Page  1
Report No. 34818-MV 
 
Maldives 
Social Protection in the Maldives:  
Options for Reforming Pensions and Safety Nets 
 
April 25, 2006 
 
Human Development Unit 
South Asia Region 
 
Document of the World Bank
 
Page  2
CURRENCY AND EQUIVALENT 
 
The external value of Maldives Rufiyaa is fixed at Rf. 12.8 per 1 US Dollar since July 25, 2001 
under a de facto fixed exchange rate policy vis-
à-vis the US dollar. 
Exchange Rate Effective = February 16, 2005 
 
Currency Unit = Rufiyaa (Rf.) 
US$1 = Rf. 12.86 
 
FISCAL YEAR
 
January 1 – December 31 
 
ABBREVIATIONS AND ACRONYMS 
 
CPI Consumer 
Price 
Index 
DC Defined 
Contribution 
DB Defined 
Benefit 
EPI 
Expanded Program of Immunization 
GDP 
Gross Domestic Product 
GoM 
Government of the Maldives 
HIES 
Household Income and Expenditure Survey 
IHD 
Integrated Human Development Policy 
ILO International 
Labour 
Organization 
LFP 
Labor Force Participation 
MDGs 
Millennium Development Goals 
MHEESS 
Ministry of Higher Education, Employment and Social Security 
MGFDSS 
Ministry of Gender, Family Development and Social Security 
PPP 
Power Purchasing Parity 
UN United 
Nations 
VPA-1 
1997/1998 Vulnerability and Poverty Household Survey 
VPA-2 
2004 Vulnerability and Poverty Household Survey 
 
Vice President  
: 
Praful Patel 
Country Director 
: 
Alastair J. McKechnie 
Sector Director 
 
: 
Julian F. Schweitzer 
Sector Manager 
 
: 
Mansoora Rashid 
Task Team Leader 
: 
Rasmus Heltberg 
Page  3
ACKNOWLEDGMENTS 
 
This report was prepared by a team led by Rasmus Heltberg (Social Protection Economist, 
SASHD) and including, as main authors, Robert Palacios (Senior Pension Economist, SASHD), 
and Kalanidhi Subbarao (Consultant/Safety Nets Specialist). Qaiser Khan (Lead Specialist, 
SASHD) led this work during its conceptualization and early stages. Yvonne Sin (HDNSP) 
contributed to the technical work on pensions, and Anna-Marie Vilamovska (Consultant, 
SASHD), Aparajita Goyal (Consultant, SASHD), and Takahiro Atsuta (Consultant, SASHD) 
provided research assistance. The team would like to thank the Government of the Maldives for 
its cooperation and assistance in the preparation of this report, and in particular Dheena Moosa. 
The team would also like to thank Alastair J. McKechnie (Country Director), Julian F. 
Schweitzer (Sector Director), and Mansoora Rashid (Sector Manager) for their guidance and 
support. 
Page  4
CONTENTS 
 
Executive Summary .......................................................................................................................................i
 
1.
 
Poverty and Vulnerability .....................................................................................................................1
 
Growth and Human Development.............................................................................................................1
 
Poverty: Level, Trends and Regional Distribution....................................................................................2
 
Vulnerability to Shocks.............................................................................................................................7
 
Who are the Poor and Vulnerable?..........................................................................................................11
 
Conclusions.............................................................................................................................................12
 
2.
 
The Pension System............................................................................................................................15
 
Current Pension System: Overview ........................................................................................................15
 
The Emerging Reform Package ..............................................................................................................16
 
Fiscal Implications..................................................................................................................................23
 
Conclusions.............................................................................................................................................33
 
3.
 
Safety Nets..........................................................................................................................................35
 
Current Safety Net System: Overview....................................................................................................35
 
Key Issues ...............................................................................................................................................37
 
Options for Reform.................................................................................................................................41
 
Targeting the Poor...................................................................................................................................46
 
Implementation........................................................................................................................................49
 
Conclusions.............................................................................................................................................51
 
References...................................................................................................................................................53
 
Annex 1: Data sources.................................................................................................................................55
 
Annex 2: Determinants of poverty..............................................................................................................57
 
Annex 3: Incidence and average size of assistance and transfers................................................................58
 
Annex 4: Targeting options.........................................................................................................................59
 
Categorical targeting...............................................................................................................................59
 
Proxy Means Test....................................................................................................................................60
 
T
ABLES
 
Table 1: Poverty Indicators 1997-2004, by Survey.......................................................................................4
 
Table 2: Poverty-inequality decomposition. Maldives 1997/98-2004..........................................................5
 
Table 3: Cross-country comparison of poverty and inequality in South Asia...............................................5
 
Table 4: Ownership of Consumer Durables by Consumption Terciles, 1998-2004 (in %) ..........................7
 
Page  5
Table 5: Household transitions in and out of poverty, 1998-2004 (panel data sub-sample, covering atolls 
only) ......................................................................................................................................................7
 
Table 6: Food scarcity...................................................................................................................................8
 
Table 7: Shocks experienced by households (past 6 years) ..........................................................................9
 
Table 8: Sector of employment of household head by tercile of per capita expenditures...........................11
 
Table 9: Incidence of Select Demographic and Household Characteristics among the Chronically Poor, 
Transient Poor, and Non-Poor (In %, Panel Sample)..........................................................................11
 
Table 10: Poverty rate and population size of select groups (in %)............................................................13
 
Table 11: Comparison of current and proposed pension system.................................................................34
 
Table 12: Expenditures on Social Assistance (pre-tsunami, and ................................................................38
 
Table 13: Public Expenditure on pensions and social assistance in selected countries...............................39
 
Table 14: Distribution of Assistance and Transfers ...................................................................................41
 
Table 15: Overview of Program Options for social assistance...................................................................43
 
Table 16: Benefit levels and benefit adequacy............................................................................................45
 
Table 17: Fiscal Cost of umbrella program (cash transfer to the poor), in different scenarios depending on 
benefit size and coverage ....................................................................................................................49
 
Table 18: Data Sources and sample sizes....................................................................................................55
 
Table 19: Poverty Correlates: Probit Regression Results Using VPA1 (1997-98) and VPA2 (2004) Data57
 
Table 20: Incidence and average size of assistance and transfers...............................................................58
 
Table 21: Simulation Results with a Fixed Budget Constraint of 0.5% of 2003 GDP ...............................59
 
Table 22: Simulation Results with a Fixed Budget Constraint of 1% of 2003 GDP ..................................59
 
Table 23: Illustration of Type I and II targeting errors ...............................................................................62
 
F
IGURES
 
Figure 1: Headcount poverty, 1997-2004......................................................................................................3
 
Figure 2: Poverty headcount, by Region.......................................................................................................5
 
Figure 3: Inequality, by Region.....................................................................................................................5
 
Figure 4: Population and expenditure shares, by region...............................................................................6
 
Figure 5: Dominance Curve..........................................................................................................................6
 
Figure 6: Growth inequality curve (Annual Percentage Growth in Per Capita Expenditure, per Percentile), 
Maldives................................................................................................................................................6
 
Figure 7: Growth
 inequality curves for Male’ and atolls, 1997/98-2004......................................................7
 
Figure 8: Demographic Dynamics ..............................................................................................................10
 
Figure 9: Replacement Rates under Different Rates of Return on Investments of the DC Scheme (for 14% 
contribution rate).................................................................................................................................18
 
Figure 10: Age Distribution of Civil Servants, 2003 ..................................................................................20
 
Figure 11: Cost of demogrant with different eligibility ages......................................................................23
 
Figure 12: Cost of pension reform package proposed by Working Group .................................................24
 
Figure 13: Comparing present value of wages to pensions for those over age 65......................................25
 
Figure 14: Dispersion in replacement rates for civil servants retiring immediately after introduction of 
mandatory retirement age....................................................................................................................26
 
Figure 15: Fiscal cost of current social safety net programs.......................................................................39
 
Figure 16: Overview of program options....................................................................................................43
 
Figure 17: Profile of proposed benefits under poverty program (example)................................................46
 
Figure 18: Targeting of a simulated proxy means test................................................................................62
 
Page  6
i
EXECUTIVE SUMMARY 
1.
 
The purpose of this report is to assist the Government of the Maldives in the 
design and implementation of social protection reforms, 
in particular for pensions and 
safety nets
.
The report includes an overview of poverty, risk and vulnerability in 
Maldives as well as analysis of the role and effectiveness of pensions and social 
assistance policies in helping poor households mitigate and cope with risks. Based on the 
analysis, the study delineates a menu of options for reform of the existing pension and 
social safety net system. The report has been prepared in close consultation with the 
Government of the Maldives
.
2.
 
The structure of the study follows the issues outlined above.
 Chapter 1 sets the 
stage with an overview of growth, inequality, poverty and vulnerability. Chapter 2 
discusses old-age pensions and pension reform options. Chapter 3 covers safety nets 
(social assistance), while annexes describe the data sources and contain the analytical 
findings underpinning the recommendations. The executive summary details the main 
findings and policy options proposed in the report.  
3.
 
The study uses both household survey and administrative data. 
The main data 
sources used for this report are the Vulnerability and Poverty Surveys conducted in 
1997/98 and 2004 and administrative data on programs and civil servants provided by the 
government. Other data sources such as the Household Income and Expenditure Survey 
and the census are also drawn upon.  
4.
 
The reform options will need to be phased in over time, depending on the 
fiscal situation.
 It should be noted, at the very outset, that the reform options proposed 
by this report involve non-negligible fiscal costs. The short term fiscal constraints facing 
the Government (as a result of the tsunami) may make it necessary to delay full 
implementation of the package until such time as budget resources are available to absorb 
these incremental costs. 
5.
 
The reforms under consideration are substantial
.
Today, Maldives has some 
1,000 recurrent recipients of social assistance (0.3 percent of the population). After the 
proposed reforms, this number could increase to perhaps 20,000 beneficiaries (6.7 
percent of the population), of which 14,000 would receive the universal old-age benefit. 
 
I. 
Growth, inequality and poverty  
6.
 
Over the past 25 years, Maldives has recorded impressive rates of economic 
growth
 
and significant accomplishments in access to health care, education and 
other social amenities
.
GDP growth averaged 7.6 percent per year in the period 1976-
2004, and exceeded 10 percent per year for most of the 1980s, surpassing the 
performance of all other South Asian economies. The Government’s drive to improve the 
Page  7
ii 
health and education systems had led Maldives to already achieve several of the 
Millennium Development Goals (MDGs). 
7.
 
Driven by this impressive growth and investment in human capital, poverty 
in the Maldives was halved from 1997 to 2004. 
Based on a poverty line of Rf. 15 per 
person per day, the share of the population living in poverty declined from 45 to 18.9 
percent of the population (or 60,000 individuals) prior to the tsunami
.
The incidence of 
poverty is fairly shallow (i.e., many, but not all, of the poor are fairly close to the poverty 
line).  
8.
 
Despite the gains from growth, the Maldivian population is vulnerable to 
aggregate and individual shocks. 
Aggregate risks include the most recent impacts of the 
tsunami that caused large-scale loss of incomes and assets; potential future environmental 
risks of global warming leading to a rise in sea levels; and population aging. The most 
important individual shocks reported by survey results are health risks and joblessness, 
with health shocks causing the largest and most frequent income losses. The main 
coping 
mechanisms
 used by the poor are informal insurance and own savings and income, and 
limited public support (government support was used for coping in about 10 percent of 
shocks). 
9.
 
The transient poor who move in and out of poverty outnumber the 
chronically poor, but the characteristics of both groups of poor are very similar
.
The 
highest incidence of poverty/vulnerability
 
is in the 
Northern atolls
,
among household 
heads 
with little or no education, a large number of children, with elderly, and without 
working members.
 Some (but not all) among widows and divorcees may also be highly 
vulnerable. However, the largest group of the poor and vulnerable are households with 
many children, and households whose heads are jobless and/or with little education 
and/or female-headed
.
Ninety percent of the poor live in the atolls; the Northern and 
North-Central regions alone now contain more than 60 percent of the count
ry’s poor
.
10.
 
The Maldivian Government would like to develop an effective social 
protection system for its country
 that is consistent with its level of income and 
implementation capacity. Economic growth has been, and will continue to be, the 
dominant factor in the reduction of poverty in Maldives, but the concentration of poverty 
in the Northern atolls and among particular population groups is of considerable concern. 
Seeking to address these issues, the government is interested in reforming its pension 
system, and meeting the constitutional mandate to provide income support to all 
employed Maldivians. It is also interested in developing an effective safety net for the 
very poor, and those who may suffer poverty from catastrophic health shocks or from 
natural disasters. The evaluation of the current system and options for reform towards 
meeting these objectives are outlined below: 
 
II. 
Old Age Security: The Pensions System  
11.
 
The pension system in the Maldives covers only public sector employees
.
It 
has two components–the Civil Service Pension (CSP) scheme and the Government 
Provident Fund (GPF). The Civil Service Pension (CSP) scheme is a non-contributory, 
Page  8
iii 
pay-as-you-go (PAYG) defined benefit (DB) arrangement covering approximately 
26,000 government employees. Pensions are financed directly from the government 
budget.  
12.
 
The Maldives’ pension system is globally unique in the sense that pensions 
are paid out every 20 years of uninterrupted government service and there is no 
requirement to retire.
 Employees can continue to work for another 20 years and earn a 
second pension, and in a few cases, even a third. Therefore, with an accrual rate of 2.5 
percent per year applied to the last wage in the 20 year period, pensioners in Maldives 
receive their benefits along with their earnings. The pensions are not indexed however, so 
their value in real terms declines over time depending on rates of inflation.  
13.
 
The Government Provident Fund (GPF) is a defined contribution (DC) 
arrangement with matching employee and government contributions of 5 percent of 
the employee’s basic wage each month
.
Participation is voluntary and withdrawals are 
allowed from their account for the education of their children, house construction and 
repair, health care and obligatory hajj. Recently, there have been massive withdrawals as 
the Government encouraged the purchase of housing. The scheme does not play a major 
role in providing a source of retirement income. On the other hand, it involves a 
significant subsidy at a cost to the government for workers that participate. 
 
Key issues 
14.
 
The system has two major deficiencies namely, inadequate retirement income and 
lack of safety net for the elderly
.
15.
 
First, the civil service pension scheme is not designed to produce a reasonable 
level of retirement income
.
Specifically, the civil service pension is calculated as 50 
percent of basic salary after every 20 years of service. The pension calculation does not 
apply to overtime and allowances which on average comprise around 40 percent of total 
remuneration. Combined with the fact that the pension is not indexed, its value is very 
low compared to average incomes when the worker reaches old age. Therefore, the 
parameters of the scheme do not result in a secure and adequate stream of pension 
income after a certain age. Given that the scheme does not stipulate a retirement age, it is 
not surprising that civil servants choose never to retire in the Maldives. A longer term 
issue is the possibility of extending coverage to the rest of the labor force. 
16.
 
The second deficiency is the absence of any safety net for the elderly in a 
country where only civil servants participate in a pension
 scheme of any kind. With 
more than two-thirds of the labor force lacking pension coverage, the only type of 
program likely to have any impact in the near future is one that involves transfers to the 
elderly not linked to their employment or contribution history. The design of this program 
would explicitly take into account reforms to the civil service pension scheme (as well as 
the reform of social assistance) given the overlap in the population being covered and the 
need to coordinate overall target pension levels. 
 
Reform options 
Page  9
iv 
17.
 
In order to develop an effective pension system, two objectives must be met:
 
(a) introducing a financially sustainable retirement savings scheme for civil servants 
(with the possibility of expanding to other formal sector workers in the future), and (b) 
providing a safety net for all elderly citizens in the Maldives.  
18.
 
The emerging reform plan 
–
guided by a GoM Inter-ministerial task force – 
would separate redistributive and savings functions in two distinct programs
:
·
The contributory plan would eventually replace current rules for civil 
servants
,
first with the introduction of a retirement age and pension credit for past 
years of service and, eventually, by a defined contribution scheme
.
Contributions 
from workers and employers would finance this retirement savings scheme. The 
consumption smoothing scheme would, in the long run, be fully funded and 
defined contribution in nature and so avoids the creation of any new liabilities. 
Past service can be credited through special recognition bonds that are redeemed 
at retirement in order to smooth fiscal flows
 
·
The non-contributory plan (often referred to as a 
‘demogrant’)
 would pay all 
elderly citizens a poverty-level benefit
.
General government revenues would 
finance the demogrant. The government could consider app
lying a ‘pension test’ 
whereby other pension income is taken into account towards achieving the 
minimum standard.  
19.
 
Within this general approach, there are a number of choices to be made that 
have both social and financial implications
.
First, the contributory scheme can be 
phased in gradually and only for younger workers or for all workers simultaneously. In 
either case, the transition rules - and exactly how past service is credited towards the new 
scheme - must be determined. For the demogrant, the benefit level and method of 
adjustment over time, clawback (targeting) mechanisms and eligibility age are all 
variables that affect the cost of the program and its ultimate impact.
 
20.
 
The annual cost of the package depends on several factors. The upper bound 
estimate of 1.5-2.0 per cent of GDP per annum
 consists of three expenditures. The first 
is an untargeted demogrant at age 65 with an estimated cost of 0.7 per cent of GDP. The 
second cost arises when the government contributes on behalf of all civil servants to a 
segregated pension fund at a rate of 7 per cent of the wage bill or about 0.7 percent of 
GDP. The third cost is ambiguous. Mandatory retirement of all civil servants age 65 and 
above would generate a net savings since the average pension would be lower than the 
average wage for these individuals. However, if the number of civil servants was kept 
constant through new hiring, this would lead to a net cost of around 0.2-0.4 per cent of 
GDP.  
21.
 
Finally, while the reform package appears sound in terms of good practice, 
the specific conditions of Maldives will require a highly customized implementation 
of these programs.
 Implementation and management of reform will require several years 
of capacity and institution building. Implementation will need to focus on issues related 
to data management (for both the demogrant and the defined contribution scheme), and 
choices regarding asset management, portfolio composition, and insurance benefits. Scale 
in particular, will be a key consideration in these areas. Finally, reforms must take into 
Page  10
v
account the overall initial conditions especially with regard to the current or prevailing 
fiscal situation.  
 
III. 
Protecting against Poverty and Vulnerability: Social Safety Net 
 
22.
 
The government has put in place several programs to protect the poor and 
vulnerable
,
including. (a) an absolute poverty scheme initiated in 2003 under which Rf. 
500 a month is given to highly vulnerable individuals who are not able to eat more than 
one meal a day, not in possession of more than two sets of clothing, or are homeless, (b) 
assistance to obtain school materials for children of low income families, (c) assistance to 
seek medical treatment from hospitals in Maldives or abroad, (d) assistance to obtain 
assistive devices for the disabled and drugs for the mentally ill, and (e) institutional care 
for orphans and mentally disabled persons. These publicly funded programs are 
supplemented by non-governmental support comprising remittances by extended family 
members and Zakat, the Islamic charitable support. Following the 
tsunami,
 the 
government disbursed emergency income support to the affected population. 
Approximately Rf. 50 million was disbursed to about 63,000 affected families. Tsunami 
assistance is temporary. 
Key issues 
23.
 
The safety net programs are ad hoc and overlapping,
 with critical vulnerable 
groups omitted, and some benefits disbursed to individuals based on unverifiable 
subjective judgments. The key objective of protecting the poor is therefore not being met. 
24.
 
The safety net system has very low coverage and funding relative to the level 
of income of the Maldives
.
Only 0.3 percent of the population and 1.6 percent of the 
poor are covered by recurrent transfers; and total safety net spending (outside of the 
tsunami benefit) is 0.2% of GDP. That said, it is important to note that adequate funding, 
without accompanying reforms in targeting efficiency will not be sufficient to ensure the 
benefits help alleviate poverty. However, without adequate funding it will not be possible 
to ensure a modern, well administered, and well-targeted safety net. 
25.
 
The distribution of publicly funded programs appears to be inequitable
 
because substantive shares of current program benefits accrue to the best-off third of the 
population. Even private transfers, when analyzed as a whole, seem more or less equal 
across income groups, and not distributionally pro-poor. 
 
Reform options 
26.
 
An effective safety net should protect the chronically and transitorily poor 
households from sudden shortfalls in income and consumption
 induced by various 
shocks. Safety nets help address chronic poverty, but also help the poor escape poverty 
by improving or protecting their human capital against a variety of vulnerabilities 
(stemming, for example, from disability, medical emergencies, macroeconomic shocks, 
and natural disasters), and provide protection to the poor in case of difficult economic 
reforms. 
Page  11
vi 
27.
 
A number of reform options have been proposed in this report for the 
consideration of the government
.
·
The government may consider a 
universal umbrella poverty targeted program
 
that folds in most of the existing benefits and provides a single transfer to the very 
poor (identified and targeted as below). For specific vulnerable groups such as the 
poor disabled and the poor among the tsunami victims, rather than establishing 
separate programs or benefits for these groups, it is proposed that they be included 
in the umbrella poverty program. 
·
The benefit can be flat lump sum, or can have two parts 
–
a
fixed ‘family 
allowance’ supplemented by a variable
 
‘child allowance’
.
The latter option 
would allow greater flexibility and fairness in addressing household vulnerability. 
A
threshold benefit level will need to be identified as an upper cap on benefits so 
that that work incentives are not compromised, and fiscal costs are sustainable 
(Rf.500-800 per month per household would seem an appropriate range). 
However, the benefit could be increased in case of a catastrophic health shock. 
The benefit should be indexed to inflation to ensure that protection is not eroded 
over time. 
·
Several targeting options are available to the government.
 In particular, the 
report explored the potential of using a proxy means test in combination with 
community-based targeting. Based on the findings of selected simulations, the 
study recommends a 
combination of geographic targeting and community-based 
identification
 of
beneficiaries aided by guidelines that emphasize household 
characteristics statistically associated with vulnerability. 
 
28.
 
Depending on options chosen by the government, the fiscal cost of the 
poverty-targeted assistance could range from 0.6-0.8 percent of GDP. 
This is in 
addition to the fiscal cost of the proposed pension reforms (demogrant and civil service 
pensions), which as mentioned ranges from 1.5-2 percent of GDP. Thus, the total fiscal 
cost of the pension and social assistance reforms discussed in this report ranges from 2-3 
percent of GDP
.
These estimates will need refining as the details of the program are 
finalized. 
Reforms should be phased in once fiscal resources become available, 
presumably once the impact of the tsunami has faded. 
29.
 
Program implementation needs careful attention
:
It is important to ensure 
proper identification and targeting of the poor, and verifying, monitoring and updating 
records need to be decided. An agency to coordinate implementation, with adequate and 
well trained staff will also need to be appointed, and the new programs should be piloted 
before being scaled up to cover the entire country. 
30.
 
Monitoring and Evaluation of the program should be built in from the start 
of the program 
to assess program performance, administration and alert the government 
to implementation problems
.
This would help officials receive feedback on program 
administration that should contribute to changes in programs or implementation 
procedures as needed to enhance efficiency of administration. 
Page  12
1
Poverty and Vulnerability 
 
To inform the design of social protection programs taken up later in this report, this first 
chapter reviews the macroeconomic and human development progress made by the Maldives over 
the past two decades, and evaluates the evolution, determinants and profile of poverty and 
vulnerability in the country. This chapter is organized as follows. The first section provides an 
analysis of growth and human development, while the second section outlines trends in poverty. 
The third section discusses vulnerability to shocks and the final section delineates who are the 
poor and vulnerable in Maldives.  
G
ROWTH AND 
H
UMAN 
D
EVELOPMENT
 
The Maldives has a very unique geography comprising a chain of Indian Ocean islands 
spread over a distance of 900 kilometers
.
The country has 1,190 islands, formed into 26 
geographic atolls; 198 islands are inhabited. Population is roughly 300,000 people. One third of 
the inhabited islands have a population of less than 500, and 70 percent of the inhabited islands 
have a population of less than 1,000 inhabitants. The altitude of most of the islands is very low, 
not surpassing a meter above sea level. The country is also characterized by extreme paucity of 
land-based natural resources, including water and agricultural land.  
 
Over the last quarter century, Maldives has transformed itself into the most prosperous 
country in South Asia. 
Notwithstanding the recent drop in GDP as a result of the tsunami,
 
GDP 
growth averaged 7.6 percent per year in the period 1976-2004, and exceeded 10 percent per year 
for most of the 1980s. As a result, real GDP in 2004 was more than eight times higher than in 
1971. The growth rate of per capita GDP averaged 3.4 percent per year from 1976 
–
2003.
Growth was led primarily by the tourism and the fisheries sector
.
Tourism and fisheries 
are now the two largest sectors of the Maldivian economy, followed by agriculture; since the mid-
1990s these two sectors also account for roughly 40 percent of government revenues. For 
example in 2003 tourism accounted directly for 33 percent of GDP
1
,
and indirectly some 60-70 
percent of GDP. Tourism provided 14,000 jobs in 2000, fisheries 9,000 in a total labor force of 
117,000, 26 percent of which is foreign (Ministry of Planning and National Development, 2003).  
 
The performance of macroeconomic indicators (e.g., inflation, public debt management) 
has also been favorable
—inflationary pressures have been under control, and measures to 
restrict the growth of public debt, on the rise since the late 1990s, have been relatively successful; 
in 2003 public debt reached 46.5 percent of GDP, up from 40.9 percent in 2000, but the debt 
service remains manageable.
2
The main factor behind rising fiscal outlays has been efforts of the 
government to expand provision of public services (especially education and health) in the atolls. 
The geography of the country militates against economies of scale, making the provision of 
public and private services and infrastructure very costly.  
 
Maldives has also made tremendous progress in delivering social service.
 
In health
:
over 
the period 1987-1997 the crude birth rate, the crude death rate and the infant mortality rate were 
halved and more than two decades were added to the life expectancy of Maldivians at birth. 
While in 1970 the infant mortality rate per thousand live births was 159, in 2002 it fell to 55. 
1
World Bank. 2004. “Maldives: Country Economic Memorandum.” World Bank. Washington, DC. 
2
IBID. Thus far the government has sought resolution to this problem by broadening the revenue base. A number of 
measures have been under consideration including the increase in the bed tax, unchanged since 1988, that became 
effective in November, 2004.  
Page  13
2
Simultaneously, the crude death rate fell from 38 to 4 per thousand. Several debilitating diseases, 
such as malaria, childhood tuberculosis, filaria and leprosy have either been eradicated or have 
transmission rates close to zero. Child immunization has become nearly universal and vaccination 
coverage against hepatitis has reached 80 percent of the population. 
In education: 
Basic 
education is virtually universal; primary and lower secondary school attendance now exceeds 95 
percent of the respective age groups, both for girls and boys, according to VPA-2. As for 
other 
services
,
far larger numbers of Maldivians today have access to potable water, sanitation and 
electricity than a decade ago 
 
Maldives has already achieved several of the MDG target indicators, but still has some 
challenges ahead.
 The gender gap in school enrollment and literacy among those aged 15 to 24 is 
virtually non existent.
3
The proportion of births not attended by skilled health personnel has been 
reduced by three-quarters.
4
MDG targets that emerge as potentially more problematic are in 
health and water: halving child malnutrition below the age of 5, halving the proportion of the 
population not having access to safe drinking water, reducing by three-quarter the maternal 
mortality ratio, and slowing the rate and reversing the spread of HIV/AIDS.  
P
OVERTY
:
L
EVEL
,
T
RENDS AND 
R
EGIONAL 
D
ISTRIBUTION
 
Consistent with economic growth and human development, poverty in the Maldives fell 
dramatically
—by more than half—from 1997 to 2004.
 Based on a poverty line of Rf. 15 per 
person per day, headcount index of poverty (the share of the population consuming below the 
poverty line) declined 
from about 45 to less than 19 percent before the tsunami
 (see Figure 
1).
5
The tsunami will have resulted in sharply increased poverty in the affected atolls, at least 
temporarily, but data on this not yet available. 
The conclusion that poverty declined substantially 
before the tsunami does not depend on any specific poverty line. The dominance curve in Figure 
5
shows that poverty incidence in VPA-2 is lower than in VPA-1 for any plausible poverty line
.
6
Most of the decline in poverty realized since 1997 took place prior to 2002/03
 (Table 1). 
In fact, some measures of poverty suggest a stagnation or slight increase in the short period from 
2002/03 to July 2004. However, as outlined in Annex 1, the sample size of HIES was not very 
large, and so it is unclear to what extent the results are fully comparable. This should certainly not 
overshadow the basic finding that poverty decline was fast, deep, and fairly widespread in the late 
1990s and early part of this decade.  
3
VPA-2 
4
Annual Reports of Ministry of Health. 
5
Annex 1 describes the data sources and the assumptions underlying these conclusions. 
6
A
statistical dominance curve involves comparing the cumulative distribution functions for consumption for each of 
the survey years. One distribution “dominates” another if the income distribution function for that year lies to the right 
of the earlier year. If one finds that first-order dominance holds between two different years, it implies that all poverty 
measures, including the headcount, poverty gap, and squared poverty gap, in the first year are higher than in the second 
year for 
all poverty lines
.
In other words, in the case of Maldives, the decline in poverty between 1997 and 2004 is not 
an artifact driven by poverty lines; the decline is noticeable for all possible poverty lines. 
Page  14
3
Figure 1: Headcount poverty, 1997-2004 
0
10
20
30
40
50
60
1997
1998
1999
2000
2001
2002
2003
2004
H
e
a
d
c
o
u
n
t
p
o
v
e
r
t
y
(
%
)
Maldives
Male
Atolls
Source: 
Preliminary Staff estimates based on VPA-1 and VPA-2.
 
The rapid improvement in living standards is reflected in a drastically expanded 
pattern of ownership of household durable goods
.
Table
 4
 shows how ownership of durable goods has expanded in each tercile of the income 
distribution in the 1997/98-2004 period, resulting in a large decline in the inequality of durable 
good ownership; indeed, households in the bottom tercile have almost caught up with households 
in the top tercile with respect to ownership of low-cost durable goods such as fans and irons. 
 
Poverty fell most in Male’—from 23 to less than 5 percent,
 and less sharply, but still 
significantly, in the atolls
—52
to 25 percent in the atolls. As a result, even more than before, 
income poverty is now concentrated in the atolls (Table 1). The depth of poverty—a measure of 
how far the poor are below the poverty line—registered an equally impressive reduction, from 
0.166 in 1997 to 0.057 in 2004 in Maldives overall. In 2004 the size of the poverty gap amounted 
to 0.9 percent of GDP—that is, assuming perfect targeting, it would cost 0.9 percent of national 
product to lift all of the poor to the poverty line. The severity of poverty also fell remarkably, 
suggesting the very poorest shared in growth.  
 
Poverty declines were much greater in the South relative to the Northern and North-
Central regions.
 The Northern region appears to have seen only modest decrease in poverty, and 
has now emerged as Maldives’ poorest region, followed by the North-Central region (Figure 2). 
This differential change in poverty has altered the geographic pattern of poverty in the country. 
Poverty now shows a declining pattern moving from North to South. In 2004, the Northern region 
with 16 percent of the country’s population accounted for just 10 percent of total expenditures 
while Male’, with 30 percent of the population accounted for more than 40 percent of 
expenditures (Figure 4). 
Page  15
4
Table 1: Poverty Indicators 1997-2004, by Survey 
Survey and year
VPA 1997/98
HIES 2002/03
VPA2004
Maldives Mean expenditure per capita
23.0
33.0
40.9
Headcount poverty
0.447
0.199
0.189
Poverty gap (P1)
0.166
0.046
0.057
Poverty depth (P2)
0.083
0.018
0.025
Gini
0.416
0.366
0.425
Male'
Mean expenditure per capita
31.8
50.1
56.0
Headcount poverty
0.229
0.040
0.045
Poverty gap (P1)
0.082
0.007
0.014
Poverty depth (P2)
0.040
0.002
0.007
Gini
0.399
0.355
Atolls
Mean expenditure per capita
19.4
26.7
34.6
Headcount poverty
0.521
0.257
0.250
Poverty gap (P1)
0.195
0.060
0.075
Poverty depth (P2)
0.098
0.023
0.033
Gini
0.397
0.431
Based on a measure of expenditures per capita that excludes housing costs.
Poverty measures based on a poverty line of Rf.15 per person per day ($1.15).
Source: Preliminary staff estimates.
The different rates of decline of poverty represent differing growth rates. 
The 
growth-
inequality curve
 (Figure 6) demonstrates that in the Maldives economic growth benefited 
households in all parts of the income distribution. Figure 6 shows that all percentiles experienced 
annualized growth rates in consumption between 9-11 percent, demonstrating that, although the 
top 40 percent of households nationally gained more than others, all regions of the country shared 
in growth
—but not to an equal extent. The regional 
growth-inequality curves
 in Figure 7, show 
that, in the atolls, the top quarter of the population gained the most while in Male’ it was the 
bottom quarter that gained the most.  
 
…and represent different changes in inequality.
 In Male’, inequality declined and 
complemented income growth in reducing poverty. In contrast, inequality rose in the atolls, but, 
the positive effect of growth substantially swamped the smaller negative effect of rising 
inequality. For the given growth rate, had inequality remained constant, poverty would have 
fallen somewhat faster in the atolls than at present (Table 2). Overall, for Maldives as a whole, 
poverty headcount would have fallen by 2 percentage points more than at present, had inequality 
not risen.  
 
Compared to other South Asian countries, Maldives has lower poverty but fairly high 
inequality
 (It is most similar to Sri Lanka). Unlike other countries in the region, Maldives 
experienced a spectacular growth performance and was therefore able to reduce poverty despite 
increasing inequality. Nonetheless, high inequality in the atolls will remain a source of concern to 
policymakers (Figure 3). Fairly high but shallow poverty means that well-targeted social 
assistance can potentially have a large impact on poverty. Table 3 attempts a cross-country 
comparison of inequality and poverty (based on national poverty lines) for the countries of South 
Page  16
5
Asia. Since national poverty lines are set at different levels
7
and expenditure measures are 
collected differently, the comparison should be treated with caution.  
 
Table 2: Poverty-inequality decomposition. Maldives 1997/98-2004 
Maldives
Male'
Atolls
Poverty rate in year 1
0.447
0.229
0.521
Poverty rate in year 2
0.189
0.045
0.25
Change in poverty        
-0.258
-0.184
-0.271
which can be decomposed into:
 Growth component
-0.277
-0.144
-0.316
 Redistribution component
0.02
-0.04
0.045
Source
:
Staff estimates using the Ravallion-Chen methodology, based on average effects. 
 
Table 3: Cross-country comparison of poverty and inequality in South Asia 
Gini index of 
inequality of 
expenditures
Poverty headcount ratio 
at national poverty line 
(% of population)
Year
India
32.5
28.6
2000
Pakistan
33
33
1999
Nepal
37
42
1996
Bangladesh
32
50
2000
Sri Lanka
47
23
2002
Maldives
42.5
18.9
2004
Source: 
World Development Indicators (except for Maldives, where estimates from this report are used) 
 
Figure 2: Poverty headcount, by Region 
Figure 3: Inequality, by Region 
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Male'
Northern
North
Central
Central
South
Central
Southern
H
e
a
d
c
o
u
n
t
p
o
v
e
r
t
y
r
a
t
e
VPA1
VPA2
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Male'
Northern
North
Central
Central
South
Central
Southern
G
i
n
i
c
o
e
f
f
i
c
i
e
n
t
1997/98
2004
7
Typically, poverty lines in South Asia are below the Rf. 15 per day used for Maldives, at least when compared at 
market exchange rates. Note that Maldives does not have a national poverty line. The Rf. 15 line used here is based on 
World Bank estimates and also conforms to earlier government practice. 
Page  17
6
Figure 4: Population and expenditure shares, by 
region 
Figure 5: Dominance Curve 
2004
0
5
10
15
20
25
30
35
40
45
Male'
Northern
North
Central
Central
South
Central
Southern
%
Population share
Share of total expenditures
0
.
2
.
4
.
6
.
8
1
P
o
v
e
r
t
y
i
n
c
i
d
e
n
c
e
c
u
r
v
e
s
f
o
r
V
P
A
1
a
n
d
V
P
A
2
0
10
20
30
40
50
60
70
80
90
100
Welfare indicator
Poverty incidence in 1997/98
Poverty in 2004
 
Expenditures per capita
 
Note:
 
The figure shows poverty incidence for a range of 
potential poverty lines (0-100 Rf. per person per day). The 
upper line refers to VPA1 and the lower line refers to VPA2
.
Source: 
Preliminary Staff estimates based on VPA-1 and VPA-2
.
Figure 6: Growth inequality curve (Annual Percentage Growth in Per Capita 
Expenditure, per Percentile), Maldives 
 
9
9
.
5
1
0
1
0
.
5
1
1
1
1
.
5
M
e
d
i
a
n
s
p
l
i
n
e
/
M
e
a
n
o
f
g
r
o
w
t
h
r
a
t
e
s
0
10
20
30
40
50
60
70
80
90
100
Percentiles
Growth rate of percentile
Mean of growth rates
Maldives
 
Source: 
Preliminary Staff estimates based on VPA-1 and VPA-2
.
Page  18
7
Figure 7
:
Growth inequality curves for Male’ and atolls, 1997/98-2004 
 
8
1
0
1
2
1
4
1
6
M
e
d
i
a
n
s
p
l
i
n
e
/
M
e
a
n
o
f
g
r
o
w
t
h
r
a
t
e
s
0
10
20
30
40
50
60
70
80
90
100
Percentiles
Growth rate of percentile
Mean of growth rates
Male'
 
8
9
1
0
1
1
M
e
d
i
a
n
s
p
l
i
n
e
/
M
e
a
n
o
f
g
r
o
w
t
h
r
a
t
e
s
0
10
20
30
40
50
60
70
80
90
100
Percentiles
Growth rate of percentile
Mean of growth rates
Atolls
 
Table 4: Ownership of Consumer Durables by Consumption Terciles, 1998-2004 (in %) 
Source
:
Preliminary Staff estimates based on VPA-1 and VPA-2. 
 
Table 5: Household transitions in and out of poverty, 1998-2004 (panel data sub-sample, 
covering atolls only) 
Category Status 
Number of  
Households
Share of Total 
(in %) 
Chronic poor 
Poor in 1998 and also in 2004  
184 
15 
Transient poor
Poor in 1998 but not poor in 2004 
419 
34.5 
Transient poor
Not poor in 1998 but poor in 2004
104 8.5 
Never poor 
Not poor in 1998 and also in 2004
509 42 
 Total 
1216 
100 
Source: 
Staff calculations based on the panel subset of the VPA-1 and VPA-2 surveys.
 
Note: 
The panel data sub-sample cover atolls only, and the poverty rates of this sample are not nationally 
representative.
 
V
ULNERABILITY TO 
S
HOCKS
 
For most households, poverty is a transient phenomenon. 
Panel survey data shows that 
15 percent of atoll households were chronically poor in the sense of remaining poor over the two 
samples under comparison, and nearly 43 percent of households in the atolls were transient poor. 
That is, they moved in and out of poverty over the study period as a result of fluctuating incomes 
 
Poorest tercile
2
nd
tercile Richest 
tercile 
Ownership 
of: VPA1 VPA2 VPA1 VPA2 VPA1 VPA2 
Fan 
26 91 33 97 45 97 
Electric 
Iron 28 91 33 96 42 95 
Fridge 
5 27 9 45 17 62 
Water 
pump 2 26 2 37 9 54 
Telephone 0.3 8 0.7 20 4 18 
Car 0 
0.3 
0 
0.6 
0.6 
1 
Dhoani 
(boat) 
1 7 3 8 5 13 
TV 
7  76  9  84 16 84 
Page  19
8
(Table 5).
8
Previous studies for other countries have also found much higher transitory than 
chronic poverty.
9
Transitory poverty reflects household vulnerability to income shocks
 that affect 
individuals or communities. Of course, if households do not have access to finance, assets, private 
networks, or public programs to cushion shocks, transient shocks can lead households to chronic 
poverty; and can also lead chronic poor deeper into poverty.
 
Food scarcity was an important recent shock experienced by households.
 Respondents in 
the VPA-2 survey were asked about food shortages they had faced within the last 12 months
—6
percent of households experienced such shortage of food (see Table 6). Those who suffered food 
shortage did so on average five times in the last year. The incidence and frequency of food 
scarcity was much higher in the lowest expenditure tercile (10% experienced shortage of food) as 
compared to the highest tercile (2.6% incidence). Incidence and frequency of food shortages was 
the same in Male’ and in atolls, suggesting that lack of income more than availability causes food 
scarcity. 
Table 6: Food scarcity 
Share experiencing 
food scarcity (%)
No. of times experienced in last 12 
months (conditional on having a crisis)
Tercile of expenditure:
 Low
10.2
6.2
 Medium
5.5
3.8
 High
2.6
4.4
Region:
 Atolls
65
 Male' 
6.4
5.8
Total
6.1
5.2
Source: 
Preliminary Staff estimates based on VPA-2. 
 
Health shocks and loss of main income earner were the most important adverse shocks 
to households over the past six-year period (the survey pre-ceded the tsunami, there is no 
doubt tsunami now would rank number one)
.
10
 Disease and injury constitute the most frequent 
type of shock (Table 7), experienced by 53 percent of those households that reported any major 
shock The five most often quoted health issues were viral fever, injury and accidents, stomach 
problems, and backbone injuries. Death of a household member is the second-most frequent 
shock (18% of reported shocks). Unemployment was reported as a major shock by 5 percent of 
8
A
small but interesting data set is available for a panel of households for 1997-98 and 2004: these are 
households that were interviewed in VPA-1 and again for VPA-2. The panel data is drawn from all areas of 
the country except Male’ 
9
Baulch and Hoddinott (2000) show that for several countries the proportions of the transient poor are 
much higher than the chronic poor.
 
10
 
The Survey Respondents were also asked about any crisis (or shock) they had faced in the last 6 years. 
The households that reported having experienced one of these shocks do not seem on average to be poorer 
or richer than other households. Only one-fifth of households reported a crisis; since one would have 
expected a higher incidence of crisis, it is possible that only relatively severe shocks were identified
.
Page  20
9
households, and the loss of boat or fishing equipment by 3 percent of households. Assorted other 
events account for the remaining 22 percent of shocks.  
Table 7: Shocks experienced by households (past 6 years) 
 
Share of all 
reported 
shocks, in %
How often 
shock occurred 
in past 6 years
Expenses to 
cope with 
shock 
(Rf.) 
Forgone 
income 
(Rf.) 
Av. expenditure of 
households that 
suffered this shock
(Rf. per capita per 
day) 
Death of bread 
winner/other 
18 0.9 
10,641 
7,671 40 
Sickness, 
injury, 
accident 
53 2.5 
26,804 
8,509 46 
Jobless 5 
1.3 
959 
5,401 
44 
Loss of fishing 
assets 
3
1.7
50,289
15,606
45
Other crisis 
22 
 
 
 
 
All crises 
100 
1.9 
22,864 
8,085 
46 
Source: 
Preliminary Staff estimates based on VPA-2. 
 
Aside from individual or household specific shocks, Maldives also faces aggregate risks:
 
·
The recent tsunami was a major natural disaster with national large-scale 
impacts
.
About 63,000 affected individuals were provided immediate cash assistance, 
many more were given in-kind relief, and some 3,700 households became homeless
 
·
The very low height of the islands render the country exposed to global warming
 
(over next 50 years) and potentially rising sea levels The government has begun 
addressing this long term risk via a focus island strategy emphasizing provision of high 
quality physical and social infrastructure at a few major regional centers. These focus 
islands would be equipped with sea walls and other protection against natural hazards
 
·
Socio-geographic vulnerability arises from the small size and large dispersion of 
the population, 
which often resides in small communities of less than 500 people, 
which translate into diseconomies of scale (unit costs of social services are 4-5 times 
higher than those found in continental developing countries) and contributes to the 
large inequalities between Male’ and the atolls in income and services.
11
 
·
The aging of the population, and associated poverty and health risks. 
The aging of 
the Maldivian society will occur as a consequence of the decreasing population growth 
rate and increasing longevity. The old-age dependency ratio
12
 currently at 5.8 percent is 
expected to stay constant until 2025, and to double between 2025 and 2050 to 11.6 
percent, though it will still be below levels found in higher income countries and Sri 
Lanka (Figure 8). Ageing also implies a heightened risk of old-age poverty, something 
which the government is planning to address through pension reform.  
11
 Currently housing more than a quarter of the Maldivian population on a total land area of less than 2 square 
kilometres. 
12
 (defined as the number of persons aged 65 and older divided by the number of persons aged 15-64, 
Page  21
10 
Figure 8: Demographic Dynamics 
Source: 
US Census Bureau
.
Households
 incur direct costs in terms of loss assets or medical fees, and forgone income 
as result of shocks. 
For example,
 
·
Households incurred direct medical costs as well as income loss as a result of 
adverse health shocks.
 Households who suffer shocks have difficulty coping and are 
much more likely to experience temporary poverty (which can sometimes become 
chronic). For example, health shocks, and especially disease, are associated with very 
high expenditures (equivalent to 3 months consumption for an average family). Of 
those who reported being sick 88.2 percent visited a doctor, 3.6 percent a trained health 
worker, 2.6 percent a traditional healer. Disability was reported by some 2.4 percent of 
respondents (because of the phrasing of the question, this is likely to cover only 
extreme disabilities). Aside from medical expenditures, sickness also resulted in a 
major loss of income during the illness.  
·
A
large rare shock such as the tsunami has resulted in massive loss of physical 
assets and forgone income.
 The total damages of the disaster are estimated at roughly 
US$470 million or close to 62 percent of GDP. About US$298 million of this are direct 
damages to physical assets, and the rest are indirect damages (such as lost tourism 
income).
13
 The cost of restoring the economy to its pre-tsunami level is estimated at 
US$406 million (WB/ADB/UN, 2005). The reconstruction phase is now getting under 
way, with efforts to rebuild damaged infrastructure and recover lost livelihoods.
13
 The indirect damage estimates need to be treated with caution particularly when it comes to tourism and livelihood 
damages. 
Page  22
11 
 
W
HO ARE THE 
P
OOR AND 
V
ULNERABLE
?
The highest incidence of poverty is experienced by households (a) with heads who do not 
work, (b) are women, (c) have little education, (d) with elderly members, and (e) have many 
children. The link between poverty and ‘not working’ is worrisome. Official unemployment, 
based in the narrow but standard definition of unemployment as “available for work, and actively 
searching for work within last week”, is low in the Maldives (2 percent). However, labor market 
participation is also very low, at 48 percent (62 percent for males and 33 percent for women). 
Many of those not employed are likely to be ‘discouraged workers’ that have given up active job 
searching. Among those working, households head working in manufacturing, fishing or 
agricultural households have a higher incidence of poverty than those in government jobs (Table 
8). 
Table 8: Sector of employment of household head by tercile of per capita expenditures 
%
of households 
Terciles of per capita expenditures
Industry in which household head is working:
Low Medium 
High 
Government 18 
29 
30 
Fishing 25 
17 
10 
Hotels/restaurants 3 
5 
6 
Other services 
2 
4 
6 
Manufacturing 28 
18 
18 
Construction 7 
5 
5 
Transport 3 
3 
4 
Agriculture 7 
4 
3 
Other/not working 
93 
85 
82 
Chronic and transient poor are largely similar, with only a few differences
:
Among the 
chronic (vs. transient) poor, there is larger share of female heads of households, a higher 
proportion of unemployed heads, large families, and households headed by widows, divorcees, or 
elderly. Unemployment and lack of education of the household head show some correlation with 
chronic poverty (Table 9). 
 
The largest group of poor is households who have a large number of children, who live 
in the atolls (especially in the Northern and North-Central region), and households whose 
head has little or no education
 (Table 10). The poorest tend to have no able-bodied members 
and receive no remittances. 
 
Estimates of the determinants of poverty (and vulnerability) confirm the above 
characteristics of the poor
.
Using regression techniques to identify the most important correlates 
of poverty reveals that residence in the atolls (especially in the north), households with a higher 
dependency ratio (ratio of children to total family size), less education, households headed by the 
elderly (65+), women, and households with widows have a higher probability of living in 
poverty. These characteristics are constant across the two survey years (see Annex 2). 
Table 9: Incidence of Select Demographic and Household Characteristics among the 
Chronically Poor, Transient Poor, and Non-Poor (In %, Panel Sample) 
Page  23
12 
Characteristics 
Chronic 
poor 
Transient 
Poor 
Never 
poor 
Female headed household 
53 
44 
44 
Household head uneducated 
28 
30 
25 
Household head unemployed 
22 
14 
16 
Household has at least one widow or 
one divorcee 
30 30 21 
Household contains elderly members 
31 
22 
27 
Household containing more than 6 
children 
11 8 6 
Household head illiterate in English 
96 
92 
86 
Any household member migrated 
18 
0.2 
21 
Source: 
Preliminary Staff estimates based on the panel sub-sample covered in both VPA-1 and VPA-2. 
 
How do households cope with poverty and vulnerability?
 Poor households use informal 
insurance as the main coping mechanism, followed by own income and government support. 
Government support was used for coping in about 10 percent of the crises cited here. In many 
countries households faced with a shock respond in a manner that undermines their human or 
physical capital
—taking children out of school, putting children to work, selling their major 
assets, incurring debt at usurious interest rates, and so on. There is little or no evidence to suggest 
that this type of behavior is widespread in the Maldives.  
C
ONCLUSIONS
 
The main findings of this chapter are the following: 
·
Poverty has declined dramatically
—by more than half—in Maldives during 1997/98-
2004 
(from 45 to less than 19 percent). The incidence of poverty is now fairly shallow. The 
reduction in poverty was caused by high and widespread economic growth, which swamped 
an increase (from already quite high levels) in inequality over the same period.  
·
Poverty in the Maldives is now amongst the lowest in the region, but inequality appears 
to be amongst the highest in South Asia 
(although comparability is problematic). However, 
the positive effect of high economic growth has swamped substantially the small negative 
effect of rising inequality on poverty.  
·
The regional incidence of poverty is now concentrated in the atolls in the Northern part 
of the country
.
The incidence of poverty is much higher in the atolls than in Male’, and more 
pronounced in the North vs. the South of the country. The Northern and North-Central 
regions contain more than 60 percent of the country’s poor. The atolls realized a modest 
growth in inequality, which reduced the impact of growth on poverty reduction.  
·
Aside from low income (poverty), the Maldivian population also suffers from 
vulnerability arising out of aggregate shocks (natural disasters, aging) and 
individual (household-specific shocks)
.
Aggregate shocks include the most recent 
impacts of the tsunami which caused enormous loss of incomes and assets and the 
potential environmental risks of global warming and a rise in sea level. The most 
important individual shocks are health risks and joblessness. 
 
Page  24
13 
 
Table 10: Poverty rate and population size of select groups (in %) 
 Poverty 
incidence 
%
Share of 
population 
%
Share of poor 
population 
%
Region 
 
Male' 4.5 
29.7 
7.1 
Northern 40.4 
16.2 
34.6 
North-Central 32.5 
16.2 
27.8 
Central 22.6 
8.6 
10.3 
South-central 14.8 
12.7 
9.9 
Southern 11.7 
16.7 
10.4 
All atolls combined 
25.0 
70.3 
92.9 
 
Household contains elderly 
19.7 
73.8 
76.9 
Household does not contain elderly
18.6 26.2 
23.1 
 
Household contains only elderly 
47.6 
0.4 
0.9 
Household does not only consist of 
elderly 
18.8 99.6 
99.1 
 
Education of household head 
 
Below primary 
23.4 
61.9 
76.7 
Primary 16.8 
9.2 
8.2 
Lower secondary 
14.2 
15.9 
11.9 
Higher secondary 
1.6 
3.8 
0.3 
Higher education or vocational 
6.0 
9.3 
3.0 
 
Household head is jobless 
24.1 
25.9 
33.1 
Household head is employed 
17.1 
74.1 
67.0 
 
Female head 
19.7 
46.3 
48.4 
Male head 
18.2 
53.7 
51.7 
 
Household contains a widow or 
divorcee 
18.1 33.3 
32.0 
Household does not contain a 
widow or divorcee 
19.3 66.7 
68.1 
 
Number of children in household 
 
0
11.1
6.8
4.0
1
10.6
13.9
7.8
2
12.1
18.1
11.6
3
18.1
20.7
19.8
4
21.8
16.1
18.6
Page  25
14 
 Poverty 
incidence 
%
Share of 
population 
%
Share of poor 
population 
%
5
28.2
9.9
14.8
6
34.5
6.8
12.4
7
20.1
3.9
4.2
8
or
more 35.4 
3.7 
6.8 
·
The incidence of poverty and vulnerability is highest among households with little 
education, with a large number of children, with elderly, and those not working
.
Some 
(but not all) among widows and divorcees may also be highly vulnerable.  
·
The largest share of the poor and vulnerable are 
households with many children
,
and households whose heads are unemployed and/or with little education and/or 
female-headed. Ninety percent of the poor live in the atolls, of which two thirds live 
in the northern atolls. 
·
The main coping mechanisms used by the poor are 
informal insurance, and own 
savings and income. Government support was used for coping in about 10 percent of 
the crises.  
 
These findings suggest that 
economic growth will be the main driver of poverty 
reduction, but not all regions and population groups may benefit from growth
.
Specifically, 
social protection policies will need to address the concentration of poverty in the Northern and 
North-Central atolls and to households without able-bodied members or other sources of income 
(such as remittances), poverty among the less educated, families with many children, and the 
elderly without family support. The design of social protection policies will have to address 
chronic poverty, but also take into account transient poverty caused by shocks to income (e.g., 
health shocks, hunger, and natural disasters), and the aging of the population.  
 
The following two chapters build on the above findings and address two key social 
protection areas of government concern
:
providing 
old age income security through pensions
 
(given high poverty rates among the elderly and an aging population), and 
safety nets for the 
poorest population groups
.
Page  26
15 
The Pension System
 
In the Maldives there is a Constitutional mandate to provide a pension to all employed 
persons, although the Constitution does not define either the type of pension or the exact 
definition of ‘employed’.
 Currently, however, most of the population is not covered by any 
pension scheme. The minority that does receive a cash payment that is labeled a pension works in 
the public sector. As discussed below, this ‘pension’ does not perform the usual function and is 
instead a kind of wage supplement for civil servants. The Government of Maldives is planning to 
replace this scheme and extend a basic or ‘social pension’ to the rest of the population. What 
follows is based on close interaction with the Government’s working group assigned to deal with 
these issues following extensive consideration of other options during 2004.  
 
Pension systems generally have two primary objectives: minimum income security, and 
smoothing consumption (and longevity insurance).
 The first is to provide a minimum level of 
income security in old age and this often entails redistribution to the elderly poor through public 
programs. The social pension that is proposed below will need to be coordinated with the 
Government’s other safety net schemes (see next chapter). The second objective is to smooth 
consumption by shifting income from individuals’ working years to their old age. Part of this 
smoothing function is often performed by an annuity that also provides longevity insurance.  
 
This chapter reviews the current pension system and presents a limited set of options 
for reform.
14
 Reforms must take into account the overall initial conditions especially with regard 
to the fiscal situation. Phasing in those parts of the package that generate savings initially may be 
necessary in light of short term fiscal constraints. At the end of the chapter, a hypothetical time 
line is presented. 
C
URRENT 
P
ENSION 
S
YSTEM
:
O
VERVIEW
 
The pension system in the Maldives covers only public sector employees
.
It has two 
components–the Civil Service Pension (CSP) scheme and the Government Provident Fund 
(GPF). The Civil Service Pension (CSP) scheme is a non-contributory, pay-as-you-go (PAYG) 
defined benefit (DB) arrangement covering approximately 26,000 government employees. 
Pensions are financed directly from the government budget. The Maldives’ pension system is 
globally unique in the sense that pensions are paid out every 20 years of uninterrupted 
government service and there is no requirement to retire.
15
 Employees can continue to work for 
another 20 years and earn a second pension, and in a few cases, even a third. Therefore, with an 
accrual rate of 2.5 percent per year applied to the last wage in the 20 year period, pensioners in 
Maldives receive their benefits along with their earnings. The pensions are not indexed however, 
so their value in real terms declines over time depending on rates of inflation.  
 
The Government Provident Fund (GPF) is a defined contribution (DC) arrangement 
with matching employee and government contribut
ions of 5 percent of the employee’s basic 
wage each month
.
Participation is voluntary and withdrawals are allowed from their account for 
the education of their children, house construction and repair, health care and obligatory hajj. 
Recently, there have been massive withdrawals as the Government encouraged the purchase of 
housing. The scheme does not play a major role in providing a source of retirement income. On 
14
 Given the advanced stage of the Government’s discussion, alternatives are presented only insofar as they conform to 
the basic structure of the Working Group’s proposal.  
15
 
Many countries are moving away from mandatory retirement ages and/or reducing disincentives to retirement, 
especially in aging OECD countries with rising ratios of pensioners to workers. Most civil service schemes however, 
have maintained this practice which is felt to be necessary from a human resource perspective.
 
Page  27
16 
the other hand, it involves a significant subsidy at a cost to the government for workers that 
participate.  
 
An assessment of current policy found two major deficiencies
:
First, the civil service 
pension scheme is not designed to produce a reasonable level of retirement income. Specifically, 
civil servants receive a pension equivalent to 50 percent of basic salary after every 20 years of 
service and regardless of age. The pension calculation does not apply to overtime and allowances 
which on average comprise around 40 percent of total remuneration. Combined with the fact that 
the pension is not indexed, its value is very low compared to average incomes when the worker 
reaches old age. The current parameters of the scheme do not result in secure and adequate stream 
of pension income after a certain age. Given that the scheme does not stipulate a retirement age, it 
is not surprising that civil servants choose never to retire in the Maldives (although it is clear that 
many are not actually working either). One interpretation of the current practice is that wages 
paid to elderly civil servants until they die are simply an expensive and inefficient substitute for 
paying an adequate pension.
16
 
The second deficiency noted by the GoM is the absence of any safety net for the elderly 
in a country where only civil servants participate in a pension scheme
.
With more than three-
quarters of the population lacking pension coverage, the only kind of program that is likely to 
have any impact in the near future is one that involves transfers to the elderly not linked to 
employment or contribution history. In other words, the safety net would be based on 
redistribution rather than savings. The design of this program would explicitly take into account 
reforms to the civil service pension scheme given the overlap in the population being covered and 
the need to coordinate overall target pension levels. 
 
The main objectives of the pension reform that have emerged from the dialogue 
between the World Bank and the GoM can be summarized as follows
:
·
To introduce a financially sustainable retirement savings scheme for civil 
servants (with the possibility of expanding to other formal sector workers in future)  
·
To provide a safety net for all elderly citizens in the Maldives. 
 
T
HE 
E
MERGING 
R
EFORM 
P
ACKAGE 
 
In this context, the Government has set up a working group to study options for reform.
 
The group includes representation from the President’s office, the Ministry of Finance and the 
Ministry responsible for social security. The emerging reform package would rely on two new 
pension schemes, one contributory and one non-contributory.  
16
 
In fact, the original motivation for the first civil service pension schemes was to replace employees whose 
productivity had started to decline while keeping them from falling into destitution.
 
Page  28
17 
 
Contributory Pension System
 
The first element of the reform package will address the consumption smoothing 
objectives of a pension system
.
As such, this part of the system will be based on contributions or 
savings and will seek to replace a significant f
raction of the worker’s own net income after 
retirement. The contributory plan will replace current rules for civil servants, starting with the 
introduction of a retirement age, under the current proposal, at age 65. This would immediately 
affect roughly about 1,400 civil servants. Those employees hired after the reform will participate 
in a defined contribution scheme. With regard to those already covered by the current scheme, 
two alternatives are being considered.  
Option 1: Fully Funded DC Scheme
 
The first option offered is an immediate switch to a fully funded DC scheme.
 This 
appears to be the preferred option (based on comments received from the Government).
 
In Option 
1, effective from the implementation date, all existing employees and new entrants to the 
government will be mandated to begin contributing, along with the Government to a new 
individual retirement account.  
 
Under the DC scheme, matching employee and government contributions totaling 14 
percent of the employee’s total monthly emoluments will be channeled to individual 
retirement accounts each month.
 The funds will then be invested by the pension fund manager 
in accordance with the investment policy and the returns to the investment will determine the 
actual pension payments of the employee upon retirement. The proposed contribution rate was 
determined based on a target replacement rate of 50 percent of lifetime revalued average wage for 
a
full career worker.
17
 
As shown below in Figure 9, this target could be achieved with a 14 percent contribution 
rate under reasonable assumptions about mortality, rates of return and wage growth
.
The 
figure also highlights that outcomes are not guaranteed by the Government. If investment returns 
were to trail the growth of wages by say, 2 percentage points over the career of a worker, the 
replacement rate would fall critically short of the required; on the other hand, better-than-
expected replacement rates would be achieved if investment returns were to exceed wage growth 
by 2 percent or more. 
17
 
The average based on a similar definition was around 56 percent for the OECD. See OECD (2005).
 
Page  29
18 
Figure 9: Replacement Rates under Different Rates of Return on Investments of the DC 
Scheme (for 14% contribution rate) 
0%
25%
50%
7
5%
100%
-3%
-2%
-1%
0%
1%
2%
3%
Difference in ROI and wage growth (in real terms)
R
e
p
l
a
c
e
m
e
n
t
r
a
t
e
(
p
e
n
s
i
o
n
a
s
%
o
f
f
i
n
a
l
w
a
g
e
)
T
arget 
R
eplacement Rate
Source: 
Bank staff calculations. 
 
For existing employees, it is proposed that the accrued years of service of each 
individual employee be valued as if they have been accrued under a DC scheme.
 Therefore, 
as recognition of accrued rights, the government will put aside a lump sum payment based on the 
individual’s accrued years of service, a total contribution rate of 14 percent, and the employee’s 
total monthly emoluments preceding the month before the implementation
 
of the new scheme. 
The lump sum is calculated as shown in Formula 1 below. 
Formula 1 
Lump sum in recognition = (14% x total monthly emoluments) x (accrued service (in months))  
Example 1 
 
If the implementation date is assumed to be 1 January 2006, then an individual with a total 
monthly emolument of Rf 4000 in Dec 2005 and 10 years of service by the end of that year, will 
have his/her accrued rights recognized as follows:  
Lump sum payment  
 
= (14% x 4000) x (120 months)  
Transferred into account  
= Rf 67,200.  
In the example in Formula 1 above, a lump sum of Rf. 67,200 will be transferred into the 
individual’s retirement account immediately before the commencement of the employer-
employee contributions under the new DC scheme. The accumulated pension after the reform 
would be based on the monthly contributions of 14 percent of the employee’s total emoluments 
each month plus investment returns, until retirement. By the time the employee reaches 
retirement the total pension accumulated would comprise of the lump sum payment and the 
accumulated balance after the reform plus returns to investment. Upon retirement, pensions will 
be paid out in the form of annuities each month, indexed to price inflation.  
Page  30
19 
 
The obvious benefit of a mandatory switch for all employees would be a quicker and 
possibly more cost-effective transition.
 The employees would benefit from the accumulation of 
a
100 percent government sponsored lump sum payment to their retirement accounts, while the 
government would benefit from making a large implicit pension debt into an explicit payment. 
The savings objective will also be met as the payment will not be accessible until retirement and 
upon retirement it will paid out as an annuity. Note that although older workers would be 
immediately included in the scheme, the value of their benefit would be almost completely 
known from the outset of the scheme since it would be derived from the valuation of their past 
service. Younger workers on the other hand, would derive their final pensions from the 
contributions and investment returns over the next four decades. In effect, there would be a very 
gradual shift from a DB to a DC environment as each cohort passes through the system.  
 
The drawback to this option, from Government’s perspective, is that a large amount of 
initial financial outlay will be required to recognize the accrued rights of all the employees 
at once.
 While there would be some savings in terms of wages not paid to those over the 
retirement age, the short term outlays could be very large if the previous service years are ‘cashed 
out’ immediately.  
 
An alternative financing approach would be to issue special bonds that would be 
deposited into the individual accounts and redeemed at the time of retirement
.
This method 
has been used in several Latin American countries where they are referred to as ‘recognition 
bonds’ since they recognize past service credits toward the pension. Under this arrangement, the 
initial portfolio would consist mainly of the special bonds which would earn a rate of interest and 
be redeemed at retirement. Gradually, the investment policy would shift the asset allocation to a 
diversified portfolio based on a target rate of return or a limited range of options provided to 
members. If desired, the transition could be accelerated by having the government buy back these 
bonds with a cash injection to the extent affordable.  
 
In contrast to much of South Asia,
18
 the civil service in the Maldives is relatively young 
and the average number of years of service is relatively low.
 This is because the Maldives is in 
the early stages of a demographic transition, and also because the public sector has expanded 
dramatically in the last 15 years. In other words, both the average age and the average number of 
years in service are quite short, making the transition to a DC scheme less costly. Unless the 
public sector continues to expand as a share of the labor force however, the system will begin to 
mature and the average age in the public sector will rise rapidly. The current situation presents a 
unique opportunity to transition to a fully funded pension scheme at a time when the transition 
costs are not very high. 
 
Figure 10 illustrates the favorable demographics that would allow for most workers to 
join a DC scheme and accumulate a sufficient balance to provide a reasonable replacement 
rate at age 65
.
The average age of a civil servant in Maldives today is approximately 34. This is 
much lower than is the case in neighboring countries. In Pakistan the average age is about 37 and 
in Indian states such as Tamil Nadu, the average age is closer to 40. On the other hand, if the size 
of the civil service relative to the overall labor force remains relatively stable, the average age of 
a
civil servant will increase dramatically. By 2040, the average civil servant would be more than 
45 years old. As the first objective of the reform package is to provide income security during 
retirement, this requires that the new pension scheme be consistent with the ageing process in 
Maldives, both in terms of the public sector workforce and the population as a whole. Lastly, a 
fully funded scheme is by definition sustainable.  
18
 
See Palacios (2005) for a regional review. 
Page  31
20 
Figure 10: Age Distribution of Civil Servants, 2003 
0%
5%
10%
15%
20%
2
5%
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80-84
85+
Age
P
e
r
c
e
n
t
d
i
s
t
r
i
b
u
t
i
o
n
Source: 
World Bank staff calculations
 
Option 2: A Hybrid Defined Benefit and Defined Contribution Scheme 
 
A
second option would involve introduction of a hybrid of a fully funded DC and a 
partially funded DB scheme, with gradual transition to a fully funded DC scheme for all 
government employees
.
Under it, as in Option One, matching government and employee 
contributions totaling 14 percent of the emp
loyee’s total monthly emoluments will be made.  
 
Effective from the date of implementation, all new entrants to the government will 
automatically be mandated to join the fully funded DC scheme and start contributing 7 
percent of their total monthly emoluments, with matching contributions from the 
government into their individual retirement accounts
.
As new entrants, they will not have any 
accrued pension rights. Thus, the accumulated pension benefits will be based only on the 
accumulated contributions and returns to investments.  
 
Existing employees could be divided into different groups based on their accrued years 
of service or age, which will in turn determine whether they will be covered under a fully 
funded DC or a partially funded DB scheme after the reform
.
For example, those employees 
with fewer than 10 years of service would be required to join the fully funded DC scheme, with 
full recognition of accrued rights as described under option one. Those with more than 10 years, 
but aged 65 or below would be given a choice between joining the DC scheme and the partially 
funded DB scheme. Recognition of accrued rights of those who choose the fully funded DC 
scheme will be treated in the same manner as that of those existing employees who were 
mandated to join the DC scheme. 
 
However, the accrued rights of those who choose the DB scheme are recognized 
differently.
 The accrued pension rights will be recognized at retirement by crediting all accrued 
years of service at the new accrual rate of 1.5 percent per annum as shown below under Formula 
2. The new accrual rate of 1.5 percent instead of the old accrual rate of 2.5 percent is used 
because when the accrued rights are credited at
 
retirement, it will be applied to a much broader 
base of total emoluments
.
Formula 2 
Women 
Men 
Page  32
21 
Accrued pension rights = 1.5% x years of service before reform  
Example 2 
A
40
year old employee with 15 years of service before the reform would have accrued pension 
rights = 1.5% x 15 = 22.5%. The accrued pension rights of 22.5% will be credited only upon 
retirement.  
Now the employee will start accumulating pension rights under the new DB scheme until 
he/she reaches retirement at age 65. Thus the accumulated pension rights for 25 years of service 
after the reform would be 37.5%.  
As is evident, the accumulated and accrued pension rights sum up to 60%, which is the target 
replacement rate for a full 40 years of service. Thus the pension amount received would be 60% 
of the reference wage at retirement.  
Finally, those employees over the retirement age at the time of the reform would have the new 
1.5 percent accrual rate applied to their years of service to that point. The pension payments 
received under the old scheme for every 20 years will be folded into the reference wage.  
 
Non-contributory pension or demogrant 
 
The second element of the new system under consideration will serve to ensure a 
minimum level of income during old age for all Maldivians (as vaguely mandated by the 
Constitution) that is not related to past contributions
.
This program will be financed from 
general budget revenues and therefore will entail redistribution from taxpayers to the elderly. 
This scheme would have to be coordinated closely with the overall safety net system (see 
Chapter 3). 
 
The Government is aware that this type of demographic targeting does not maximize 
poverty reduction for a given level of resources. The program is seen as complementing 
rather than substituting for the targeted safety net. The parallel approach is considered 
useful for several reasons. 
First, it fulfills the constitutional mandate, although not necessarily 
in the way anticipated by the language of this document. Second, viewed from a long term 
perspective, it completes the overall pension system by providing redistribution and a minimum 
pension outside the contributory pension scheme. In the long run, as coverage in the 
contributory scheme expands along with the size of the formal sector, the combination of the 
two components would achieve the consumption smoothing and poverty alleviation objectives. 
It would also reduce the long run costs of the demogrant should the suggested pension test be 
applied. This separation of redistributive and savings functions has been recommended by the 
World Bank and others for various reasons.
19
 Finally, there may be additional arguments for 
separating this cash transfer based on the permanence of the condition (old age), the 
simplification of administration and therefore reduction of costs, and even questions of 
intrahousehold allocation.
20
 
19
 
See World Bank (1994).
 
20 Another possibility would be to merge the non-contributory old-age pension into the proposed new umbrella poverty benefit. In this option, the targeting 
mechanism and benefit scheme developed for the proposed umbrella benefit would also apply for the old-age benefits. This option would provide room for 
Page  33
22 
Option 1: Demogrant
 
The 
proposed non-
contributory pension or ‘demogrant’ would use the same benefit level 
as the current Absolute Poverty Scheme (Rf. 500 per month). 
In order to avoid the 
administrative burden of means-testing the non-contributory pension, the eligibility condition 
could be based purely on age and residence. The eligibility age might be coordinated with the 
reformed contributory scheme, presumably at 65. As shown in Figure 14 below, the cost of the 
program is very sensitive to the eligibility age chosen.  
Option 2: Pension Tested Demogrant  
 
One option to reduce costs and improve targeting at the margin, the proposal includes a 
‘clawback’ based on a pension test
.
In other words, if an individual already received a pension 
at or above the Rf. 500 amount paid from the demogrant, that individual would not receive the 
transfer. On the other hand, if the contributory pension generated less than Rf. 500, the individual 
would receive a ‘top-up’ to 500. Furthermore, the MHEESS makes a monthly payment of Rf.500 
to those identified to be living in absolute poverty and at the moment there are 1026 people 
registered in the program, of which 51 percent are over 65 years. Thus once the demogrant is 
introduced, most
21
 who are over 65 years will automatically be exempt from receiving the 
payment under the absolute poverty program.
2
2
Another important policy choice involves the adjustment of benefit levels in the future
.
In Figure 11, the benefit is projected based on the assumption that it maintains its level relative to 
per capita income in the future. An alternative that would reduce costs is to either link it to an 
absolute poverty line or to inflation. The important principle in this regard is to adjust benefits in 
a
transparent way linked to objective and explicit criteria rather than on an ad hoc basis.
 
significant fiscal savings (depending on how large a share of the elderly the system would aim to cover). For a review of the arguments for and against non-
contributory old age pensions or ‘social pensions’, see Palacios and Sluchynsky (2006). 
21 A few elderly in co-residence situations with very poor households may still quality for safety net benefits.  
22 Another possibility would be to merge the non-contributory old-age pension into the proposed new umbrella poverty benefit. In this option, the targeting 
mechanism and benefit scheme developed for the proposed umbrella benefit would also apply for the old-age benefits. Under this scenario, Maldives would have a 
truly unified social assistance system. However, not all the elderly population would then be covered, and it is unclear if the Constitutional mandate would be 
achieved. This option would provide room for significant fiscal savings (depending on how large a share of the elderly the system would aim to cover). 
Page  34
23 
Figure 11: Cost of demogrant with different eligibility ages 
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
2006
2011
2016
2021
2026
2031
2036
2041
2046
2051
s
p
e
n
d
i
n
g
a
s
%
o
f
G
D
P
6
5
70
75
Eligibilty age for demogrant
F
ISCAL 
I
MPLICATIONS
 
Assuming that the package proposed by the Working Group was to be fully 
implemented as of January 1, 2006, the overall fiscal cost could be on the order of 1.5-2 
percent of GDP
.
23
 The new pension system introduces three types of new cost to the government 
as shown in Figure 12 below. 
 
First, in the base case shown above, the demogrant would increase the number of 
elderly receiving a Rf. 500 per month transfer from 500 to around 14,000 at a cost of 
approximately 0.7 percent of GDP
.
As the proportion of the population above age 65 rises in 
the long run, so would the cost as a share of GDP. A cost saving option discussed in Chapter 3 
would be to target the non-contributory pension to the elderly poor. The second cost is the 
contribution that the GoM would now be making on behalf of its employees into the defined 
contribution scheme. This is equivalent to 7 percent of the wage bill under Option 1 and around 
0.4 percent in Option 2 since only workers with fewer than 10 years of service would be 
mandated to join the DC scheme. Over time, the contribution in both schemes would be made on 
the entire wage bill so the cost of this component would converge in the two options. Another 
way of reducing the cost of the demogrant in the long run is to index it to prices rather than adjust 
it in line with incomes over time. This however, would gradually reduce the value of benefits 
relative to average incomes and would make more sense in the context of coverage expansion of 
the contributory scheme. 
 
The third element is the cost of paying out pensions to those that retire in the new 
contributory scheme after the reform is implemented
.
In both Option 1 and Option 2, 
approximately 1400 civil servants over the age of 65 would retire immediately and be paid 
pensions based on the respective methods of valuing past service. Option 2 costs less in this 
regard however, because there is no provision for survivor benefits while in Option 1, a refund of 
the balance would be provided to the survivor in case of death. In this sense, Option 1 is more 
23
 See Annex 1 for detailed explanation of methodology and data used for the projections. 
Page  35
24 
generous with regard to the benefit payout; the inclusion of a survivor benefit for Option 2 is one 
of several outstanding issues (see below).
 
The reform package has significant fiscal implications on a cash flow basis. As always, it 
is crucial to make an assessment of the tradeoffs (higher deficits, increased taxes or reduced 
spending in other areas) that are always present in the shift to a funded pension scheme
.
It 
is important however to keep in mind that the current practice is essentially to pay workers their 
full emoluments until they die rather than shifting to a pension that represents some fraction of 
that amount as is the case in all other countries. The relevant comparison then is between the 
present value of future wage payments and the present value of future pensions under the reform. 
This calculation has been made for the current workers over age 65 as shown in Figure 13 below. 
The same argument applies to the existing workforce. The key difference is that for those already 
above age 65 the stream of pension payments will not be backed by a fund. For new hires, 
pensions would be prefunded and for workers already in the old system, the proportion of funding 
will increase with each retiring cohort.
 
Figure 12: Cost of pension reform package proposed by Working Group 
 
Source: 
Preliminary staff calculations subject to revision pending data clarifications. 
 
The caveat is that to the extent that hiring practices lead to the full replacement of all 
retiring workers, these savings will not materialize
.
In our cost estimates, we have ignored the 
savings by maintaining constant the size of the civil service (proportional to the population). This 
is a pessimistic assumption
—there is general consensus that older workers do not contribute 
much to the functioning of government, and some do not work at all. If they are not replaced, the 
impact on what the public sector produces may not be affected and if they are replaced, public 
sector output would increase. In either case, the cost estimates presented here are on the high side. 
0%
1%
2%
3%
1
2
12
12
C
o
s
t
a
s
%
o
f
G
D
P
Payment
Contribution
Demogrant
2006
2015
2030
0%
1%
2%
3%
1
2
12
12
C
o
s
t
a
s
%
o
f
G
D
P
Payment
Contribution
Demogrant
2006
2015
2030
Page  36
25 
Together with the assumption that the benefit level of the demogrant would be raised with 
incomes, the cost estimates presented here should be seen as an upper bound. 
 
It is also important to keep in mind that the alternative of introducing a partially 
funded or unfunded defined benefit scheme would result in an unfunded liability. 
Although 
the cash flow implications of the government’s contribution to the scheme are significant, the 
avoidance of a new liability is an important advantage of the proposed reform. To the extent that 
the funds are invested abroad (see discussion below), funding also provides an opportunity to 
diversify the source of workers’ retirement savings, an important consideration in a small, 
undiversified economy with extremely limited avenues for investment.
 
Figure 13: Comparing present value of wages to pensions for those over age 65 
-
100,000,000
200,000,000
300,000,000
400,000,000
500,000,000
600,000,000
NPV of wage bill
NPV of pension payments
C
o
s
t
i
n
R
f
Source
:
Preliminary staff calculations subject to revision pending data clarifications. 
Note
:
NPV=Net present value 
 
How are individuals affected by the proposed reform? 
 
Clearly the demogrant unambiguously raises the incomes of households with elderly 
members
.
As discussed in Section 2, households with elderly members are somewhat poorer 
than other households in Maldives, and a transfer of this size can be expected to make some 
contribution to poverty reduction. The impact of the program would grow over time as the 
population ages assuming that benefits keep pace with incomes. 
 
The impact of introducing the contributory pension scheme is much more 
complicated
.
Three categories of individuals can be separated for the analysis 
–
new 
entrants, current civil servants under age 65 and civil servants over age 65. We take each in 
turn. New entrants to the civil service will contribute to their individual accounts and will 
therefore have net compensation 7 percent lower than would have been the case otherwise. 
They will also not be eligible for lifetime employment but will have to retire at age 65. 
Instead of receiving the compensation of an active employee until they die, a pension will be 
awarded and its value would be a function of the number of years he/she contributed and the 
investment return that was earned. 
Page  37
26 
 
For workers already in service at the time of the reform, the impact will depend on 
their current age and when they joined
.
For older workers that had joined at a young age, 
say 21 and who will subsequently retire at age 65, the method of calculating their past service 
credits described above would result in relatively high replacement rates 
–
in this case, 
around two-thirds of pre-retirement income. In other cases, where workers for some reason 
have few years of service at ages 40 or above, replacement rates would be modest. As a 
transition mechanism, a minimum replacement rate could be applied. This would obviously 
increase the cost of the overall package, albeit marginally for reasonable target benefit levels. 
 
The same logic applies to workers that must retire immediately after the reform is 
introduced due to the mandatory retirement age
.
In their case, the effect on their current 
income is immediate and again depends on the number of years of previous civil service 
employment. Figure 14 shows the dispersion of outcomes in terms of replacement rates for 
this group. Whether workers are better or worse off under the reform depends on their 
subjective valuation of leisure versus work. Of course, for those individuals that were not 
actually working currently and still received pay, there is a clear loss
.
Figure 14: Dispersion in replacement rates for civil servants retiring immediately after 
introduction of mandatory retirement age 
0
100
200
300
400
500
600
700
800
25%
50%
75%
100%
150%
200%
More
Replacement rate
F
r
e
q
u
e
n
c
y
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
More analysis should be done prior to the implementation of mandated retirement to 
confirm the individual effects
.
For example, there are many older workers with relatively few 
years of service. This is why the first bar to the left of Figure 14 is so large. It is possible, 
however, that some of the data includes errors and that in fact, older workers would receive 
higher pensions than what is shown here.  
Overall assessment
 
The basic reform proposal appears sound taking into account the mandate of the 
Working Group. 
It provides a universal minimum pension in the form of the demogrant and 
establishes a sustainable consumption-smoothing scheme for civil servants. The latter scheme 
Page  38
27 
provides for adequate replacement rate targets and holds several advantages over alternative 
models. In particular, it avoids the creation of new, open ended liabilities for the government, 
allows for international diversification of retirement savings sources and results in a platform that 
could readily be extended to the rest of the population. 
 
Introducing a new pension scheme of any kind implies a range of implementation 
challenges.
 These challenges are attenuated when a fully funded defined contribution system 
with private asset management is involved. A brief review of the various implementation issues 
that arise are discussed next. 
Implementation of reforms  
 
Implementation of the proposed reform package requires actions in at least four 
different areas
:
Data management, asset management, investment policy, and insurance benefits. 
The last three are related to the new defined contribution scheme and some are likely to take 
several years to fully implement. Interim arrangements will be necessary in some cases. (It is 
recommended that the recordkeeping for the demogrant be integrated with that of the safety net 
benefits as discussed in Chapter 3). This section looks at each of these issues in turn. 
Data Management
 
Both the parametric and systemic reforms require computerized information on 
earnings histories
.
The gradual shift to a longer averaging period for the reformed defined 
benefit scheme requires that earnings histories are constructed and available in electronic form to 
the pension department so that pensions can be accurately and efficiently calculated.  
 
The data requirements for the new DC scheme are more demanding.
 Individual account 
contributions must be tracked monthly with adjustments to status of the employee and any 
changes in remuneration taken into account on a regular basis to ensure that deducted amounts 
are correct. Workers should receive regular statements and a mechanism for resolving any 
discrepancy through reconciliation of information should be possible. Individual workers may be 
provided with choices as to fund managers and portfolios; an information system that allows for 
efficient processing of these selections may have to be designed. Depending on investment 
policy, it is likely that valuation should be done on a daily basis. 
 
In principle, systems that provide some of this information should already be available 
since individual accounts already exist in the form of the GPF
.
However, it is likely that the 
GPF accounting is subject to significant delays, errors and even fraud (in this sense, improving 
recordkeeping would provide benefits whether or not the DC scheme was being introduced). The 
first step towards implementation would be
 an assessment of current systems and a ‘gap analysis’ 
that lays out a plan to migrate to the new information system necessary. Existing GPF staff could 
be retrained to administer the new scheme’s recordkeeping although new software and upgraded 
hardware may be required. 
 
The international experience in this area suggests that significant delays can be 
expected.
 In the US for example, there have been problems with information systems purchased 
to track the DC accounts of several million civil servants that belong to the Thrift Savings Plan 
(TSP). The DC scheme for civil servants in Thailand, introduced in 1997, has yet to completely 
centralize their recordkeeping process and is forced to rely on separate submissions by ministries. 
Depending on when the scheme is implemented in the Maldives, an interim arrangement that 
ensures that records are kept from the inception of the scheme until the final system is in place are 
Page  39
28 
likely to be necessary. The starting point would be the current records kept for the purposes of the 
GPF, but this would not offer a long term solution for a system that should be accessible on a 
timely basis and with (eventually) daily valuation of assets. 
 
Another relevant case is that of India. 
In India, civil servants hired after January 1, 2004 
are automatically enrolled with the new DC scheme. While significant background work has been 
done towards establishing a new recordkeeping agency, this entity had not been contracted after 
more than two years.
24
 In the meantime, ten percent of the salary of approximately 100,000 
workers is being deducted in the same way as had always been done with GPF contributions. The 
balances are bookkeeping entries just as the GPF had always been and returns are prescribed.
25
 
This interim arrangement has allowed the scheme to start operating on a notional basis but only 
when the new Central Recordkeeping Authority is established will monies be invested outside of 
government.  
 
The final set of implementation issues relate to the demogrant program. This will 
require the registration of all Maldivians age 65 and over from the time the scheme is 
introduced
.
A
database to monitor them and to stop payment after death will have to be created 
and staff assigned. The same database would have to be cross-referenced against civil service 
pensioners in order to apply the pension test. The Social Security Section is expected to 
administer this program. An efficient system of payment of benefits would also have to be 
introduced in conjunction with other cash transfer programs. If means tested, the demogrant will 
require an effective targeting mechanism to the poorest elderly, which will have its own 
informational requirements (see next chapter). 
Governance and Asset Management 
 
A
new institutional structure and governance arrangement will have to be put into place 
to provide oversight and in some cases, administer the new system.
 Collection and 
recordkeeping, discussed above, would be centralized. However, some services such as asset 
management and eventually certain insurance coverage could be contracted out. The decision-
making apparatus for this system would therefore need to be carefully crafted to safeguard this 
significant pool of long term savings. 
 
The method by which the asset manager(s) are selected, monitored (and removed) will 
be an important part of the design of this new scheme
.
Three broad approaches are possible 
–
(i) contracting out asset management on behalf of all members, (ii) time bound concessions to 
specific firms, and (iii) individual choice of provider. The last option has two variants. Individual 
workers can choose from all market participants or a subset as limited by the government based 
on some criteria. Conversely, the concession option (ii) could result in the selection of several 
external managers. 
 
The international experience includes examples of each of these approaches. For 
example, Malaysia’s Employee Provident Fund (EPF) has shifted the management of a 
small part of its members’ assets to external, private fund managers
.
The government 
dominated Board of the EPF selects the external managers and their performance affects the 
investment return of all the members. A similar approach is taken by Thailand’s GPF which 
manages the assets of the DC scheme for civil servants in that country. In both cases, individuals 
24
 See IIEF (2004). 
25
 The Central Pay and Accounts Office in India has been assigning 16-digit unique identification numbers for this 
purpose.  
Page  40
29 
play no role in choosing the external asset managers and only part of the portfolio is managed 
externally.  
 
The second approach is to solicit bids from the private sector and select one or more 
providers based on a set of objective criteria
.
In the case of Bolivia and Macedonia, this 
approach was used for all workers, including civil servants. In both cases, an important criterion 
for the government granting the concession was the proposed level of fees. The five year 
concessions granted in Bolivia and Panama resulted in low fees, although there are significant 
caveats in each case. The concession period in Macedonia was longest at 10 years. Panama has 
also used the concession approach to contract the firm that manages the defined contribution 
accounts of its civil servants that was established in 1998. Like Panama, India plans to solicit 
separate bids for recordkeeping and asset management. Unlike Panama, however, the approach in 
India is to select multiple asset managers (perhaps as many as six) and to allow individuals to 
choose from among them. The main advantage of the concession approach is to reduce costs, but 
at the expense of worker choice. 
 
The third approach is to allow workers to choose from among all licensed pension fund 
managers
.
This is the approach taken in all of the Latin American and Eastern European 
countries where private DC schemes have been introduced with the exceptions already mentioned 
of Bolivia and Macedonia.
26
 These twenty or so countries allow civil servants (along with private 
sector workers) to choose their asset managers from among all market participants. It should be 
noted however, that market concentration is high and that the number of providers averages fewer 
than a dozen.  
 
The process used to select the manager(s) is as important as the ultimate arrangement
.
Except in the case that workers are allowed to choose from the entire universe of licensed 
providers, it will be crucial that the selection of one or several managers is handled through a 
transparent process based on objective and explicit criteria. The concession approach involves 
competitive bidding with scoring on predetermined criteria, similar to procedures used in other 
large government procurement. In India, the entity responsible, the Pension Fund Regulatory and 
Development Authority is also the supervisor. In Panama and the U.S., the selection is done by a 
special Council or Board. The latter approach may be preferable if the fact that the supervisory 
agency has been involved in the selection of the asset manager makes the agency more reluctant 
to expose problems with the firms that it chose. On the other hand, ensuring the independence and 
professionalism of the special Board that would be created to oversee the selection is not a trivial 
task.
27
 
The appropriate strategy depends, among other things, on the scale of the system, the 
quality of market providers available and the tradeoff between costs of administration and 
the objective of allowing individual choice
.
In a system such the Maldivian where the number 
of workers contributing to the DC scheme is very small in absolute terms and where there is little 
or no experience upon which to base the selection of asset managers, safety and administrative 
simplicity may trump competition and choice. As the assets grow however, the benefits of 
transparent competition may grow relative to the costs and local capacity can be developed in this 
area.  
26
 Out of the ten countries that have introduced mandatory, privately managed DC schemes in Latin 
America, only Mexico has not extended this system to the civil service.  
27
 For a review of the issues and recent attempts to resolve the conflicts of interest in governance of public 
pension funds see Palacios (2002).  
Page  41
30 
Investment Policy
 
The potential for defined contribution schemes to produce reasonable pensions is 
largely determined by the rates of return that individuals achieve on their investments
.
The 
mandatory nature of the Maldivian scheme will almost certainly lead the Government to propose 
portfolio restrictions, especially with regard to the overall risk.  
 
Countries with mandatory private defined contribution schemes usually impose explicit 
limits on the schemes’ investment portfolios
 (Australia and Sweden are notable exceptions.) 
Investment rules for mandatory funded schemes take several forms, including asset class limits 
and minimums, issuer limits, and concentration limits. Rules also specify valuation requirements 
and reporting of investment practices and results. Regulations that restrict how pension funds 
exercise shareholder rights can also be included in this category.  
 
Limits are likely to be appropriate in countries where a relatively small number of 
instruments trade frequently
.
This is especially important in terms of achieving transparency 
which requires accurate valuation of assets. From the perspective of domestic investments, the 
tradeoff between diversification and liquidity will pose a challenge to policymakers. Limits can 
impose costs, however, by restricting the efficient frontier of investments.
28
 A frequent restriction 
imposed on pension fund investments (especially for publicly managed funds) involves investing 
abroad. 
 
The situation in the Maldives in this regard is extreme. There are essentially no 
domestic instruments that are regularly traded where pensions could be invested. 
The issue 
of scale suggests that this would not (and probably should not) change in the longer run. The 
Government is in the process of studying options for issuing debt and, prior to the tsunami, was 
developing the new mortgage bank with potential for housing backed securities. These are never 
likely to be liquid markets and in any case, would not serve a more general diversification 
objective.  
 
It is clear that international diversification can provide significant benefits for investors, 
especially in small countries facing significant country-specific risks
.
29
 While foreign 
investment may introduce additional costs and exchange rate risk, there are no real alternatives in 
the Maldives. International investment has the additional benefit of diversifying a part of 
Maldivian workers’ assets beyond its economy that also happens to be highly dependent on two 
sectors – fishing and tourism. 
 
Another policy question to be determined is whether to allow portfolios to vary across 
workers
.
It would be possible, for example, to allow some workers to choose a portfolio that 
included a higher proportion of equities and others to choose a higher share with fixed income 
instruments. 
 
The majority of mandated DC schemes around the world have forced all workers to 
hold the same portfolio
.
Most of these cases were national systems which included civil 
servants. The desire of policymakers to limit dispersion in outcomes and to make the system 
simpler to administer and supervise was an important factor supporting this approach. The 
drawbacks are also clear however, and the recent trend appears to be towards more (but limited) 
28
 See Srinivas and Yermo (2001) for a review of portfolio limits in Latin America and their impact on risk-adjusted 
returns. 
29
 For a discussion of portfolio regulations and their effects, see Davis (2000), Reisen (1997) and Srinivas, Yermo and 
Whitehouse (2000). 
Page  42
31 
portfolio choice. Hong Kong’s reform of 2001 included a range of options for workers (including 
newly hired civil servants) ranging from very conservative money market funds to much more 
aggressive equity funds. Thailand’s GPF plans to introduce portfolio choice within a year or two 
and is currently studying the issue. India plans to allow three portfolio options ranging with 
higher/lower concentration of bonds and equities, respectively. 
 
An int
eresting variant on the question of portfolio choice is the use of ‘life-cycle’ or ‘life-
style’ investing strategies
.
In Chile, after more than 20 years with a single portfolio system, 
regulators introduced five options with varying degrees of fixed and variable income asset limits. 
Since 2002, about 3 million workers have declared their portfolio choice. Members that do not 
make a choice are assigned according to age with younger workers being assigned to funds with 
higher variable income shares and vice versa. Workers approaching retirement age are not 
allowed to invest in Fund A, the portfolio with the highest equity share. Chile provides a good 
example of how it is possible to transition from one design to another. The Chilean design relies 
on the age-based default to allocate workers to funds that better reflect their ability to absorb 
market volatility.  
 
The examples in Chile, the U.S., India and Thailand illustrate two important points 
about investment policy choices for DC pension schemes
.
First, it is clear that investment 
options can (and probably should) change over time. As administrative issues are resolved, 
workers become more familiar with the scheme and capital market developments proceed in 
parallel, more flexibility can be offered to participants.  
Insurance Benefits
 
The insurance benefits of the current scheme can be transposed to the DC scheme.
 This 
can be done in several ways and policymakers would have to determine which way is most 
suitable in the context of Maldives. From a scheme design perspective, the insurance provisions 
can be separated into those that apply before and after retirement.  
 
Provisions for death and disability during service fall into the first category.
 Assuming 
that the intent was to maintain the prevailing target benefit levels, at least two options for 
financing exist. First, the government could directly insure the members of the scheme, charging 
an actuarial fair premium and paying out the pensions directly. In this case, the balance of the 
individual would revert t
o
the government’s budget. A second method, variants of which have 
been implemented in Latin America, is to purchase group insurance from a private insurer. In this 
case, a premium is paid and the insurer is responsible for making up the difference between the 
amount required to purchase the prescribed annuity and the balance in the individual’s account at 
the moment of death or disability. 
 
At retirement, withdrawals could be taken in lump sums, annuities or scheduled 
withdrawals
.
Most countries that rely heavily or exclusively on a DC scheme have restricted the 
amount that can be taken as a lump sum for the same reasons that justify less than full 
commutation of benefits in the DB scheme. On the other hand, once target replacement rates have 
been achieved, some Latin American countries allow for some withdrawals in this form. The last 
two types of withdrawal typically rely on private providers. Many types of annuity can be offered 
including price indexed annuities and variable annuities (where payments depend partly on 
investment returns). Another key decision is whether to require that joint survivor annuities are 
Page  43
32 
purchased in order to protect the spouses of pensioners.
30
 Scheduled withdrawals are an attempt 
to avoid the rapid dissipation of DC balances after retirement and are sometimes handled by the 
same firms that managed assets during the accumulation period. They do not, however, offer 
longevity insurance. The problems during the early stages of the system include the dearth of 
annuity products in the market (partly due to the lack of hedging instruments), an underdeveloped 
insurance sector and difficulties in pricing in the absence of reliable mortality tables.
31
 
Sequencing the Reform
 
In addition to these design and implementation issues, there is also an important 
question of sequencing the reform especially in light of the fiscal impact of the package. 
Facing serious short term fiscal constraints, the GoM may want to stagger the reform into at least 
two phases. In phase 1, the contributory scheme could be introduced by retiring workers over age 
64 immediately and deducting seven percent of total compensation for all civil servants. 
Preparations would be made for investment of DC contributions and for the payout of the 
demogrant in phase 2 of the reform. Assuming a rapid economic recovery and rising revenue 
base, it may be feasible to implement phase 2 of the reform package in early 2007. In the 
meantime, account balances would be recorded but would not be segregated.
 
In short, while the proposal is fundamentally sound, good sequencing and 
implementation will require substantial effort.
 There are no ‘losers’ in the demogrant program, 
but in the case of the contributory program, some individual workers will receive lower take-
home income for two reasons: (i) for workers forced to retire, pensions will not generally be as 
high as wages (which they now receive until death). Some of these workers would have preferred 
to continue to work. (ii) deduction of the seven percent contribution rate would reduce the take-
home wage of civil servants. Finally, short term fiscal constraints may make it necessary to delay 
full implementation of the package until such time as budget resources are available to absorb the 
incremental costs. 
 
Longer term considerations for coverage expansion 
 
The system contributory system described above has the further advantage of providing 
a
solid platform for a long run expansion to other segments of the labor force. 
By nature, it is 
self-financed and portable. Self-financing would allow workers outside government to ‘plug in’ 
to the DC infrastructure and harness the economies of scale involved in administration, 
recordkeeping and asset management. This is important in a small country where per contributory 
costs can be high in a decentralized system. Moreover, portability would allow account balances 
to be maintained when individuals moved in an out of the public sector over their working lives. 
 
The DC framework is also conducive to the inclusion of workers in the informal sector 
where regular wages are not the norm.
 If extended to the informal sector, contributions can be 
made on a flat basis and can vary according to factors such as seasonality of employment or 
incomes. In addition, a few countries have recently harnessed DC schemes to attract lower 
income informal sector workers with matching contributions from government, although this 
obviously implies additional fiscal costs. To the extent that a pension or other income test were to 
be applied to a demogrant, these matching contributions could help defray the costs of the 
program as the population ages in the coming decades. 
 
30
 Recall that in our simulations of replacement rates of the DC scheme under different assumptions, the annuity was a 
joint and survivorship annuity that replicates the current DB provisions. 
31
 See Palacios and Rofman (2000) for a review of evidence in Latin America. 
Page  44
33 
 
C
ONCLUSIONS
 
The main findings of this chapter are the following: 
·
The current practice with regard to pensions in the Maldives 
does not meet the twin 
public policy objectives of consumption smoothing and minimum income security in 
old age
.
·
Lifetime employment 
for civil servants
 is a poor substitute for a retirement scheme 
that pays reasonable benefits.
 
·
The proposed reform package includes a
 non-contributory element that would provide 
a
safety net for all elderly Maldivians as well as a contributory scheme 
for civil 
servants that would be financially sustainable and mitigate income loss after retirement 
(i.e., smooth consumption). 
This component will need to be carefully coordinated with 
Maldives safety net programs (see next chapter)
 
·
The reform package developed by the Working Group is generally sound
 and the 
fully funded approach to the contributory element of the new system holds several 
potential advantages including sustainability, the possibility of international 
diversification and a platform that can be readily extended to a greater proportion of the 
labor force if necessary.  
·
The proposed system also lends itself to the long run expansion of coverage to the 
rest of the labor force in Maldives. 
This is due to its self-financing nature and the 
portability of accounts. A seamless DC system would be able to take advantage of 
economies of scale and thereby reducing the costs that otherwise may reduce ultimate 
retirement account balances. 
·
However, reforms must take into account the overall initial conditions especially with 
regard to the fiscal situation.
 Phasing in those parts of the package that generate 
savings first may be necessary given short term fiscal constraints
.
·
Perhaps most importantly, both elements of the new pension system require 
strong 
implementation measures and new management structures.
 These will be a challenge 
in light of existing institutional capacity constraints. 
·
The table below summarizes the main changes and their implications for social protection 
and fiscal policy. 
Page  45
34 
 
Table 11: Comparison of current and proposed pension system 
 
C
urrent 
system 
P
roblems with 
current system 
P
roposal Advantages 
Fiscal 
Costs 
Social 
Security 
Civil 
servants 
receive 
unindexed 
pension of 
50% of basic 
wage every 
20 years 
Pension concept 
is really a long 
tenure wage 
supplement. 
Initial level and 
subsequent 
indexation not 
sufficient to 
provide support 
in old age 
Introduce contributory 
funded scheme for 
civil servants with a 
14% contribution rate 
split between worker 
and government for a 
mandated retirement 
age of 65 with price 
indexed annuity 
Self-financing in 
the long run. 
Allows workers to 
retire and 
government to 
hire younger staff 
as needed  
0.4-0.7 % of GDP for 
contributions as 
employer per annum. 
0.5-1% of GDP per 
annum to pay for 
retroactive service 
credit. (Overall cost to 
government depends 
on hiring practices in 
response to retirement) 
Demogrant 
About 500 
elderly 
recipients of 
social 
assistance at 
Rf. 500 per 
month  
Does not meet 
mandate for 
universal 
pensions in 
Constitution 
Introduce demogrant 
paying all elderly 
above age 65 (roughly 
14,000 beneficiaries) a 
benefit of Rf. 500 per 
month; Pension test 
applied 
Meets 
constitutional 
mandate for 
universal pension 
coverage; 
Provides safety 
net for all elderly 
in Maldives; 
Rf. 85,000,000 per 
annum (0.7% of 
GDP), rising in the 
very long run  
Overall 
system 
No system in 
place to 
achieve 
primary 
objectives of 
any pension 
system. 
Very limited 
extent of pension 
coverage; key 
objectives not 
met. 
Redistributive plus 
savings programs 
would be introduced 
Meets objectives 
of pension system 
in coherent 
fashion with 
feasible 
administrative 
strategy. 
Around 1.5-2 percent 
of GDP; reduction in 
NPV of future wage 
bill for over 65 
workers relative to 
pension bill. 
Page  46
35 
 
Safety Nets 
 
Maldives has a 
range of safety net programs, instituted with the objective of alleviating 
poverty and vulnerability
.
As noted below, these programs are fragmented, overlapping, and are 
not well targeted to the poor.  
 
The objective of safety net programs is to help the poor and vulnerable cope with 
income shocks and other risks
.
Safety nets can help address chronic poverty, targeting the very 
poorest who are often unable to work. Safety nets can also help individuals maintain or develop 
their human capital, e.g., by helping the poorest participate in education or health care, or 
improve their employability. Safety nets can also help individuals avoid transient poverty brought 
about by unpredictable life cycle risks such as unemployment, sickness, disability, or old age 
poverty. If well-designed and flexible to contract and expand when needed, safety nets can also 
help cope with large covariate risks arising from natural disasters (e.g., tsunami) or 
macroeconomic shocks (e.g., price changes) and help facilitate enaction of difficult policy 
changes (e.g., worker retrenchment). In this way, safety nets not only protect against poverty, but 
also help promote growth. 
 
Informal safety nets also help individuals cope with poverty, but often break down in 
time of large covariate shocks
.
These transfer, include, for example, remittances and intra-
household transfers that help individuals share risk and smooth consumption. To the extent these 
transfers are redistributive in nature, it is important that government programs be so designed as 
to not crowd out informal transfers. However, informal safety nets are rarely sufficient, and rarely 
redistributive (the poor may not have friends in high places), and they often break down in the 
face of large covariate shocks.
32
 
The objective of this chapter is to evaluate existing safety net programs and propose 
potential options that would improve their effectiveness
.
The first section of the chapter 
describes current programs and their key parameters, including expenditures. The next section 
discusses the key issues facing the delivery of safety net to the poor (including, inter alia, 
targeting. incentives, and administration). Reform options are discussed in the final section, 
including the choice of targeting method and reform of implementation procedures.  
 
C
URRENT 
S
AFETY 
N
ET 
S
YSTEM
:
O
VERVIEW
 
The current social safety net in Maldives is comprised of several formal programs
.
A
brief description of current programs, including the program launched immediately after the 
tsunami, follows.  
32
 In the wake of financial crisis in 1997, informal family support in Korea has collapsed leading to significantly higher 
than average poverty among the elderly, necessitating a revamped and largely expanded publicly funded safety net. 
(See K. Subbarao, 1999).  
Page  47
36 
 
Coping with Chronic Poverty  
 
The Absolute Poverty Scheme, initiated in 2003, is the main cash transfer program for 
the poor in the Maldives
.
The program is administered by the Social Security section of the 
Ministry of Higher Education, Employment and Social Security (MHEESS), previously located at 
the erstwhile Ministry of Gender, Family Development and Social Security (MGFDSS). 
Eligibility is based on individuals meeting at least one of the following criteria: not able to eat 
more than one meal a day, not in possession of more than two sets of clothing, or currently 
homeless. The level of benefit is Rf. 500 per month, disbursed through the atoll offices. In 2004, 
program had 1026 beneficiaries. Program expenditures were Rf. 6 million, less than 0.1 percent 
GDP.  
 
The Government also 
provides vouchers to defray textbook costs, one uniform, shoes, 
and socks 
for low income families. While school enrollment is high in the Maldives, some low 
income families find it hard to afford the cost of school books and uniforms. The voucher 
program allows poor families to buy textbooks for their children. The program is administered by 
the Ministry of Education but the selection of eligible children is made by the island and atoll 
offices. Island offices also handle the distribution of the vouchers. Eligibility is at the discretion 
of island authorities, e.g., the absence of a 
father’s support is one reason for granting textbook 
vouchers. 
 
Coping with Vulnerability  
 
Assistance for health care in Maldives or abroad
.
The Government provides two types of 
assistance—medical treatment against catastrophic illness (payment of fees/travel costs). It also 
provides assistance to disabled (hearing aids/glasses) and critical drugs (for mentally ill).
 
·
Assistance to seek medical treatment in the Maldives
.
Eligibility is open to 
everyone, based on a petition supported by a doctor. Most applications are from 
patients in need of medical treatment in Male’ or on atolls other than their home atoll. 
The benefits cover travel and treatment costs. In 2004, the average benefit (or cost of 
treatment) was Rf. 1530. A total of 2093 people received this benefit, at a cost of Rf. 
3.2 million.
 
·
Assistance to medical treatment abroad
.
The Government also pays for medical 
treatment abroad (treatment and airfare for patient and one helper), as many types of 
advanced treatment are unavailable in the country. The Social Security section of 
MHEESS administers the benefit. Eligibility is open to all individuals, subject to doctor 
certification (from the Indira Gandhi Memorial Hospital (IGMH)) of their need for 
medical treatment and agreement of the Medical Board. In 2004, the average benefit 
was Rf. 7670 (1173 people were provided this assistance, at a total cost of Rf. 9 
million). 
 
·
Assistance for disabled
.
The Government provides assistive devices for the disabled 
and critical drugs for the mentally ill.
 
(a) Disabled are provided with key assistive 
devices (hearing aids, crutches, etc.) Some 317 disabled benefited last year, at a 
moderate cost; (b) mentally ill patients are provided assistance to pay for critical 
needed drugs. These benefits are administered by the MHEESS). In addition to 
monetary assistance, Maldives has institutions for children whose mothers are unable to 
Page  48
37 
provide adequate care, for example, due to
 
parental drug abuse; a home for the 
mentally disabled; and a rehabilitation centre for drug abusers. The responsibility for 
the latter was transferred to the special task force on drug a
buse under the President’s 
Office.  
·
Ad hoc petitions for special assistance
.
In addition to programs described above, any 
citizen can petition the President’s Office for assistance under special family 
circumstances. The number of cases and amount of expenditures under this assistance 
are not known 
 
Targeted cash assistance to tsunami victims was used to help tsunami victims cope with 
natural disaster
.
In the months following the December 2004 tsunami, a program of cash 
transfers was developed to support the affected population. The Security Forces, relying on Inter-
Ministerial teams, were responsible for program implementation. 
Eligibility was open to all 
families whose homes were damaged by the tsunami
.
Multi-sectoral teams from Male’ 
together with island committees identified the beneficiaries, registered them, and disbursed the 
assistance as a one-time cash transfer. The level of benefit was Rf. 1500 per family member to 
those who completely lost their houses; Rf. 1000 per person to those whose houses were damaged 
and needed repairs; and Rf. 500 per family member to those whose houses were flooded and who 
therefore lost household items and belongings. The program covered 
63,000 beneficiaries and 
disbursed approximately Rf. 50 million
.
A
follow-up income support program for the poorest 
among the tsunami affected (i.e., the elderly and people with disabilities) has also been proposed. 
Non-Governmental Support
 
Charity, remittances, and inter-household transfers comprise the informal safety net
.
Zakat, the Islamic charity, also helps poor families obtain text books and uniforms for their 
children. The charity collects a modest amount each year (Rf. 7.4 million in 2002), which is 
distributed based on eligibility determined by island chiefs, Approximately 60,000 individuals 
receive benefits. Benefit levels are very modest. Aside from Zakat, the informal safety net 
consists of remittances and assistance provided by (extended) family members and friends (either 
in Male’ or from abroad). 
Total Beneficiaries and Expenditures
 
Maldives 
spends a total of 0.2% of GDP annually on safety net programs
 (not counting 
the 0.5% of GDP spent on income support for tsunami victims in early 2005). Table 12 provides a 
summary of programs. 
K
EY 
I
SSUES 
 
There are several issues that constrain the effective functioning of the current safety net. 
These issues include the ad hoc and fragmented nature of safety net programs, limited coverage, 
low expenditures, and targeting efficiency. Each issue is discussed in turn below. 
 
Page  49
38 
 
Table 12: Expenditures on Social Assistance (pre-tsunami, and  
excluding ad hoc petitions) 
 2004 
2003 
2002 
Type of Assistance 
No. of 
beneficiaries
Expenditure 
(Rf.) 
No. of 
beneficiaries
Expenditure 
(Rf.) 
Expenditure 
(Rf.) 
Income support 
 
Absolute Poverty Scheme 
1,026 
6,156,000 
 
7,968,990 
- 
Assistance to obtain school 
materials for low-income 
families 
921 730,676 452 386,312 
 
Coping with vulnerabilities 
 
Assistance to seek medical 
treatment from abroad 
1,173 
9,093,728 
834 
5,637,054 
5,221,944 
Assistance to seek medical 
treatment in Maldives 
2,093 
3,231,946 
1,311 
2,237,523 
1,830,273 
Assistance to obtain 
assistive devices for the 
disabled 
317 173,563 355 114,840 
156,034 
Assistance to obtain drugs 
for the mentally ill 
624 
96,003 
 
 
74,455 
Other assistance 
 
9,310 
 
 
 
Total (excluding tsunami relief) 
19,491,228  16,344,720 
7,282,708 
Source: 
Maldives government.
 
Program development and management 
 
The majority of safety net programs in Maldives are administered by different agencies, 
some without formal qualifying criteria or a regular schedule of benefits
.
There are a number 
of disadvantages associated with fragmented social policy frameworks of this nature, for 
example: 
·
Lack of coherent strategy and planning: Most programs are run on an ad hoc basis 
without an overall strategy. 
·
Administration is uncoordinated which is not surprising given that there is as yet 
no overall strategy.  
·
Lack of coordinated monitoring and evaluation. There is no way of knowing who 
is accessing which programs and for how long. 
 
Program coverage, benefit levels, and expenditures
 
Program coverage is quite low
,
relative to total poor population of Maldives. However, the 
exact coverage is difficult to ascertain because of ad hoc petitioning and in the absence of a good 
monitoring and evaluation framework. Overlapping target groups for each individual program 
may lead to some individuals or households accessing multiple benefits from different ministries 
while some vulnerable groups may be excluded from coverage. 
Page  50
39 
 
Benefit levels are low relative to average incomes of households
.
For example, Rf. 500 
benefit for the destitute is only 25 percent of the poverty threshold expenditure level of a family 
of four. Such benefit levels would be even lower per capita for large families.  
 
Potential problems of work disincentives are likely to be limited given the low level of 
benefits. However, there is currently no incentive for families to move out of programs 
when economic conditions change.
 The GoM is aware that many of the beneficiaries no longer 
meet the original eligibility criteria but essentially remain enrolled in the program for lack of a 
regular procedure for updating beneficiaries’ lists. Lack of clear and transparent eligibility criteria 
mean that atoll offices currently have little incentive to strike people off the lists. The existence of 
ad hoc payments from multiple sources may contain unverifiable subjective judgments and also 
decrease the transparency and accountability of safety net program expenditures. 
Figure 15: Fiscal cost of current social safety net programs 
Spending on select social programs (in Rf million)
0.0
5.0
10.0
15.0
20.0
25.0
1999
2000
2001
2002
2003
2004
Source: 
Government of Maldives
.
Though very modest, expenditures have increased over time
.
The fiscal cost has more 
than doubled, however, over the last couple of years. This reflects the introduction in 2003 of the 
Absolute Poverty Scheme, and the increased number of beneficiaries and amount of assistance 
under medical treatment assistance (international). As a result, the absolute amount spent has 
risen from just about Rf. 7 million (0.1% of GDP) in 2002 to Rf. 20 million (0.2% of GDP) by 
2004 (Figure 15), and further increased to 0.7% of GDP after the tsunami (when adding tsunami-
related social assistance expenditures).  
 
Program expenditures are lower than many other middle income countries, which often 
spend the equivalent of at least 2 to 3 percent of GDP on safety nets
,
while in OECD 
countries the expenditure levels are even higher.
33
 Even a low income country such as India 
spends 2 to 2.5 percent of GDP on safety net programs; Sri Lanka spends around 1.3 percent of 
GDP (Table 13). However, it should be noted that it is not the level of expenditure, but whether it 
effectively reaches the poor, without distorting work incentives that really matters. Maldives 
needs to address this issue of reaching the desired target groups effectively as it attempts to 
reform the safety net program framework. 
 
Table 13: Public Expenditure on pensions and social assistance in selected countries 
33
 See K. Subbarao et. al. (1997) for details.  
Page  51
40 
(% of GDP)
 
Country
 
Public pensions*
Social Assistance
Pakistan 1.7 
0.5 
India 2.2  2-3 
Sri Lanka 
1.8 
1.3 
Bangladesh
0.3 0.8 
Maldives 0.2 0.2 
pre-tsunami 
0.7 post-tsunami 
*
Does not include large implicit debt. 
 
Targeting of Transfers
 
Government transfers 
–
social assistance and pensions – appear to have a pro-rich 
distribution
.
(See Table 14, based on VPA-2, and also Annex 3).
 
In fact more than half of all 
payments go to the richest tercile. It is difficult to identify which social safety net programs that 
make “government assistance” so pro-rich because no disaggregated information is available in 
the vulnerability survey. That pensions are not pro-poor is not surprising, as they are received by 
civil servants (who are not poor) and also linked to last salary of the individual. 
 
Weak targeting of other programs may reflect inequitable access to programs
,
given ad 
hoc eligibility criteria, use of petitions, and lack of program information for those poor living 
some distance from Male’ or an atoll capital. Highly vulnerable groups may not be aware of 
benefits or services that can support their urgent needs.
34
 
Intra-household transfers are roughly equitable between terciles.
 Private transfers, 
remittances, and alimony are also not pro-poor. When all private transfers are combined together, 
they are about equitable across terciles. This is not surprising. Evidence for other countries is 
similar: the poor cannot be expected to have friends in “high places”. The poor help the poor via 
private transfers. Government assistance including pensions, on the other hand, is significantly 
inequitable inasmuch as the share of the richest consumption tercile is highest.  
34
 In the VPA-2 survey all respondents were asked whether and how much they received of the following types of 
transfers: (a) government assistance, (b) pensions, (c) zakat, (d) remittances from Male', (e) remittances from abroad, 
(f) alimony payments and (g) other transfers. Zakat payment has only one observation in the sample; so we deleted this 
from analysis
—it is possible that there is significant under-reporting of zakat receipts. Annex 4 shows another way of 
breaking down the information on receipts by source and by tercile. 
Page  52
41 
Table 14: Distribution of Assistance and Transfers
 
35
 
Consumption Tercile 
 
Type of transfer 
Low
Medium
High
Total
Government assistance (a) 
26.1
21.7 52.2 
100 
Pension (b) 
14.9
29.6 55.5 
100 
All government transfers (a+b) 
19.4
26.4 54.2 
100 
Intra-household remittances (c) 
30.9
37.8 31.3 
100 
Other assistance from family and friends in Maldives (d)
27.3
25.2 47.6 
100 
Assistance from family and friends from abroad (e) 
18.8
63.3 17.8 
100 
Alimony (f) 
28.3
41.8 29.9 
100 
All private transfers (c + d + e + f) 
30.0
35.8 34.3 
100 
Other Sources 
27.3
25.2 47.6 
100 
Source
:
VPA-2. The table shows how transfer receipts are distributed across terciles. 
 
O
PTIONS FOR 
R
EFORM
 
International experience shows that developing an effective and financially viable safety 
net system requires several building blocks
:
(i) identification and targeting of benefits to the 
poor (and protecting them against vulnerability); (ii) earmarking a sustainable fiscal envelope for 
safety nets, which is both affordable to the country and ensures sufficient resources to maintain 
effective and well-managed programs; (iii) tailoring the program to the administrative capacity of 
the country; (iv) paying adequate attention to work incentives, ensuring that beneficiaries do not 
reduce their labor supply while claiming safety net transfers; (v) complement informal forms of 
support; and (vi) be based on the preferences of society about the level and type of protection it 
would like to offer those in poverty.  
 
Knowledge about the correlates of poverty 
–
which groups of the poor are the most 
sizable in the society – is necessary for targeting
.
Chapter 1 provides this information 
–
families with many children, households headed by individuals with little education or by 
women, households with no able-bodied member, etc. Perhaps most importantly, safety nets 
reforms require considerable Government commitment in legislating and implementing reform 
measures. Finally, it should be noted that the most effective safety net is economic growth and 
employment creation, and the resulting growth in household incomes, which will continue to be 
the main vehicle for poverty reduction. 
 
The Government is aware of the problems facing its safety net programs (noted above) and 
would like to develop a single, rational system with clear eligibility criteria;
 to develop 
protection against “uninsurable risks” such as chronic poverty and health shocks, and in so doing, 
help reduce poverty and promote growth. Given that Maldives is now a middle income country, 
the Government has expressed its preferences for the basic building blocks of the safety net: 
·
Address chronic poverty as the basic building block of the system (including safety net 
for the elderly), with adequate incentives to work 
 
·
Improve human capital of the poor (allow poorest children to participate in education 
programs)
 
35
 (Annex 5 presents the same information broken down in a different manner, namely as incidence and average size of 
transfers received in each tercile). 
Page  53
42 
·
Adaptability of the safety net to cope with vulnerability (expansion of the safety net to 
handle disability, medical emergencies, macro shocks, and natural disasters). The 
government has also expressed its intent to extend health insurance to the entire 
population in a later phase (this has not been covered by this report).  
·
Ensure a safety net to protect the poor in case of difficult reforms (retrenchment, or 
increase in prices of essential goods). 
Developing a Unified Safety Net  
 
The most optimal way of achieving the above objectives would be to develop an 
‘umbrella’ or unified safety net program. 
The program could fold in several safety net 
programs currently under operation (e.g., absolute poverty benefit, health benefit, disability 
benefits, and textbook vouchers) under a unified targeting and administrative system, and under 
management of a single Ministry.  
 
Program unification has several advantages for program design and implementation.
 It 
would allow: More unified strategic development, harmonized benefit levels, improved 
incentives, lower costs of administration, and improved monitoring and evaluation. Transparent 
and uniform eligibility criteria will go a long way to facilitate entry into and exit out of the 
program. 
 
Page  54
43 
 
The reform options are illustrated in 
Figure 16
 and are elaborated below
:
Figure 16: Overview of program options 
Group to be covered
Retired civil servants:
Civil service pension 
(benefits depending on 
contributions/work history)
+
Demogrant, pension tested 
(will top up civil service 
pension to Rf.500 per month)
All other elderly 
above age 65:
Demogrant (Rf.500 per 
month)
Moderate and severe 
poor:
Poor among tsunami 
victims:
Disabled poor:
+
Free assistive 
devices
Mental patients:
Free drugs
Umbrella poverty benefit 
(family allowance + child 
allowance + textbook 
vouchers)
+
Health benefits (exemptions 
from user fees)
Program (benefits)
By folding in all benefits, the umbrella program would cover both the chronic and 
transitory poor
.
That is, the single poverty program would cover all of the chronic poor 
(including disabled, poor widows, potentially the elderly above a particular age, and poor children 
who need funds to meet school expenses), but could also expand to help the transient poor, 
particularly those who experience health shocks or natural disasters. For example, the program 
would provide a basic level of protection to all poor households, the benefit would also be 
increased to allow for participation of poor children in school (through textbook vouchers), and in 
the case of a catastrophic health shock. The poor among households affected by the tsunami could 
also be included in the umbrella program (avoiding the need to establish parallel programs). The 
interaction of benefits, i.e., umbrella poverty benefit with school and medical benefits need to be 
specified clearly (see Table 15 for more detail--issues of how to target are discussed below).  
Table 15: Overview of Program Options for social assistance 
Program and benefits 
Targeted to the poor  
(using new targeting approaches, 
see below)
 
Universal 
Old age benefits 
 
Universal untargeted old-age 
pension (See Chapter 2; for 
retired civil servants apply 
‘pension testing’)
 
Umbrella Poverty-targeted Program comprising the following elements
 
Family allowance 
(fixed tier) 
Cash transfers to the moderate and severe 
poor—revised version of existing Absolute 
Poverty Scheme (example: Rf. 400-500 . 
per month in fixed tier) 
 
Child allowance(s)  
(variable tier) 
·
Add a ‘child allowance’ of Rf. 100 
(example) per child below 18 in 
eligible families, up to a ceiling, for 
example Rf. 800 per month. 
 
Page  55
44 
·
Add assistance for textbooks for poor 
children (or simply convert the 
existing voucher program into cash 
a
nd fold into variable tier) 
Disabled poor 
Poor individuals among the disabled are 
included in umbrella program, subject to 
targeting, but with less strict targeting 
criteria, and would receive the same family 
and child allowance as everybody else 
enrolled in the umbrella program. 
·
Add Assistive devices for disabled
—
existing program, consider targeting 
 
Poor among tsunami 
victims 
Include poor among tsunami victims 
(those not benefiting from rehabilitation 
grants) in umbrella program (benefits 
not
 
additive) 
 
Health benefits 
Assistance for obtaining 
medical treatment abroad 
and domestic—existing 
program, propose move to 
targeted waiver of user 
fees: 
User fees waived for poor (using same 
targeting approach as for umbrella benefit) 
Program of medical treatment 
abroad and domestic 
continued, but with user fee 
charged to nonpoor 
Mental patients 
 
Drugs for mental patients—
continue existing program, 
untargeted (benefits can be 
additive if mental patient 
happens to receive poverty or 
pension benefits) 
Two tiered benefits, with fixed and variable components. 
The government is considering 
various options regarding the benefit levels, and considering a system with fixed and variable 
tiers. A two-tiered benefit, a base benefit 
(‘family allowance’) and another (‘child allowance’) 
varying with the number of children (or with family size) is used in many countries because it 
results in fairer distribution of assistance when larger families receive higher benefits than small 
families (for example, Figure 17 shows the scenario of a base benefit of Rf. 400 and a variable 
benefit of Rf. 100 per child, up to a limit of Rf. 800 per month). Alternatively, the basic benefit to 
a
poor
household could be increased if there was a special vulnerability, say a disabled member 
or a widow. This variation would allow for basic poverty relief, while giving flexibility of 
covering additional vulnerable members. One option to consider would be to scrap the textbook 
vouchers, converting this benefit into cash and absorbing it into the child allowance (for eligible 
households).  
 
At the household level 
some upper threshold level of benefit will need to be determined
.
There would need to be some limit as to the amount of benefit that a household could receive 
from different categories of benefit for which it was eligible. This is both for fiscal reasons but 
mainly because too high a benefit may lead to work disincentives (by making a job unattractive 
for household members who are able to work). Table 16 shows how this would work using as 
example the scenario of Rf. 400 in the fixed tier and Rf. 100 per child, with a ceiling of Rf.800 
per household per month: For a ‘typical’ poor family with 3 children, this schedule of benefits 
would result in a benefit of Rf. 700, covering 31 percent of the poverty threshold, and 
corresponding to 11 percent of average expenditures. As a rule of thumb, this level of assistance 
Page  56
45 
would seem adequate (poor households tend to have other income sources apart from social 
assistance), and would be unlikely to jeopardize work incentives. Moreover, given that the 
intended target group for all of these benefits is the chronically poor, with emphasis on those 
unable to work, problems of work disincentives can also be minimized by sticking to this target 
group. Benefit levels would be uniform across the country. 
 
Coverage would be extended to (transient poor) victims of natural disaster and the poor 
among the disabled by including these groups in the umbrella poverty program. The 
targeting criteria for these groups could be less strict
—for example, if in general the umbrella 
poverty program is intended for moderate-to-severely poor households, the eligibility rules for the 
disabled (and potentially also for natural disaster victims) could be softer, accepting also the 
slightly poor from these groups into the program.
36
 In case of any future climatic or economic 
shock, a good precedent would be established that the umbrella program can simply extend 
coverage to the affected, thereby providing a system of flexible protection. It is always preferable 
to pilot new programs, and one possibility would be to pilot the umbrella program in some of the 
islands affected by the tsunami.  
Table 16: Benefit levels and benefit adequacy 
 (example of fixed tier benefit=Rf. 400, variable tier= Rf. 100 per child, and ceiling of Rf. 800) 
Number 
of 
children 
Proposed 
assistance 
%
of
poor
households 
that have this 
size 
Assistance as 
%
of poverty 
threshold 
(family with 
2
adults) 
Assistance as 
%
of poverty 
threshold 
(family with 
1
adult) 
Assistance as 
%
of
average 
expenditures 
(2 adult family)
0
400
8
44.4
88.9
16.3
1
500
12 37.0 
55.6 13.6 
2
600
20 33.3 
44.4 12.2 
3
700
22 31.1 
38.9 11.4 
4
800
16 29.6 
35.6 10.9 
5
800
11 25.4 
29.6 9.3 
6
800
6
22.2
25.4
8.1
7
800
3
19.8
22.2
7.2
36
 This option would seem administratively easier than a separate disability pension, for which a system of medical 
certification of impairment of working ability has to be established. 
Page  57
46 
Figure 17: Profile of proposed benefits under poverty program (example) 
Proposed assistance
0
200
4
00
600
800
1
000
01
234
5
67
89
10
11
Number of children
B
e
n
e
f
i
t
s
(
R
f
p
e
r
m
o
n
t
h
)
Add-on benefits, in case of a specific shock or vulnerability.
 The benefit could be 
increased in case of a shock. For example, any poor household realizing a 
catastrophic health 
shock
 needing treatment abroad or in Maldives would be provided this treatment free of charge. 
For the non-poor, it is proposed that the existing untargeted program begins charging user fees 
(also known as co-pay).
37
 The existing program of free assistive devices for the disabled could 
continue unchanged, and would be additive as well.  
 To 
maintain 
work incentives, 
as noted above
,
the government should ensure threshold 
benefit levels for households are not in excess of minimum wages
—one
rule of thumb or 
benchmark is about 30% of average wage. Another incentive is to restrict benefits to only those 
households who do not have able bodied individuals. For those households with able bodied 
individuals, the government could instead consider a public workfare program, or undertake paid 
community service, rather than a poverty-targeted cash transfer program. Requiring that able-
bodied perform work ensures incentives for leaving the program, and is also more empowering to 
households. A program of providing work at low wages has been implemented in many countries 
in South Asia and globally, and if designed well, can be successful. However, the scope for this 
option appears to be limited in Maldives, largely because most of the poor happen to be among 
households with no able bodied members such as the elderly and women with many children and 
because demand for labor is high in the country, which has a large expatriate work force.  
 
Benefit indexed to inflation.
 Regardless of options chosen, it would be preferable to index 
benefits to inflation to avoid erosion over time.  
T
ARGETING THE 
P
OOR 
 
A
key issue for the government is whether some existing or newly proposed benefits (e.g., 
demogrant, health benefits) should have 
universal or targeted coverage
.
For example, the 
government must decide whether all elderly over 65 would receive an old age pension, as 
opposed to only poor elderly over 65. In making this decision it will be important to consider the 
costs of providing a universal vs. targeted pension. While a universal benefit is easier to 
administer than targeted benefits, it can be more costly because it has a higher coverage. 
However, the cost of targeting is not negligible, and needs to be factored into the decision as well. 
37
 The program of free treatment abroad or in Maldives for specific groups should not be confused with general health 
insurance, something which the GoM is seeking to introduce later on. 
Page  58
47 
The study has worked on the understanding that a universal old-age pension is desired by the 
government and mandated by the Constitution.  
Targeting Options 
 
No safety net can ever be perfectly targeted
.
There will always be targeting errors, large or 
small, which program design should take into account and seek to minimize. The literature refers 
to two different kinds of targeting problems: inclusion errors
—some non-poor are included in the 
program and exclusion errors—some poor are left out of the program. In addition to inclusion and 
exclusion errors, there are also other factors that reduce the share of program resources being 
transferred to the poor, which program design also needs to take into account: 
Corruption/leakage—resources not reaching any beneficiary. Administrative costs of targeted 
programs need to be kept low in order for targeting to be cost-effective.  
 
Safety net schemes worldwide use several different methods of targeting cash benefits: 
Means testing involves testing an individual’s incomes and assets,
 but is not relevant for a 
largely informal economy, where incomes are hard to verify (as in the Maldives). In this situation, 
many countries, e.g., Armenia, Russia, and many countries in Latin America, use proxy means 
testing
—which uses key characteristics correlated with poverty to identify the poor (e.g., area of 
residence, housing characteristics, etc.). This mode of targeting may be one possibility for the 
Maldives. Categorical targeting is a simpler targeting method that uses a single characteristic 
(e.g., girls, or old age or disability), and is also a possible means of targeting a poverty benefit. 
Geographical targeting is another means of targeting benefits, particularly in countries where 
poverty is very concentrated in particular regions, such as the Maldives (northern atolls). 
Community based targeting has been found to perform very well in analysis of global targeting 
experience, particularly in countries where communities have a long tradition for collaborating in 
various spheres of life (e.g., Uzbekistan and Albania). Maldives also has a tradition of strong 
local communities, which could be used to identify and target the poor.  
 
Based on the above discussion, the following options for targeting the umbrella poverty 
program appear promising candidates:  
Option 1: Community-based targeting 
 Implementing 
a 
new system of community-based targeting
 in Maldives in combination 
with geographical criteria and a set of guiding principles. Geographical criteria could be used to 
ensure more benefits flow to the poorer atolls. For example, one option would be to limit the 
umbrella poverty program to the atolls (or even to high-poverty atolls). However, the political 
feasibility of this option will have to considered. 
 
Communities (island councils) would have to nominate poor households for the number of 
slots they have been allocated. People could apply to the program and their applications be 
decided at community meetings, carefully avoiding stigmatization of applicants. Communities 
could be provided with a list of guiding (but not binding) criteria to be used when deciding on 
eligibility: many children, no breadwinner, little education, no formal sector job, does not receive 
remittances or other assistance, and so on. Such identification can be done at meetings of the 
community as is being done in India and Ethiopia (and should 
not
be done by just a few members 
of a Committee). To avoid exclusion and inclusion errors, transparency is key, and it can be 
ensured only if decisions are taken at community meetings. This is not altogether a difficult 
proposition: in small islands “every one knows everyone”, and consensus on selecting the most 
vulnerable should not be difficult, especially since Maldives has considerable experience in this 
Page  59
48 
regard. However, it should be noted that international experience shows that sometimes 
community targeting can sometimes lead to poor being excluded if they do not follow community 
norms, and this issue will have to be addressed through a grievance addressal mechanism (e.g. at 
Ministry level) if this approach is followed in the Maldives.  
Option 2: Proxy means test vetted by communities 
 
In this approach, 
applicants would be visited by a social worker (say, from the atoll 
administration) who would interview the household and solicit information on income (i.e., 
a
self reported means test) and on the variables of the proxy means test
 (demographics, 
housing characteristics, and occupation). The proxy means test should not stand alone, though, 
because simulations (described in Annex 4) show that in the Maldives proxy means test by itself 
is likely to result in a substantial number of exclusion and inclusion errors. Proxy means test 
needs vetting by communities: (i) communities could be asked to approve those applicants which 
the proxy means test has indicated to be potentially eligible; and (ii) rejected applicants who feel 
they were wrongly excluded could appeal to the community for inclusion. 
 
Option 3: Categorical targeting only 
 Under 
this 
proposal, 
all households with four or more children would be eligible for 
benefits, as would recent divorcees and widows, and elderly over 65. 
This approach fits well 
with the proposed universal old age pension (which is also categorically targeted) and would 
probably be the simplest to implement. However, simulations suggest that this option is also 
guaranteed to have fairly high inclusion and exclusion errors, so there appears to be no alternative 
to relying on vetting by communities and instituting a good mechanism for monitoring and 
handling complaints. 
Fiscal cost 
 
Table 17 illustrates the potential costs associated with the proposed umbrella program, based 
on seven alternative sets of assumptions about benefit size and coverage. Based on the 
assumptions spelled out in the table, the umbrella program (covering cash transfers to the poor, 
including disabled poor) could cost the government around 0.3-0.6 percent of GDP per year. An 
important assumption which determines total costs is the number of poor households to be 
covered. When other parts of the proposed safety net package (demogrant, and medical benefits) 
are factored in, the entire package of safety net and reform outlined here would cost the 
equivalent of around 1-1.5 percent of GDP in the first year (using a 2006 estimate of GDP). More 
than half the cost stem from the universal old-age pension, unless it is also means- or pensions 
tested. The table is illustrative and is based on assumed benefits and target groups. 
 
These calculations are very tentative
—many aspects of the design of the programs are 
yet to be decided by the government
,
and the total cost obviously depends on the decisions 
made. As noted, the government has several options to consider: to keep some programs universal 
or means-tested; it could vary the benefit levels. Further work is needed to determine the actual 
cost of safety net reform package preferred by the government based on the parameters chosen.  
Page  60
49 
Table 17: Fiscal Cost of umbrella program (cash transfer to the poor), in different scenarios 
depending on benefit size and coverage  
Number
As % of all 
poor 
households
scenario
1
400
100
800
6000
68
48.2
0.43
2
500
100
800
6000
68
52.7
0.47
3
500
50
800
6000
68
46.8
0.42
4
400
100
800
8845
100
71.1
0.63
5
400
100
800
4423
50
35.6
0.32
6
400
100
600
6000
68
42.2
0.38
7
500
0
500
6000
68
36.0
0.32
Cost 
(million 
Rf.)
Cost, % of 
GDP
Households covered
U
pper 
threshold (Rf. 
per month per 
household)
Basic benefit 
(variable tier, 
per child)
Basic benefit 
per month per 
household 
(fixed tier)
Note: 
Table is illustrative. Actual costs will depend on decisions regarding benefit levels, coverage, and 
benefit interactions. The table does not include the cost of the demogrant and medical benefits. 
Administrative costs are excluded.
 
I
MPLEMENTATION
 
Implementation of safety net programs raises a number of challenges. 
Implementing new 
administrative procedures, targeting systems, flow of funds, methods for delivering benefits, 
M&E, and communicating with communities, potential and existing beneficiaries will pose 
complex challenges for the government.
 Significant capacity building (and probably additional 
staff) to ensure successful implementation will be required. The new programs should be piloted 
before being scaled up to the entire country (those pilots could experiment with different targeting 
methods and benefit levels to determine what works best). The implementing agencies currently 
have very limited staff and capacity and will need substantial capacity building before scaling 
new programs up at the national level. A few key issues of critical relevance to the 
implementation are highlighted below. An implementation manual describing the detailed 
processes of the new programs would be useful.  
 
Given that MGFDSW does not have implementation capacity in the islands, it will be 
critical to involve atoll and island chiefs in the implementation of programs
 (Ministry of 
Atolls Development has indicated its willingness to assist). Atoll administrations, in particular, 
could be charged with the responsibility to verify, monitor and record claims. It could also be 
engaged in some random checks from time to time. Alternatively, implementation could also be 
carried out by social workers in social service units that the Ministry of Gender and Family is 
planning to establish at island level. In addition, the implementing agency could use the baseline 
and follow-up information to define an 
exit strategy
:
which households need to be taken out of 
the program because the family situation has improved. It is necessary to periodically revisit 
beneficiaries to determine if they should exit the program. Conversely, the implementing agency 
needs to be open to include a hitherto non-poor household into the program if their household 
situation has worsened.  
 
The unit of analysis and assistance should be the household:
 A household is usually 
defined as a group of people that share living and eating arrangements. Multiple family units 
Page  61
50 
sometimes form a single household (say, in polygamous marriages and in extended families with 
multiple siblings and their families). For social assistance purposes, the preferred unit is generally 
the household, whether or not the constituent family units pool income. There is good experience 
from a number of countries with providing benefits to the women/mothers of the household. 
 
Transparency is key to good performance, and in particular, it is important to examine 
complaints received from the population.
 To monitor such complaints, an institutional 
mechanism to handle complaints must be in place before programs are implemented. The persons 
involved in the selection of beneficiaries and administering of programs would arguably not be 
the right person to handle complaints. Another way to also enhance transparency is to post lists of 
beneficiaries in public places on the islands.  
 
Emerging technologies hold promise to simplify and fool-proof benefit delivery. 
Smart 
cards (with or without chips) can help identify beneficiaries, ensure benefits are claimed only by 
eligible beneficiaries, and automate financial management functions, when combined with a good 
central data base of beneficiaries and flow of funds. Smart technologies are especially promising 
since Maldives has very good access to electricity and internet and high literacy.  
 
Public information campaigns will be required to inform the public of the new 
programs
.
Except in a large emergency situation like the tsunami, it is relatively easy to transmit 
information in Maldives, thanks to the availability of television, radio, and widespread use of cell 
phones. However, when a new program is being introduced, or an old program is being re-
designed or re-targeted, it is important to communicate the changes at island meetings. In this 
respect Island Chiefs have an important role to play.  
Monitoring and Evaluation 
 
M&E should be built in from the start of the program to assess program performance, 
administration and alert the government to implementation problems. Evaluation should 
help the government determine how well programs are meeting their objectives
 of reducing 
poverty and vulnerability, and make reforms to programs as needed. It is a good practice to 
continuously make changes to programs rather than relying on one-off design solutions that may 
or may not yield the best results. Strategies for evaluation could focus on (i) assessing impact 
using a before-after approach where poverty/vulnerability before and after program 
implementation is assessed; (ii) a control group approach where gradual program rollout is used 
to create a control group of households that are similar to the early beneficiaries but who did not 
benefit initially due to the phased rollout; (iii) inclusion of detailed questions on benefits received 
from each of social assistance programs in future household surveys (that is, future VPA or 
similar surveys), and (iv) any of the above could be supplemented with a participatory evaluation 
of beneficiaries and non-beneficiaries. This would help officials receive feedback on program 
administration that should contribute to changes in programs or implementation procedures as 
needed to enhance efficiency of administration. 
 
Monitoring requires that a computerized database is maintained for all safety net programs as 
well as the demogrant described in Chapter 2. Dedicated personnel at the Ministry of Higher 
Education, Employment and Social Security could be responsible for maintaining the records of 
beneficiaries, analyzing the data and publishing basic statistics. Basic database software and 
hardware required would be purchased. Using unique identification numbers for all citizens of 
Maldives, a record would be created for a new beneficiary containing relevant data on age, sex, 
household size and composition, location and other relevant characteristics. A systematic process 
for reporting deaths and births in the beneficiary households would be developed. Updates, 
Page  62
51 
including exit from the program, would be submitted periodically by local authorities. Eventually, 
the database would facilitate periodic auditing and random spot-checks on beneficiaries.
3
8
C
ONCLUSIONS
 
This chapter reviewed the existing set of social assistance programs in the Maldives, 
identified the main issues in current safety net programs, delineated the options for reform, and 
discussed some issues in targeting and implementation. The following conclusions emerge: 
·
Maldives implements a large number of safety net program
s
ranging from a 
cash benefit targeted to the ultra poor, health benefits against catastrophic shocks, 
disability, as well as education vouchers. As noted above, the Government is also 
considering an old age benefit to help alleviate poverty for the elderly 
·
However, the programs do not meet their main objectives of protecting the 
poor. 
The programs are ad hoc and overlapping, lack an underlying strategic 
framework, some critical vulnerable groups are omitted, and some benefits are 
disbursed to individuals based on unverifiable subjective judgments, and appear 
to be under-funded. (That said, it is important to note that adequate funding, 
without accompanying reforms in targeting efficiency will not be sufficient to 
ensure the benefits help alleviate poverty.) 
·
The distribution of publicly funded programs appears to be inequitable
 
because substantive shares of current program benefits accrue to the best-off third 
of the population. Even private transfers, when analyzed as a whole, seem more or 
less equal across income groups, and not distributionally pro-poor. 
·
While rapid economic growth has been and will continue to be the main vehicle for 
poverty reduction, the GoM also would like to protect those least likely to benefit 
from growth, and those vulnerable to shocks
.
The Government is therefore interested 
in an effective safety net program to protect the chronically and transitorily poor 
households from sudden shortfalls in income and consumption induced by various 
shocks. Safety nets can also help address chronic poverty, improve human capital of the 
poor, cope with a variety of vulnerabilities (stemming, e.g., from disability, medical 
emergencies, macroeconomic shocks, and natural disasters), and provide protection of the 
poor in case of difficult economic reforms. 
·
A
number of reform options have been proposed for the consideration of the 
government
.
The government may consider a 
universal umbrella poverty 
targeted benefit
,
which folds in most of the existing benefits (and makes away 
with ad hoc petitioning) and provides a single transfer to the very poor (identified 
and targeted as below). The benefit can be flat lump sum, or, as recommended, 
can have two parts (a fixed 
‘family allowance’ and a variable ‘child allowance’ 
per child, up to some threshold) to allow flexibility in addressing household 
vulnerability. Benefits could be increased with textbook vouchers for poor 
students and also to help poor families undergoing a catastrophic health shock to 
pay for medical treatment abroad or in Maldives (for the nonpoor, it is proposed 
to at least introduce user fee under the existing program).
 
38
 
The implementation of a pension test as described in Chapter 2 would be facilitated if the database was 
compatible and easily compared with individual records from the pension office of the contributory pension 
scheme for civil servants.
 
Page  63
52 
·
For specific vulnerable groups such as the disabled and the poor among the 
tsunami victims, 
rather than establishing separate programs or benefits for these 
groups,
 it is proposed that they be included in the umbrella poverty program 
to the extent possible. Eligibility requirements could be less strict for those groups 
–
for example, the government could decide that the slightly poor among the 
disabled and the tsunami victims be eligible, while for others, only the more 
severely poor are accepted into the umbrella program. 
·
A
threshold benefit level under the umbrella poverty program will need to be 
identified
 so that work incentives are not compromised, and fiscal costs are 
sustainable (Rf. 500-800 per month per household would seem an appropriate 
range). The benefit should be indexed to inflation to ensure that protection is not 
eroded over time. 
·
It will be important to target benefits effectively to the poor
.
In particular, the 
chapter explored the potential of using a proxy means test in combination with 
community-based targeting. Based on the findings of selected simulations, the 
study recommends a combination of geographic targeting and community-based 
identification of beneficiaries aided by guidelines that emphasize household 
characteristics potentially associated with vulnerability, including the absence of 
able-bodied household members. 
 
·
The costs of various benefit options and their interactions will need to be 
assessed prior to the final design and implementation
.
Depending on the 
option chosen by the government (especially the size of benefits and the size of 
the population to be covered), the fiscal cost of the umbrella poverty benefit could 
be in the range of 0.6-0.8 percent of annual GDP (assuming some 6,000 
beneficiaries, or 10 percent of all households, more than half of all poor). This is 
in addition to the cost of the demogrant, estimated at 0.7 percent of GDP. These 
estimates will need refining as the details of the program are finalized. Reforms 
should be phased in once fiscal resources become available, presumably once the 
impact of the tsunami has faded. 
·
Finally, 
implementation issues need careful attention
:
It is important to ensure 
proper identification and targeting of the poor, and responsibilities for verifying, 
monitoring and updating records need to be decided. An agency to coordinate 
implementation will also need to be decided, and this agency will need significant 
capacity building. New programs and systems should always be piloted before 
being scaled up to cover the entire country. 
·
Programs should be designed with adequate monitoring and evaluation to ensure 
that policy makers are aware of their impact and programs can be adjusted to 
improve their performance. 
Page  64
53 
References 
Baulch and Hoddinott (2000) “Economic Mobility and Poverty Dynamics in Developing 
Countries”, Journal of Development Studies v36, n6 (August 2000): 1-24. 
Davis, E.P. (2000) “Portfolio Restrictions of Life Insurance and Pension Funds”, mimeo. 
Holzmann, Rober, Richard Hinz, et all. 2005. “Old Age Income Support in the 21
st
 
Century”. World Bank. Washington, DC. 
Kingdon, Geeta and John Knight (2000) “Are searching and non-searching 
unemployment distinct states when unemployment is high? The case of South 
Africa”. Center for the Study of African Economies, University of Oxford. 
MDNP (Ministry of Planning and National Development). 2004. “Household Income and 
Expenditure Survey 20002/03”.
 
———. 
Statistical Yearbook
,
various issues. 
———. 
National accounts, 
various years 
———. 
Government budgets, 
various years 
———. 
Population Census, 1990, 1995 and 2000. 
———. 
Supply and Use Table 1997. 
Ministry of Planning and National Development (2003), Statistical Yearbook of Maldives 
2003.
 
MoH (Ministry of Health), 
Health Report 1999 and 2002. 
———. 
Maternal Death Review Reports 1997-2001. 
———. 
Vital Registration Data 2004. 
MoE (Ministry of Education), 
Educational Statistics
,
various issues.
 
———. 
Administrative records, 
various years 
MPND (Ministry of Planning and National Development) and UNDP (United Nations 
Development Programme. 1999. “Republic of Maldives: 1998 Vulnerability and 
Poverty Assessment”. Stara Design Team, Singapore. 
Murgai, Rinku. 2004. “Poverty Measurement and Poverty Profile: Estimates Based on the 
2002/03 Household Expenditure Survey”. World Bank, Washington, DC. 
(unpublished manuscript). 
Reisen, H. (1997) “Liberalizing Foreign Investments by Pension Funds: Positive and 
Normative Aspects”, World Development
,
Vol. 25, no. 7. 
Srinivas, P., J. Yermo and E. Whitehouse (2000) “Regulating Private Pension Funds’ 
Structure, Performance and Investments: A cross country comparison”, World 
Bank Pension Reform Primer series, 
www.worldbank.org/pensions
.
Subbarao, K., and others. 1997. Social Assistance Programs and Poverty Reduction: 
Lessons from Cross-Country Experience. World Bank, Directions in 
Development. Washington, D.C 
UN Theme Group on HIV/AIDS. “A Situational Assessment of HIV/AIDS in the 
Maldives for the year 2000.” 
US Census Bureau, IDB population pyramids: 
http://www.census.gov/ipc/www/idbpyr.html. 
World Bank-Asian Development Bank-UN System. 2005. “Joint Needs Assessment on 
the Tsunami Impact and Recovery for the Republic of the Maldives”. 
Washington, DC: World Bank. 
Page  65
54 
World Bank.
 2005. “Post-Tsunami Recovery and Reconstruction Project”. Washington, 
DC: World Bank. 
———. 2004. “Maldives Country Economic Update: Sustaining Robust Development”. 
Washington, DC: World Bank. 
———. 2004. “Integrated Human Development Project”. Washington, DC: World Bank. 
WB/ADB/UN (2005) “Tsunami: impact and recovery. Joint Needs Assessment”. 
Page  66
55 
 
ANNEX 1: DATA SOURCES 
The analysis in this report draws upon three household surveys conducted at different points in 
time (see Table 18 for details of sample sizes): 
·
The 1997/1998 Vulnerability and Poverty Survey (VPA-1) 
·
The 2002/2003 Household Income and Expenditure Survey (HIES) 
·
The 2004 Vulnerability and Poverty Survey (VPA-2) 
Table 18: Data Sources and sample sizes 
 Maldives
Male
Atolls 
VPA-1 2597 
311 
2286 
HIES 834 
246 
588 
VPA-2 2728 
307 
2421 
The VPA-2 data was collected in June and July of 2004, some six months prior to the tsunami, 
partly financed with World Bank assistance. VPA-2 gives a baseline for analyzing pre-tsunami 
household welfare. The government has plans to conduct a post-tsunami nationally representative 
household survey, to assess household level impacts of the natural disaster. If carried out, 
Maldives will have comparable survey data from before and after the tsunami, and be in an 
excellent position for studying the impacts of a major natural disaster. For the time being, 
however, VPA-2 is the latest survey data available. 
 
A
question arises: Are pre-tsunami data such as VPA-2 useful for understanding poverty and 
vulnerability and designing safety net reform? At the time of writing the answer seems to be yes, 
by and large: the economic recovery is progressing well; people are out fishing or working in the 
tourist resorts; many islands were not affected severely; those who were poor or vulnerable before 
the tsunami are likely to remain so. However, the tsunami created new vulnerable groups, i.e. 
those who lost their assets (in kind or cash) or livelihoods. Especially hard affected are families 
that lost their house and are still suffering from lack of permanent housing; the elderly (having 
sometimes lost lifetime savings held in cash), people with disabilities, pregnant women and small 
children and their mothers. While jobs appear to recover relatively quickly, those livelihoods that 
depend on physical assets such as fishing vessels, perennial crops, fish processing equipment, or 
other tools will require public assistance to restore to the pre-tsunami situation.  
The VPA surveys contain modules on household demographics, occupation, sources of income, 
health, education, housing amenities, assets, transfers, loans, expenditures, and distance to 
facilities. The surveys enable one to understand the transitions in and out of poverty, and the 
characteristics of poor and non-poor households. The recent VPA-2 survey will also serve as a 
valuable baseline for the situation prevailing immediately prior to the tsunami.  
The comparison of poverty trends over time utilizes mostly VPA-1 and VPA-2. These surveys 
have a much larger sample size than HIES, and are therefore considered more reliable. Moreover, 
some households, though few in number, have been surveyed in both VPA-1 and VPA-2 which 
enabled us to construct a panel and trace transitions in and out of poverty during this period, as 
well as the characteristics of the respective households.  
Page  67
56 
All three Maldivian surveys measured consumption in a broadly comparable manner. However, 
in the VPA-1 survey, the expenditure module had fewer food items pre-printed, and households 
were instead asked to fill in items based on recall, whereas VPA-2 contained a longer list of pre-
printed food items. This could potentially cause some (upward) bias when comparing changes in 
average household expenditures over time (government officials familiar with survey 
implementation of both VPA-1 and VPA-2 however do not see this as a major concern). 
Another issue when comparing poverty over time is inflation. As over the period 1997-2004 the 
inflation rate (measured by the Consumer Price Index) in the Maldives was virtually zero, a 
choice was made to use nominal expenditures as the provisional welfare measure, in line with 
other researchers and Government of Maldives working on the same data.  
Due to the small change over time and across regions, no indexation of price differentiation was 
adopted. While conceptually it is the right thing to do, the empirical reality is such that the results 
do not change much by such adjustments.
39
 Rent is excluded from the consumption aggregate 
used in the analysis as it is only relevant for a very small part of the sample, largely in Male’.
4
0
Maldives has no official poverty line. Prior government reports used poverty lines of Rf. 7.5, 10, 
and 15 per capita per day. World Bank estimates based on HIES using the Cost of Basic Needs 
Approach suggested Rf. 14.8 as the consumption level at which an average poor household meets 
its food and non-food needs (Murgai, 2004). This is very close to the upper poverty line used 
previously by the government (Rf. 15). We therefore chose to use the Rf.15 line to maintain 
consistency with earlier work. As in the case of the expenditure measure, no inflation adjustment 
was made to the poverty line in light of zero inflation during the period. 
An important limitation of VPA-2 data is that information on transfers is collected only “for the 
previous month”. To the extent many transfers in Maldives are one-off grants determined on a 
case-by-case basis, and not regular monthly payments, the transfer module misses out the ad hoc 
transfers received by households during the previous 11 months. Also “government assistance” is 
not disaggregated into different types of payments that households receive. 
39
 Previous attempts to estimate regional price indices based on an average standard consumption basket proved 
impossible because there appeared to be only a few items that could be included in such a basket that met the criteria of 
homogeneity of the goods and availability and use throughout the country. Luxury goods and consumer durables were 
excluded since there is only one shopping centre in the country for these goods. Further, of the remaining items that are 
homogenous and available and consumed throughout the country, three important food items: wheat flour, rice and 
sugar are imported and sold throughout the country at a common fixed price. 
40
 The small sample of renting households makes it impossible to estimate with any reasonable degree of confidence 
the imputed value of owner-occupied accommodation. 
Page  68
57 
Annex 2: Determinants of poverty 
This annex presents the regression results on the determinants of poverty referred to in 
Chapter 1. Table 19 presents the results of a probit regression equation where P
i
is the 
probability of being non-poor, and X
i
are household characteristics:  
P
i
=
i
+
X
i
+
u
i
Table 19: Poverty Correlates: Probit Regression Results Using VPA1 (1997-98) and VPA2 
(2004) Data
 
VPA1  
 VPA2 
headcount  
 
headcount 
Male
’
0.28  
0.159 
(8.60)**  
(5.89)** 
Any Household member migrated in the last 3 years 
-0.04  
0.004 
 
(-1.25)  (0.23) 
Household head more than 65 
-0.088   
 
(-2.48)*    
Ratio of elderly (more than 65 yrs) to the size of the household 
 
-0.185 
 
(-3.80)** 
Ratio of children (less than 16) to the size of the household 
-0.319  -0.096 
 
(-5.52)**  
(-2.17)* 
Ratio of wage earners to the size of the household 
0.172  0.195 
 
(2.74)**  
(4.60)** 
Ratio of the number of members who had education /size of the 
household 
0.103  0.065 
 
(2.26)*  (2.04)* 
Female household head 
-0.059   
 
(-2.66)**  
 
 Household head unemployed 
-0.036  -0.025 
 
(-1.08)  (-1.18) 
 Household contains at least one widow 
-0.045  -0.024 
 
(-1.52)  (-1.23) 
 Household has a loan to be repaid 
 
0.114 
 
(5.76)** 
Household contains at least one divorcee 
 
-0.018 
 
(-0.78) 
Household head literate in English 
 
0.117 
 
(5.08)** 
Size of the household 
 
-0.012 
 
(-4.78)** 
Observations 2619 
 
 
2548 
Absolute value of z statistics in parentheses 
*
significant at 5%; ** significant at 1% 
Dependent Variable: 1=Nonpoor, 0=poor 
Note
:
Marginal effects (evaluated at the mean) are shown. 
Page  69
58 
 
Annex 3: Incidence and average size of assistance and 
transfers 
 
Table 20: Incidence and average size of assistance and transfers 
Consumption 
tercile
Government 
assistance
Pension
Intra-
household 
remittances
Other 
assistance 
from family 
and friends 
in Maldives
Assistance 
from family 
and friends 
from abroad
Alimony
Other 
sources 
Incidence of assistance and transfers (in %)
Low
1.8
2.7
25.5
8.2
0.5
1.4
8.2
Medium
1.5
3.3
23.3
8.9
0.7
1.8
8.9
High
2.4
7.1
17.3
9.7
0.7
0.6
9.7
Total
1.9
4.4
22.0
8.9
0.6
1.3
8.9
Average size of transfers and remittances (for receiving and non-receiving). In Rf.
Low
19
16
514
117
6
13
117
Medium
16
32
627
108
21
19
108
High
38
61
520
204
6
14
204
Total
24
37
554
143
11
15
143
Page  70
59 
ANNEX 4: TARGETING OPTIONS 
As discussed in Chapter 3, assessing incomes is a difficult task in Maldives, and targeting of 
social programs has to rely on either observable categories or community identification, or a 
combination. This annex reports on statistical experiments with two targeting approaches: 
categorical targeting and a proxy means test. These are discussed below.  
C
ATEGORICAL TARGETING
 
Four potential groups for targeting assistance are considered: widows and divorcees, the elderly, 
families with 4+ children, and families with no earning member. These categories of households 
are distinct and easily identifiable. The group of all poor households is included as benchmark 
(i.e., assuming perfect targeting, how large a reduction in headcount poverty could be achieved by 
these cash transfers). 
The simulations estimate the reduction in poverty that would result from targeting social 
assistance to these identified groups. The simulations work with fixed budget constraints
—0.5 
percent and 1 percent of 2003 GDP, which is divided equally among all the households in the 
identified groups.
41
 We further assume that the policy objective is reduction of poverty not only 
among specific identified group but also nationally. The results are presented in Table 21 
(transfer budget of 0.5% of GDP) and Table 22 (budget of 1% of GDP). 
Table 21: Simulation Results with a Fixed Budget Constraint of 0.5% of 2003 GDP 
 
Initial poverty rate 
 
Poverty rate after simulated 
transfer 
Target group 
In 
group 
National
In group 
National 
Households with widows and divorcees 
17.1 
18.4 
 
15.4 
17.8 
Households with widows and divorcees and 
elderly 
18 18.4 
 16.5 
17.8 
All households with 4+ children 
26 
18.4 
 
21.9 
16.8 
All households with no earning member 
30.2 
18.4 
 
20 
17.1 
All households below the poverty line 
100 
18.4 
 
71 
13.1 
Source: 
Staff simulations based on VPA-2
.
Note
:
2003 GDP is Rf. 9157 million. 
Table 22: Simulation Results with a Fixed Budget Constraint of 1% of 2003 GDP 
 
Initial poverty rate 
 
Poverty rate after simulated 
transfer 
Target group 
In 
group 
National
In group 
National 
Households with widows and divorcees 
17.1 
18.4 
 
12.1 
16.7 
Households with widows and divorcees and 
elderly 
18 18.4 
 14.2 
16.8 
All households with 4+ children 
26 
18.4 
 
18 
15.2 
All households with no earning member 
30.2 
18.4 
 
12.3 
16.3 
All households below the poverty line 
100 
18.4 
 
48 
8 
Source: 
Staff simulations based on VPA-2. 
 
The tables show the current rate of headcount poverty, in the groups considered and nationally. 
The next columns show the rate of headcount poverty, in each group and nationally, that would 
41
 An equal sized transfer is “given” to all households in the category, regardless of size. The amount of the transfer is 
calculated so as to exhaust the budget. 
Page  71
60 
result after the hypothetical transfer, assuming no change in households’ income in response to 
the transfer, and assuming (for convenience) that the entire transfer is consumed. 
Both simulations suggest that targeting households with 4+ children and households with no 
earning member would yield the highest reductions in poverty, both group-specific and national. 
For example, if 1 percent of GDP were to be transferred in equal amounts to all households with 
four or more children, the poverty rate of this group would decline from 26 to 18 percent, 
whereas the national poverty rate would fall from 18.4 to 15 percent. If 1 percent of GDP was 
transferred to all households with no earning member, poverty headcount of this group would fall 
from 30 to 12 percent, whereas national headcount would fall from 18.4 to 16.3 percent. When 
interpreting these numbers, one must keep in mind that national poverty always falls less than 
group poverty because each of the groups exclude many of the poor in Maldives. Also, no 
categorical targeting achieves anywhere near as good a targeting performance as the benchmark 
case where the poor are perfectly targeted. It does surprise, though, that even perfect targeting of 
transfers amounting to 1 percent of GDP only achieves reductions in national poverty from 18.4 
to 8 percent; the reason for this is that some of the poor consume well below the poverty line, and 
would need quite sizeable income transfers to reach a consumption level implied by the poverty 
line.
42
 
These simulations suggest that there is really no easy way to target assistance in Maldives: both 
poor and non-poor households co-exist in each category. While targeting households with 4+ 
children and no earning members may achieve better poverty reduction than targeting other 
categories, a policy of targeting assistance exclusively to specific categories would be 
inappropriate and undoubtedly lead to serious exclusion errors: too many ultra poor households 
would be missed by categorical targeting and not receive any assistance.  
Considering that categorical targeting could be problematic, we now discuss the feasibility of a 
proxy means test.  
P
ROXY 
M
EANS 
T
EST
 
We considered how well a proxy means test in Maldives could identify the poor. A proxy 
means test is a composite score or formula that for given values of observable household 
characteristics (demographics, education, housing and so on) estimates the welfare level 
of the household and based on this determines whether the household qualifies to 
participate in social programs. Households have an obvious incentive to misreport their 
characteristics once the purpose of the information being collected is known. Therefore, 
proxy means tests need to be based on characteristics that can be verified fairly easily in a 
home visit. In most countries, household size, household demographic composition, land 
holdings and other assets, housing characteristics, ethnic group, and geographical area are 
some of the variables that can be useful for targeting. 
Evidence from various countries suggests that proxy means tests can be fairly precise and are able 
to improve targeting substantially relative to untargeted transfers. In Chile, for example, such a 
formula is routinely used to assess whether households applying for various welfare programs 
qualify and at what level. Information is collected by social workers who visit the applicants and 
fill out a questionnaire with information on the variables used in the proxy means testing formula 
42
 Median consumption among the poor is Rf. 10.8 per capita per day. One-quarter of the poor consume below Rf. 8.1. 
The poverty line is Rf. 15 per capita per day. 
Page  72
61 
(Grosh and Baker, 1995). Proxy means tests are also used in many other countries, in Latin 
America, the Caribbean, Eastern Europe, Middle East, and Asia. 
We worked with both VPA-2 and HIES to assess a potential proxy means test for Maldives, 
taking the following steps: 
·
A
regression is run in which the welfare measure, per capita expenditures, is 
regressed on a number of household characteristics 
·
The estimated parameters of the preferred expenditure regression are used to 
predict the welfare level of all households
—the estimated regression parameters 
essentially comprise the proxy means testing formula. 
·
The predicted welfare level is compared to a cutoff point (in this case, the bottom 
20% of households) to determine if the household is eligible for social assistance.  
·
The targeting efficiency of the proxy means testing formula is evaluated—using 
the same survey data—by comparing households deemed eligible for assistance 
according to the formula and households with actual welfare below the cutoff. A 
targeting success is when a household is placed on the same side of the welfare 
cutoff by both the formula and the survey data. 
 
Targeting errors occur when formula and observed data disagree on which side of the 
cutoff to place a household. There are two kinds of targeting error. Households that are 
predicted non-poor by the formula but whose actual survey welfare is below the cutoff 
are considered to be undercovered. Households predicted eligible for assistance by the 
formula but actually living above the cutoff are said to represent leakage. The calculation 
of undercover age and leakage rates is illustrated in Table 23. The rate of under coverage 
is calculated by dividing the number of cases of Type I error by the total number who 
should actually be receiving benefits [e
1
/n
1
]. The rate of leakage is calculated by dividing 
the number of Type II errors by the number of persons served by the program [e
2
/m
1
]. 
Page  73
62 
 
Table 23: Illustration of Type I and II targeting errors 
 
Target group 
Non-target group 
Total
Eligible (as predicted by 
formula) 
Targeting success 
(s
1
)
Type II error (e
2
)
m
1
Ineligible (as predicted by 
formula) 
Type I error (e
1
)
Targeting success 
(s
2
)
m
2
Total 
n
1
n
2
n
Results are disappointing: a formula based on observable characteristics cannot predict 
very accurately the 20 percent poorest households. Variables used include household 
size; number of children and old persons in the household; size, type and rental costs of 
house; type of job; education; assets and consumer durables; whether household receives 
a
pension; and whether household lives in Male’.
43
 Targeting results are shown in Figure 
18 (for HIES data). 
 
Figure 18: Targeting of a simulated proxy means test 
Simulated proxy means test
65.2
34.0
22.4
12.1
4.3
34.8
0
20
40
60
80
Poorest
2
3
4
Richest
%
o
f
H
H
s
i
n
q
u
i
n
t
i
l
e
Target success--poor
and eligible
Inclusion error--nonpoor
but eligble
Exclusion error--poor and
noneligible
Source: 
Staff simulations based on HIES. 
The simulated formula could correctly predict around two-thirds of households in the 
bottom quintile. The formula erroneously include more than one-third of households in 
the second quintile, and more than one-fifth of households in the thirds quintile among 
the eligible. This is a fairly high rate of error. Moreover, analysis conducted using VPA-2 
resulted in worse targeting performance. 
43
 
We experimented with both HIES and VPA2, and also ran the regressions separately for Male’ and atolls.
 
Page  74
63