A review of international experience

% of households
0
10
20
30
40
50
60
age 18–34 
age 35–54 
age 55+ 
2002 
2004 
2005 
2006 
2007 
2008 
2009 
7%
9%
3%
14%
17%
4%
20%
23%
8%
28%
31%
10%
36%
37%
14%
42%
38%
12%
48%
41%
19%
SOURCE: Authors’ calculations based on data from the SAVE panel, households eligible for subsidies. 
NOTE: Brackets show 95 percent confi dence intervals. Figures are weighted.
FIGURE 4.7  Uptake of Riester pensions by number of children
0
10
20
30
40
50
60
70
80
no children 
1 child 
2 children 
3+ children 
% of households
2002 
2004 
2005
2006 
2007 
2008 
2009 
5%
8%
12%
16%
10%
12%
22%
23%
14%
19%
30%
38%
19%
26%
41%
58%
23%
36%
45%
60%
28%
33%
49%
57%
29%
38%
56%
69%
SOURCE: Authors’ calculations based on data from the SAVE panel, households eligible for subsidies. 
NOTE: Brackets show 95 percent confi dence intervals. Figures are weighted.

92 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
FIGURE 4.8  Private pension instruments by number of children in 2009
27% 
38% 
56% 
68% 
25% 
29% 
30% 
25% 
14% 
14% 
18% 
17% 
0
10
20
30
40
50
60
70
80
% of households
no children 
1 child 
2 children 
3+ children 
Riester pension 
occupational pension
private pension 
SOURCE: Authors’ calculations based on SAVE 2010, households eligible for subsidies. 
NOTE: Brackets show 95 percent confi dence intervals. Figures are weighted.
FIGURE 4.9  Private pensions by monthly household disposable income in 2009
0
10 
20
30
40
50 
60
70 
€0–999
€1,000–€1,999
€2,000–€3,999
> €4,000
% of households
Riester pension 
occupational pension 
private pension 
19%
4%
4%
30%
15%
13%
47%
35%
18%
48%
53%
23%
SOURCE: Authors’ calculations based on SAVE 2010, households eligible for subsidies. 
NOTE: Brackets show 95 percent confi dence intervals. Figures are weighted.

4.  RIESTER PENSIONS IN GERMANY: DESIGN, DYNAMICS, TARGETING SUCCESS, AND CROWDING-IN 
93
Demand for supplemental pensions increased among all income groups since 2002 
(figure 4.10). Starting from an already lower initial level, the increase in percentage 
points is least apparent in the lowest income bracket. This effect is less pronounced when 
expressed in relative percentage terms. Particularly striking is the impressive growth of 
occupational pensions and unsubsidized private pension instruments among high-income 
households.
FIGURE 4.10  Uptake of Riester pensions by quintiles of monthly household disposable income
0
10
20
30
40
50
60
quintile 1 
quintile 2 
quintile 3 
quintile 4 
quintile 5 
2002 
2004
2005 
2006 
2007 
2008 
2009 
% of households
5%
6%
9%
6%
11%
6%
10%
14%
17%
19%
10%
19%
24%
24%
21%
12%
23%
27%
34%
35%
16%
28%
36%
40%
41%
18%
30%
37%
42%
42%
25%
26%
46%
45%
51%
SOURCE: Authors’ calculations based on data from the SAVE panel, households eligible for subsidies. 
NOTE: Brackets show 95 percent confi dence intervals. Figures are weighted.
An important finding from figure 4.10 is the unbroken increase in the share of 
households with Riester pension plans in the lowest income quintile since 2002. Low-
income earners actually had a slightly higher percentage rate of increase in their uptake 
than the middle of the income distribution. Hence, the jury is still out with regard to how 
well Riester pensions will end up serving low-income households.
Displacement Effects or Crowding In?
As the earlier discussion of fiscal costs shows, Riester pensions are an expensive govern-
ment program. The subsidies increase direct government spending and thus potentially 
reduce aggregate national savings. In addition, the successful uptake of Riester pensions 
may have simply displaced other saving, in particular saving within other types of private 

94 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
funded pensions. The macroeconomic effect of subsidizing Riester pensions may thus be 
zero or even negative.
Corneo, Keese, and Schröder (2009, 2010) and Pfarr and Schneider (2010, 2011) 
provide econometric analyses that cannot refute the hypothesis that subsidizing Riester 
pensions produces only displacement effects. In all those papers, however, strong implicit 
assumptions must be made in order to overcome the problem of a missing counterfactual 
(thanks to the design of the Riester scheme, virtually everyone is eligible, so there is no 
natural control group). They assume, for example, that having a Riester pension and hav-
ing other savings are independent decisions.
The causal effect of the subsidies on total saving cannot be unambiguously isolated, 
as most households are eligible. Cross-sectional analyses are difficult, because ineligible 
households (for example, certain groups of self-employed workers) have quite different 
socioeconomic characteristics. Comparisons over time are more credible, although a clean 
counterfactual is not available. Evidence is thus more circumstantial than causal.
One way to address the issue is to ask households about the extent to which the sub-
sidies stimulated new savings. Coppola and Reil-Held (2010) analyze responses to these 
types of questions. Responses to questions about changes in behavior may be subjective 
and contain elements of wishful thinking or ex post justification. Nevertheless, the results 
shown in figure 4.11 are unambiguous: only a minority of households reports saving less 
in total since enrolling in a Riester pension plan, and most households report saving more. 
Particularly striking is the fact that a very large proportion of low-income households 
indicate that they saved more. The two upper quintiles of income provided similar pat-
terns of responses.
As a less subjective piece of evidence, figure 4.3 shows that the increase in Riester 
pensions was not accompanied by a decline in the use of other pension instruments: after 
introduction of Riester pensions, occupational pensions also increased, and unsubsidized 
pension plans stayed essentially flat between 2004 and 2009. Perhaps most important, the 
aggregate household saving rate rose, increasing from 9.4 percent of disposable income in 
2001 to 11.4 percent in 2010. Moreover, the new pensions did not result in an increase in 
aggregate dis-saving. Until the financial crisis and its large-scale fiscal stimulus packages, 
government deficits declined.
Börsch-Supan, Reil-Held, and Schunk (2008) provide an econometric analysis of 
supplemental pensions, taking account of many socioeconomic variables. They provide 
evidence on displacement effects through two channels. First, they employ a bivariate 
regression model in which the decision to take up a subsidized Riester pension plan and 
the decision to enroll in other unsubsidized private pension plans are modeled simulta-
neously. A negative correlation between the two equations would indicate displacement. 
Second, they include competing motives for saving as explanatory variables. A negative 
coefficient on such variables would also indicate displacement.
Table 4.3 presents selected results of two specifications. Specification A describes 
disposable income as a set of four quintile indicators. Specification B employs a quadratic 
function of disposable income. In both specifications, the first dependent variable (col-
umns 1 and 3) indicates whether a household has a Riester pension plan, and the second 
dependent variable (columns 2 and 4) indicates whether a household has an unsubsidized 
private pension plan. All variables are as of the end of 2005.

4.  RIESTER PENSIONS IN GERMANY: DESIGN, DYNAMICS, TARGETING SUCCESS, AND CROWDING-IN 
95
FIGURE 4.11  Change in total saving after enrolling in a Riester plan 
43.3%
33.3%
23.3%
40.0%
36.4%
23.6%
40.0%42.0%
18.0%
50.0%
39.6%
10.4%
58.5%
38.5%
3.0%
0
20
40
60
quintile 1
quintile 2
quintile 3
quintile 4
quintile 5
a. Household income quintiles
46.8%
29.8%
23.4%
36.0%
41.9%
22.1%
40.2%
46.7%
13.1%
59.2%
31.7%
9.2%
56.1%
41.2%
2.6%
0
20
40
60
quintile 1
quintile 2
quintile 3
quintile 4
quintile 5
b. Per capita income quintiles
% of households
% of households
saving the same
saving less
saving more
SOURCE: Coppola and Reil-Held 2010, based on SAVE 2008 households with Riester pensions.
NOTE: Data are weighted.
The regressions have a very satisfactory fit (see McFadden R  
2
 values at the bottom 
of table 4.3). Moreover, the estimated model exhibits a positive correlation between the 
two equations, or, more precisely, between the unobservables in the decision to take up 
a subsidized Riester pension plan and the decision to enroll in unsubsidized private pen-
sion plans. Although the correlation is small and not statistically significant, it does not 
support the hypothesis of crowding out of unsubsidized private pension plans by Riester 
pensions.
Table 4.3 shows a set of variables in the regression that reflect the importance of dif-
ferent saving motives for respondents. Of interest are three saving motives: acquiring real 
estate, bequeathing wealth, and pocketing state subsidies. There is evidence pointing to 
a possible displacement effect between old-age provision and the purchase of real estate. 
This effect is apparent when looking at households that report a particular interest in sav-
ing for the purchase of real estate property. The more important this saving motive is, the 
less likely respondents are to have taken up a Riester pension plan. A second significantly 
negative coefficient is attached to the variable indicating the desire to save for a bequest. 
This result may suggest that the requirement that Riester plans be paid out as annuities 
acts as a disincentive for households for which making a bequest is an important saving 
motive.

96 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
TABLE 4.3  Selected determinants of the demand for Riester and other private pension products
BIVARIATE PROBIT ESTIMATES
Variable
Specification A
Specification B
Riester
(1)
Other private 
pensions
(2)
Riester
(3)
Other private 
pensions
(4)
Intention to buy real estate
0.001
0.143
−0.001
0.147
(0.01)
(1.67)*
(0.01)
(1.71)* 
Saving motives
Buy real estate
−0.090
−0.057
−0.089
−0.058
(2.11)*
(1.39)
(2.08)**
(1.43)
Provide for unforeseen events
−0.096
−0.057
−0.086
−0.052
(1.44)
(0.86)
(1.28)
(0.79)
Pay off debts
−0.055
−0.041
−0.054
−0.043
(1.24)
(0.99)
(1.22)
(1.03)
Old-age provision
0.229
0.694
0.218
0.691
(3.06)***
(7.87)***
(2.92)***
(7.86)***
Holiday
0.009
−0.068
0.012
−0.069
(0.18)
(1.49)
(0.25)
(1.51)
Finance major purchases
0.042
0.039
0.035
0.037
(0.81)
(0.77)
(0.68)
(0.72)
Finance (grand)child education
−0.038
−0.091
−0.038
−0.094
(0.81)
(2.02)**
(0.80)
(2.09)**
Inheritance
−0.124
0.090
−0.128
0.090
(2.32)**
(1.80)*
(2.39)**
(1.80)*
State subsidies
0.264
−0.015
0.269
0.008
(6.03)***
(0.38)
(6.13)***
(0.20)
Alternative instruments
Other form of supplementary
0.469
0.462
0.466
0.466
pension provision (dummy)
(6.27)***
(6.56)***
(6.25)***
(6.64)***
McFadden R2
0.137
0.136
Rho [Chi2(1)]
0.055 [1.32]
0.060 [1.54]
Number of observations
2,255
2,255
SOURCE: Börsch-Supan, Reil-Held, and Schunk 2008.
NOTE: Absolute value of the z statistics in parentheses. Regression also includes a large set of socioeconomic control vari-
ables. ***p < 0.01, **p < 0.05, *p < 0.1.

4.  RIESTER PENSIONS IN GERMANY: DESIGN, DYNAMICS, TARGETING SUCCESS, AND CROWDING-IN 
97
In addition to Riester plans and unsubsidized private pension plans, occupational 
pension plans and whole life insurance products have been very popular as instruments 
to provide supplemental retirement income (Bundesministerium für Arbeit und Soziales 
2006). The coefficient of a variable indicating the presence of such instruments is statisti-
cally significant and positive: households already covered by one of these pension types 
are more likely to have a Riester pension plan. One interpretation of this phenomenon is 
that households that think ahead and invest in old-age provision at all tend to use several 
instruments for this purpose. As opposed to saving for real estate acquisition or bequests, 
this result is consistent with a form of “crowding in” among pension products.
Filling the Pension Gap
The ultimate measure of success is the extent to which Riester pensions fill the pension 
gap created by the reduction of benefits provided by the pay-as-you-go financed public 
pension system. Börsch-Supan and Gasche (2010) simulate pension payments from the 
public system and pension payments from a Riester annuity for a typical employee saving 
exactly the amounts required to obtain the maximum subsidy according to table 4.1. Fig-
ure 4.12 shows the policy-induced decline in public pension benefits and the increasing 
annuity from Riester savings.
Individuals retiring in 2008 had accumulated very little Riester savings, as the pro-
gram started only in 2002. Individuals retiring later have a longer time span in which to 
accumulate Riester savings, until an equilibrium will be reached in about 2047. Although 
the combined pension benefits will be larger than the current public pension benefits 
FIGURE 4.12  Benefits from public and Riester pensions 
0
10
20
30
40
50
60
2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 2032 2034 2036 2038 2040 
% of average income 
year of retirement
pension level with Riester 
gross pension level without Riester 
SOURCE: Börsch-Supan and Gasche 2010. 

98 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
from about 2030 onward, the transitional generation will not be able to fill the pension 
gap completely (figure 4.13).
FIGURE 4.13  Filling the pension gap 
−2
0
2
4
6
8
10
12
14
16
pension gap 
Riester pension
2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 2032 2034 2036 2038 
% of average income 
year of retirement
2040
SOURCE: Börsch-Supan and Gasche 2010.
Actual saving for old age may be less than suggested by table 4.1, as uptake is well 
below 100 percent and individuals may not save the full amount or may interrupt contri-
butions (because of unemployment or other episodes of financial distress, for example). 
It may also be greater than suggested by table 4.1, as individuals may have other old-
age savings plans, such as unsubsidized individual account pension plans or occupational 
pensions.
Börsch-Supan, Essig, and Wilke (2005) provide a simulation exercise based on actual 
household data that sheds some light on this issue. They measure current financial wealth 
and the current financial net saving rate for all households in the SAVE panel. Assuming 
constant net saving rates and a real return of 2.8 percent, they compute financial wealth 
at retirement and convert it into a real annuity. This annuity is then compared with the 
pension gap.
On average, Germans have more than closed the emerging pension gap. More than 
two-thirds will be able to fill the gap between what they would have received in retirement 
income under the old and new public systems. The coverage rate for this gap (the Riester 
annuity divided by the pension gap) is 120 percent for the average household. The distri-
bution, however, is skewed. Predictably, results are very different for people with different 
socioeconomic characteristics, especially income. Although the median household will be 
able to compensate for the cuts in public pensions, about 27 percent of the households 
represented by the SAVE panel will not be able to fill the pension gap.

4.  RIESTER PENSIONS IN GERMANY: DESIGN, DYNAMICS, TARGETING SUCCESS, AND CROWDING-IN 
99
Conclusions
The most obvious lesson from Germany’s experience with Riester pensions is that new 
forms of subsidized savings need time before they are able to take off. It took more than 
10 years in the United States before IRAs were accepted by households in the upper 
60 percent of the income distribution (Venti and Wise 1990). In Germany, Riester pen-
sion plans exhibited more rapid and dynamic growth. Growth in the first few years was 
very steep. It then stalled before rebounding after the 2005 legislative changes. Overall, 
growth rates are much higher than those experienced in the United States. The higher 
take-up rate may reflect the direct subsidies or the urgency of off-setting the benefit reduc-
tions in the wake of the public pension reforms.
There are several explanations for the discontinuous development of Riester pen-
sions. Poor product design probably accounted for the initially lower rates of acceptance. 
In addition, the learning process regarding the need and way to invest in old-age provision 
took time, despite heavy advertising. As people learn from their social environment, the 
pace of this development depends on how widespread such pensions are in the population 
at large (Ruprecht 2004), generating exponential growth until saturation is reached.
It is not possible to identify whether the dynamic spread of Riester pensions reflects 
the financial incentives, the availability of information, or marketing efforts by govern-
ment and providers, mainly insurance companies. However, it is striking that the accel-
eration in Riester saving kicked in only after substantial simplifications were made to the 
scheme. Therefore, an important lesson is to avoid complex savings plans, which are not 
immediately understood by customers. Further simplifications are possible without jeop-
ardizing the aims of the program (see, for example, Westerheide and others 2010).
The target groups of the Riester reform in 2001 were mainly families with children 
and individuals with low income. Among these groups, parents with more than one child 
seem to be the group that was most easily reached. The evidence is less compelling for the 
other target group, low-income households. Although take-up continues to rise, its level 
is still low, with just 25 percent in the lowest income quintile participating in the system. 
A key lesson is that high subsidies alone are not enough to reach low-income house-
holds: information and social acceptance appear to be crucial in reaching this group. 
Coppola and Gasche (2011) find that the level of knowledge about eligibility for Riester 
subsidies is directly related with household income: households in the bottom quintile 
of the income distribution have a higher probability to wrongly believe that they are not 
eligible for the subsidies than households in the upper quintiles.
This insight is strengthened by the regression results of Börsch-Supan, Reil-Held, 
and Schunk (2008) and Coppola and Reil-Held (2010), which show that holding other 
characteristics, such as income and family size, constant, households with higher levels of 
educational attainment are more likely to enroll in Riester pensions than households that 
have not completed any vocational training. This finding is consistent with the broader 
experience indicating that knowing future pension replacement levels correlates positively 
with enrolling in private pension schemes. Another lesson therefore is that information 
and knowledge about arrangements relating to old-age pension provision are vital to 
achieving high take-up rates.

100 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
Although the SAVE data do not allow for a cleanly designed experiment, they pro-
vide circumstantial evidence about possible displacement effects. The results indicate 
that households that want to purchase residential real estate are significantly less likely to 
enroll in a Riester pension plan. The clumsy withdrawal rules of the Riester regulations 
do not provide these households with sufficient liquidity to persuade them to make pro-
vision for old age alongside with saving to acquire real estate property. There is also evi-
dence that stated bequest motives displace Riester pension plans, with their strict annuity 
rules. These findings do not indicate that these restrictions should be removed. On the 
contrary, they are needed, as Riester pensions were designed to fill the gap in public pen-
sion benefits after pension reform, and pay-as-you-go pensions are paid out as life-long 
annuities.
The desire to purchase property and bequeath assets are saving motives that appear 
to compete with taking up Riester pensions. Occupational pensions and whole life insur-
ance, however, are complements rather than substitutes for Riester pensions. This finding 
provides another important lesson. Households that think ahead and invest in old-age 
provision tend to use several instruments for this purpose. In this sense, there are “crowd-
ing in” effects of fostering retirement saving. Although subsidization of Riester pensions is 
expensive, the net effect (that is, the additional savings created after subtracting subsidies 
and reductions in other savings vehicles) appears to be positive.
The ultimate measure of success is the extent to which Riester pensions will fill the 
pension gap created by the reduction in benefits provided by the pay-as-you-go financed 
public pension system. On average, Germans will be able to more than close the growing 
pension gap. The distribution is skewed, however. Although the median household will be 
able to compensate for the cuts in public pensions, about 27 percent of German house-
holds will not be able to fill the pension gap.
Riester pensions are not a panacea for the problem of inadequate retirement income; 
preventing poverty in old age requires different instruments. This new pension instru-
ment can be regarded as a success story for the middle of the income distribution in 
Germany, however. It has moved about €9.4 billion from consumption or other saving 
instruments not related to old-age provision into savings earmarked for old age, making 
the system of retirement income more robust and stable. At the same time, it has reduced 
pressure on younger generations by permitting lower pay-as-you-go contribution rates 
than before reform. Whether this change justifies the Riester scheme’s tax and subsidy 
costs of €3.5 billion a year—equal to 1.5 percent of the total pay-as-you-go pension bud-
get—remains a subjective judgment.
Notes
1.  In essence, the formerly monolithic German pay-as-you-go system is transiting to a multipillar 
system of public, occupational, and private pensions. For a detailed description of the pension 
reform process in Germany, see Börsch-Supan and Wilke (2004).
2.  The Commission for the Sustainable Financing of the German Social Security Systems (2003) 
suggested including an automatic inflation adjustment. Currency values in this chapter are 
given in euros; €1 = $1.29.
3.  For an analysis of the initial phase, see Dünn and Fasshauer (2003).

4.  RIESTER PENSIONS IN GERMANY: DESIGN, DYNAMICS, TARGETING SUCCESS, AND CROWDING-IN 
101
4.  In 2008, 57.1 percent of beneficiaries were women, and 24.4 percent of people receiving 
subsidies were from states in the former German Democratic Republic. About 9.8 million 
people received a basic subsidy and 3.7 million people an additional child’s subsidy (Stolz and 
Rieckhoff 2011).
5.  The interview is conducted with the individual who best knows the household’s finances.
6.  As in all surveys that deal with sensitive topics such as household finances, nonresponse to sen-
sitive questions cannot be ignored (see Essig and Winter 2003 and Schunk 2006 for discussion 
and documentation). To prevent biased inference based on an analysis of complete cases only, 
researchers applied an iterative multiple imputation procedure to the SAVE data (Schunk 
2008). All results reported in this chapter use the fully imputed SAVE data.
References
Börsch-Supan, A. 2004. “Mind the Gap Incentives: The Effectiveness of Incentives to Boost Retire-
ment Savings in Europe.” OECD Economic Studies 39: 111–44.
Börsch-Supan, A., ed. 2003. Life-Cycle Savings and Public Policy: A Cross-National Study of Six 
Countries. London: Academic Press.
Börsch-Supan, A., M. Coppola, A. Eymann, L. Essig, and D. Schunk. 2010. “The German SAVE 
Study. Design and Results.” MEA Study No. 6, Mannheim Research Institute for the Eco-
nomics of Aging, University of Mannheim.
Börsch-Supan, A., L. Essig, and C. Wilke. 2005. “Rentenlücken und Lebenserwartung. Wie sich 
die Deutschen auf den Anstieg vorbereiten.” Deutsches Institut für Altersvorsorge, Köln.
Börsch-Supan, A., and M. Gasche. 2010. “Kann die Riester-Rente die Rentenlücke in der gesetz-
lichen Rente schließen? ” Diskussionspapier Mai, Mannheim Research Institute for the Eco-
nomics of Aging, Mannheim University.
Börsch-Supan, A., A. Reil-Held, and D. Schunk. 2008. “Saving Incentives, Old-Age Provision and 
Displacement Effects: Evidence from the Recent German Pension Reform.” Journal of Pension 
Economics and Finance 7 (3): 259–319.
Börsch-Supan, A. and C. Wilke. 2004. “The German Public Pension System: How It Was, How It 
Will Be.” NBER Working Paper 10525, National Bureau of Economic Research, Cambridge, 
MA.
Bundesministerium für Arbeit and Soziales. 2006. Renten- und Alterssicherungsbericht 2005. Berlin. 
http://www.bmas.de.
Commission for the Sustainable Financing of the German Social Security System. 2003. “Final 
Report.” German Federal Ministry for Health and Social Affairs, Berlin.
Coppola, M., and M. Gasche. 2011. “Die Riester-Förderung: Das unbekannte Wesen.” Diskus-
sionspapier Mai, Mannheim Research Institute for the Economics of Aging, Mannheim 
University.
Coppola, M., and A. Reil-Held. 2010. “Jenseits staatlicher Alterssicherung: die neue regulierte pri-
vate Vorsorge in Deutschland.” In Die Alten der Welt. Neue Wege der Alterssicherung im globalen 
Norden und Süden, ed. Lutz Leisering. Frankfurt: Campus.
Corneo, G., M. Keese, and C. Schröder. 2009. “The Riester Scheme and Private Savings: An 
Empirical Analysis Based on the German SOEP.” Schmollers Jahrbuch 129 (2): 321–32.
—. 2010. “The Effect of Saving Subsidies on Household Saving: Evidence from Germany.” 
Ruhr Economic Paper 170, Ruhr-Universität Bochum, Bochum.

102 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
Deutsche Bundesbank. 2002. “Funded Old-Age Provision and the Financial Markets.” Monthly 
Report July: 25–39, Frankfurt.
Deutsche Rentenversicherung. 2011. “Rentenversicherung in Zahlen 2011.” Deutsche Rentenver-
sicherung Bund, Berlin.
Disney, R., C. Emmerson, and M. Wakefield. 2001. “Pension Reform and Saving in Britain.” 
Oxford Review of Economic Policy 17 (1): 70–94.
Dünn, S., and S. Fasshauer. 2003. “Ein Jahr Riesterrente: Eine Übersicht aus Sicht der gesetzlichen 
Rentenversicherung. ” Deutsche Rentenversicherung 1–2.
Essig, L. 2005. “Methodological Aspects of the SAVE Data Set.” MEA Discussion Paper 80-05, 
Mannheim Research Institute for the Economics of Aging, University of Mannheim.
Essig, L., and J. Winter. 2003. “Item Nonresponse to Financial Questions in Household Sur-
veys: An Experimental Study of Interviewer and Mode Effects.” MEA Discussion Paper 39-03, 
Mannheim Research Institute for the Economics of Aging, University of Mannheim.
Gale W. J., and J. K. Scholz. 1994. “IRAs and Household Saving.” American Economic Review 84: 
1233–60.
Gerber, U., and M. Zwick. 2010. “Daten zur kapitalgedeckten Altersvorsorge: Die Riesterrente.” 
Deutsche Rentenversicherung, Heft 2/2010, S. 197–207.
Pfarr, C., and U. Schneider. 2010. “Angebotsinduzierung und Mitnahmeeffekt im Rahmen 
der Riester-Rente: eine empirische Analyse.” SOEP Paper on Multidisciplinary Panel Data 
Research 341, DIW, Berlin.
—. 2011. “Anreizeffekte und Angebotsinduzierung im Rahmen der Riester-Rente: Eine 
empirische Analyse geschlechts- und sozialisationsbedingter Unterschiede. ” Perspektiven der 
Wirtschaftspolitik 12 (1): 27–46.
Ruprecht, W. 2004. “Automatische Entgeltumwandlung in der betrieblichen Altersversor gung: 
Eine Replik. ” Wirtschaftsdienst 10: 651–56.
Schunk, D. 2006. “The German SAVE Survey: Documentation and Methodology.” MEA Discus-
sion Paper 109-06, Mannheim Research Institute for the Economics of Aging, University of 
Mannheim. 
—. 2008. “A Markov Chain Monte Carlo Algorithm for Multiple Imputation in Large Sur-
veys.” Advances in Statistical Analysis 92 (1): 101–14.
Skinner, J., and R. G. Hubbard. 1996. “Assessing the Effectiveness of Saving Incentives.” Journal of 
Economic Perspectives 10 (4): 73–90.
Stolz, U., and C. Rieckhoff. 2005. 2008. “Förderung der zusätzlichen Altersvorsorge für das Bei-
tragsjahr 2005 durch die ZfA.” RVaktuell 9/2008.
—. 2009. “Beitragsjahr 2006: Erstmals mehr als eine Milliarde Euro Zulagenförderung durch 
die ZfA.” RVaktuell 10/2009.
—. 2010. “Beitragsjahr 2007: Zulagenförderung nochmals um mehr als ein Viertel gestie-
gen.” RVaktuell 11/2010.
—. 2011. “Förderung der Riester-Rente für das Beitragsjahr 2008: Mehr als neun Millionen 
Personen mit Zulagen.” RVaktuell 12/2011.
Venti, S., and D. Wise. 1990. “Have IRAs Increased U.S. Savings? Evidence from Consumer 
Expenditure Surveys. ” Quarterly Journal of Economics 105: 661–98.
Westerheide, P., M. Feigl, L. Jaroszek, J. Leinert, and A. Tiffe. 2010. Transparenz von privaten 
Riester- und Basisrentenprodukten. Abschlussbericht zu Projekt Nr. 7/09, Zentrum für Europä-
ische Wirtschaftsforschung, Mannheim.

103
CHAPTER 5
New Zealand’s Experience with the 
KiwiSaver Scheme
Geoff Rashbrooke
New Zealand’s automatic enrollment KiwiSaver retirement scheme, launched in July 2007, 
has proven very effective in achieving relatively high levels of participation in supplemen-
tary retirement saving. In a relatively short period, participation has been extended to 
nearly half the population under age 65 and an even larger proportion of the labor force. As 
a result, New Zealand now has one of the highest rates of coverage for supplementary retire-
ment saving in the world. The level of coverage likely reflects the combination of automatic 
enrollment and substantial matching incentives. Consistent with the power of defaults and 
inertia, the proportion of people who opt out after automatic enrollment is relatively low. 
The fact that many members continued to make contributions at the default rate even after 
that rate was reduced confirms the stickiness of default parameters and reinforces the impor-
tance of the default rate in establishing participation patterns.
N
ew Zealand has had a comprehensive system of public, tax-funded, flat-rate old-age 
pensions since at least 1938. In addition, until 1989, participation in formal private 
retirement saving arrangements was encouraged through a number of tax incentives, such 
as tax-free accrual of investment returns and tax exemptions on contributions, plus some 
scope for obtaining benefits as lump sums free of tax.
1
In 1989, all concessions for private retirement saving arrangements were removed, 
based on the belief that workers could be left to make sensible choices in their individual 
best interests, without any encouragement from the state beyond light-handed regulation 
and provision of some education on saving. This stance remains a strong current in the 
ongoing policy debate in New Zealand.
Coverage of workers in occupational schemes had never been particularly high. It 
fell markedly after concessional tax treatment was withdrawn (figure 5.1).
As a response to this trend and other concerns about preparation for retirement and 
saving behavior in general, in the mid-2000s the government commissioned a working 
party and subsequently an external report. Its results led to the setting up, on July 1, 2007, 
of an automatic enrollment national retirement saving scheme known as KiwiSaver.
The automatic enrollment (“soft compulsion”) feature of KiwiSaver is its key ele-
ment and an innovation for national retirement schemes worldwide. As a new approach, 
This chapter draws heavily on a presentation by Peter Whiteford, of the University of New South 
Wales, at the World Bank conference, The Potential for Matching Defined Contributions (MDC) 
Design Features in Pension Systems to Increase Coverage in Low- and Middle-Income Countries, 
held June 6, 2011. The author thanks Teneti Ririnui, of New Zealand’s Department of Inland Rev-
enue, for providing the Excel spreadsheets from the 2011 KiwiSaver evaluation report.

104 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
it has been of keen interest both in New Zealand and internationally, as a policy falling 
between compulsion and voluntary saving.
Understanding the design and potential lessons from the KiwiSaver system requires 
placing this new element of the retirement income system into the appropriate context. 
In particular, it requires understanding the role played by New Zealand’s long-established 
public pension system, New Zealand Superannuation. 
This chapter is organized as follows. The first section describes the history, structure, 
and outcomes of New Zealand’s public pensions system and raises questions about its 
viability. The second section describes the motivations, development, and structure of the 
KiwiSaver scheme. The last section summarizes the chapter’s conclusions and discusses 
the policy debate on retirement saving as it has unfolded in New Zealand over the past 
decade, highlighting concerns raised by the introduction of KiwiSaver.
New Zealand’s Public Pension System
HISTORY AND STRUCTURE
New Zealand’s public system originated in 1898, when the first old-age pension was intro-
duced.
2
 This seminal system provided a means-tested benefit of one-third of the average 
wage for an individual and two-thirds for a couple. About a third of people over age 65 
qualified for benefits. The scheme had other qualifying conditions, including a require-
ment to provide evidence of good character and provisions designed to exclude criminals, 
drunkards, and wife deserters.
In 1938, a substantial increase in means-tested payments and a new small universal 
benefit for people not qualifying for the means-tested benefit was introduced, establishing 
FIGURE 5.1  Coverage in occupational retirement schemes, 1990–2003
0 
5 
10 
15 
20 
25 
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 
% coverage
private 
government 
total 
SOURCE: Author, based on data from New Zealand Government Actuary 2006. 
NOTE: Figures for 1991 and 1992 are estimates.

5.  NEW ZEALAND’S EXPERIENCE WITH THE KIWISAVER SCHEME 
105
universal eligibility for some type of old-age income support. In 1977, a system of national 
retirement was introduced, providing a taxable, universal benefit of 80 percent of the aver-
age wage for a couple (48 percent for single person) payable at age 60 after 10 years’ resi-
dence. Over the ensuing two decades, incremental changes were made to modify benefit 
levels, link benefit formulas to after-tax earnings measures, and add a surcharge that effec-
tively created an income test, which was later removed. The age of eligibility was increased 
to 65, a change that was phased in between 1989 and 2001.
The current system reflects this evolution, incorporating the basic design principles 
and benefit structure of its antecedents. It provides a flat benefit payable at age 65 to all 
eligible recipients. There is no income or means testing; eligibility is based solely on a 
residency test. The system is funded through general taxation, with no earmarked social 
security contributions. Public pension entitlements that have been earned in other coun-
tries are, however, deducted from the total benefit.
The level of this flat-rate public pension (sometimes called a demogrant in other 
settings) is indexed to prices. However, the benefit level is subject to a floor and ceiling 
that are linked to movement in wages. For a couple, current legislation requires that the 
net-of-tax value of the benefit as of April 1 is not less than 65 percent and not more than 
72.5 percent of a net-of-tax surveyed weekly earnings measure.
3
 The net-of-tax rates for 
single people are set at 65 percent (living alone) and 60 percent (sharing accommodation) 
of the rate for couples. If movements in prices remain consistently below movements in 
the net-of-tax surveyed weekly earnings, net-of-tax surveyed weekly earnings effectively 
become the index. The public pension is subject to personal income tax.
The  OECD Factbook 2011 indicates that public pension spending in New Zea-
land was 4.3 percent of gross domestic product (GDP) in 2007, the sixth lowest 
among Organisation for Economic Co-operation and Development (OECD) countries 
(OECD 2011a). This figure does not include the contributions of about 2 percent of 
GDP made to a prefunding arrangement, the New Zealand Superannuation Fund 
(http://www.nzsuperfund.co.nz/), established in 2000 to smooth the emerging increase 
in the cost of public pensions that will result from the aging of the population. Contribu-
tions to this reserve fund come out of government fiscal surpluses. Its size in April 2012 
was $NZ 19.46 billion.
4
 It is projected to peak at about 40 percent of GDP in 2035, when 
draw-downs are expected to begin. The time-weighted annualized return on investment 
since 2000 has been 7.65 percent before tax, some 2.34 percent higher than the return 
on 90-day treasury bills. The current government ceased making contributions in 2010, 
however, citing the increase in public debt; it does not plan to resume payments until it is 
running fiscal surpluses.
Tax claw-backs from New Zealand Superannuation—resulting from taxation of ben-
efits as ordinary personal income and a broad-based consumption tax—are relatively high 
(about 28 percent of the value of benefits). Hence, the effective fiscal cost of the system is 
considerably less than the value of the gross benefit payments. The deduction of public pen-
sion entitlements accrued in other countries also reduces expenditure, by some 2 percent.
OUTCOMES
New Zealand has the lowest relative poverty rate among people age 65 and over in OECD 
countries: only 1.5 percent of people this age have incomes that are less than half the 

106 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
median income. New Zealand has more people age 65 and over between 50 and 60 per-
cent median income than in any other OECD country, however. It has the highest mini-
mum pension among OECD countries (39 percent of average earnings). But together 
with the United Kingdom, the net pension replacement rate for an average worker is the 
second lowest (after Ireland) in the OECD (OECD 2011b).
As a share of the population, the average income of people in New Zealand over 
age 65 is 68 percent, the second lowest in the OECD (after Australia). Although inequal-
ity among people of working age is the eighth highest in the OECD, among people age 65 
and older it is the seventh lowest.
The driver of all these results is the high dependence on New Zealand Superan-
nuation as the primary source of retirement income. Some public servants have govern-
ment defined benefit pensions, but this scheme was closed to new entrants in 1992. New 
Zealanders have varying degrees of capital assets, including high (but declining) rates of 
homeownership, bank deposits, and other financial assets.
Because most assets held by older New Zealanders are not formally classified as 
retirement savings, they have proved difficult to quantify. Social surveys have suggested 
hardship levels of about 6–8 percent among older New Zealanders, associated mostly with 
living in rental housing. Supplementary means-tested disability allowances and accom-
modation supplements are available through the welfare system, as they are for all New 
Zealanders.
VIABILITY OF NEW ZEALAND SUPERANNUATION
The main issue faced by New Zealand Superannuation relates to the sustainability of its 
financing. The increasing ratio of the number of people receiving pensions to the number 
of people in the labor force will put pressure on the financing of New Zealand Superan-
nuation, whose principal source of support is taxation of labor income. The New Zealand 
Superannuation Fund will smooth but not fully prefund tax expenditure, and contribu-
tions to the fund are currently frozen.
In its Long-Term Fiscal Report (New Zealand Treasury 2009), the Treasury consis-
tently draws attention to steadily rising costs, recommending increases in the age of enti-
tlement, changes in indexing, and/or the introduction of means testing. A paper presented 
at the 2010 New Zealand Retirement Income Conference (Rashbrooke 2010) proposed 
establishing a fixed-rate pension contribution and maintaining it through progressive 
increases in the age of entitlement and revenue-targeted means testing. The current gov-
ernment holds that no changes need to be made and that the pension remains sustainable 
within the time frame over which it is considering policy.
KiwiSaver
DEVELOPMENTS LEADING TO ADOPTION OF THE SCHEME
The KiwiSaver system emerged as a result of a sequence of efforts to expand supplementary 
private pension coverage to achieve income replacement rates that would enable the broad 
population to sustain its standard of living in retirement. Private provision for retirement 
had followed a pattern in New Zealand similar to that in other Anglo-American countries: 

5.  NEW ZEALAND’S EXPERIENCE WITH THE KIWISAVER SCHEME 
107
a mixture of defined benefit and defined contribution arrangements, with tax concessions 
for contributions and the investment accrual. In 1974, a Labour Party government intro-
duced a compulsory accumulation scheme, intended to collect 4 percent of earnings from 
employees and 4 percent from employers. The National Party government elected in 1975 
disbanded it, strengthening the tax-funded state pension instead.
Some lack of control over eligibility for the concessional tax treatment for the vol-
untary regime still in place had become a concern in the 1980s. Rather than plugging the 
gaps, in the course of introducing a number of market-led reforms, a reforming Labour 
government instead took the view that retirement saving was no different from any other 
forms of saving and removed all tax concessions in 1989.
In 1992, the government ordered a review to determine whether compulsory retire-
ment saving, tax concessions, or both should be reconsidered. The review rejected both 
options but recommended that greater resources be put into educating people about the 
need to save for retirement. In response, the government established the position of Retire-
ment Commissioner (now Commissioner for Financial Literacy and Retirement Income).
In 1998, as part of a coalition agreement, a referendum was held on whether to 
introduce a compulsory retirement scheme. The scheme put forward was positioned as a 
replacement of the existing tax-funded state pension. It received only minimal support.
In the mid-2000s, a Labour government set up a working party to consider whether 
New Zealanders had a full range of options for retirement saving. The working party 
assessed the conventional wisdom that deductions from salary were a relatively painless 
method of forgoing immediate consumption to save for retirement. Its report provided 
the impetus for KiwiSaver development (Harris 2004).
An early policy decision was to use the pay as you earn (PAYE) tax system as the 
collection mechanism for employees. Doing so placed Inland Revenue in the lead role as 
administrator. Policy development was shared by the Treasury, the Ministry of Economic 
Development (which had regulation of financial services as one of its responsibilities), and 
the Department of Building and Housing (as there was to be a homeownership compo-
nent). The government acknowledged the need for adequate funding for evaluation, and 
Inland Revenue was tasked with conducting it.
In order to identify issues for evaluation, and to confirm the direction of policy 
development generally, Inland Revenue commissioned a literature review from the Urban 
Institute (Toder and Khitatrakun 2006). This comprehensive report set out various 
insights from developments in behavioral economics, as well as other material generally 
supportive of the KiwiSaver concept. Although the arguments for KiwiSaver were not uni-
versally accepted, they were sufficiently persuasive for the government to enact legislation 
establishing KiwiSaver in 2006, with a commencement date of June 1, 2007.
FEATURES OF THE SCHEME
The KiwiSaver system was intended to provide a supplementary source of retirement 
income that would extend the basic New Zealand Superannuation benefit in a manner 
that would achieve a high level of participation without fully mandating coverage. The 
objective of KiwiSaver, as set out in its governing legislation, is “to encourage a savings 
habit and asset accumulation among individuals who may not be in a position to enjoy 
standards of living in retirement similar to those in preretirement.” A secondary goal is 

108 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
to support local capital markets by providing an additional source of investment funds. 
The principal design feature of the system is automatic enrollment; other aspects of the 
design include specified contribution rates, incentives, and provider provisions. KiwiSaver 
is explicitly not guaranteed by the government.
Automatic Enrollment
On taking up a new job, all new employees are automatically enrolled in a KiwiSaver 
scheme; they must opt out if they do not wish to remain a member. The opt-out period 
is two to eight weeks. If new employees do not opt out within their first eight weeks, they 
are automatically enrolled. Some safeguards allow later opt-out in specific circumstances, 
such as not being given an information pack or investment statement in a timely fashion. 
People aged under 18 or over 65 are excluded from the automatic enrollment procedures. 
Anyone under age 65, including children under 18, can, however, voluntarily opt in to 
join a KiwiSaver scheme. The age at which withdrawals of account balances are permitted 
is age 65 or five years after joining the scheme, whichever comes later.
Members can also withdraw their balances upon permanent emigration, death, seri-
ous illness, or significant financial hardship. In addition, all KiwiSaver members who have 
been making contributions from their pay for 12 months or more are permitted to elect 
to take a contribution holiday without providing a reason. A contribution holiday can last 
up to five years. There is no limit on the number of times a contribution holiday can be 
taken, and it can be renewed at any time. While a member is taking a contribution holi-
day, the employer is not required to make compulsory employer contributions; individu-
als can still make voluntary contributions.
Required Contribution Rates
Employee contributions are deducted from pay at the rate of 2 percent, 4 percent, or 
8 percent of earnings, as the employee chooses.
5
 Originally, the default contribution rate 
was set at 4 percent; it was reduced in 2009 to 2 percent, although it is scheduled to 
increase to 3 percent on April 1, 2013. The employer must contribute at least 2 percent of 
earnings, which will also increase to 3 percent in 2013.
Employee and employer contributions are collected through the income tax sys-
tem and reconciled monthly. Inland Revenue passes on the contributions to providers. 
It currently takes about three months for a KiwiSaver contribution to reach a KiwiSaver 
account, but interest is paid while the contributions are held.
People who are self-employed or not working arrange directly with their KiwiSaver 
provider how much to contribute; they make payments directly to the financial institution 
managing the account. Members who are working who want to make voluntary contribu-
tions while on a contribution holiday from their wage or salary deductions can do like-
wise, making the minimum contribution required for the member tax credit (described 
below) if they wish. (This design feature does not appear to have been intentional.)
Financial Incentives 
The range of financial incentives has changed somewhat even within the short time 
KiwiSaver has been in existence (for more information, see http://www.kiwisaver.govt.nz/):

5.  NEW ZEALAND’S EXPERIENCE WITH THE KIWISAVER SCHEME 
109
 
• $NZ 1,000 “kick start.” The government kick starts accounts with a tax-free 
contribution of $NZ 1,000. Originally, this amount together with a $NZ 40 a 
year fee subsidy was intended to be the sole incentive, but others were added just 
before the 2007 launch.
 
• Member tax credit. From the inception of the system through July 1, 2011, the 
government matched individual contributions made by people between age 18 
and the earliest retirement age (age 65 or five years after joining) by up to $NZ 
1,043 a year ($NZ 20 a week). This amount has been reduced to a 50 percent 
match on the first $NZ 20 of contribution, capping the value of the match at 
$NZ 521 a year.
 
• Compulsory employer contributions. For contributing member employees, 
employers are required to contribute an amount equal to at least 2 percent of 
earnings, although they may contribute more. This minimum employer contri-
bution will increase to 3 percent on April 1, 2013.
 
•