Chapter 18 improving the equity, transparency, and solvency of pay as you go pension systems

nan-424 ◾ Pension Fund Risk Management: Financial and Actuarial Modelinga notional rate of return.. Improving the PAYG Pension Systems ◾ 425where c t is the contribution rate at moment

18

C H A P T E R

Improving the Equity,

Transparency, and Solvency

of Pay-as-You-Go

Pension Systems: NDCs,

the AB, and ABMs

Carlos Vidal-Meliá, María del Carmen

Boado-Penas, and Ole Settergren

CONTENTS

18.2 N otional Defi ned Contribution Accounts (NDCs) 422

420 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

instruments for i mproving equity, t ransparency, a nd solvency i nto the pay-as-you-go (PAYG) pension system Th is is in line with the trend seen in some countries of applying actuarial analysis methodology to the

we explain and analytically develop various aspects of notional defi contribution accounts (NDCs), the actuarial balance (AB), and automatic balance mechanisms (ABMs) Th e main conclusion reached is that these tools are not simply unrealistic theoretical concepts but a response to the growing social demand for transparency in the area of public fi nance management, the need to minimize the political risk faced by PAYG sys-tems, the desire to set the pension system fi rmly on the road to long-term

confi dence in the system in the sense that promises of pension payments will be respected

JEL Classifi cations: H55, H83, J26, M49.

to public pension systems

One aspect that has received much less attention in the literature cerns the instruments that can be a pplied to deal with one of the main problems faced by traditional defi ned benefi t (DB) pay-as-you-go (PAYG) systems: po litical r isk De aling w ith t his p roblem c an o ft en, according

con-to Boado-Penas (2008), bring about clear i mprovements i n t he system’s equity, transparency, and solvency

Political ri sk s hould b e u nderstood b asically as r eferring t o t he decisions t aken b y po liticians t ied t o t heir t raditional p lanning h ori-zon (oft en only 4 years), which is clearly far less than that of the PAYG

Improving the PAYG Pension Systems ◾ 421

pension s ystem Valdés-Prieto (2006a) po ints o ut t hat D B PAYG s tems tend to require periodic adjustments due to demographic and eco-nomic uncertainties Relying on discretionary legislation for these social security modifi cations c reates political r isk for both contributors a nd benefi ciaries

ys-Cremer and Pestieau (2000) argue that economic and demographic tors play a r elatively small role in the PAYG pension system’s problems; political factors are far more important, and the process of reforming the pension system is mainly a political problem Financial (solvency) prob-lems caused by fl uctuations in fertility rates, population ageing, increas-ing longevity, and declining productivity growth can easily be addressed

fac-by the experts, but social security systems are established and reformed through the political process Consequently, the outcome is likely not to

be socially optimal

terms “populism in pensions.” Th is can be defi ned as a form of tion between politicians in which voters are off ered subsidies and benefi ts without their appreciating that it is they themselves who will pay through higher taxes, higher contributions, higher infl ation, or reduced economic growth

competi-Populism in pensions is a phenomenon usually seen in countries with pension systems that are fi nanced by PAYG; it is aggravated if a country suff ers from a weak democratic structure and could also be increased by

a low level of education Where the fi nancial method is capitalization, it is more diffi cult for populism in pensions to appear given that the pensions are fi nanced in advance and there is an obligation to compile an actuarial balance (AB) sheet every year, from which the corrective measures to be applied are derived when necessary

Finally, H olzmann a nd P almer ( 2008) st ate t hat soc ioeconomic changes—basically greater participation by women in the labor market, changes in family structures, and increasing globalization, which together imply greater integration of the goods and services markets, factors of pro-duction and knowledge—call for a r eformulation of the basic ideas gov-erning pension system design, some of which have remained unchanged for more than a century

Th e aim of this chapter is to show the advisability of introducing ments to improve the equity, the transparency, and the solvency of PAYG pension systems Th is is in line with the trend seen in some countries of

instru-422 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

applying ac tuarial a nalysis m ethodology t o t he fi eld o f p ublic PAYG pension system management

Aft er this introduction, in Section 18.2 we give a brief description of ous aspects of notional defi ned contribution (NDC) pension plans In Section

sys-tem’s actuarial balance (AB), focusing especially on some of the features of

the term automatic balance mechanism (ABM) as applied to a pension tem, and includes a brief presentation of those in place in Sweden, Canada, Germany, Japan, and Finland Th is chapter ends with the main conclusions,

sys-a bibliogrsys-aphy with references, sys-and four sys-appendices in which we sys-ansys-alyti-cally develop the relationship between the main formulae for calculating the retirement pension in PAYG systems, the contribution asset, and the sys-tem’s liabilities as shown in the AB for both Sweden and the United States

ACCOUNTS (NDCs)

NDCs,* as a component of modern multi-pillar pension systems in some countries, ha s be en o ne o f t he ma in i nnovations o f t he la st dec ade a s regards pension reform Th ey c an be f ound i n Italy (1995), Kyrgyzstan (1997), L atvia ( 1996), P oland ( 1999), S weden ( 1999), B razil† ( 1999),

Leliebre 2005), Finland (Lassila and Valkonen 2007a), Portugal (Barrías 2007), and Norway (Stensnes and Stølen 2007) have also incorporated ele-ments of notional philosophy to assist in calculating or indexing the initial retirement pension

Nisticò (2006) point out, the original idea of the NDC was present in two

* For an international perspective see the books by Holzmann and Palmer (2006, 2007), and Holzmann, Palmer, and Uthoff (2008).

† Th is does not have all the characteristics of a notional account system.

‡ See Hauner (2008).

§ Th e author calls this a “notional defi ned benefi t system” (NDB) as all the contributions made are registered in a not ional account and amount to 1.78% of t he annual contribution base and are revalued in line with the average increase in the contribution bases If the individual retires at age 65 aft er 45 years of contributions, his entitlement will be (1.78 × 45) = 80.1% of the contribution base.

Improving the PAYG Pension Systems ◾ 423

papers published in the 1960s by Buchanan (1968) and Castellino (1969), which were rediscovered in the late 1990s by Gronchi (1998) and Valdés-Prieto (2000) Th e latter traces the origin of the concept back t o France

in 1945, in what was known as the points system* (PS), and to the United States in the 1980s, when Boskin et al (1988) proposed a reform of the pension system based on ideas in which the concept of notional accounts was implicit

According to Vidal-Meliá et al (2004), a notional account is a v irtual account r efl ecting t he i ndividual contributions of e ach pa rticipant a nd the fi ctitious returns that these contributions generate over the course of the participant’s working life In principle, the contribution rate is fi xed Returns are calculated in line with a notional rate that may be the growth

When t he i ndividual retires, he or she ( henceforth, he) receives a pen sion that is derived from the value of the accumulated notional account, the expected mortality of the cohort retiring in that year, and, possibly, a notional imputed future indexation rate In this way, the notional model combines PAYG fi nancing w ith a pens ion formula t hat depends on t he amount contributed and the return on it

-At fi rst glance, NDC plans simply appear to be an alternative way of culating the amount of retirement pension, but in fact the notional method goes beyond what might be imagined from seeing the collection of formu-lae in Appendix 18.A.1 Th e account is called notional because it exists only

cal-on paper Mcal-oney is not deposited in any real account Nevertheless, the amount of the pension is based on the fund accumulated in the notional

account (K) Contributions made to notional accounts are capitalized at

* Th e points system for re tirement was developed in France w ithin t he framework of c plementary regimes for salaried workers (Association pour le régime de retraite complé- mentaire des salariés: ARRCO) and management (Association générale des institutions de retraite des cadres: AGIRC) Pension entitlement for p eople in these regimes was based on the accumulation of retirement points throughout working life See Appendix 18.A.1.

om-† A somewhat over-simplifi ed “truth” is that the most appropriate rate of return, from a fi cial stability point of view, is the growth rate of the covered wage bill, which refl ects not only the variation in contributors but a lso the variation in contribution bases (productivity) In practice, this index is not always used—in Sweden, for instance, the index used, disregard- ing periods when the automatic balance mechanism is activated, is the growth in average earnings, partly because it is considered to be less volatile Mechanisms are in place to deal with any negative fi nancial consequences that could arise from using this index or fi nancial imbalances in the pension system deriving from demographic or economic shocks or turbu- lence in fi nancial markets Th ese are looked at in Section 18.4

nan-424 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

a notional rate of return Th is hypothetical return* is normally linked to some external index set by law

When the individual retires, the notional account—in all countries

by dividing the value of the notional account (K) by a conversion factor (g) that depends on life expectancy at the chosen retirement age and

the interest rate, which will indirectly bring about a r eduction in the degree of variation in the internal rate of return (IRR) between gen-erations The base for calculating the conversion factor should be se t

by law and decisions need to be made as to which mortality table and interest rate should be u sed in the calculation and how the mortality

the conversion factors for men and women should be sepa rated, as is common with pension plans with real capitalization, or whether some common c onversion factor should b e u sed to average out l ife e xpec-tancy for men and women, as is generally the case in traditional PAYG systems

Following Vidal-Meliá et al (2006), in order to calculate the initial sion of an individual at retirement age in notional account models, the contributions made and valued at the date of retirement are made equal to the pension that that individual will receive until his death, which is also calculated at the date of retirement Hence the initial pension at normal

pen-retirement age will be t he product of the conversion factor (CvF), g, and the notional capital (K):

* It is worth pointing out that the volatility of the return on notional accounts is usually less

or much less than the return on a pension plan under the capitalization system, which will depend on the choice of portfolio and the investment market.

† Th is i s one w ay it d iff ers f rom a s ystem of i ndividual c apitalization a ccounts, i n mo st of these c ountries, opt ions ot her t han l ife a nnuities a re av ailable ( lump s um p ayments a nd programmed withdrawals).

‡ In Sweden and Brazil, demographic parameters undergo an adjustment process every year following observed survival rates In Italy, the review should be every 10 years According to Diamond (2005), the best way of putting an end to unwanted political manipulation would

be to carry out annual adjustments with real data instead of projections.

Improving the PAYG Pension Systems ◾ 425

where

c t is the contribution rate at moment “t”

moment “t”

r i is the real notional rate applied to capitalize the contribution

g i s t he predetermined conversion fac tor, wh ich i s t he i nverse o f a n

actuarial annuity

In practice an estimated va lue can be g iven for the relevant variable so

as to make the initial pension as high as possible, and the indexation of the pension in payment is adjusted annually in line with the behavior of the relevant variable If the variable actually behaves as forecast, pensions remain constant in real terms; if growth is greater than predicted, pen-sions grow in real terms; and if growth is lower than forecast, then pensions decrease in real terms A mechanism similar to this is applied in the case of Sweden

Arguably, NDC systems have stronger immunity against political risk than more traditional DB PAYG systems According to Valdés-Prieto (2005), t he n otional acco unt s ystem i s a u seful wa y i n wh ich t o m ini-mize the political risk associated with PAYG systems as it increases the long-term fi nancial solvency of the system, although it also increases the explicit economic risk aff ecting contributors.* Marin (2006) believes that the notional account system is a better way of managing and diversifying risk in comparison to all other pension paradigms as it creates no false expectations about pensions to be received in the future, makes it diffi cult for contributors to be tempted to behave opportunistically, and is not subject to the fi nancial risk of capitalization systems

As regards fi nancial su stainability, Valdés-Prieto (2000, 2 002) shows that, e ven wh en a pplying t he m ost fa vorable f ormula ( model), N DCs can only achieve this in a rather unrealistic steady state Hence notional account systems always require other fi nancial adjustment mechanisms—such as government guarantees and repeated recourse to legislation—to

be imposed in the same way as traditional benefi t systems, or according

to Settergren (2001), special measures such as ABMs, which will be looked

426 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

intervention as long as economic growth is high enough For Lindbeck and Persson (2003), a quasi-actuarial system with an exogenous contribution rate (i.e., a contribution-based system) will increase the fi nancial stability

of the pension system in the sense that politicians will not have made any promises concerning future pension benefi ts, although, as Börsch-Supan (2005) points out, only time will tell whether the political risk of NDC systems is really much less than traditional DB systems as some of the system’s parameters could always be modifi ed

As fa r a s t he i ntergenerational a spect i s co ncerned, f ollowing K nell (2005a,b), NDCs are more “forward looking” while DB PAYG systems have a more “ backward lo oking” c haracter “For ward lo oking” s ystems are more in line with principles of intergenerational fairness and respon-sibility, wh ereas “ backward l ooking” s ystems i ndicate t hat pe ople a re obliged to shoulder the burden of changes in the size of cohorts that have been determined before they were even born, or before they were part of the electorate or the labor force

As Diamond (2006) and Barr and Diamond (2006) correctly point out, almost all the advantages attributed to NDCs could be obtained with a well-designed DB system, although of course this is precisely the diffi culty

inherent in such systems: the ease with which erroneous political decisions

convert them into badly designed systems.* For Marin (2006), the ority of the NDC system over the DB system lies not in the theory but in the practice and application Th e formulae that determine the initial pen-sion for DB and NDC systems, respectively, may produce a v ery similar

PS is its highly discretional nature, similar to DB systems in this aspect, which, according to Valdés-Prieto (2000), is built into the system Th is can

be seen from the way the authority in charge of the system can arbitrarily adjust the cost of buying new points, the contribution rate, and the value

of points sold to obtain a pension every year Börsch-Supan (2006) states that discretional deviations have been frequent in the French PS, and the German system is not free of them either

Börsch-Supan (2005) points out that NDC systems have a high level of transparency and, at least potentially, a deg ree of credibility that are not usually found in DB systems Th is is because the basic elements that deter-mine the amount of the pension appear naturally in notional accounts,

* On this subject, see the papers by Palmer (2006) and Williamson (2004).

Improving the PAYG Pension Systems ◾ 427

whereas they do not in the more complex formulae needed for calculating pensions in DB systems Marin (2006) argues that NDC systems encour-age actuarial fairness and stimulate the contributors’ interest in the pen-sion system as they bring to light any improper or hidden redistribution of benefi ts to privileged groups and reveal who really benefi ts from the legis-lation It also forces contributors to think about the relationship that exists between their contributions, the option to retire at diff erent ages and the amount of pension in the form of a life annuity that they will eventually receive, a nd a ll t hese t hings ma ke pe ople more i nterested i n a nd more knowledgeable about the way the pension system works.*

Th e NDC system allows for special circumstances to be taken into eration—it enables married couples to share notional accounts during certain periods if they have children to look aft er, for example, or periods of military service—but the funding for this must come from the State and general taxa-tion, and the appropriate entries must be made in the notional accounts.According to Holzmann (2006, 2007), these positive features of the notional acco unt s ystem a re su ffi cient r easons f or p utting i t f orward

consid-as a f undamental referent for t he f uture u nifi ed pe nsion s ystem of t he European Union

Finally, following Vidal-Meliá et al (2004), although the notional account system has many positive elements, it also has some characteristics, which

in most cases it shares with the traditional (DB) PAYG system:

1 It d oes n ot f ully de al w ith t he p roblem o f dem ographic cha nge Although it takes the evolution of mortality into account, there is a delay before it does so Pensions are generally calculated only once—

at the time they are awarded—and improvements in life expectancy are not t aken i nto account when pensions t hat have a lready be en

2 If the contributor is free to choose his age of retirement, this may result in an excessive number of early retirements, which, in turn, may cause pressure on the authorities to increase the amount of the guaranteed m inimum pens ion De spite t he ac tuarial ad justment incorporated into the NDC system, Palmer (1999), there is empirical

* Every year in Sweden, all affi liates are sent what is known as the “orange envelope” ing information about the notional account and capitalization, plus a projection of expected benefi ts at ages 61, 65, and 70.

contain-† Th is point is also shared by capitalized pension systems.

428 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

evidence that contributors use a higher personal discount rate than that applied in the actuarial adjustment and tend to retire as soon as they are allowed to, and for this reason a great deal of care needs to

be taken when it comes to establishing the minimum retirement age

In a sc enario with a fi xed contribution rate and a persistent rise in longevity, the size of the pension tends to decrease For this reason, Barr (2006), the minimum retirement age needs to be raised in line with the increase in life expectancy.*

3 If the return on the contributions using the chosen index were less than the return on the c apitalization fu nds—this would be more likely in mixed systems sharing notional and individual capitaliza-tion accounts—then the individual might consider that there was an implicit cost (tax) in the notional accounts equivalent to the diff er-ence in return Whether this would be a co rrect perception or not depends on t he contributor’s deg ree of r isk aversion a nd t he r isk-adjusted return Nominally positive diff erence in y ield could a ft er risk adjustment be negative.†

4 According to Boado-Penas et al (2007), contributors take on the risk

of how the index evolves and are subject to a r isk-return ratio they did not choose, that is, their risk aversion is not taken into account, whereas it is—or at least has the potential to be—in private capital-ization funds It should also be pointed out that the index or indices chosen as notional rates can be h ighly volatile and submit the con-tributor-benefi ciary to more risk than they would willingly take on

5 Th e practical a pplication of t his notional account s ystem t o t he retirement contingency needs to be combined with traditional formulae or i nsurance formulae i n order to cover d isability a nd survivor benefi ts

* Th is is shared by capitalized pension systems.

† Th e Swedish experience shows that since 1995 the average rate of return in the notional type (Inkomstpension) system, measured as the capital-weighted rate of re turn, has been 3.1%

Th e average annual variation in the rate of re turn, as measured by t he standard deviation, has been 1.1 percentage points Since the fi rst payments into the premium pension system

in 1995, t he ave rage re turn of t he pre mium p ension s ystem a ft er t he d eduction of f management fees has been 5.8% Th e annual variation in this rate of re turn, as measured

und-by the standard deviation, has been 14.3 percentage points Th e risk-adjusted return for the Inkomstpension system would be 2.81% and barely 0.41% for the premium pension system

If the return were measured by means that took into account the degree of risk aversion, the comparison would still be more favorable toward the notional account system.

Improving the PAYG Pension Systems ◾ 429

6 Political risk still exists insofar as the system’s parameters could still

be altered Th is risk will be g reater the lower the level of legislative regulation

7 Finally, just like traditional DB systems—as well as most DC systems—NDCs have in practice established a u niform actuarial factor where there is expected income redistribution from those with a shorter than average life expectancy to those with longer life expectancy

It can be concluded from the points mentioned earlier that, despite the fact that it is a clear improvement on the traditional DB PAYG system, the NDC system should be applied whenever possible in conjunction with other instruments such as those that will be studied in the following sections

Th is section looks exclusively at what will be de scribed as the AB of the PAYG system, with particular attention given to the so-called Swedish and U.S models

According to the BOT (2008), a very detailed annual AB—an actuarial report, in fact—has been compiled in the United States since 1941, and from

2002 it has included stochastic methodology Th e so-called U.S AB drawn

up by the U.S social security—similar to that published by the authorities

in Japan every 5 years, Sakamoto (2005) and Canada every 3 years, OSFIC (2005, 2007)—is not a balance sheet in the traditional accounting sense of the term, with a list of assets and liabilities that consider an indefi nite horizon.Compiling a n offi cial AB sheet has been normal practice in Sweden since 2001 Th is AB sheet—in the form it takes in Sweden—has attracted little attention from academics Th is is surprising given that in the litera-ture there are a great many methodologies applied to analyze the viability

or the sustainability of pension systems or to forecast aggregate spending, and this fi eld is of special interest to a number of researchers As far as we are aware, only in the cases of Japan* and Spain† has the AB sheet with its

* Takayama (2005) u ses t he actuarial balance sheet a s a n element to a nalyze proposals for reforming the pension system, although the list he presents of the items it comprises is not very developed.

† Boado-Penas et al (2008) compile an actuarial balance sheet to a ssess the solvency of t he Spanish system a nd compare it to t he Swedish system Th ey also develop the concepts of the contribution asset and the average turnover duration for defi ned-benefi t PAYG systems Vidal-Meliá et al (2009) use the balance sheet of the Spanish system to support the introduc- tion of an automatic balance mechanism (ABM).

430 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

typical structure of assets and liabilities been used by researchers, and on

an offi cial level it has not been used outside Sweden

Based ma inly on i nformation o btained f rom L efebvre (2007), T EPC (2007), and Jimeno et al (2008), the most commonly used methodologies for making aggregate projections of spending on pensions are

1 Aggregate or growth accounting models: According to Domenech and M elguizo (2008), t he a ggregate acco unting a pproach r elies

on making a variety of assumptions regarding the economy as a whole, taking into account future trends in demography (fertility rates, migration flows, and life expectancy), economic conditions (participation an d e mployment r ates, p roductivity, w ages, an d interest rates), and institutional factors (coverage and pension lev-els) Th ese are used mainly for making aggregate projections of spend-ing on pensions Despite the fact that these models are becoming more and more complex as they are made heterogeneous, their main advantage is that they are easy to apply and accurately reproduce the reality of t he pension system T hey a re often referred to by some authors as actuarial models They are also frequently used

by pu blic aut horities a nd or ganizations; t he A geing W orking Group, the technical working group of European Union’s TEPC, which is responsible for spending forecasts, follows this basically deterministic a pproach a lthough n ot a ll t he countries i nvolved apply it

2 Micro-simulation mo dels or pr ojections b ased on i ndividual l ife cycle profi les: Th e working lives of a group of individuals are used

to project how t heir pensions w ill evolve Z aidi a nd R ake (2001) explain t hat t here a re a n umber of va riants: dy namic a nd st atic, micro-simulations with behavior, etc Linked to these micro-simu-lation models are the generational accounting models, such as the one applied in Slovenia It is oft en diffi cult to distinguish certain hybrid models that combine features of this model with those of aggregate accounting models M icro-simulation models a re u sed

in France and Sweden.*

* Projections as to the possible future evolution of the system are made in the annual report on the Swedish pension system, although the results have no infl uence on the parameters that pilot the pension plan Th e actuarial balance sheet, as seen in the next section, is the basic source of information for the Swedish pension system.

Improving the PAYG Pension Systems ◾ 431

3 General equilibrium models: Th e pension system is placed w ithin

an economic environment of general equilibrium with endogenous prices, which generates explicit models of demographic and macro-economic e volution Th e ma in d rawbacks of t hese models a re t he computational complexity, the sensitivity to hypotheses, and a clear shift away from the reality of the pension system, which means that

it is rarely applied by offi cial organizations Holland is an exception where this type of model is applied

4 Indirect models: Based mainly on the IRR or the transfer nent, and usually applied to study intergenerational and intragenera-tional fairness

compo-18.3.1 The Swedish Model

fi t into any of the methods briefl y described earlier It can be described as a

fi nancial statement listing the pension system’s obligations toward tors and pensioners at a particular date, with the amounts of the various assets (fi nancial, real, and through contributions), which back up these obligations

contribu-As B oado-Penas e t a l (2008) a nd Valdés-Prieto (2002) have pointed out, the main aim of the AB sheet is to give a t rue and fair view of the pension system’s capital at the end of each fi scal year and, by comparing these fi gures, to determine t he cha nge in net worth It a lso contributes

to management and external information as it is useful not only for the authority governing the system but also for contributors and pensioners in general, and also for whichever body guarantees payment, that is, for the State along with the contributors it represents

Th e ma in en tries o n t he ba lance sh eet a re ba sically t hose sh own i n Table 18.1

In general terms, it can be said that a PAYG pension system is

reason-ably solvent as l ong as ( fi nancial assets + contribution a sset) ≥ (liability t o

TABLE 18.1 Main Entries on the Actuarial Balance Sheet

of a PAYG System

Contribution asset Liability to contributors

Actuarial losses for the period Actuarial profi ts for the period

432 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

pensioners + liability to contributors) At the date of the balance sheet, if this is the case, participants should have a realistic expectation of receiving the ben-efi ts that have been foreseen, without the system’s sponsor (the State) having

to make periodic contributions Solvency is clearly never completely assured

in the long term as neither assets nor liabilities are known in their entirety

Th e diff erence between the concepts of solvency and sustainability is not self-evident According to K nell et a l (2006), t he ter m sustainability ha s many defi nitions, though it almost always refers to the fi scal policies of a government, the public sector, or the pension system One of the most widely accepted defi nitions in the area of pensions is that of “a position where there

is no need to increase the pension contribution rate in the future.” Th is defi nition applies perfectly to NDC systems but not to DB systems

-The n ovel en try o n t he PAYG ba lance sh eet i s t he o ne c alled t he

“contribution asset.” This is derived from linking the pension system’s assets and liabilities and is the result of a formula that shows the size

of both the assets and the liabilities when the pension system is arially b alanced a nd f inanced by pu re p ay-as-you-go, i n a s cenario defined by the economic, demographic, and legal (pension legislation) conditions of t he accounting period It can be i nterpreted intuitively

actu-as the maximum level of liabilities that can be financed by the existing

contribution rate, without periodic supplements from the sponsor, in a

steady state

Both the assets and the liabilities are valued on the basis of verifi able cross-section facts, that is, no projections are made For example, current longevity is used even though it is expected to increase If and when that expectation ma terializes i n n ew m ortality t ables, t his w ill be i ncorpo-rated into the information on the balance sheet on a y ear-to-year basis Similarly the calculation of the contribution asset does not anticipate that contributions will grow in line with real salaries due to expected economic growth Th is should not be interpreted as a belief that all the basic param-eters determining the items on the balance sheet will remain constant in time, but as a r esult of t he policy of using only verifi able factors in the compilation of the balance sheet Changes are not included until they hap-pen a nd c an be v erifi ed Th e Swedish National S ocial I nsurance B oard (Försäkringskassan 2002) argues that another advantage of this principle

is that it avoids the manipulations and biases that could aff ect any jections Th e Försäkringskassan (2002) also indicates that the economic and demographic forecasts that have to be made i n order to predict the system’s IRR and future variations in average salary are not very accurate

pro-Improving the PAYG Pension Systems ◾ 433

of prediction in the short term with an acceptable degree of certainty or accuracy According to their criteria, the ability to make this type of pre-diction for the long term with the degree of reliability required by the pen-sion system is even more limited

As already mentioned, producing an AB sheet is a practice that has been carried out in Sweden since 2001 Th e main data for the period 2001–2007 are shown in Table 18.2 Th e retirement contingency of the Swedish pension system is mixed, with 86.49% of the contributions being allocated to the PAYG system, N DC t ype, a nd t he other 13.51% to t he defi ned contribu-tion capitalization system Th e balance sheet refers only to the PAYG part, notional type (Inkomstpension), and to the commitments deriving from the old pension system, earnings related benefi ts, known as Allmän Tillägs pension (ATP)

Th e “fi nancial asset” is the value of the fi nancial assets owned by the Swedish pension system at the date of the balance sheet It is valued accord-ing to internationally accepted principles, that is, based on the fi nancial prices of the securities held It is large considering it is a PAYG system, amounting to 29.3% of GDP in 2007

Th e value of the contribution asset is the product of the turnover tion (T D) a nd t he va lue of t he contributions made i n t hat per iod Th e

dura-TD is the time that is expected to pass from when a monetary unit enters the system as a co ntribution u ntil it leaves i n t he form of a pens ion It

is equivalent to t he sum of t he weighted pay-in and pay-out durations*

of one monetary u nit i n t he s ystem for t he year’s contributions, a nd is based on population data obtained from a cross-section, not a projection

In S weden, t o l imit fl uctuations i n t he pe nsion s ystem’s a nnual r esult, the contribution fl ow used in the calculation of the contribution asset is smoothed If the population declines (increases), there is a risk that the accounts will (slightly) overstate (understate) the system’s assets in rela-tion to its liabilities, since in such a case the TD is (slightly) overestimated (underestimated) However, as t he ba lance sheet is compiled every year according to verifi able data, it tends to provide a t rue and fair view Th e stationary demographic and economic state is for sure not ex post fac to true, but because successive changes are included as they are registered in successive balance sheets, the solvency indicator remains reliable.†

† See Auerbach and Lee (2006).

TABLE 18.2 Balance Sheet of the Swedish Pension System at December 31 of Each Year (ATP and Inkomstpension) for the Period 2001–2007,

in Million of Swedish Krona (SEK)

Assets

Financial asset (F) 898,472 857,937 769,190 646,200 576,937 487,539 565,171 Contribution asset (CA) 6,115,970 5,944,638 5,720,678 5,606,592 5,465,074 5,292,764 5,085,252 Actuarial losses ( Table 18.3 ) 81,607 — — 49,029 — 166,762 —

Total assets 7,096,049 6,802,575 6,489,868 6,301,821 6,042,011 5,947,065 5,650,423

Liabilities

Liabilities to contributors (AD) 4,909,569 4,750,749 4,612,959 4,486,030 4,313,706 4,157,021 3,942,873 Liabilities to pensioners (DD) 2,086,915 1,952,261 1,848,517 1,757,979 1,670,493 1,571,637 1,489,143 Accumulated surplus 99,565 28,392 8,783 57,812 51,645 218,407 218,407

Total liabilities 7,096,049 6,802,575 6,489,868 6,301,821 6,042,011 5,947,065 5,650,423 GDP (in millions of SEK)

GDP a 3,070,591 2,899,653 2,735,218 2,624,964 2,515,150 2,420,761 2,326,176

Liabilities to contributors/

liabilities (%) 70.2 70.9 71.4 71.8 72.1 72.6 72.6

Sources: Th e Swedish Pension System Annual Report 2001 Ed O Settergren, National Social Insurance Board (Försäkringskassan), Stockholm, Sweden,

2002; Th e Swedish Pension System Annual Report 2002 Ed O Settergren, National Social Insurance Board (Försäkringskassan), Stockholm, Sweden, 2003; Th e Swedish Pension System Annual Report 2002 Ed O Settergren, National Social Insurance Board (Försäkringskassan), Stockholm, S weden, 2004; Th e S wedish P ension S ystem Ann ual Rep ort 2002 Ed O S ettergren, N ational S ocial I nsurance B oard (Försäkringskassan), Stockholm, Sweden, 2005; Th e Swedish Pension System Annual Report 2002 Ed O Settergren, National Social Insurance Board (Försäkringskassan), Stockholm, Sweden, 2006; Th e Swedish Pension System Annual Report 2002 Ed O Settergren, National Social Insurance Board (Försäkringskassan), Stockholm, Sweden, 2007; Th e Swedish Pension System Annual Report 2002 Ed O Settergren, National Social Insurance Board (Försäkringskassan), Stockholm, Sweden, 2008 and own.

Note: Totals do not necessarily equal the sums of rounded components.

a Th is fi gure is introduced to give the reader some idea of the size of the pension system in relation to the size of the Swedish economy.

436 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

Th e “ liability to contributors” is t he notional c apital acc umulated i n the contributors’ accounts and that deriving from commitments to con-tributors under the old system (see Appendix 18.A.3), and the “liability to pensioners” is the present value of the amount of all pensions in payment

to current pensioners, taking into account current life expectancy and the technical interest rate that was applied* (1.6%) when the amount of the initial pension was calculated and which is subsequently deducted from the yearly indexation of the pension Th e liability to contributors amounts

to 70.2% of total liabilities

As can be seen from the balance sheet (Table 18.2), the Swedish system’s

degree of capitalization (F/(AD + DD)) is high, amounting to 12.84% of the

liabilities i n 2 007 Th is enables a ny pos sible yearly i mbalances be tween the s ystem’s i ncome a nd ex penditure to be de alt w ith by sel ling fi nan-cial assets, which would make the need for outside funding from the state unlikely

Th e accumulated surplus is the “accumulated profi t” or net worth of the pension system, which is owned by the system’s sponsor, in this case the State As can be seen in Table 18.3, the system’s annual profi t or loss is the diff erence between the increase in assets and the increase in liabilities during the period Th e loss is also identical to the increase in the accu-mulated defi cit or the reduction in the accumulated surplus, depending

on the situation It is important not to confuse this profi t or loss with the annual cash defi cit or surplus In Table 18.3, the cash defi cit or surplus is the diff erence between the contributions received and the pensions paid

In 2 007, t he c ash su rplus a mounted t o (190,416–185,653) 4 ,763 m illion Swedish Krona, approximately 0.16% of GDP for that year Th e system had

“losses” in 2002, 2004, and 2007, and “profi ts” in 2003, 2005, and 2006

the diff erence between all the assets and liabilities as a whole

Th e main reason for va luing assets and liabilities w ithout taking t he future into consideration is that the system’s fi nancial solvency does not depend on the amount of assets and liabilities taken separately, but on the ratio between them as measured by the solvency ratio

* Th e classic actuarial discount factor is not u sed Th e Försäkringskassan (2002) shows that

a so-called economic divisor is used which takes into account the amount of pensions to be paid at each age for each individual, and which in the case of Sweden supplies a slightly diff er- ent value to that of the classic actuarial discount factor Th e economic divisor is coherent with the defi nition of the TD, in which the ages of the working and retired affi liates are weighted

by their economic amounts (contributions and pensions) See more in Appendix 18.A.3.

Improving the PAYG Pension Systems ◾ 437

Th e so lvency r atio i ndicator u sed (Table 18.2) em erges f rom t he A B sheet and is expressed as

Assets(Financial Contribution)Solvency ratio

TABLE 18.3 Actuarial Income Statement at December 31, 2007

(Inkomstpension + ATP), Millions of SEK

Fund Assets (Changes) 40,535 293,474 Pension Liabilities (Changes)

credit and ATP points a Pension

Source: Own based on Th e Swedish Pension System Annual Report 2002 Ed O Settergren,

National Social Insurance Board (Försäkringskassan), Stockholm, Sweden, 2008.

Note: Totals do not necessarily equal the sums of rounded components.

a Th is does not coincide exactly with the contributions made due to the fact that the sion points will co ntinue to be earned in t he old (DB) system (ATP) until 2018 F rom then on the value of contributions will equal the value of pension credits, with only small diff erences due to administrative reasons.

pen-b Th is is derived basically from changes in the average salary as measured by the income index us ed f or indexin g no tional acco unts a nd p ensions (t he la tter wi th t he 1.6% reduction).

c Th is is derived from changes in life expectancy as measured by the so-called economic divisors See Appendix 18.A.3

d Th e pension balances of deceased persons (inheritance gains arising) are distributed to survivors of the same age Th is distribution is made as a p ercentage increase in pension balances by an inheritance gain factor Due to lags in t he recording and distribution of inheritance gains and the fact that, for persons over 60, statistically measured rather than actually incurred mortality is us ed, there are small dis crepancies between inheritance gains arising and inheritance gains distributed.

438 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

Th e so lvency r atio u sed i n S weden ha s a d ouble p urpose: t o m easure whether t he s ystem c an f ulfi ll its obligations to its contributors a nd to

is solvent i n t he sense t hat, at t he d ate of t he ba lance sheet, t he sion l iabilities c an reasonably be co vered by t he fl ow of i ncome f rom contributions It is therefore clear that if only the obligations had be en valued, the diagnosis would have been quite diff erent, showing a bank-rupt, i nsolvent s ystem.* Judging f rom t he ba lance sh eet, co ntributors and pensioners have reasonable expectations that they will receive the foreseen pensions

pen-Th e e volution o f t he so lvency r atio o ver t he la st f ew y ears ha s be en slightly negative Should this trend continue, then the automatic adjust-ment mechanism will activate within the next few years Th is would be a real test of the system’s political solidity

Projections of the system’s possible future evolution are in fact made

in the Swedish pension system’s annual report A projection is made of the ba lance sheet itself, of t he a mount in t he reserve or “ buff er fund” and the cash defi cit or surplus, on the basis of three possible scenarios—normal, pessimistic, a nd optimistic—which provide va luable i nforma-tion However, this information is not used in the preparation of the AB sheet It would be diffi cult to justify a r eduction in pensions in real or nominal terms or a decrease in the expected value of contributions made

on the basis of a projection (or projected balance sheet) that may or may not be accurate

As can be se en in Figure 18.1, in t he base scenario t he ba lance ratio

is never less than one, and the fi nancial position of the system ens f rom ye ar to ye ar A ft er 2 037, t he ba lance r atio exceeds 1.1, a l evel which, as proposed by the government report “Distribution of Surpluses

strength-in the Inkomstpension System,” means that there is a distributable surplus between contributors and pensioners In the pessimistic scenario, the bal-ance ratio falls below 1.0 in 2025, and consequently balancing is triggered With balancing, the system’s liabilities accrue interest at the same rate as the growth in the system’s assets As a result, the balance ratio stabilizes

at around 1.0

* As Barr and Diamond (2008) point out, any analysis that looks only at future liabilities (i.e., future pension payments) while ignoring explicit assets and the implicit asset arising from the government’s ability to levy taxes is misleading.

Improving the PAYG Pension Systems ◾ 439

18.3.2 The U.S Model

pro-jection model of aggregate accounting spending on pensions or an ial model Basically it i nvolves using t he forecast demographic scenario to determine the future evolution of the number of contributors and pensioners according to the rules of the pension system Th e macroeconomic scenario that determines the amounts of future contributions and pensions is exogenous

actuar-Th e AB of the U.S old-age and survivors insurance (OASI) and disability insurance (DI) social security programs is aimed at measuring the system’s

fi nancial solvency with a 75-year time horizon It measures the diff erence

in present va lue, d iscounted by t he projected y ield on t rust f und a ssets, between spending on pensions and income from contributions, expressed

as a percentage of the present value of the contribution bases for that time horizon, taking into account that the level of fi nancial reserves (trust fund)

at the eff ective date reaches a m inimum value Th e value summarizes the system’s fi nancial defi cit or surplus for the 75-year horizon and only for that horizon, and it therefore allows for a sharp jump in the contribution rate or

in pension payments at the end of the 75-year period, and the winding-up

of the trust fund on that date If the balance is negative, the fi gure can be interpreted as the increase that would need to be applied to the contribution rate—immediately from that moment—in order to fi nance predicted ben-efi ts until the end of the 75-year period Th e balance can also be expressed

as the decrease in pensions, to be applied immediately, that would be needed for the contribution rate not to change within the next 75 years

Balance ratio (historical evolution and scenarios)

440 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

TABLE 18.4 Elements of the 75-Year Actuarial Balance under Intermediate Assumptions (2008–2082) Present Value as of January 1, 2008, in Billions

of Dollars a

Details in BOT (2008) Deterministic actuarial balance

Stochastic AB (TH 75 years) −1.79%

Source: Board of Trustees, Federal old-age and sur vivors insurance and

dis-ability insurance trust funds (BOT), 2007 Annual Report, Government Printing Offi ce, Washington, DC, 2008 and own.

Note: Totals do not necessarily equal the sums of rounded components.

Bold values are the main elements of the U.S actuarial balance.

a A billion dollars is equal to $1 × 109, and a billion euros would be 1 × 1012 €

A trillion dollars is equal to $1 × 1012, and a trillion Euros would be the much greater fi gure of 1 × 1018 €.

b Present value of the debt that would have to be incurred to fund the payments that have been promised, wiping out all the fi nancial assets Th is should not be confused with the implicit debt of the system at a particular date.

c Th e calculation of the actuarial balance includes the cost of accumulating a get trust fund balance equal to 100% of annual cost by the end of the period.

tar-d Th is represents 32.56% o f GDP in 2008 ($14,445 b illion), or 0.61% o f the present value of the GDPs for the period 2008–2082 ($768.4 trillion).

Th e report from which this AB is compiled is actually much more detailed I t co ntains a co mplete a nalysis o f t he a ssumptions u sed, t he underlying methods and the long-term sensitivity of the main assump-tions, and a stochastic balance sheet is drawn up too

According to the AB on December 31, 2007 (see Table 18.4), the tem could regain fi nancial solvency in 75 years if a 1.70-point increase to

sys-Improving the PAYG Pension Systems ◾ 441

the contribution rate were to be implemented immediately, applied to able earnings Th e AB (see Appendix 18.A.4) is the diff erence between the summarized income rate and the summarized cost rate over a given valu-ation period Th e adjusted summarized income rate (13.94%) equals the ratio of (a) the sum of the trust fund balance at the beginning of the period plus the present value of the total income from taxes during the period,* to (b) the present value of the taxable payroll for the years in the period Th e adjusted summarized cost rate (15.63%) is equal to the ratio of (a) the sum

tax-of the present value tax-of the cost during the period plus the present value tax-of the targeted ending trust fund level, to (b) the present value of the taxable payroll during the projection period

Similarly, all expected payments could be made up to 2082 if an the-board cut of 11.5% were imposed on benefi ts or an allocation of $4.3 trillion were made to the “trust fund.” Naturally, a combination of both these measures could be made i nstead In terms of the annual defi cit or surplus, a cash defi cit is forecast to appear in 2017 and the reserve fund is expected to be exhausted in 2041

unbaked liabilities are estimated at 13.6 trillion dollars Th e AB estimated using stochastic methodology for a 7 5-year t ime horizon g ives a r esult

of −1.79% for t he 50 th percentile, u nbaked l iabilities total $ 4.6 t rillion, and the reserve fund is forecast to be exhausted in 2041, that is, everything

is very similar to the result of the determinist AB in the case of ate assumptions Th e confi dence interval of 95% indicates that the value

intermedi-of the AB swings between −3.52% and −0.28%, and this range is smaller than would be the case for the best and worst assumptions for the system (−4.66% y + 0.52%)

Figure 18.2 shows in the second vertical axis the evolution of the AB over the last 26 years Although the value of the balance for 2008 is the best for the last 15 years, comparisons between values are not homogeneous due to the fact that practically every year methodological improvements are i ncorporated, wh ich prevent d irect comparison, a lthough details of the changes and their year-on-year eff ects are supplied in the report from which the AB is drawn Even so, the chart gives an eff ective summary of each year’s expectations as to the evolution of the system’s fi nancial health over the following 75 years

* Th is represents 15.49% of GDP in 2008.

442 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

Table 18.5 shows the past and forecast evolution for the intermediate assumption of some of the key elements that have an eff ect on determin-ing the AB

Th e income rate is the ratio of income from tax revenues on a liability basis (payroll tax contributions and income from the taxation of scheduled

Historical evolution of OASDI actuarial balance estimates (1982–2008)

1982 1984 1986 1988 1990 1992 1994

Year Income Costs Actuarial balance

1996 1998 2000 2002 2004 2006 2008

0 –0.50 –1 –1.50 –2 –2.50

TABLE 18.5 Past Figures and Forecast Evolution (Intermediate Assumption)

of Some of the Basic OASDI Indicators

Years Rate (%) Income Rate (%) Cost Balance (%) Annual Contributors/ Benefi ciaries Ratio (Years) Trust Fund

Source: Board of Trustees, Federal old-age and survivors insurance and disability insurance

trust funds (BOT), 2007 Annual Report, Government Printing Offi ce, Washington, D.C., 2008 and own.

Improving the PAYG Pension Systems ◾ 443

benefi ts) to the OASDI taxable payroll for the year Th e cost rate for a year

is the ratio of the cost of the program to the taxable payroll for the year

Th e annual balance is the diff erence between the income rate and the cost rate in a given year As can be seen in the table, the annual balance wors-ens considerably in that the ratio of contributors to benefi ciaries decreases due to the retirement of the large baby-boom generation between about

2010 and 2030 Aft er 2030, increases in life expectancy and the relatively low fertility rates since the baby boom cause the level of the accumulated

fi nancial reserves to fall noticeably until they become exhausted, as shown

in the value of the “trust fund ratio,” defi ned as the assets at the beginning

of a year (including advance tax transfers if any) expressed as a percentage of the cost during the year Th e trust fund ratio represents the proportion of a year’s cost, which could be paid with the funds available at the beginning

of a year

If no action were taken until the combined trust funds become exhausted

in 2041, then the eff ects of changes would be more concentrated in fewer years and fewer cohorts For example, payroll taxes could be raised to fully

fi nance scheduled benefi ts in every year starting in 2041 In this case, the payroll tax would be increased to 15.94% at the point of trust fund exhaus-tion in 2041 and continue rising to 16.60% in 2082 Similarly, benefi ts could

be reduced to the level that is payable with scheduled tax rates in each year beginning in 2041 Under this scenario, benefi ts would be reduced by 22% at the point of trust fund exhaustion in 2041, with reductions reaching 25% in

2082 Either of these examples would eliminate the shortfall for the 75-year period as a whole by specifi cally eliminating annual defi cits aft er trust fund exhaustion C onsequently, ensu ring t he s ystem’s so lvency be yond 2 082 would probably require further changes beyond those expected to be needed for 2082 All this can be seen in Figure 18.3

To summarize, the main diff erences between the Swedish and U.S AB sheets are the following:

1 Projections of demographic, economic, a nd fi nancial va riables a re made for a 75-year period in the United States, whereas in Sweden a valuation system based on verifi able facts at the eff ective date of the balance sheet is used

2 In Sweden, t he co ntribution a sset i s q uantifi ed for a h ypothetical steady state derived f rom present economic a nd demographic fac-tors, while in the United States contributions are estimated for the next 75 years

444 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling

3 Th e AB in the United States is dependent on the market interest rate, but the Swedish balance sheet, being independent, is not

4 Th e S wedish ba lance sh eet f ollows t he t raditional st ructure o f t he accounting balance sheet deriving from principles of double-entry bookkeeping and has a very strong actuarial profi le as it includes as lia-bilities its commitments to both pensioners and contributors, while the U.S balance has a more fi nancial profi le as its commitments to contrib-utors are not quantifi ed until those contributors become pensioners

5 Every year the information deriving from the Swedish balance sheet has an eff ect on the indexation of the contributions registered in the notional accounts and on the rate of variation of pensions in pay-ment, while the U.S AB has no immediate eff ect but serves as an ele-ment to provoke thought and analysis for possible legislative reform

of the pension system

6 Th e solvency indicators for the system emerging from both balance sheets can be considered as complementary, although they are con-ceived a nd co mposed i n cl early d iff erent wa ys Th e U.S so lvency indicator is no more than the sum in the present value of the cash defi cits o r su rpluses f or t he t ime per iod co nsidered, t o wh ich a re added the value of the fi nancial assets at the beginning of the valua-tion period Th e Swedish indicator concentrates more on evaluating the commitments it has with both pensioners and contributors

OASDI income rates under intermediate assumptions

Trust fund exhausted (2041)

Payable benefits as % of scheduled benefits 2008–2040; 100%, 2041: 78%, 2082: 75%

Year Income Cost: scheduled and payable benefits Cost: not fully payable benefits

FIGURE 18.3 OASDI income and cost rates under intermediate assumptions as

a percentage of taxable payrolls

Improving the PAYG Pension Systems ◾ 445

18.4 AUTOMATIC BALANCE MECHANISM (ABMs)

According to the American Academy of Actuaries (AAA 2002), the fi rst proposal for an ABM for the PAYG pension system came from actuary Robert J Myers in 1982, while he was the head of the National Commission for Social Security Reform in the United States

Vidal-Meliá et al (2009) explain that the ABM is a set of predetermined measures established by law to be applied immediately as required accord-ing to the solvency or sustainability indicator Its purpose, through suc-cessive application, is to re-establish t he fi nancial equilibrium of PAYG pension systems with the aim of making those systems viable without the repeated intervention of the legislators In other words, they are used to depoliticize the management of the PAYG system by adopting measures with a long-term planning horizon, which will bring about greater inter-generational fa irness a nd re-establish t he sustainability or t he fi nancial solvency of the system Th e aim of the ABM is to provide what could be called “automatic fi nancial stability,” which Valdés-Prieto (2000) defi nes

as “the capacity of a pension system to adapt to fi nancial turbulence out legislative intervention,” it being understood that turbulence can be caused by economic and/or demographic shocks that have an eff ect on the system’s solvency or fi nancial equilibrium

Automation: Decisions that are taken to deal with possible

situa-•

tions of insolvency need to be automatic given the political ease of increasing benefi ts or reducing contributions and the political dif-

contribution r ate i ncreases a nd f uture b enefi t dec reases o n t he

books lowers the political cost of preserving balance, since it is

eas-ier to legislate future pain than current pain For Diamond (2004),

relying on f ully automatic adjustment rather t han assuming t hat there w ill be per iodic new legislation be ars some similarity to a familiar distinction from macroeconomics: rules versus discretion for monetary policy

Short-term: As Valdés-Prieto (2000) has pointed out, the strength of