A review of international experience

 
• Communications. Without appropriate communications, individuals are not 
going to make effective decisions and will not normally develop their capabilities 
to make future decisions.
 
• Complexity. Individuals are typically able to make simpler decisions but often 
have difficulty with more complex decisions.
 
• Community. Friends and family are often a critical part of decision making and 
an important socioeconomic enabling factor behind the success of matching 
programs.

318 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
 
• Credibility. The system needs to be credible in order to be sustainable and to 
function.
 
• Complements. The system needs to work in tandem with other vehicles that will 
produce retirement income.
 
• Culture. The system needs to be consistent with social norms and the historical 
development of other social protection arrangements.
 
• Connection. Intermediaries play an important role in connecting an individual 
with products and in aiding his or her decisions.
All of the above-mentioned elements are important to defined contribution plans 
in general, not just to matching programs. It is therefore helpful to focus attention on 
what is special about matching programs in terms of the difficulties and issues likely to be 
encountered. 
 
• Matching programs that are publicly sponsored involve the credibility of the gov-
ernment in visible ways; hence, the expectations of the populace are higher. In 
particular, as matching systems aim to increase participation in the system, any 
failures will be particularly poorly received. The pressure to get people to partici-
pate is likely to be very intense.
 
• Matching can introduce complexity. If individuals have difficulty making deci-
sions without matching, they may have even more difficulty factoring in the extra 
challenges of calculating incentives from matching programs. This complexity 
increases the administrative load and makes governance arrangements that much 
harder to get right.
 
• Matching is more vulnerable to agency issues than systems without matching. 
Any time the government or another party is paying for something, there is likely 
to be some part of the benefit that accrues to parties other than the consumer. For 
example, intermediaries might charge more, or employers may constrain wage 
increases or cut benefit programs in response to the provision of contribution 
matches.
Matching therefore imposes additional implementation challenges and makes the 
enabling environment that much more significant. The issues outlined above are of greater 
importance within this environment; the capacity for choice, administrative systems, and 
effective governance are discussed in more depth below along with general issues of imple-
menting defined contribution plans.
INFORMED CHOICE
Consider the problem of getting an individual to make an informed choice about match-
ing contributions when individuals seem to have difficulty making any sort of complex 
decision about retirement saving. In 2005, Watson Wyatt conducted a survey of U.K. 
individuals in which half were presented with a graph of annuity payouts as a function of 
age and given a choice between a level annuity and an indexed annuity. The other group 
was presented with the same choice in tabular form. Figure 16.2 and table 16.1 present 
the choices given to individuals.

16.  IMPLEMENTATION ISSUES IN LOW- AND MIDDLE-INCOME COUNTRIES 
319
The results are shown in figure 16.3. Sixty-five percent of those shown the table 
picked a level annuity, whereas only 48 percent of those shown the graph picked a level 
annuity. Clearly, the form of presentation of options matters.
One could infer a similar dichotomy to choice on contribution levels with match-
ing contributions. The impact of a decision of an individual’s choice of contribution level 
on the flow of funds into an account can be illustrated either graphically or as a table. 
Although the numbers in a graph and table could be the same, individuals could draw 
different conclusions.
TABLE 16.1  Tabular presentation of annuity 
choices
Age
Annuity (£)
Inflation-linked 
annuity (£)
65
7,000
4,900
70
7,000
5,500
75
7,000
6,300
80
7,000
7,100
85
7,000
8,000
90
7,000
9,000
95
7,000
10,300
100
7,000
11,600
SOURCE: Watson Wyatt Limited 2006.
FIGURE 16.2  Graphical presentation of 
annuity choices
4,000 
5,000 
6,000 
7,000 
8,000 
9,000 
10,000 
11,000 
12,000 
65
70
75
80 85  90
95 100
income (£)
age
fixed-income
(option A)
inflation-linked
(option B)
Retirement income
SOURCE: Watson Wyatt Limited 2006.
FIGURE 16.3  Choice between annuities
48% 
41% 
11% 
65% 
26% 
9% 
0 
10 
20 
30 
40 
50 
60 
70 
level real 
don't 
know 
%
graph
table
level real 
don't 
know 
SOURCE: Watson Wyatt Limited 2006.

320 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
The importance of communication in decision making is illustrated by work Tow-
ers Watson has done with employee/employer data over time. This research found that 
communication plans often have more impact on decisions than matching programs. An 
analysis of individual-level data for more than 306,000 workers at 48 firms indicates that 
communication has a significant impact on contribution and participation rates. In gen-
eral, high levels of communication increase participation quite significantly over a baseline 
case of low communication, although low communication will clearly be better than no 
communication. There does not appear to be a significant age gradient in how communi-
cation affects participation rates in defined contribution plans. In the context of matching 
plans, there is little evidence to suggest that the effectiveness of communication will be 
age specific, which in turn implies that communication issues are pervasive across all age 
groups instead of specific to one age segment (Towers Watson 2011). The likely cause of 
this outcome is that without meaningful assistance individuals typically get the answers to 
complex questions wrong. For example, when in 2003 Towers Watson asked individuals 
which investment offered the best protection against inflation, only 19 percent were able 
to correctly identify an index bond as that vehicle; this result was in the United Kingdom, 
where indexed bonds had been in place for two decades.
There is clear evidence across many markets that individuals have difficulty making 
highly technical decisions, and matching is inherently technical particularly if the matching 
structure is nonlinear. While individuals cannot make technical decisions very well, there is 
a fair amount of evidence that individuals do a better job with intuitive decisions. Evidence 
of this can be found in a three-country study in which individuals in Germany, Holland, 
and the United Kingdom were able to adjust their retirement choices using a dashboard. 
The survey format was unique in that it started with a set of standard questions on desired 
income, retirement age, and risk tolerance, and used these to populate a dashboard with 
which individuals could explore trade-offs involving risk and levels of benefits. All decisions 
were tracked, and individuals were subsequently asked why they had made the changes 
they did. In the three countries, the percentages of respondents who rated the survey as 
good or average on these parameters of length, understanding, relevance, and interface were 
99 percent, 83 percent, 96 percent, and 96 percent, respectively. The decisions emerging 
from the exercise were quite reasonable, and indicated that individuals did have reasonably 
heterogeneous preferences, with a heavy bias for safety (Towers Watson 2009).
The social context in which decisions are made is also important. This finding is 
indicated by a quasi-experimental survey undertaken in Australia, Germany, the Neth-
erlands, the United Kingdom, and the United States. Individuals were given different 
default options for their savings levels. Some were given no default; others were given a 
low, medium, or high default with no explanation of who had determined that default. 
Others were told that the default level was recommended by an employer, friends/family, 
or a financial advisor. The results were striking. In contrast to the findings of other stud-
ies, defaults did not have much impact on saving rates (mean or median) in the absence 
of context as to who was providing the choice. For example, in the United States, the 
mean saving rate was 6.3 percent in the absence of a default and 6.1, 5.6, and 6.7 percent, 
respectively in the presence of low, medium, and high defaults with no default context 
provided. In no country was the difference between default results and decisions without 
defaults significant. However, the difference became very significant in the presence of 

16.  IMPLEMENTATION ISSUES IN LOW- AND MIDDLE-INCOME COUNTRIES 
321
context—for instance, when the individual was told that an employer had recommended 
the default. Interestingly, outcomes varied considerably by country setting, possibly 
reflecting differing levels of credibility and trust. In some countries, financial advisers are 
highly trusted; in others, recommendations of high saving rates from financial advisers are 
viewed with considerable doubt.
In general, where individuals are searching for credible information in a sea of con-
fusing choices, a trusted party providing implicit guidance though a default or the pro-
vision of a matching contribution becomes an effective reference point that can have a 
powerful influence on outcomes. In the context of matching programs, this suggests that 
the entity providing the match is particularly important. For example, if employers pro-
vide matching contributions and are trusted, the impact on saving may be considerably 
greater than provision of contributions by a government that is mistrusted. A reason-
able precondition for implementing matching contributions would be a credible party to 
design the matching provisions and other choices to be made by participants.
If individuals were proficient at making informed choices, there would of course be 
little need for matching contributions. Therefore, the condition that individuals have the 
ability to make informed choices will never be satisfied fully, but will always be a matter 
of degree. It is important to keep in mind that a key advantage of matching contributions 
over strategies such as default options is that the former aim to engage the individual in 
active decision making and to help the individual learn. In this sense, an important imple-
mentation requirement is the capacity to monitor and evaluate choice behavior to ensure 
progress is being made. It is also important to allow the system to evolve over time.
ADMINISTRATION
Matching contributions impose significant challenges for recordkeeping and adminis-
tration. The additional payments into the system must be carefully tracked, effectively 
accumulated into individual accounts, and invested properly; and information about con-
tributions and account balances communicated accurately to participants. Any errors will 
reduce confidence in the system. If systems cannot handle the complexity of the choices 
required to be made or can only implement these expensively, the benefits of defined con-
tribution and matching programs will be significantly diluted.
Administration costs are a significant component of the total costs of running a 
defined contribution system. While much literature and thinking have focused on how to 
minimize investment costs through index funds and other approaches, in practice a large 
share of costs arise from administration, recordkeeping, and marketing. Consequently, 
minimizing administration costs is a key element of the enabling environment for match-
ing defined contribution scheme implementation.
One common approach to reducing administrative costs is reliance on the informa-
tion technology platform. However, individuals consistently exhibit a strong preference 
for receiving paper rather than electronic statements, and in all but a few environments 
electronic communication is not feasible for the relevant populations. In higher-income 
settings, customers consistently refuse to buy packaged financial products that are long 
term in nature—such as pensions and insurance—through the Internet. Thus any attempt 
to mitigate the administrative cost issue by establishing information technology as the sole 
mechanism for administration is doomed to experience significant problems.

322 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
 Individuals clearly want direct service: they want to be able to call someone up who 
understands their questions and can help them sort through complex issues and decisions. 
The presence of credible institutions with experience in handling these sorts of issues for 
long-term financial products is thus very helpful for successfully implementing a matched 
defined contribution system. Where customer service is outsourced or “off-shored,” indi-
viduals will be particularly critical when things go wrong. The cost-benefit of outsourcing 
and off-shoring thus needs to be weighed against the risk that a perception of poor or 
remotely provided service could undermine the system.
While it would be a mistake to look only at the information technology system 
in developing a matched defined contribution system, the quality of that system is 
extremely important. Misallocated or missing transactions and payments are a major 
source of participant anxiety and stress. The integrity of contribution records is particu-
larly an issue with matching systems, as changes in contributions need to be linked to 
any changes in the matching contributions. Matching contribution payments need to be 
calculated accurately and tracked, and any error has the possibility of resulting in adverse 
publicity—especially if the customer service arrangements are not up to an appropriate 
standard. 
GOVERNANCE
A high-quality governance system is required to deliver good choices and decent manage-
ment systems. If choices are not well aligned with the preferences and behavior of target 
groups and do not evolve in response to changes in these, the system will not be sustain-
able. If management systems and processes are unreliable, confidence will be undermined. 
When a matching defined contribution system is first put in place in a country with some 
significant experience with defined contribution systems in general, the governance issues 
may be relatively incremental. Still, matching contributions pose unique governance 
issues that non-MDC systems do not. First, with matching contributions, there is a need 
for adequate oversight and strategy on the matching contribution rate. There is also a need 
to continuously monitor take-up of matching contributions and the overall effectiveness 
of the strategy, communicate the matching contribution design, oversee administration 
arrangements, and monitor costs. It is especially important to ensure the extra contribu-
tions are not being appropriated by intermediaries or service providers such as fund man-
agers through higher charges.
Addressing these challenges requires resources and significant data, system-specific 
regulations, and the collaboration of many parties whose incentives may be limited. There 
are several ways to help address these governance challenges:
 
• Define in advance success metrics (such as coverage) that can be tracked and 
evaluated over time in an appropriate way
 
• Establish human capital metrics for the development of management and gover-
nance expertise around matching contribution programs
 
• Define appropriate service standards for components of provision such as 
administration
 
• Set up appropriate mechanisms for comparing/measuring best practice from 
other markets

16.  IMPLEMENTATION ISSUES IN LOW- AND MIDDLE-INCOME COUNTRIES 
323
 
• Create mechanisms for participant feedback on a regular basis
 
• Define in advance a planning and review cycle to discuss issues that may arise.
The common thread in this regard is advance planning; thus, there should be sig-
nificant discussion of governance issues in the run-up to implementing a MDC program.
Governance of a defined contribution pension system requires many specific types 
of expertise that are often underappreciated. First, there needs to be an understanding of 
investment options, determining options for individuals to select from, and expertise in 
selecting and evaluating the performance of fund managers. Second, there needs to be a 
good understanding of how the advisory process works and monitoring to ensure individ-
uals are making informed choices and intermediaries are acting in accordance with good 
market practice. Third, there needs to be a good understanding of administrative systems 
and how to make them work well for defined contribution plans. In general, the provision 
of choice that is inherent in the matching design makes administration of defined contri-
bution plans more complex. This complexity is exacerbated by the fact that individuals 
have much less clarity than in traditional pension plans regarding the anticipated nature 
of their benefits.
Initial Parameters
The relevant initial conditions that affect the best design choice for an MDC scheme 
vary by country. Table 16.2 shows three stylized cases that can be helpful in thinking 
about these conditions. The first group of countries is low-income countries with young 
populations and low coverage rates. The second group is middle-income countries with 
older populations and higher coverage in which, nonetheless, a significant and persistent 
gap still exists. The last group is transition socialist economies in Eastern Europe and the 
former Soviet Union. These countries are at an advanced stage in their demographic tran-
sition and have coverage rates higher than would be expected given their income levels. In 
many transition economies, coverage rates are still declining, as state employment levels 
fall and the informal sector grows.
These conditions have implications for the short- and long-run prospects of MDCs. 
One concern is that the matching contribution incentive may lead workers to move from 
TABLE 16.2  Initial conditions affecting the design choice of a matching defined contribution 
scheme 
Country type
Purchasing power–adjusted 
income per capita, 2008 ($)
Coverage ratio (%)
Ratio of age 20–59/60+ 
population (%)
Low income
> 4,500
17
7.6
Middle income
4,500–15,000
51
6.3
Transition economy
2,000–20,000
66
3.7
SOURCE: Author’s calculations, based on World Development Report tables (http://data.worldbank.org/data-catalog/
world-development-indicators).

324 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
the formal to the informal sector. This threat is greater in middle-income countries than 
in low-income countries, especially when the MDC is targeted to the lower end of the 
income distribution. In middle-income countries and transition economies, the potential 
to move in and out of the formal sector and game the system tends to be greater, as cov-
erage rates extend well into the bottom half of the distribution. In fact, special attention 
must be given to the potential for a transition from MDC status to traditional contribu-
tory status, as there may be frequent shifts in status and “graduation” out of the targeted 
population. In low-income countries, in contrast, there is little interaction between the 
labor force in the bottom deciles and the top deciles where most of the pension coverage 
occurs, much of it in the public sector (see chapter 12).
Along similar lines, the match that is fixed in a low-income country environment 
may be set much lower relative to the average wages covered in the contributory scheme 
for formal sector workers. The MDC plan in Colombia has this feature.
3
 In China, too, 
rural pension levels are not at all comparable to formal sector pensions in urban areas.
In middle-income countries, the match should not be set in such a manner as to 
influence the decision at the margin as to whether to move to the informal sector. Such a 
situation could emerge among self-employed workers or through collusion between work-
ers and employers in small firms. One way to address this potential problem is to target 
occupations that are either de jure or de facto excluded from the contributory scheme. 
Farmers are one such category in many countries. Another approach is to target house-
holds covered by cash transfer programs, where eligibility has already been determined 
and a recordkeeping infrastructure exists. This option may be particularly attractive for 
programs that graduate households after a certain number of years or after they reach a 
certain level of income. In these cases, the subsidy period is finite, and workers can con-
tinue to contribute to the scheme after they leave the program, hopefully having become 
accustomed to the contribution process.
Initial conditions such as the current poverty line and life expectancy also affect key 
parameters of the MDC scheme, including the target benefit level, the contribution level, 
the age of withdrawal, the rules for the accumulation phase, and indexation. Each of these 
parameters is discussed below.
TARGET BENEFIT LEVEL
The target benefit level is the starting point for determining the parameters of the scheme, 
as the MDC is, by definition, fully funded through its contributions, including the match. 
The contribution level will therefore depend on the target benefit.
From a public policy viewpoint, a reasonable target would be that someone who 
participated for most of his or her working life would accrue a benefit sufficient to provide 
old-age income above the poverty level.
4
 How poverty is defined varies across countries. 
An important distinction is absolute versus relative poverty. In low-income countries, it 
may be more reasonable to target absolute poverty; in richer countries, relative poverty is 
often the metric of interest. In either case, the target benefit could allow for other sources 
of income or could place the entire burden of achieving the target on the MDC scheme. It 
may be useful to take into account the different needs of the elderly as well as intrahouse-
hold dynamics.

16.  IMPLEMENTATION ISSUES IN LOW- AND MIDDLE-INCOME COUNTRIES 
325
It is good practice to set the target benefit level using objective criteria such as 
income per capita or an empirically based poverty line (as opposed to politically deter-
mined criteria such as the minimum wage). So doing allows for a coherent pension sys-
tem design, as the parameters can be set to be consistent with both existing contributory 
and noncontributory schemes or social assistance programs (see discussion below). The 
absence of a clear and objective rationale for target benefit levels opens the program to 
arbitrary changes, political manipulation, or both, undermining the original goals of the 
scheme, including its sustainability, and reducing its credibility in the long run. 
CONTRIBUTIONS
Contributions of informal sector workers cannot be easily linked to income, which is 
not easily verifiable. Instead, a flat amount or several levels of flat contributions would 
be specified. The match would apply only to these flat amounts up to a ceiling consistent 
with the target benefits of the scheme. Indexation of contributions is discussed below 
along with benefit indexation.
The total contribution required to achieve this benefit level—which includes the 
contributions of both the worker and the government—must be calculated. This calcula-
tion is somewhat complicated, as it requires a set of assumptions about the rate of return 
on investments and the annuity factor. (Even if an annuity is not required, a calcula-
tion that translates the projected value of accumulations in an account to a target benefit 
stream is useful.) Returns (specifically, net returns after fees) depend on what contributors 
are charged for administering their accounts and the performance of the assets in which 
they are able to invest. This calculation is necessarily an estimate with a fairly large disper-
sion of possible outcomes. Shah (2005) presents a good example of how this calculation 
could be made using historical asset prices and a Monte Carlo simulation approach. In 
each country, this set of calculations will be different and will result in a different con-
tribution level according to local circumstances.
5
 Budget constraints will also have to be 
taken into account in setting matching levels, determining the target population, or both 
(as discussed below).
AGE OF WITHDRAWAL
The term age of withdrawal, as opposed to retirement age, recognizes the fact that many 
if not most participants in MDC schemes are likely to continue working in one form or 
another. Typically, however, their ability to maintain their income level is limited. MDC 
funds can supplement their partial retirement and may eventually provide their only 
source of income. Age of withdrawal is also a more appropriate term to use with regard to 
people who were never employed outside the home.
In low-income countries, there may be substantial differences between the life expec-
tancy and health status of people covered by formal pension schemes and people covered 
by an MDC scheme, particularly targeted programs. Restricting withdrawals until the for-
mal system retirement age may not make sense for informal sector workers who are likely 
to need their savings earlier, as their productivity and therefore their incomes decline. In 
middle-income countries, where there is greater overlap between people in and outside the 
formal sector, these differences may not justify different age restrictions.

326 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
RULES FOR ACCUMULATION 
The main set of rules for the accumulation phase relate to investment options. Low 
financial literacy is common even in rich countries; a paternalistic view may be even 
more justified for informal sector workers. As discussed in chapter 15 and elsewhere in 
this volume, default options are highly influential; limited investment choice with simple 
portfolio choices that do not require significant financial literacy are preferable. Low-risk 
alternatives such as government bonds are preferable to government guarantees. At the 
same time, low risk implies relatively low returns. A possible alternative would be life-
cycle defaults, which gradually reduce risk automatically as workers approach retirement 
age.
RULES FOR WITHDRAWAL 
A number of policy choices in addition to age must be made regarding withdrawal of the 
MDC accumulation upon death or during old age. Funds could be withdrawn in a lump 
sum, after full or partial annuitization, or in some other form, such as a phased with-
drawal (which mimics an annuity but does not pool longevity risk across the group). The 
choice should take into account the qualification criteria for other programs, such as social 
assistance. There may be practical reasons for withdrawing as a lump sum, such as bal-
ances that are too small to annuitize to consider. An innovative option would be universal 
pooling of mortality risk to avoid problems of adverse selection in annuity conversions. 
A simple solution is for the government to offer an actuarially fair annuity based on the 
mortality rates of the group in question. The annuity could be managed directly or con-
tracted out on a competitive basis.
Allowing at least partial withdrawals for unpredictable expenditures, such as health 
shocks, would make the program more attractive. These contingencies are difficult and 
expensive to monitor and enforce, and may inevitably be based on subjective criteria 
related to needs. The potential need for liquidity suggests that a holistic approach should 
consider the overall set of risks faced by workers and to use complementary programs, 
such as health and other types of insurance (for example, crop or cattle insurance), wher-
ever possible.
INDEXATION
In a growing economy with inflation, the initial parameters of the system will soon 
become inadequate without some form of indexation. Even the absolute poverty line will 
require adjustment for inflation, even if not for relative income or expenditure. The target 
benefit level should thus increase over time, at least in nominal terms. As in the case of 
the initial value, it is important to avoid arbitrary, intercohort variation of the level of the 
target benefit by indexing it to an objective indicator. The most obvious indicator may be 
inflation, but other measures that reflect changes in the definition of absolute poverty in a 
particular country or take into account structural changes (such as the extension of health 
insurance coverage) can be used.
Any change in the target benefit level implies a change to the contribution level. 
In order to maintain the initial target in real terms, the contribution could be gradu-
ally increased as accumulated price inflation leads to a discrete increase (for example, 

16.  IMPLEMENTATION ISSUES IN LOW- AND MIDDLE-INCOME COUNTRIES 
327
10 percent). Alternatively, the contribution may be indexed to changes in per capita 
income levels, particularly in fast-growth environments, so that it does not become irrel-
evant to large portions of the contributor population over a few decades. Indexation is the 
bridge between current and future policy and the role of the MDC in the overall system.
INSTITUTIONAL FACTORS
It is not possible to discuss parameters without considering institutions and other relevant 
local conditions, especially with regard to the financial sector. Some countries may have 
specialized providers and regulatory capacity that could make MDC schemes more viable 
and less costly to establish. In others, local conditions may not be conducive to the MDC 
approach. This discussion is beyond the scope of this chapter, but has been covered in the 
broader policy debate over the potential role of defined contribution schemes (see Rocha 
and Rudolph 2009).
Planning for the Long Run and Internal Consistency
By its nature, pension policy has to be formulated for the long term. The conditions 
discussed above change as incomes grow, the size of the formal sector expands, and the 
population ages. These changes mean that the role of the MDC should evolve. Ideally, 
contributors to the MDC scheme should be seamlessly joined with contributors to the 
formal contributory scheme. Doing so requires careful planning of the transition path 
and the portability of benefits. For example, administrative recordkeeping could be har-
monized and common identifiers used to allow for portability. In some cases, the same 
delivery infrastructure and information systems could be used from the outset, reducing 
set-up costs, as was done in Colombia and India and proposed in Mexico.
Perhaps the most important intersection of policy for MDC schemes is with 
noncontributory or social pensions. A growing number of countries have introduced or 
expanded cash transfer programs aimed at the elderly. In some cases, they are targeted 
to discrete populations or means tested; in other cases, they feature universal age-based 
categorical benefits. Bolivia, Botswana, Brazil (rural areas), Kosovo, Maldives, and New 
Zealand provide universal age-based benefits. They provide a floor for old-age income 
support and a short-term, simple solution to an important policy challenge—the coverage 
gap. As the population ages, these programs may become expensive, especially if their role 
goes beyond providing solely a poverty alleviation benefit.
6
Viewing pension policy in a dynamic manner, MDC schemes (and contributory 
schemes in general) do little or nothing to address the coverage gap in the near term, 
because pension income is generated only after decades of accumulation followed by a 
payout. For people currently reaching old age or who are already old, only a social pension 
can address the gap in pension coverage. For younger cohorts, expansion of MDC and 
contributory schemes generally can help address the gap.
In the example in figure 16.4, the changing role of social pensions over time for a 
hypothetical low-income country is shown. For simplicity, the initial period has zero cov-
erage for the contributory scheme. The figure shows the case of a universal social pension 
worth 40 percent of income per capita in its first year. Figure 16.4a shows the replacement 
rate of the social pension value to people with one-third to three times income per capita. 

328 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
Figure 16.4b shows that, as a share of income per capita, the benefit is equivalent for all 
recipients across the income spectrum.
FIGURE 16.4  Replacement rates from universal flat pension for hypothetical worker by income 
level 
b. Replacement rate compared
to average income per capita
a. Replacement rate compared
to own income per capita
individual income, % of own
income per capita 
individual income, % of
average income per capita 
0.0 
0.5 
1.0 
1.5 
2.0 
2.5 
gross relative pension level 
social
0.5
1.0
1.5
2.0
2.5
3.0
social
0.5
1.0
1.5
2.0
2.5
3.0
0.00 
0.25 
0.50 
0.75 
1.00 
1.25 
gross replacement rate 
Figure 16.5 shows the situation after 40 years. The simplified example shows what 
would happen if the initial social pension value was linked to a real absolute poverty line 
and indexed accordingly. Under reasonable assumptions of real wage growth, the social 
pension provides a much lower replacement rate of only about 10 percent of income 
per capita. People who participated in the contributory scheme (partly financed through 
matching contributions) have earned sufficient pension incomes to make up the difference. 
FIGURE 16.5  Role of social pension after maturation of contributory scheme
0.5
1.0
1.5
2.0
2.5
3.0
individual income, % of own
income per capita 
gross relative pension level 
individual income, % of
average income per capita 
0.0 
0.5 
1.0 
1.5 
2.0 
2.5 
contributory
social
0.00 
0.25 
0.50 
0.75 
1.00 
1.25 
gross replacement rate 
contributory
social
0.5
1.0
1.5
2.0
2.5
3.0
b. Replacement rate compared
to average income per capita
a. Replacement rate compared
to own income per capita

16.  IMPLEMENTATION ISSUES IN LOW- AND MIDDLE-INCOME COUNTRIES 
329
In this case, the contribution is assumed to be 10 percent of income per capita during the 
entire period (that is, effectively indexed to income growth). Taking into account the net 
rate of return and annuity factors for the stylized low-income country case, the result is a 
replacement rate of 40 percent for the average-income person and about 60 percent for 
the lowest income category.
This microlevel view can be translated into a set of long-run fiscal projections for the 
cost of the MDC scheme. The starting point is the initial year. Assuming a target benefit 
level of 40 percent replacement rate and an informal sector that makes up 80 percent of 
the labor force, half of the 10 percent of income per capita contribution would require 
0.8 percent of gross domestic product (GDP), assuming that the match would result in 
complete take-up. Relaxing this unrealistic assumption and aiming for take-up rates that 
are more in line with those shown in figure 16.1, about half the informal sector popula-
tion would receive a match, at a cost of 0.4 percent of GDP. This level of spending could 
also be achieved by targeting the subsidy.
7
 Other variants would involve different match-
ing rates. As discussed earlier, however, there is almost no evidence as to the elasticity of 
take-up of MDC schemes, especially in developing countries.
The path of these costs would depend on the interaction of the MDC with the 
other components of the pension system. If, for example, the formal contributory scheme 
were able to absorb a growing share of the labor force as incomes rose and formalization 
spread, costs could be gradually reduced without sacrificing the overall objective in terms 
of pension adequacy. A design that allowed for an offset to the social pension—such as 
the scheme recently implemented in Chile—could further reduce the total cost to govern-
ment. In this case, the MDC subsidy reduces future social pension costs.
Conclusions
Coverage in formal pension systems has failed to expand (and in some cases has con-
tracted) in low- and middle-income countries for decades. In many cases, lack of cover-
age is linked to increasing levels of informal labor or the failure to formalize much of the 
economy as it develops. Financial incentives linked to income taxes imposed on earnings 
are associated with higher participation rates in voluntary pensions in high-income coun-
tries—although experience in a variety of high-income settings indicates that the relation-
ship between the value of tax preferences and supplemental savings for retirement is not 
particularly strong. Such incentives are largely irrelevant and may even be regressive in 
their distributional outcomes in developing countries. In these settings, matching contri-
butions may provide a greater incentive that is more effective in drawing low-income and 
informal sector workers into the pension system.
Implementation of an MDC pension system in any setting, but especially in the low- 
and middle-income environments where it has increasingly being presented as an option, 
should only be considered after careful review of enabling conditions. These include a 
reasonable capacity of individuals to make informed choices to respond to incentives, 
administrative capacity to deal with the complexity and challenges of processing contribu-
tions and maintaining individual accounts, a reliable governance mechanism, and a sus-
tainable commitment that includes a political environment that will support the system 
over the decades required for it to mature and pay benefits. Operating a system requires a 

330 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
developed and well-supervised financial system with the appropriate long-term products 
to reliably accumulate savings and convert these into secure retirement income.
Establishing an MDC pension scheme in a low- or middle-income country will also 
require effective choices to be made regarding basic parameters so as to align the system 
with individual needs and objectives. Key factors that should be taken into account in 
designing an MDC scheme include anticipated contribution flows and individual levels, 
the size of the match, (although the empirical evidence on how much coverage could 
be increased through different levels of matching does not yet exist
8
), targets for benefit 
levels, and rules for withdrawals. In settings in which MDCs are viewed as a useful policy 
instrument for addressing the coverage gap (taking into account the opportunity cost of 
the resources required), policy makers must also consider the effect of MDCs on existing 
contributory and noncontributory pension schemes over the long run. MDCs could oth-
erwise encourage evasion of contributory schemes, and the cost of social pensions could 
be greater than required to meet public policy objectives as contributory schemes mature. 
The combination of all these factors makes the decision to pursue a matching con-
tribution scheme and its design a complex process that will require considerable analysis 
before it can be effectively implemented.
Notes
1.  In fact, in the countries in which public employment has been reduced least, such as Belarus, 
coverage remains much higher than would be predicted by the level of income per capita (see 
Pallares-Miralles, Romero, and Whitehouse 2011).
2.  The study cited most often, by Duflo and others (2005), is for low-income workers in the 
United States.
3.  The Colombian scheme is called Beneficios Económicos Periódicos (Periodic Economic Ben-
efits), as it could not legally be termed a pension, which according to the country’s constitu-
tion cannot be lower than the minimum wage.
4.  In practice, there will be contribution gaps (contribution density will be less than 100 per-
cent), which will vary widely across workers. Allowing some flexibility in the timing of the 
match may be warranted. For example, matching could be cumulative over a multiyear period 
rather than annual. Such a feature would add to the complexity of system administration, 
however, and has limitations.
5.  See chapter 12 for a discussion of India.
6.  The second key objective for pension policy is consumption smoothing. Most countries use 
contribution-based schemes to achieve this objective. New Zealand is one of the few countries 
that provided only a universal pension until recently, when it made efforts to increase volun-
tary contributory pensions to address the consumption-smoothing objective. See chapter 5 for 
a detailed description.
7.  Alternatively, a budget envelope could be established and take-up rationed on a first-come, 
first-served basis.
8.  See Palacios and Robalino (2009) for simulations of different take-up elasticities.

16.  IMPLEMENTATION ISSUES IN LOW- AND MIDDLE-INCOME COUNTRIES 
331
References
Duflo, E., W. Gale, J. Liebman, P. Orszag, and E. Saez. 2005. “Savings Incentives for Low- and 
Middle-Income Families: Evidence from a Field Experiment with H&R Block.” NBER Work-
ing Paper 11680, National Bureau of Economic Research, Cambridge, MA.
Godfrey, P. C. 2011. “Toward a Theory of the Informal Economy.” Academy of Management Annals 
5 (1): 231–77.
ISSA (International Social Security Administration). 2011. MBAO Pension Plan. Good Practices in 
Social Security Series, ISSA, Geneva.
Loayza, N. 1997. “The Economics of the Informal Sector.” Policy Research Working Paper 1727, 
World Bank, Washington, DC.
Palacios, R., and D. A. Robalino. 2009. “Matching Defined Contributions: A Way to Increase Pen-
sion Coverage.” In Closing the Coverage Gap: The Role of Social Pensions and Other Retirement 
Income Transfers, ed. R. Holzmann, D. A. Robalino, and N. Takayama, 187–202. Washington, 
DC: World Bank.
Pallares-Miralles, M., C. Romero, and E. Whitehouse. 2011. “International Patterns of Pension 
Provision II: Worldwide Overview of Facts and Figures.” Draft, World Bank, Washington, 
DC.
Rocha, R., and H. Rudolf. 2009. “Enabling Conditions for Second Pillars of Pension Systems.” 
Policy Research Working Paper 4890, World Bank, Washington, DC.
Shah, A. 2005. “Pension Outcomes Associated with Alternative Asset Allocation Under the New 
Pension System.” Unpublished.
—. 2009. Directions: A New Lens on Retirement Preferences. http://www.watsonwyatt.com/
pubs/directions/.
—. 2011. “Clear Direction in a Complex World: How Top Companies Create Clarity, Con-
fidence and Community to Build Sustainable Performance. 2011–2012.” http://www.tower-
swatson.com/united-kingdom/research/6639.
Watson Wyatt Limited. 2006. “The Pension Research Forum Research Results: DC Investment 
Choice—Can Employees Make Appropriate Investment Decisions?” Watson Wyatt Limited, 
London. 

---