Rethinking the Contribution-Benefit Linkage in

Pay-As-You-Go Retirement Systems Based on

Algerian Case
FARID FLICI* Original scientific paper
Department of Human Development & Social Economics UDK: 368.914(65)
Research Center in Applied Economics for Development doi: 10.3935/rsp.v29i2.1822
- CREAD Received: April 2021
Algiers, Algeria

In Pay-As-You-Go systems, retirement benefits result from the multiplica-

tion of the reference wage by the contribution duration by an annuity rate.
This formulation can undergo some adjustments to ensure sustainability
within a changing environment. In this paper, we showed that the current
contribution-benefit linkage might generate a large gap between individuals
in terms of profitability and thus in terms of wealth accumulation over long
periods. Considering the possible career scenarios in the Algerian retirement
system, we found that return rates vary from 9.6% to 13.2%. To reduce in-
equality, we propose a new retirement benefit formula partially based on the
actuarial fairness principle.
Keywords: retirement, PAYG, equality, IRR, scenario, Algeria.

INTRODUCTION must prioritize some aspects compared

From an individual perspective, pen-
sions are an efficient way to prevent late-
life risks and smooth consumption over
the life cycle (Barr & Diamond, 2006;
Bovenberg & Van Ewijk, 2011a). For gov-
ernments, they represent a tool to alleviate
poverty among the elderly through fund-
ed, unfunded Pay-As-You-Go (PAYG),
hybrid, or multi-pillar schemes. Indeed,
pension systems design in a given country
depends on a set of economic, political,
cultural, and social circumstances (Bov-
enberg & Van Ewijk, 2011b). Experience
suggests that a perfect pension system
filling all objectives in all situations does
not exist (Barr, 2002). Thus, policymakers
to others when designing a pension sys-
tem. PAYG retirement systems are more
efficient than funded systems in reducing
poverty (Cousins, 2007; Weyland, 2009),
but are more sensitive to demographic
risks. In such situations, state interven-
tion is necessary (Barr & Diamond, 2006)
to correct financial imbalances and keep
poverty reduction on target.
In recent decades, population aging
weakened the efficiency of state interven-
tion in maintaining the financial balance
of pension systems and resulted in lower-
ing their generosity. Following the World
Bank [WB] report in 1994 (WB, 1994),
many reforms worldwide were imple-

*
Farid Flici, Research Center in Applied Economics for Development - CREAD, PB. 197, Srt Djamal Eddine
El-Afghani, Rostomia Algiers 16011, Algeria, farid.flici@gmail.com

281
Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022
mented to lighten the load on the classical
PAYG schemes, mainly by introducing ad-
ditional funded pillars. Indeed, a multi-pil-
lar system is the most suitable for a mul-
ti-objectives strategy (Holzman & Hinze,
2005). Some countries, such as Mexico,
have experienced a shift from the PAYG to
a fully-funded system. Under fully funded
systems, the linkage of pension benefits
to contributions obeys actuarial fairness
based on financial market return rates and
life expectancy evolution. What a retiree
receives during retirement is proportional
to the contributions accumulated during
his working career (Castanheira & Galas-
so, 2011). Within the PAYG defined ben-
efits systems, the current working genera-
tion contributes to the current generation
of retirees against a promise to receive
pension benefits once at retirement (Barr &
Diamond, 2006). Keeping these promises
relies on the volume of contributions that
future working generations will be able to
pay. In some circumstances, maintaining
the income-outcome balance will require a
few adjustments, mainly by asking future
workers for more contributions or making
future retirees receive less than promised.
Such parametric reforms result in a sort of
inequality between the successive genera-
tions (Morris, 2022; Piirits & Võrk, 2019;
Miles & Iben, 2000). On the other hand,
various types of inequality can rise within
the same generation, and different types
and sources of inequality exist. Unequal
wages and unequal employment chances
translate directly into unequal retirement
benefits (Klos et al., 2022; Castanheira &
Galasso, 2011).
Although the literature had addressed
most inequality sources in PAYG systems,
issues related to the system design and the
way pension benefits are linked to con-
tributions were not considered. In PAYG
systems, the first pension benefit equals
the product of the replacement rate by
282
Flici F.: Rethinking the Contribution-Benefit Linkage in...
the reference wage. The replacement rate
refers to the years of contribution multi-
plied by an annuity rate. For example, 40
contributed years with an annuity rate of
2% result in a replacement rate of 80%.
Usually, the reference wage corresponds
to the final salary. This formulation, even
though very common in defining pension
benefits under the PAYG system, remains
very debatable since its definition does not
obey well-defined rules.
The first objective of this paper consists
of studying pension inequalities due to the
classical design of PAYG retirement sys-
tems referred to here as “systemic inequal-
ities” that are considered from an actuarial
fairness point of view. To this end, we use
the Algerian public retirement system of
salaried workers as a case study. Assum-
ing that wages are equal for all workers
aged x in time t and that contribution rates
for retirement are constant over time, we
estimate and compare the internal return
rate (IRR) of all the possible scenarios of
contribution-retirement. High dispersion
of the IRRs distribution translates as an
inequality.
The use of the IRR for pension valu-
ation within PAYG systems dates back to
Samuelson (1958) and Aaron (1966). It re-
mains widely used for analyzing pension
generosity (Diseny, 2006; Reznik et al.,
2009; Ben Brahem, 2009; Luthen, 2016),
studying inter- and intra-generational re-
distribution (Klos et al., 2022), and com-
paring PAYG and fully funded systems
returns (Turner, 1998). Because PAYG
systems, unlike funded plans, do not obey
financial market investments, using the
IRR for its evaluation seems inappropriate.
However, such a tool is very performant in
comparing the value of the perceived ben-
efits against the paid contributions. Indeed,
the IRR in PAYG systems translates as the
demographic and economic returns (Ga-
lasso, 2019; Settergren & Mikula, 2005;
 contributions in PAYG OF
THE PROBLEM retirement systems,
SYSTEMIC we
INEQUAL
contribution-retirement scenarios within the system.
To analyze the inequalities related to the cla
IRR translates as an inequality. We consider the
Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022 contributions in thePAYG
Flici F.: Rethinking retirement
Contribution-Benefit systems,
Linkage in... we est
workers managed by the “Caisse Nationale des
contribution-retirement scenarios within the system. As
Brunner 1996). When the population and considered. First,
simplification we give
reasons, a short
only men,overview
who constitute m
salaries grow over time, retirement benefits IRRthetranslates
of Algerian as an inequality.
retirement system.We consider the Al
Then,
will be more generous in the future, mean- theweCNR,
defineare
workers all considered.
the possible
managed First,
by the scenarios,
“Caisse we Nationale
give a short
pres- des ove
Re
ing a high IRR. But, when the population ent the methodology for estimating the
is decreasing or is aging and salaries are Then, we
and define
simplification
IRR, analyze all
thethe
reasons, possible
only
results. men, scenarios,
who constitutepresent
more
stable or are decreasing, one would expect andtheanalyze
CNR, are
Overview theofconsidered.
results.
the Algerian First, we give a short overvi
retirement
future benefits to be less generous. system
The second objective of this paper is to Then, we define all the possible scenarios, present the
propose a new approach for linking retire- TheOverview
retirement of the
system Algerian retirement
in Algeria is system
and analyzeofthe
composed results.
three schemes, one for mil-
ment benefits to contributions aiming at
itary personnel, one for the high ranking
reducing inequalities and improving fair- The retirement
Overview the system inretirement
Algeria system
is compo
ness. Then, we propose a new retirement civil servants andofgovernment
Algerianstaff, and
formula using simple models based on the one for theone
personnel, workers in thehigh
for the civilranking
sector. The civil servants
final wage, the years of contribution, and latter comprises two sub-schemes:
The retirement system in one for is compose
Algeria
workers in the
the salaried civil sector.
employees that isThe latter comprises
managed
the retirement age. personnel, one for the high ranking civil servants an
by “La Caisse
employees that isNationale
managed des Retraites
by “La Caisse Nationa
Section 2 starts with a short overview
of the Algerian PAYG retirement system workers and
(CNR)” in the onecivilfor sector.
the self-employed
The latter comprises tw
for the salaried workers, followed by a
self-employed workers thatbyis“La
workers that is managed Caisseby “La Cais
managed
employees thatSociale
d‘Assurance is manageddes by “La Caisse Nationale d
Non-Salaries
presentation on the definition of the possi- The CNRcoverscovers almost 90%
ble contributions-benefits scenarios. Then,
(CASNOS)”.
(CASNOS)”.
self-employed The CNR
workers that is almost
managed 90% by “Laof the
Caisse
of the
focuses retirees of
on the salariedthe civil scheme.
employees’ This ar-
regime
we expose and implement the methodolo-
focuses The
(CASNOS)”.
ticle on the CNR coversemployees’
salaried almost 90% only.
of the reti
gy of estimating the IRRs for the different
regime
focusesonly.
on the salaried employees’ regime only.
scenarios, and we discuss inequalities due
to the classical way of linking benefits to In In
thethe salaried
salaried workers
workers scheme,scheme,
whichwhich this p
contributions. In Section 3 we propose a thisis paper
age focuses
60Inyears
the old.on,Thethe
salaried statutory
workers
first retire-
scheme,
retirement which
benefitthis
is pap
calc
new method to link benefits to contribu- ment age is 60 years old. The first retire-
age isbenefit
ment 60 years old. The first
is calculated by:retirement benefit is calcul
tion while reducing inequality. We present
and discuss our results in Section 4. Final- 𝑹𝑹𝑹𝑹𝒙𝒙 � 𝟐𝟐. 𝟓𝟓% ∗ 𝒏𝒏 ∗ 𝑾𝑾�𝒙𝒙�𝟓𝟓,𝒙𝒙�
ly, we end with a Conclusion. 𝑹𝑹𝑹𝑹𝒙𝒙 � 𝟐𝟐. 𝟓𝟓% ∗ 𝒏𝒏 ∗ 𝑾𝑾�𝒙𝒙�𝟓𝟓,𝒙𝒙�
(1)
With:With:
With:
THE PROBLEM OF SYSTEMIC • RBx is the retirement benefit paid to a
INEQUALITY 𝑅𝑅𝑅𝑅� isretired
● newly the retirement
person; benefit paid to a newly re
● 𝑅𝑅𝑅𝑅� is the retirement benefit paid to a newly retire
To analyze the inequalities related to • ● x𝑥𝑥represents
represents thethe
ageage
of retirement;
of retirement;
the classical method of linking benefits to ● 𝑥𝑥 represents
• n refers to the years of the age of contribution,
retirement;
contributions in PAYG retirement systems, 𝑛𝑛𝑛𝑛refers
● ●with to
15 years the years of contribution, with 15 y
we estimate the IRR for all the possible refers to theasyears
a minimum to open with
of contribution, 15 year
the right to pension
retirement a retirementand32 pension
32years and
yearsas asa maximum;
a maximum
contribution-retirement scenarios within 32retirement
years as apension
maximum; and
the system. As previously said, high disper-
sion of the IRR translates as an inequality. • ● ●𝑊𝑊𝑊𝑊 isthethe
����,��isis
����,�� average
average
the average wage wage
of theofof
wage 5 thethe5 5years
yea
We consider the Algerian retirement sys- years before retirement (final wage); ∗ ∗
simplification
simplification
for simplification issues,
issues, ititit
issues, isisis
noted
noted
notedasas
as𝑤𝑤𝑤𝑤. .
tem for salaried workers managed by the
“Caisse Nationale des Retraites (CNR)” w.*

as a case study. For simplification reasons, Interested readers can refer to Flici and
only men, who constitute more than 83% Planchet (2020) for a more detailed over-
of the direct pensioners of the CNR, are view of the Algerian retirement system.

283
Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022 Flici F.: Rethinking the Contribution-Benefit Linkage in...

Defining the possible scenarios pected retirement benefits at a given time

Since we are not willing to consider t. Time t is assumed to be the year of start-
the effect of the year of retirement on pen- ing work within each scenario, and IRR(i)
sion inequality, all the hypothetical retir- to be the IRR corresponding to the ith sce-
ees considered here are supposed to retire nario ranging from 1 to 483. All payments
are supposed to occur at the beginning of
in the same year (even though they are of
the year. The contributions for retirement
different ages), which is supposed to be
are paid annually as a part of the annual
2020. Then, the possible scenarios for a
wages. Wages are assumed to evolve line-
contribution-retirement career have been
arly with time and follow a quadratic func-
defined based on two variables:
tion over
baseage.
wage Theused
baseinwage used in cal-
calculating the first retire
a) The duration of contribution, which culating the first retirement benefit is the
varies from 15 to 40 years; salary.
average of theThen, retirement
last five benefits
years’ salary. Then, are annuall
base wage used in calculating the first retirement b
retirement benefits are annually revaluated
b) The age of retirement, which goes
from 50 to 70 years; with a factor
salary. Then, (1+rev).
retirement benefits are annually reval
Even if the statutory retirement age
For each
For each scenario, we scenario,
estimate thewepay-estimate the
is 60 years, many options have allowed mentexpected
schedule ofretirement
the contributions and theWe adopt s
benefits.
getting retired starting from the age of 50 expected For each scenario,
retirement benefits. we
Weestimate
adopt the paym
years. Also, many workers remain occu- some hypotheses
involved
expected inabout the different We
the retirement
retirement benefits. param-
system:
adopt some h
pied beyond 60 and even 65 years in some eters involved in the retirement system:
cases (Flici & Planchet, 2020). involved in the retirement system:
a) Wagesa) Wages
The combination of the two variables
gives a total of 26x21 = 546 scenarios. Since we are not addressing the effect
a) Wages
Some constraints have been imposed: of wage inequalitySincehere,
we are all workers aged
not addressing the effe
x in year t receive the same salary. As-
a) The age of starting contributing is Since we are
between 18 and 55 years;
sumingyearthat wages
𝑡𝑡 receive thenot
will keepaddressing
same growing the
salary. in effect ofthw
Assuming
the future, linearly, following their his-
b) The year of starting contributing to yearfollowing
torical
receive the
𝑡𝑡evolution, same
their
implies
salary. Assuming
historical evolution,that
that the average
wag
implie
be later than 1977; wage (noted their
following w) at the time (t+1)
historical equals theimplies that
evolution,
The first constraint comes from the in- (𝑡𝑡 �wage
average 1) equals the (t)
at the time average wage
multiplied by at the time
teraction between the minimum required (𝑡𝑡evolution
an � 1) equals the(1+wgr),
factor averagewith wage at de-
wgr the time (𝑡𝑡) m
years of contribution (15 years) and the notingwith wagesdenoting
the 𝑤𝑤𝑤𝑤𝑤𝑤 growth rate.theThis
wages
can growth
be rate. T
with 𝑤𝑤𝑤𝑤𝑤𝑤
written as: denoting the wages growth rate. This ca
maximum age for retirement, i.e. 70 years.
The second constraint relies on the fact 𝑤𝑤��� � 𝑤𝑤� � �1 � 𝑤𝑤𝑤𝑤𝑤𝑤�
that the mortality data of Algeria, which is 𝑤𝑤��� � 𝑤𝑤� � �1 � 𝑤𝑤𝑤𝑤𝑤𝑤� (2)
used here as a survival function in the con-
tribution-retirement cycle, is only availa- We note We that, note
for each
that,scenario,
for eachthescenario, th
We note
average wage of thethat, for each
first year scenario, the aver
of contribu-
ble starting from 1977.
is
tion 𝑤𝑤����� is assumed
assumed toto equal
equal
𝑤𝑤����� is assumed to equal to 1.
to 1.
to 1.
After removing the impossible combi-
nations, the final number of scenarios is On the other hand, wages vary with
age. Flici and Planchet (2020) used the
483.
OnofOn
distribution the the other
otherby
salaries largehand,
hand, wages
age wages
inter- vary a
vary with
Assumptions vals distribution
provided by of thesalaries
household incomeage intervals
distribution by large
The IRR represents the discount rate survey by the of salaries
Algerian by large
Office age intervals provi
of National
which allows equalizing the sum of the Algerian
Statistics
Algerian [ONS]
Office Office
in of
2011 of National
(ONS,
National 2014). Statistics
Un- [ONS][ONS
Statistics in 20
expected contributions and that of the ex- fortunately, no data is available for more
available
available for more
for more recent
recent years.years.
Here, Here,
the agethe a
dist
284 same for the
same for whole period
the whole of analysis.
period of analysis.

FliciFlici
and and
Planchet (2020)
Planchet fitted fitted
(2020) the crude
the d
 We note that, for each scenario, the shared
average wage the
between of the firstperson
insured year of contribution
himself (6.75%), the e
𝑤𝑤����� is assumed to equal to 1. (0.5%). Historically, the CRR evolved from 5% in 1985
On the other hand, wages vary with age. Flici and Planchet (2020) used the
with Then it underwent successive revisions in 1999, 2000 a
butionage. Flici
of salaries and
by
Rev. soc. polit.,
Planchet
Onlarge
god.the age
other
29, br.
(2020)
2, str.intervals
used
provided
hand,Zagreb
281-295, wages
the with
2022 varyby the Flici
household
age. Flici and
F.: Rethinking income survey
Planchet by the
(2020)
the Contribution-Benefit Linkage used the
in...
and 17.25%, respectively.
provided
ian Officebyofdistribution
the household
National of income
Statistics
salaries by survey
[ONS]
large in by
age 2011the(ONS,
intervals 2014).
provided by Unfortunately,
the household incomeno data is by the
survey
recent years. Here, the age distribution of inequality in this article, the CRR is as-
in 2011
able (ONS,
for more 2014).
Algerian
recent
wages Unfortunately,
Office
years.
is assumed oftoNational
Here, the the
remain nosame
age datafor
Statistics is sumed
[ONS]
distribution Since
inof2011
to(ONS,
wages weassumed
do not aim
2014).
be isconstant totoremain
address the
atUnfortunately,
18.25%
the
noeffects
over the data isof para
the
efordistribution whole
of period
wages
available of analysis.
is assumed
for more to remain
recent years. the age1977-2019
Here, the indistributionperiod.
this article, of Thus,
thewages
CRR we can calculate
isisassumed
assumed totoberemain
constant
theat 18.25
the whole period of analysis. the the annual contribution to retirement
Flici and Planchet (2020) fitted the we can calculate the the annual contribution to retirement
same for
crude the wholeofperiod
distribution wagesofby analysis.
large age (CR) as follows:
Flici and Planchet (2020) fitted the crude distribution
intervals with a quadratic function, which of wages by large age intervals
allowed Flici them and Planchetthe(2020)
to estimate wage fittedpatternthe crude � 𝑤𝑤�,� ∗ 𝐶𝐶𝐶𝐶𝐶𝐶
𝐶𝐶𝐶𝐶�,�distribution of wages by large age(5) intervals
rude distribution
a quadratic by function, of wages
detailedwhich
by
ages ranging
large
allowed age
from them intervals
18 to to estimate
64. the wage pattern by detailed ages
with a quadratic function, which allowed them to Additionally,
estimate
Additionally, thethewage contribution
the pattern by period
contribution detailed
period isages
assumed
em to estimate Here, the thewage
same patternfunctionby is used
detailed to extendages is assumed to be continuous and concen-
ng from 18 the to 64.
ranging Here,
wage from pattern the
18until same
to 64. theHere,function
age ofthe 69same is used
years.function to extend
is used the
to wage
extend pattern
the wage until
patternthe until the
towardtoward
trated the endthe of end
the working age. Also,
of the working age.it is not assum
tion
f 69 is usedHowever,
years. to extend
However, ourthe
our wageofpattern
variable
variable interest until
of interestbecomes the Also,
becomes the it ratio
is not of the
assumed wage to beat age
limited 𝑥𝑥 and
to 32 𝑥𝑥 and
ageratio
the of 69 ofyears.
the wage However,
at age xour andvariable
time t onof interest currently becomes regulated, the ratio
but can of the wage
reach at age
40 years.
years as it is currently regulated, but can
𝑡𝑡erest becomes the ratiowage of the wage atatage asas𝑥𝑥 𝑡𝑡𝑤𝑤and
𝑤𝑤𝑥𝑥,𝑡𝑡 𝑤𝑤𝑥𝑥,𝑡𝑡
on the average
the 𝑡𝑡wage
timeaverage on the at year
at year
average (noted
t (noted
𝑡𝑡wage year ) instead
(noted as 𝑤𝑤of
reach 40) 𝑤𝑤 �,� . This
instead
years. of 𝑤𝑤can be written
�,� . This can be as written as
𝑤𝑤𝑥𝑥,𝑡𝑡 𝑡𝑡 c) 𝑡𝑡 Survival functions
as ) instead
insteadof of 𝑤𝑤 . This Thiscan can be be written
written as fol- as
ws: 𝑤𝑤𝑡𝑡 follows: �,�
lows: c) Survival functions
The payment of contributions and retirement bene
The payment of contributions and re-
𝑤𝑤 � 𝑤𝑤� � �� �� � � 𝑥𝑥 � � � 𝑥𝑥 � � (3) payment dates.
tirement benefits Longevity is assumed
is conditional (3)
to keep improvi
� 𝑤𝑤� � �� � ��,� � 𝑥𝑥 � � � 𝑥𝑥 � (3) on sur-
viving to the payment dates. Longevity
females on one side and the global population and the ret
Combining Equations 2 and (3)3 allows is assumed to keepthe improving at different
Combining Equations
for estimating the year-to-year evolution 2 and 3 allows for estimating year-to-year evolution of
Equations 2 and 3 allows for estimating paces for
the males and
year-to-year females on
evolution one side
Combining of wages
wages at at
allall ages.
ages. ThatThatgives:
gives: and theBecause of the unavailability
global population and the retirees ofofspecific life t
ws
s atfor estimating
all ages. That gives: the year-to-year evolution of on the other
Algeria, side. the mortality of the contributing wo
we assume
� ∗ �1 � 𝑤𝑤���(4)
������∗�∗��
𝑤𝑤����,���� � 𝑤𝑤��,�� ∗ �1 � (4)
����∗���∗� � �
Because
population. To of this
the end,
unavailability
we use theofcoherent
spe- forecast o
With:
With: 𝑎𝑎 � 0.5244, � � 0.0196, 𝑎𝑎𝑎𝑎𝑎𝑎 � � � 0.00016.
cific life tables for the insured population
during
in Algeria, the contribution
we assume the phase. For theofpayment
mortality the phase, we
a=0.5244, b=0.0196, and c=- 0.00016.
The annual wage growth rate is assumed to keep at 9.5%. contributing
For more details,workers see Flici is similar to that of
life tables estimated by Flici and Planchet (2019) based
The annual wage growth rate is as- the global population. To this end, we use
and Planchet (2020).
sumed to keep at 9.5%. For more details, males
the and females
coherent forecastagedof50Flici years(2021)
and older.
as a
seeb)Flici and Planchet (2020).
Contributions survival function during the contribution
phase. For the payment phase, we use the
b) Workers
Contributions
pay the contributions for retirement as a part of the contribution to social
retirees-specific prospective life tables
security. Since late 2015, the contribution rate for retirement (CRR) has equaled 18.25%
Workers pay the contributions for re- estimated by Flici and Planchet (2019)
tirement
shared between as athepart of the
insured personcontribution
himself (6.75%), totheso- employerbased
(11%), and on the thegovernment
mortality data of the CNR
cial
(0.5%).security. Since
Historically, lateevolved
the CRR 2015,from the 5% contribu-
in 1985 to 11% for males
in 1991 and infemales
and 12.5% 1998. aged 50 years and
tion
Then rate for retirement
it underwent (CRR)inhas
successive revisions 1999, equaled
2000 and 2006,older. taking it up to 14%, 16%,
18.25%
and 17.25%, shared between the insured person
respectively. Estimating the IRR passes through dis-
himself (6.75%), the employer (11%), and counting contributions and benefits to a
Since we do not(0.5%).
the government aim to address Estimating
the effects of parametric
Historically, the Estimating
reforms onthe IRR
pension
unique time reference, the passes
IRR in
inequality through
passes
our case
discounting contri
through
the yeardiscounting
CRR evolved
in this article, the CRRfrom 5% into1985
is assumed to 11%
be constant at 18.25% in over of the 1977-2019 period. Thus,
starting work. Two different
1991 and 12.5%
we can calculate in 1998.
the the annual Thentime
contribution reference,
time
toitretirement
under- (CR) in
reference,
as follows: ourincase ourthe case yearthe of
year ofsurvival
starting work.work.
starting Two diff
Tw
functions are used for this issue, one spe-
went successive revisions in 1999, for this 2000
forissue, onetospecific
this cific
issue, the
onecontributionto thetocontribution
specific phase and another
the contribution phasephaseand anoan
𝐶𝐶𝐶𝐶�,� � 𝑤𝑤�,� ∗ 𝐶𝐶𝐶𝐶𝐶𝐶 (5)
and 2006, taking it up to 14%, 16%, and one to the retirement phase. Considering
17.25%, respectively. � �
Additionally, the contribution periodConsidering Considering
is assumed 𝑞𝑞�,� to
to be continuous to𝑞𝑞�,� be the
be
and tothebe age-specific
the age-specific
age-specific
concentrated mortality
mortality rate rate of
mortality theo
rate
Since
toward the endweof the
doworking
not aim to address
age. Also, the to beoflimited
it is not assumed the global
to 32 years population
as it is at age x and year
�
effects of parametric
currently regulated, but can reachreforms
40 years.on year 𝑡𝑡year
pension and𝑡𝑡 𝑞𝑞tandandto
�,� 𝑞𝑞�,�� be the mortality rate for the retired p
to be bethethe mortality
mortality rate forratethefor the reti
c) Survival functions probability to survive
probability from from
to survive the first
the age
firstofage
contribution (𝑥𝑥)
of contributio
285
phasephase
(𝑥𝑥 � 𝑛𝑛)
(𝑥𝑥 as
� follows:
𝑛𝑛) as follows:
The payment of contributions and retirement benefits is conditional on surviving to the
payment dates. Longevity is assumed to keep improving at different paces for males and
females on one side and the global population and the retirees on the other side. �����,�����
�����,�����
 in our case the year of starting work. Two different survival functions are used
ce, in our casefor
the this
year issue, one work.
of starting specific
Two todifferent
the contribution
survival phase and
functions
𝑅𝑅𝑅𝑅� � 2.5% are another
𝑛𝑛𝑛 𝑤𝑤 one
∗used ∗ to the retirement phase.
ne specific to the contribution phase and another one to the retirement phase.
𝑅𝑅𝑅𝑅� � 2.5% ∗ 𝑛𝑛𝑛𝑞𝑞𝑤𝑤
Considering
� ∗ be the
�,� tophase andage-specific mortality rate ofphase.
the global population at age (6)𝑥𝑥 and
�e, one specific to the contribution another one to the retirement
,� �to be the Rev.
age-specific mortality rate of the global population
�br. 2, str. 281-295, Zagreb 2022 With: at age 𝑥𝑥 and
𝑞𝑞�,� to be theyear and
age-specific
𝑡𝑡
soc. polit., to
mortality
𝑞𝑞
god. 29, be the
rate ofmortality
the global rate for
populationthe
Flici retired
atRethinking
F.: age 𝑥𝑥 and population,
the weLinkage
Contribution-Benefit canin...estimate the
With: �,�
� �to be the mortality rate for the retired population, we can estimate the
𝑞𝑞�,� to be theprobability
mortality rate for the from
to survive retiredthepopulation,
first age of we contribution
can𝑟𝑟:estimate
The age theof
(𝑥𝑥) toretirement;
any age of the contribution
survive fromretiredthe
𝑟𝑟:theThe
population, we can estimate the
first age
age of contribution
ofofretirement; (𝑥𝑥) to any r: The age of retirement;
age of the contribution
o survive from probability
phase first
(𝑥𝑥 �toage
𝑛𝑛)survive contribution
from the first
as follows: (𝑥𝑥) age
to any
of agewof * the𝑤𝑤contribution
: represents
∗ the finalthe
: represents wage averaged
final on
wage averaged on the
s follows: contribution (x) to any age of the contribu- ( (𝑤𝑤 ∗ � � ∑���
𝑛𝑛) as follows: 𝑤𝑤 : represents the final wage averaged on the last
∗
last 55years
years 𝑤𝑤 ;
);
tion phase (x+n) as follows: �����,�����𝑛𝑛: The duration�of contribution;
����� �
n: The duration of contribution;
𝑛𝑛: The duration �����,����� of contribution; �
� 𝑝𝑝�,� � �Then, the �1retirement
� 𝑞𝑞�,� � benefit of the next
�����,�����
��
� 𝑝𝑝 𝑝𝑝�,���
��,� �� �1��𝑞𝑞𝑞𝑞�,�
�1 �,�� � Then,
year
���,��� is the retirement
calculated benefit of
by applying an the next year is ca
annual
Then, the retirement benefit
���,���
���,��� of the next year is calculated
revaluation by
rate applying
(rev) of 5%an annual
(Flici & revaluation
Plan-
Similarly, the probability of surviving ratefrom
chet, (𝑟𝑟𝑟𝑟𝑟𝑟)
2020).theof 5%age
Itfirst
gives: (Flici
of & Planchet, 2020).
contribution (𝑥𝑥) to It
anygives:
age
larly, rate Similarly,
(𝑟𝑟𝑟𝑟𝑟𝑟) of 5% the
(Flici probability
& Planchet,of surviving
2020). It gives:
y, thethe probabilityofofsurviving
probability
from
surviving from from the
the first
first age
age ofofcontribution
contribution (𝑥𝑥)(𝑥𝑥)
to any
to ageage
any
(𝑥𝑥 � the 𝑠𝑠) offirsttheage of contribution
retirement phase, (x) withto 𝑥𝑥 � 𝑟𝑟 is the age of retirement and 𝑠𝑠 � 𝑟𝑟, can be
𝑅𝑅𝑅𝑅��� � 𝑟𝑟,𝑅𝑅𝑅𝑅can
he retirementphase,
phase,
any with of
age(x+s) 𝑥𝑥�the is the
the age
� 𝑟𝑟𝑟𝑟retirement of
phase, retirement
with and � ∗ �1be �𝑟𝑟𝑟𝑟𝑟𝑟�
retirement𝑅𝑅𝑅𝑅��� � with
𝑅𝑅𝑅𝑅 ∗ �1 𝑥𝑥 �𝑟𝑟𝑟𝑟𝑟𝑟� is age of retirement and𝑠𝑠 𝑠𝑠� � 𝑟𝑟, can be
written
x+r is the�as:
age of retirement and s > r, can
be written as: Estimating the IRR by scenario
Estimating the IRR by scenario � Estimating
�����,����� the IRR� by scenario
� �����,����� � 𝑝𝑝�,� � �� 𝑝𝑝�,� ∏�����,����� �1 � 𝑞𝑞�,� �
� 𝑝𝑝�,� � � 𝑝𝑝 ��,� ∏�����,����� �1 � 𝑞𝑞 �
�,� � Following the definition of the IRR, it IRR, it rep
� 𝑝𝑝�,� � � 𝑝𝑝�,� ∏�����,�����
�����,�����
�1 � 𝑞𝑞�,� � Following the definition of the
represents the discount rate, which equal-
ement benefits Followingd) Retirement the definition
benefits of the IRR, it represents the the
izes sumsum
the discount rate, which equalizes
of of thethe expected
expectedcontributions
contributions (SEC) and t
ent benefits d) Retirement benefits
the sum of the expected contributions (SEC) and(SEC) the sum and the of the sumexpected retirement
of the expected benefits
retire-
rement benefitThe isThe first retirement
first
calculated retirement
as follows: benefit
benefitisiscalculated
calcula- as ment follows: benefits (SERB). Considering the
ment benefit is tedcalculated
as follows:as follows: year of starting work as the reference year
for all the financial flows within a retire-
% ∗ 𝑛𝑛𝑛 𝑤𝑤 ∗ 𝑅𝑅𝑅𝑅� � 2.5% ∗ 𝑛𝑛𝑛 𝑤𝑤 ∗ (6) (6)
(SERB). Considering the year of starting work ment
(6) as the contract,year
reference we can
forwrite:
𝑛𝑛𝑛 𝑤𝑤 ∗ (6)all the financial flows
With:
within a retirement
With: contract, we can write:
e age of retirement;
𝑟𝑟: The age of retirement;
𝒙𝒙����,𝒕𝒕����
𝒈𝒈
ge of retirement;
presents the final wage∗averaged
� 𝑪𝑪𝑪𝑪on
𝒎𝒎,𝒌𝒌the last 5 years
∗ 𝒎𝒎�𝒙𝒙,𝒌𝒌�𝒕𝒕 𝒑𝒑𝒙𝒙,𝒕𝒕 (𝑤𝑤 ∗ ������
∗ ��� ∑��� ��𝒌𝒌�𝒕𝒕�
����� 𝑤𝑤� ); �
𝑤𝑤 : represents the final wage averaged � on the last 5 years (𝑤𝑤 ∗ � ∑��� 𝑤𝑤� );
𝒎𝒎�𝒙𝒙,𝒌𝒌�𝒕𝒕 � � �����
esents the final
duration of contribution; the last 5 years (𝑤𝑤 ∗ � ∑���
wage averaged on𝒙𝒙����,𝒕𝒕���� ����� 𝑤𝑤� );
�
𝑛𝑛: The duration of contribution;
uration of contribution; � � 𝑹𝑹𝑹𝑹𝒍𝒍,𝒗𝒗 ∗ 𝒍𝒍�𝒙𝒙��,𝒗𝒗�𝒕𝒕�� 𝒑𝒑𝒆𝒆𝒙𝒙��,𝒕𝒕�� ∗ �� � ������𝒗𝒗�𝒕𝒕�
tirement benefit of the next year is calculated by applying an annual revaluation
𝒍𝒍�𝒙𝒙��,𝒗𝒗�𝒕𝒕��
Then, the retirement benefit of the next year is calculated by applying an annual (7) revaluation
5% (Flici & Planchet, 2020). It gives:
ment(7) benefit of the next year is calculated by applying an annual revaluation
rate (𝑟𝑟𝑟𝑟𝑟𝑟) of 5% (Flici & Planchet, 2020). It gives:
%� (Flici
𝑅𝑅 & Planchet,
∗ �1 �𝑟𝑟𝑟𝑟𝑟𝑟� The IRR of each
2020). scenario was estimat-
It gives: nario: start working (contribution) at age
ed through an optimization problem using 51 and retire at age 70 with 19 years of
The 𝑅𝑅𝑅𝑅IRR
the ��� �
optimx of package
each
𝑅𝑅𝑅𝑅 scenario
� ∗ �1 R.was
�𝑟𝑟𝑟𝑟𝑟𝑟�
under estimated through
The obtained an optimization
contribution. problem
On the other hand, those using
who the
mating the IRR
�1 �𝑟𝑟𝑟𝑟𝑟𝑟� by scenario
results are shown in Figure 1. began contributing at age 35 and retired at
optimx package under R. The obtained results are shown in Figure 1.
age 50 have the highest IRR. This differ-
Estimating
The results show thethehigh
IRRvariability
by scenario of
wingthe
ing theIRR
definition
the IRRofbythescenario.
by scenario IRR, it Even
represents
if thethe discount rate,
average encewhich
of 3.62% equalizes
in terms of the IRR can lead
to huge disparities in wealthifaccumulation
he expectedThe IRRresults
contributions show
is 11.49%,(SEC)
Following theand
the high
minimum
the the variability
sumand
definition
maxi-
of the of the
of expected
the IRR,IRR by scenario.
retirement
it represents theEven
benefits discount the average
rate, IRR
mum are 9.59% and 13.21%, respectively. over the long run. For example, inwhich
30 equalizes
ngisthe definition
11.49%, ofminimum
thestandardthe IRR, it represents
and maximum thedistri-
discount
are 9.59% rate,
and which equalizes
The
the sum deviation
of the expectedof the IRR
contributions (SEC) anda13.21%,
years, return
the ofrespectively.
sumrate of 1%
the expected The
generates standard
34%
retirement benefits
bution is equal to 8.77E-03. The lowest of the initial investment.
expected contributions
deviation IRR of the (SEC) and the sum of the expected retirement benefits
IRR distribution is equal to 8.77E-03. The lowest IRR is associated with the
is associated with the following sce-
following scenario: start working (contribution) at age 51 and retire at age 70 with 19 years of
286 On the other hand, those who began contributing at age 35 and retired at age 50
contribution.
have the highest IRR. This difference of 3.62% in terms of the IRR can lead to huge
Rethinking the Contribution-Benefit Linkage in Pay-As-You-Go Retirement Systems Based on
Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022 Flici F.: Rethinking the Contribution-Benefit Linkage in...
Algerian Case
Figure 1
1 IRR Vs. age of starting contributing and retirement age. The subplot (a) shows the IRRs in the function
s. age of starting contributing and retirement age. The subplot (a) shows the IRRs in the function of the year of starting contribution a
of the year of starting contribution and the retirement age. The subplot (b) shows the distribution of the
ment age. The subplot (b)according
scenarios shows thetodistribution
their IRR. of the scenarios according to their IRR.

In
In the
the next
next section,
section, we
we propose
propose and
and evaluate
evaluate new
new
reduce
reduce the
the gap
gap in
in terms
terms of
of IRR
IRR as
as much
much as
as possibl
possib
simple,
simple, understandable,
understandable, and
and easily
easily implementable.
implementable.

THE
THE PROPOSED
PROPOSED MODELS
MODELS

To
To reduce
reduce the
the inequalities
inequalities generated
generated from
from th
th
formulation
formulation isis proposed
proposed partially
partially based
based on
on the
the
contributions
contributions and
and the
the expected
expected retirement
retirement benefits.
benefits.
relationship
relationship in
in Equation
Equation 77 using
using generalized
generalized linear
linear mo
m

First
First pension
pension benefit
benefit Vs.
Vs. Final
Final wage
wage

We
We start
start by by writing
writing the
the first
first term
term of of Equation
Equation
In the next section, we propose and w* and the second term as a function of the
evaluate new ways to calculate retirement the
the second
first second term
retirementterm as
as aa function
benefit function of
of the
(RBr). This the first
leads to retirement
first retirement be
b
benefits to reduce the gap in terms of IRR write:
as much as possible while keeping the cal- 𝐴𝐴𝐴∗∗ �
𝐴𝐴𝐴𝐴 𝐴𝐴𝐴 � 𝑅𝑅𝑅𝑅 ∗∗ 𝑅𝑅𝑅𝑅
𝑅𝑅𝑅𝑅�� (8)
culation method simple, understandable,
and easily implementable. A and B resulted from dividing the two
𝐴𝐴 and
parts of 𝐴𝐴 and 𝐵𝐵𝐵𝐵 resulted
Equation 7resulted from
by w* and RBdividing
from and arethe
dividing the two
two parts
part
r
named
named“coefficient
named “coefficientofof
“coefficient ofcontributions”
contributions”
contributions” and
and
and “coefficie
“coefficie
THE PROPOSED MODELS “coefficient of retirement benefits”, res-
To reduce the inequalities generated The
The first
first retirement
pectively. retirement benefit
benefit can
The first retirement can simply
simply
benefit be
canbe obtained
obtained by
by
from the design of the PAYG systems, simply be obtained by:
a new formulation is proposed partially 𝑅𝑅𝑅𝑅
𝑅𝑅𝑅𝑅�� �
��
𝑤𝑤∗∗ ∗∗ �
� 𝑤𝑤 (9)
based on the actuarial fairness between the �
paid contributions and the expected retire-
Simplifying the Contributions –
ment benefits. To this end, we simplify the
benefits equation
actuarial relationship in Equation 7 using
generalized linear models (GLM). Next, we model the first retirement be-
nefit using GLMs. We compare two mo-
First pension benefit Vs. Final wage
dels. The first model (noted as L1) tries to
We start by writing the first term of write the ratio of the first retirement bene-
Equation 7 as a function of the final wage fit on the final wage as a linear function

287
 𝐿𝐿1: 𝑤𝑤 ∗∗ �
𝐿𝐿1: ����� �
�� ��� ∗∗ 𝑛𝑛
𝑛𝑛
plifying the Contributions – benefits equation 𝑤𝑤
𝑅𝑅𝑅𝑅
𝑅𝑅𝑅𝑅�� � � � ∗ 𝑛𝑛 � � ∗ 𝑟𝑟
𝐿𝐿2:
𝐿𝐿2: 𝑤𝑤 ∗∗ �
� ��� � ��� ∗ 𝑛𝑛 � ��� ∗ 𝑟𝑟
, we model the first retirement benefit using GLMs.
𝑤𝑤 We compare two models.
Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022 Flici F.: Rethinking the Contribution-Benefit Linkage in...
odel (noted as L1) triesThe proposed
to write
The proposed models
the ratio
models of attempt
the firstto
attempt make
make the
retirement
to IRRs
IRRs resulting
the benefit from
from the
on the final
resulting the different
different scenarios
scenarios
converge
inear functionofconverge
ofthe to
theduration a
to a of
duration central value
ofcontribution
contribution
central better
(n).
value better than
The
(𝑛𝑛). The
than under
se-under
second the
one,
the current
ter-decilenoted
current system.
ranges
as L2,
system. Thus,
to measure
Thus, thethe models
themodels are
dispersion
are
cond one,
compared noted
based as L2, (the current system of the IRRs distribution under each model.
system is labeled based on
L0) considers onthethe concentration
retirement age (𝑟𝑟)of the obtained
of in obtainedto IRRs
n. Thearound
three thethe mean.
mean. We We use
use the
iscompared
labeled L0) considers the concentration
the retirement agetheaddition IRRs around the
interquartile
(r) inone-
addition andto n.inter-decile
The ranges to measure the Eliminating
dispersion of the
the extreme
IRRs scenarios
distribution under
interquartile
uding the current and
are written as: threeranges
inter-decile models to -inc-
measure the dispersion of the IRRs distribution under
luding
each the
model.current one- are written as: Since the proposed models rely partial-
each model.
𝑅𝑅𝑅𝑅� ly on the actuarial fairness equation, the
𝐿𝐿0: ∗ �the2.5%
Eliminating ∗ 𝑛𝑛 resulting IRRs might strongly converge to
Eliminating 𝑤𝑤 the extreme
extreme scenarios
scenarios the central IRR. However, some scenari-
𝑅𝑅𝑅𝑅 os consistently lead to aberrant IRRs, far
Since
𝐿𝐿1: the
Since the
� proposed models rely partially
proposed
� �� � �� models
∗ 𝑛𝑛 rely partially on
on orthe actuarial
theabove
actuarial fairness
fairnessInequation,
equation, the
the
𝑤𝑤 ∗ below the mean. such a case,
resulting
resulting IRRs
IRRs might might strongly
strongly converge
converge to to the central
it will
the IRR.
IRR. However,
be necessary
central However, some
to introduce scenarios
some some re-
scenarios
𝑅𝑅𝑅𝑅
consistently quirements for getting retired, such as a
𝐿𝐿2: ∗ lead � ��to to�aberrant
�� ∗ 𝑛𝑛 �IRRs,
�� ∗ 𝑟𝑟far
far below
below or above
above the mean.
mean. In Inofsuch aa case,
case, it will be
�
consistently lead aberrant IRRs, orminimum theduration such
contribution it or
will be
min-
𝑤𝑤
necessary
necessary to to introduce
introduce some some requirements
requirements for for getting
imumretired,
getting such
retirement
retired, such age.as
as aa minimum
minimum duration
duration ofof
proposed models The proposed
attempt to make models
the attempt
IRRs to make
resulting from the different scenarios
contribution
contribution or minimum
or minimum retirement
retirement age.
age.sce-
the IRRs resulting from the different
a central value
narios converge to a central value system.
better than under the current better Thus, RESULTSthe modelsAND are DISCUSSION
than
ased on the concentrationRESULTS
under the
RESULTS of theANDAND
current DISCUSSION
system.
obtained Thus, the
IRRs around the mean.
DISCUSSION We GLMs
use themodels have been esti-
The two
models are compared based on the con- mated using the lm function under R. The
and inter-decile ranges
centration to measure themodels
dispersion ofbeen
the IRRs distribution under
Theoftwo
The two theGLMs obtained
GLMs IRRs
models have
havearound
been estimated
estimated
estimated using
using the
the lm
models lmcanfunction
be written
function under
as:R.
under R. The
The
the mean.
estimated
We
models
use the
can
interquartile
be written as:
and in-
estimated models can be written as:
𝑅𝑅𝑅𝑅
𝑅𝑅𝑅𝑅�� 32.9% � 3.05% ∗ �𝑛𝑛 � 15�
minating the extreme scenarios 𝐿𝐿1: 𝑤𝑤 ∗∗ �
𝐿𝐿1: � 32.9% � 3.05% ∗ �𝑛𝑛 � 15�
𝑤𝑤
e the proposed models rely partially𝑅𝑅𝑅𝑅
𝑅𝑅𝑅𝑅�� on the actuarial fairness equation, the
𝐿𝐿2: 𝑤𝑤 ∗∗ �
𝐿𝐿2: � 14.55%
14.55% � � 2.83%
2.83% ∗∗ �𝑛𝑛
�𝑛𝑛 �
� 15�
15� �
� 1.97%
1.97% ∗∗ �𝑟𝑟�𝑟𝑟 �
� 50�
50�
RRs might strongly converge to 𝑤𝑤the central IRR. However, some scenarios
The
lead to aberrant
Theresulting
TheIRRs, farSSEs
resulting
resulting below(sums
SSEs
SSEs (sums of
of squared
or above
(sums of squared errors)Inequal
errors)
the mean.
squared errors) equal
equal 8.08
8.08
such8.08 and1.36
and
a case,
and it1.36
1.36 forbe
for
will
for L1and
L1
L1 andL2,
and L2,respectively.
L2, respectively.
respectively.

o introduce some requirements

Figure
Figure 22 2
Figure for getting retired, such as a minimum duration of
������ ������
Vs.
Vs. duration
age.ofofcontribution.
duration
� ∗ � ∗2 Vs. duration
or minimum Figure
retirement The
ofcontribution.
contribution. subplot
The
The (a)(a)
subplot
subplot shows
(a) thethe
shows
shows fit fit
the of
fit of
of using the the
using model L1. The
model L1. The subplot (b)
� ∗ � ∗ using the model L1. The
shows
subplot
𝑅𝑅𝑅𝑅
𝑟𝑟
subplotthe(b)
(b) fitofshows
shows
Vs. duration
based on
thethe
contribution.
the
fit fit
based model
on on
based
The the
subplot
L2.
model
(a)the
shows theL2.
model fit L2.
of
𝑅𝑅𝑅𝑅
𝑟𝑟
using the model L1. The subplot (b) shows the fit based on the model L2.
∗
𝑤𝑤 𝑤𝑤∗

ULTS AND DISCUSSION

two GLMs models have been estimated using the lm function under R. The
odels can be written as:
𝑅𝑅𝑅𝑅�
𝐿𝐿1: � 32.9% � 3.05% ∗ �𝑛𝑛 � 15�
𝑤𝑤 ∗
𝑅𝑅𝑅𝑅�
𝐿𝐿2: � 14.55% � 2.83% ∗ �𝑛𝑛 � 15� � 1.97% ∗ �𝑟𝑟 � 50�
𝑤𝑤 ∗
𝑅𝑅𝑅𝑅𝑟𝑟𝑅𝑅𝑅𝑅𝑟𝑟
g SSEs (sumsFigure 2 compares
ofFigure thethe
2 compares
squared errors) observed ratio
observed
equal 8.08ratio
and to the fitsfits
for obtained
∗ to the
𝑤𝑤∗ 𝑤𝑤1.36 and using
L1obtained
L2, models
using L1L1
models andand
respectively. L2.L2.

According to to
According thethe
results, thethe
results, model L1L1
model provides a replacement
provides raterate
a replacement of of
nearly 33%
nearly forfor
33%
thethe
required 15 15
required years of of
years contribution plus
contribution 3%3%
plus forfor
each additional
each year
additional beyond
year thisthis
beyond minimum.
minimum.
The current
The system
current would
system lead
would to to
lead a slightly higher
a slightly replacement
higher raterate
replacement (37.5%) forfor
(37.5%) thethe
same
same
288
contribution period.
contribution OnOn
period. thethe
other hand,
other 32 32
hand, contributed years
contributed would
years provide
would a replacement
provide raterate
a replacement
of of
84.76% against
84.76% 80%
against under
80% thethe
under current system.
current The
system. model
The L2L2
model provides merely
provides 15%
merely of of
15% a a
flat-rate replacement
flat-rate augmented
replacement by by
augmented nearly 2.8%
nearly forfor
2.8% each contributed
each year
contributed beyond
year 15 15
beyond years
years
 Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022 Flici F.: Rethinking the Contribution-Benefit Linkage in...

Figure
Figure 2 2 compares the observed ratio cording to the model L2, retiring at age 60
��� ��
� ∗ to
Vs. the fits
duration ofobtained using
contribution. models
The subplot (a) L1
shows thewith
fit of 20∗� using
� contributed years would result in
the model L1. The
and L2.
subplot (b) shows the fit based on the model L2. a replacement rate of 48.4% against 50%
According to the results, the model L1 under the current system. However, for a
provides a replacement rate of nearly 33% retirement age of 60 years and 32 years of
for the required 15 years of contribution contribution, an 82.36% replacement rate is
plus 3% for each additional year beyond provided, which is a little higher than the
this minimum. The current system would 80% provided by the current system.
lead to a slightly higher replacement rate Comparing the two models based on
(37.5%) for the same contribution period. the goodness-of-fit does not answer the
On the other hand, 32 contributed years wo- research question we are addressing in
uld provide a replacement rate of 84.76% this article, which is ensuring more equ-
against 80% under the current system. The ality in terms of profitability. Instead, we
model L2 provides merely 15% of a flat-ra- prefer using dispersion measures for such
te replacement
Figure augmented
2 compares the by 𝑅𝑅𝑅𝑅
observed ratio nearly
𝑟𝑟
to 2.8%
the fits comparison.
obtained using models Figures
L1 and L2.3 and 4 show the new
𝑤𝑤∗
for each contributed year beyond 15 years distributions of the different scenarios ac-
and 2% for each
According delayed
to the yearmodel
results, the beyond the a replacement
L1 provides cording torate
their IRR. 33% for
of nearly
minimum retirement age of 50 years. Ac-
the required 15 years of contribution plus 3% for each additional year beyond this minimum.
Figure
The 3
current system would lead to a slightly higher replacement rate (37.5%) for the same
IRR Vs. age of starting contributing and retirement age (model L1). The subplot (a) shows the IRR accord-
contribution
Figure 3 to the year
ing period. On the contribution
of starting other hand, 32 andcontributed years age.
the retirement wouldTheprovide
subplota replacement ratedistribution of the
(b) shows the
IRR Vs. age of starting contributing and retirement age (model L1). The subplot (a) shows the IRR according to the year of starting contribution
scenarios
of 84.76% according
against 80% to the
underIRR.
the current system. The model L2 provides merely
and the retirement age. The subplot (b) shows the distribution of the scenarios according to the IRR. 15% of a
flat-rate replacement augmented by nearly 2.8% for each contributed year beyond 15 years
and 2% for each delayed year beyond the minimum retirement age of 50 years. According to
the model L2, retiring at age 60 with 20 contributed years would result in a replacement rate
of 48.4% against 50% under the current system. However, for a retirement age of 60 years
and 32 years of contribution, an 82.36% replacement rate is provided, which is a little higher
than the 80% provided by the current system.

Comparing the two models based on the goodness-of-fit does not answer the research
question we are addressing in this article, which is ensuring more equality in terms of
profitability. Instead, we prefer using dispersion measures for such comparison. Figures 3 and
4 show the new distributions of the different scenarios according to their IRR.

Figure 3
IRR Vs. age of starting contributing and retirement age (model L1). The subplot (a) shows the
IRR according to the year of starting contribution and the retirement age. The subplot (b)
shows the distribution of the scenarios according to the IRR.

Figures 1 and 3 show high similarity,
meaning that the model L1 does not redu-
ce the dispersion of the IRRs compared to
the current situation (model L0). Table 1
exposes this comparison with more evi-
dence. The variances of the two distri-
butions are almost similar, with 7.7e-05
and 7.55e-05 for L0 and L1, respectively.
Compared to model L0, the minimum and
maximum of the IRR distribution have sli-
pped down but without affecting the mean,
the median, and the interquartile and inter-
decile ranges of the distribution.

289
Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022 Flici F.: Rethinking the Contribution-Benefit Linkage in...

Table 1 The model L2 allowed a large por-

The dispersion characteristics of the IRRs distri- tion of the scenarios to converge toward
bution under the different models
the central IRR. As shown in Table 1, the
L0 L1 L2 variance of the distribution of IRRs fell to
Max(%) 13.21 12.73 12.42 2.79e-05. 80% of the examined scenari-
Min (%) 9.59 8.81 8.10 os result in an IRR ranging from 10.9%
Mean (%) 11.49 11.49 11.45 to 11.8%, while 50% of them lead to an
Variance 7.70E-05 7.55E-05 2.79E-05 IRR between 11.32% and 11.74%. Con-
Q10 (%) 10.15 10.24 10.91 sequently, the interquartile distance nar-
Q25 (%) 10.82 10.84 11.32 rowed to only 30% compared to its value
Q50 (%) 11.60 11.66 11.60
under model L0. Also, results show that
60% of the IRRs range between 11.25%
Q75 (%) 12.20 12.21 11.74
and 11.75%.
Q90 (%) 12.56 12.51 11.78

Figure 4
IRR Vs.
Figure 4 age of starting contributing and retirement age (model L2). The subplot (a) shows the IRR accord-
IRR
ing Vs. age year
to the of starting contributing
of starting and retirement
contribution andage
the(model L2). Theage.
retirement subplot
The(a)subplot
shows the (b)IRR according
shows the to the year of starting
distribution of thecontribution
and the retirement age. The subplot (b) shows the distribution of the scenarios according to their IRR.
scenarios according to their IRR.

What remains evident is that model L2
has significantly reduced the dispersion of
the IRRs, which means better equality in
terms of profitability for the retirees. Un-
der some scenarios, however, even with
model L2, excessively high or low IRRs
are still obtained. In 10% of cases, IRR is
above 11.8% and it is below 11% in near-
ly 12% of the examined scenarios. Elimi-
nating these extreme scenarios returns to
setting some eligibility conditions about
the retirement age and the duration of con-
tribution. To this end, we have to define
a targeted IRR variation range. For exam-
ple, if the targeted variation range is [11%,
12%], we will need to eliminate all the
contribution-retirement scenarios leading
to an IRR outside the defined interval.
Figure 5 shows the distribution of IRR
as a function of the retirement age and the

290
 Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022 Flici F.: Rethinking the Contribution-Benefit Linkage in...

duration of contribution. Such visualiza- cap the IRRs at 11.85% against 12.42%
tions would help greatly in deciding about for 15 years. Figure 6 shows the distribu-
the conditions to impose as a requirement tion of the IRRs by the retirement age af-
for accessing retirement. Imposing a min- ter eliminating the periods of contribution
imum contribution period of 20 years will shorter than 20 years.
Figure 5Figure 5
IRR Vs. retirement age (subplot a) and Vs. duration of contribution (subplot b), under model L2
IRR Vs. retirement age (subplot a) and Vs. duration of contribution (subplot b), under model L2

Figure 6
Figure 6
IRR Vs. retirement age for a duration of contribution of 20 years or longer
IRR Vs. retirement age for a duration of contribution of 20 years or longer

291
Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022
On the other hand, eliminating the
IRRs below a fixed minimum (11.2% as
an example) returns to impose a retirement
age constraint, which must vary from 56
to 66 years old. Then, we can improve
the concentration of the IRRs by making
the retirement age close to 60 years. In
the best scenario, retiring at age 60 with
at least 20 years of contribution would
result in an IRR ranging from 11.66% to
11.76%. Similarly, for a targeted IRR
range of [11%, 12%], the duration of con-
tribution required for getting access to
retirement needs to be set at 18 years or
longer with a retirement age from 56 to
68 years. Different combinations are also
possible depending on the target range of
the IRR.
CONCLUSION
Poverty-reducing is the ultimate ob-
jective behind maintaining the long-run
financial sustainability of a retirement sys-
tem. In Pay-As-You-Go retirement sys-
tems, pension benefits are defined based
on the reference wage, the duration of
contribution, and an annuity rate. Follow-
ing economic and demographic changes,
a readjustment of the pension parameters
becomes necessary to keep the system
sustainable. Such a formulation, as its par-
ametric adjustment, can weaken the con-
tribution-benefit linkage and, most impor-
tantly, distort equality between individuals
within the system.
In this article, we have shown how the
retirement formula in PAYG systems can
lead to huge inequalities between the dif-
ferent contribution-retirement scenarios
taken from the profitability point of view.
Firstly, we defined all the possible combi-
nations allowed in the Algerian retirement
system considering the retirement age and
the duration of contribution. We assumed
that: 1) there is an automatic contribution
292
Flici F.: Rethinking the Contribution-Benefit Linkage in...
to social security when starting working,
2) the contribution period is continuous
and concentrated to the end of the work-
ing career, 3) all individuals aged x during
the year t gain identical salaries, and 4)
contribution rates are constant throughout
the analysis. We assumed wages to keep
growing in the future at 9.5% annually.
The different contribution-retirement sce-
narios led to internal return rates (IRRs)
varying between 9.6% to around 13.2%
and an average value of 11.5%. Such a
min-max gap of 3.6% might result in a sig-
nificant gap in wealth accumulation over
long periods.
The main objective of this article was
to propose a new contribution-benefit link-
age formula allowing a reduction of these
observed inequalities. The idea was to use
the actuarial fairness equation connecting
the paid contributions and the expected re-
tirement benefits and to derive a simplified
expression of the first retirement benefit as
a function of the final wage, the contribu-
tion duration and, eventually, the retire-
ment age. To this end, we evaluated and
compared two generalized linear models
(GLMs). The first model relies on the du-
ration of contribution to explain the ratio
of the first pension benefits. The second
one includes the retirement age in addition
to the contribution duration.
Compared to the model solely based
on the years of contribution, the one con-
sidering the retirement age decreased the
dispersion of the IRRs distribution and
narrowed the interquartile distance to only
30% of its initial length. Also, under the
second model, a large proportion of the
scenarios had an IRR between 11.2% and
11.85%. Then, regulating the requirements
to access retirement will disallow the con-
tribution-retirement careers leading to
IRRs outside the desired interval. In our
case, imposing 20 years as a minimum for
the contribution duration and encouraging
Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022
retirement at age 60 or nearly would help
concentrate further the IRR toward 11.7%
for all the retirees.
The methodology proposed in this
paper can be generalized to other coun-
tries and contexts. In addition to its effi-
ciency in reducing pension inequality,
it represents a practical tool for defining
the contribution-benefit linkage in PAYG
retirement systems and designing para-
metric reforms. Indeed, the current prac-
tice regarding pension reforming suggests
generally postponing the retirement age or
reducing the annuity rate to balance the fu-
ture incomes and outcomes of the system.
Usually, a set of reform scenarios are con-
sidered to select the most favorable one.
However, acting this way, there is a risk
of not including the best possible reform
combinations in the comparison process.
Our method allows us to define the ade-
quate re-parametrization of the system for
any changes in the economic and demo-
graphic environments and a given (target-
ed) return rate.
In the end, it is worth recalling that this
article did not address the effect of wage
disparity on pension inequality nor that
of parametric reforms. Also, we note that
our findings build on a high inflation case,
which implies a high salary evolution. Thus,
the gap between the minimum and maxi-
mum return rates is high enough to make a
significant difference regarding wealth ac-
cumulation over long periods. The proposed
methodology needs further assessment in a
low salary evolution context.
REFERENCES
Aaron, H. (1966). The social insurance paradox. Cana-
dian Journal of Economics and Political Science,
32(3), 371–374. https://doi.org/10.2307/139995
Barr, N. (2002). Reforming pensions: Myths, truths,
and policy choices. International Social Security
Review, 55(2), 3–36. https://doi.org/10.1111/1468-
246X.00122
Flici F.: Rethinking the Contribution-Benefit Linkage in...
Barr, N., & Diamond, P. (2006). The economics of pen-
sions. Oxford review of economic policy, 22(1),
15–39. https://doi.org/10.1093/oxrep/grj002
Ben Brahem, M. (2009). Pension system generosity
and reform in Algeria, Morocco and Tunisia. Inter-
national Social Security Review, 62(2), 101–120.
https://doi.org/10.1111/j.1468-246X.2009.01331.x
Bovenberg, A. L., & Van Ewijk, C. (2011a). Designing
the pension system: Conceptual framework. Ne-
therlands: Netspar.
Bovenberg, A. L., & Van Ewijk, C. (2011b). The future
of multi-pillar pension systems. Netherlands: Nets-
par.
Brunner, J. K. (1996). Transition from a pay-as-you-go
to a fully funded pension system: The case of dif-
fering individuals and intragenerational fairness.
Journal of Public Economics, 60(1), 131–146. htt-
ps://doi.org/10.1016/0047-2727(95)01517-5
Castanheira, M., & Galasso, V. (2011). Which reforms
for a fair and sustainable pension system?. Reflets
et perspectives de la vie économique, 50(3), 187–
202. https://doi.org/10.3917/rpve.503.0187
Cousins, M. (2007). Welfare policy and poverty. Du-
blin: Combat Poverty Agency.
Disney, R. (2006). Actuarial-based public pension
systems. In G. L. Clark, A. H. Munnell & J. M.
Orszag (Eds.), The Oxford Handbook of Pensions
and Retirement Incomes. Oxford: Oxford Univer-
sity Press.
Galasso, V. (2019). Designing a pension system. Rivista
di estetica, (71), 204–221. https://doi.org/10.4000/
estetica.5747
Flici, F. (2021). Coherent mortality forecasting for the
Algerian population. Assurances et gestion des ri-
sques/Insurance and Risk Management, 87(3–4),
209–231. https://doi.org/10.7202/1076125ar
Flici, F., & Planchet, F. (2019). Experience prospective
life tables for the Algerian retirees. Risks, 7(2), 38.
https://doi.org/10.3390/risks7020038
Flici, F., & Planchet, F. (2020). Financial sustainability
of the Algerian retirement system: A perspective
analysis of the 50 coming years. In M. Peris-Or-
tiz, J. Álvarez-García, I. Domínguez-Fabián & P.
Devolder (Eds.), Economic Challenges of Pension
Systems. Cham: Springer.
Holzmann, R., & Hinz, R. (2005). Old age income su-
pport in the 21st century: An international perspe-
ctive on pension systems and reform. Washington,
DC: The World Bank.
Klos, J., Krieger, T., & Stöwhase, S. (2022). Measuring
intra-generational redistribution in PAYG pension
schemes. Public Choice, 190(1), 53–73. https://
doi.org/10.1007/s11127-021-00914-w
293
Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022
Lüthen, H. (2016). Rates of return and early retirement
disincentives: Evidence from a German pension
reform. German Economic Review, 17(2), 206–
233. https://doi.org/10.1111/geer.12070
Miles, D., & Iben, A. (2000). The reform of pension
systems: Winners and losers across generations in
the United Kingdom and Germany. Economica,
67(266), 203–228. https://doi.org/10.1111/1468-
0335.00204
Morris, T. (2022). The unequal burden of retirement
reform: Evidence from Australia. Economic In-
quiry, 60(2), 592–619. https://doi.org/10.1111/
ecin.13034
Office National des Statistiques. (2014). Revenus sala-
riaux et caractéristiques individuelles -Enquête sur
les dépenses de consommation et le niveau de vie
des ménages 2011. Collections Statistiques Série S
: Statistiques Sociales, N 189. Alger: ONS. Availa-
ble at http://www.ons.dz/IMG/pdf/pubfinsal13.pdf
Piirits, M., & Võrk, A. (2019). The effects on intra‐ge-
nerational inequality of introducing a funded pen-
sion scheme: A microsimulation analysis for Es-
tonia. International Social Security Review, 72(1),
33–57. https://doi.org/10.1111/issr.12194
294
Flici F.: Rethinking the Contribution-Benefit Linkage in...
Reznik, G., Weaver, D. A., & Biggs, A. G. (2009). Soci-
al security and marginal returns to work near reti-
rement. Social Security Issue Paper No. 2009–02.
Samuelson, P. A. (1958). An exact consumption-loan
model of interest with or without the social con-
trivance of money. Journal of Political Economy,
66(6), 467–482. https://doi.org/10.1086/258100
Settergren, O., & Mikula, B. D. (2005). The rate of
return of pay-as-you-go pension systems: A more
exact consumption-loan model of interest. Journal
of Pension Economics & Finance, 4(2), 115–138.
https://doi.org/10.1017/S1474747205002064
Turner, J. A. (1998). Retirement income systems for
different economic, demographic and political
environments. Ageing Working Papers 3.8. Paris:
OECD.
Weyland, K. (2009). Bounded rationality and policy
diffusion: Social sector reform in Latin America.
Princeton, NJ: Princeton University Press.
Wilke, C. B. (2009). Internal rates of return of the Ger-
man PAYG System: A comparison of deterministic
and stochastic rates. In German Pension Reform:
On Road Towards a Sustainable Multi-Pillar
System. Bern: Peter Lang. https://www.jstor.org/
stable/j.ctv9hj6rz.7
Rev. soc. polit., god. 29, br. 2, str. 281-295, Zagreb 2022 Flici F.: Rethinking the Contribution-Benefit Linkage in...

Sažetak

PROMIŠLJANJE O POVEZANOSTI DOPRINOSA I NAKNADA U MIROVINSKIM

SUSTAVIMA MEĐUGENERACIJSKE SOLIDARNOSTI NA TEMELJU
PRIMJERA ALŽIRA

Farid Flici
Research Center in Applied Economics for Development - CREAD,
Algiers, Algeria

U sustavima međugeneracijske solidarnosti (Pay-As-You-Go), mirovine proizlaze iz

množenja referentne plaće s trajanjem doprinosa i stopom anuiteta. Ova se formulacija
može podvrgnuti određenim prilagodbama kako bi se osigurala održivost u okruženju koje
se mijenja. U ovom smo radu pokazali da trenutna veza doprinosa i naknada može stvoriti
veliki jaz između pojedinaca u smislu profitabilnosti, a time i u smislu akumulacije bogat-
stva tijekom dugih razdoblja. Uzimajući u obzir moguće scenarije karijere u alžirskom mi-
rovinskom sustavu, otkrili smo da stope povrata variraju od 9,6% do 13,2%. Kako bismo
smanjili nejednakost, predlažemo novu formulu za mirovine koja se djelomično temelji na
načelu aktuarske pravednosti.
Ključne riječi: umirovljenje, sustav međugeneracijske solidarnosti (PAYG), jedna-
kost, IRR, scenarij, Alžir.

295