wiser – The eco- nomics of public pensions, ISSA and The Urban Institute, Washington. The Social Insurance Office: The Swedish Pension System Annual report 2003, Stockholm, 2004. Holzmann, R. and Hinz, R. 2005. Old age Income Support in the 21st Century: The World Bank’s Perspective on Pension System and Reform (fi- nal version of 18 February 2005), World Bank, Washington.