The Projected Development of the Gender Pension Gap in Switzerland: Introducing MIDAS_CH

The Gender Pension Gap (GPG) measures the relative difference between the retirement income of both women and men. In a general form, the GPG(l, x) can be written as $1-\frac{{l\left(x\right)}_f}{{l\left(x\right)}_m}$ ; usually l is the mean of the variable of interest, x, e.g. gross pension income, including survivors’ pensions of the 65+, though l(.) can be any measure of location. The GPG therefore reflects how much the pension benefit of women lags behind that of men.

The GPG is usually much higher than the gender pay gap,¹ since the latter is a view at any moment in time, whereas the former is the result of how gender differences in in the prevalence of part-time work, unemployment, withdrawals from the labour market, and the gender pay gap cumulate over the working lives of men and women (; Bettio et al., 2013; Jolly, 2014, 37 and 50). These differences in earnings may be related to other gendered behaviour, such as the impact of parental leave on wages after return (e.g. Lequien, 2012; Thévenon and Solaz, 2013), sectoral segregation, and so forth. However, since pension insurance systems include redistributive elements, the relationship between lifetime earnings and pension income after retirement is far from linear (Kirn and Dekkers, 2022). Thus, the impact of gender-specific aspects on pension income depends in part on the design of the pension system.

In Switzerland, the GPG for the 65+ group fluctuates between 40.2% in 2008 and 30.9% in 2018 (European Commission, 2021a) and averages 33.6% between 2007 and 2019. Using data from SESAM,² Fluder et al. (2016) find that women’s total pensions are 37 percent- or nearly 20,000 Swiss francs (henceforth CHF) per year lower than men’s pensions, which is roughly €20,510.³ However, as we will see in more detail later in this paper, the GPG varies dramatically with the pillar of the Swiss pension system.⁴ The 1^st pillar pensions have a negligible GPG of just under 3 percent while 2^nd pillar pensions have a GPG that exceeds 60 percent. The differences in the GPG of the respective pillars are due to the redistributive elements of the 1^st pillar, which predominantly benefit women. In the 2^nd and 3^rd pillars, where pension entitlements are directly linked to contributions and thus to income, the differences are large and reflect gender differences in lifetime income. Kuhn (2020) reports that women’s pension entitlements in the 1^st pillar are higher than men’s despite their lower lifetime income. However, men’s 2^nd pillar entitlements are more than twice as high as women’s, as are 3^rd pillar entitlements.

Measures of the current GPG, such as Fluder et al. (2016), Kuhn (2020), Kuhn et al. (2021) and the European Commission (2021), are retrospective since the gender pension gap observed among today’s retirees is shaped by labour market behaviour over the past decades. Since labour market participation of women has increased from 68% in 1991 to 79% in 2014 and in addition, part time work rates among active women have increased (Kuhn and Ravazzini, 2017), the question arises how the changing labour market participation of women will affect the GPG in the future.

Measures of future GPG simulated with static microsimulation models are valuable for assessing specific aspects, such as the impact of informal care on future pension income (Kirn and Thierbach, 2020), but do not incorporate dynamic aspects, such as changes in labour market participation at the extensive and intensive margins or migration.

This paper will present a newly-developed dynamic microsimulation (henceforth DMSM) model for Switzerland and use it to make projections of the Gender Pension Gap in 2070 for the cohort entering the labour market in 2018, as well as for immigrants entering and emigrants leaving the labour market afterwards. These projections will be as consistent as possible with official demographic and economic projections produced by various Swiss institutions, including the Swiss Federal Statistical Office (FSO). Furthermore, we will study how key parameters (such as the conversion rate) and the effect of these exogenous projections change the simulation results. This will allow us to assess how the GPG would develop if the statutory retirement age of women would increase, or if the conversion rate of the 2^nd pillar pension income would be reduced.

Our study follows Bonnet et al. (2006), who use a DMSM to study the effects of pension reform on the GPG for a cohort of men and women in France. Their analysis is however limited to private-sector pensions, whereas this study includes both the 1^st and 2^nd pillars of the Swiss pension system. Halvorsen and Pedersen (2019) also use a microsimulation model to study the projected GPG of one cohort for the reformed Norwegian pension system. Kirn and Thierbach (2020)

Our study follows Bonnet et al. (2006), who use a DMSM to study the effects of pension reform on the GPG for a cohort of men and women in France. Their analysis is however limited to private-sector pensions, whereas this study includes both the 1^st and 2^nd pillars of the Swiss pension system. Halvorsen and Pedersen (2019) also use a microsimulation model to study the projected GPG of one cohort for the reformed Norwegian pension system Kirn and Thierbach (2020).

The study by Dekkers et al. (2022) is closely related to this paper. It presents a more general view on the same Swiss simulation results, however in a comparative context with simulation results from Belgium, Luxembourg, Slovenia and Portugal. This paper is complementary to theirs in that it focusses on the Swiss case, and presents the microsimulation model that was developed to this aim.

We contribute to this literature as we are the first to produce projected GPGs of the Swiss Pension system, which provides a gender-specific perspective on how the design of the Swiss pension system might affect within- and between-gender inequalities of pension income in a context of net-immigration and demographic ageing. Furthermore, we extend the analysis from the gap measured at the mean, to gaps in the lower and upper ends of the income distribution. Our analysis concentrates on one birth-cohort – individuals born in 2000 – and we focus on two perspectives. First, we project the future GPG by mapping our own projections about the future labour market participation at the extensive and intensive margin. In a second step, we simulate the impact of the most recent pension reform, which increased the statutory retirement age (SRA) of women from 64 to 65. Third, we simulate the impact of a reduction of the conversion rate in the 2^nd pillar - which is a quite likely policy scenario – and analyse its impact on the future GPG.

An important caveat is that our analysis is limited to the 1^st pillar of the Swiss pension system and the mandatory part of the 2^nd pillar. This means that the voluntary and more defined-contribution parts of the pension system are not included. The reason for this is the theoretical and empirical difficulty in the modelling of private savings (Tedeschi et al., 2013) and the trade-off with social security in dynamic microsimulation modelling (Rein and Turner, 2001). This would bring us too far from the ambitions of the paper, but it also means that we probably underestimate the GPG in both 2018 and in our projections.

The remainder of the article is organized as follows. The next section introduces the dynamic microsimulation model MIDAS_CH and puts it in the classification of existing models. This discussion will include the data used as the starting dataset as well as the auxiliary information and the alignment methods by which this is done. Section 3 describes in more detail the simulation of lifetime events in the field of demographics (birth, death, migration, etc.), labour market states, retirement and disability, as well as immigration and emigration. Section 4 provides information about the Swiss pension system and the simulation of pension income of 1^st and 2^nd pillar pension income. Finally, section 5 will present the simulation results of the model and discuss various projected gender pension gaps, in the base variant and the policy scenarios. Finally, we draw a conclusion and provide an outlook to the next steps in the development of the model.

MIDAS_CH (which is an acronym for ‘Microsimulation for the Development of Adequacy and Sustainability’) is a discrete-time DMSM using dynamic cross-sectional ageing. It is developed to simulate the distribution of pension income, and its underlying processes, including labour supply, migration, family formation and dissolution, and many others. As such, the model creates a synthetic panel dataset out of a cross-sectional observed starting dataset, by simulating the life spans of the existing individuals and their households while adding new-borns and immigrants, and removing emigrants and deceased individuals. Older – and further developed – models of the same group include MIDAS_BE in Belgium, MOSART in Norway, SESIM in Sweden, DESTINIE in France, T-DYMM in Italy and CBOLT in the US (see Li et al., 2014, for a discussion).

MIDAS_CH belongs to the MIDAS group of DMSM, which also includes models in Belgium, Luxembourg, Portugal and Hungary. It has taken the MIDAS_BE model as an example and source of inspiration (Dekkers et al., 2009; Dekkers and Van Den Bosch, 2016) and runs on LIAM2 (De Menten et al., 2014). In doing so, MIDAS_CH has benefitted considerably from the years of knowledge and experience of other users and developers of models in the MIDAS group (Bryon et al., 2018; Bryon et al., 2018; Dekkers et al., 2022; Dekkers and Van den Bosch, 2021; Li et al., 2014; Liégeois, 2021 Moreira and Wall, 2021Moreira and Wall, 2021), as have the other models in this list.

MIDAS_CH is set up as a discrete-time, cross-sectional ageing model, thereby allowing for the modelling of interactions between individuals, as between partners. This is relevant for the modelling of survivors’ benefits.

As the model follows a closed population approach, partners in the family formation process are selected sporadically from individuals in the dataset. Furthermore, following Dekkers (2015), individuals are cloned in the case of immigration, and removed from the sample in the case of death or emigration. Finally, discrete time models are easier to implement in stepwise computer simulations and often require fewer variables than continuous-time models (Li et al., 2014). But the main advantage of the combination of cross-sectional aging and discrete time simulation is that it allows to define exogenous constraints on the number of objects or events within each time unit, which assures, that the simulation results are consistent with exogenous projections and assumptions (Andreassen et al., 2020; Dekkers et al., 2009). This is explained in the next section.

2.1 Simulation and alignment

The purpose of a DMSM like MIDAS_CH is to produce long-term projections of pensions. As these pensions in some shape or form are the result of previous earnings or labour market participation, or even of the person being present in the country, this requires the simulation of the intertemporal development of a large variety of characteristics, such as demographic ageing, the household size and marital states, activity rates, etc. In its simplest form, this is mostly done using separate discrete-dependent variable models, such as logistic regressions. However, if these estimated models are considered sub-optimal,⁵ or if the researcher wishes to make use of projections that were developed by others in order not to reinvent the wheel and to paint a consistent picture, many models make use of so-called alignment techniques. Various alignment methods have been developed (Morrison, 2006), but Li et al. (2014); Li and O’Donoghue (2014) conclude that, for those simulation models that focus on the impact on distributions, the sidewalk hybrid and the SBDL⁶ methods achieve the best results. However, as sidewalk hybrid methods have a low computing performance, the “alignment by sorting” by the SBDL method is widely used in complex microsimulation models, and MIDAS_CH is no exception.

In addition to the alignment of the states, consistency with exogenous macroeconomic forecasts is achieved through the indexation of earnings and the adjustment of macroeconomic variables. In so doing, the simulated micro-level earnings are adjusted to be equal to macroeconomic wage growth projections. Furthermore, the minimum 1^st pillar pensions and the related thresholds of the 1^st and 2^nd pillar pension are adjusted by the arithmetic mean of the consumer price index and nominal wage index (the so-called mixed index). The short-term forecast of the wage-price index is based on the estimate of SECO (2022). In the absence of official estimates of long-term future wage and price indices, it is assumed that future nominal wage growth rates would correspond to the average annual growth rates of the past twenty years, based on the nominal wage index (FSO, 2022b) and that the real wage growth is equal to the inflation rate (full inflation compensation). Figure 1 illustrates the nominal and real wage indexes based on official statistics (base=1993), as well as the projected nominal and real wage indexes, and the related consumer price index⁷ and mixed index.

Figure 1

Download asset Open asset

Additionally, a range of interest rates are projected, as these affect the central parameters of the 2^nd pillar (see Figure 2). Based on the current inflation trend, it is assumed that the Central Bank rate will rise in the coming years (KOF, 2022), which will also result in an increase in the yield on 10-year federal bonds,⁸ as well as a decline in inflation rates. Accordingly, the expected gross return on pension wealth will increase, as in accordance with Art. 3 of Professional Guideline 4 (FRP 4), the upper limit is calculated as the interest rate of the 10-year federal bonds, increased by a surcharge of 2.5%⁹ and reduced by a discount (at least 0.3% -points) for the increase in longevity. Additionally, the minimum interest rate on accumulated assets of the 2^nd pillar, which is set by the Federal Council according to the recommendation of the Federal Commission for Occupational Pensions, is adjusted according to our own forecasts on the return on federal bonds, as well as the expected gross returns. Hence, the yield on 10-year federal bonds and the expected gross return delimits the interest corridor on accumulated pension assets.

Figure 2

Download asset Open asset

Another key parameter of the 2^nd pillar pension system is the conversion rate, which is the percentage at which an insured person’s retirement assets are converted into a life annuity. Since the conversion rate depends on the interest rate levels and life expectancy, assuming increasing life expectancies and low interest rates, a lower conversion rate is likely in the future. Based on a life expectancy at age of 65, which is projected by the FSO (2022c), the conversion rate is adjusted according to the annuity formula¹⁰ to achieve an implicit interest pledge, which is close to the minimum interest rate on accumulated assets of the 2^nd pillar. Since widows' and orphans' pensions are not taken into account in this simplified approach, the resulting implicit interest rate tends to be underestimated.¹¹ However, since the projected conversion rate is estimated using the minimum interest rate, which represents the lower boundary of the interest corridor on accumulated pension assets, the assumption is plausible. Thus, the estimated future conversion rate is 5.2%, which corresponds to the average conversion rate for the supplementary second pillar pension (currently 5.28%) for retirements in five years (OAK, 2021) seems plausible.

However, since a reduction in the conversion rate leads to a significant reduction of the pension income of the 2^nd pillar (For example, a reduction of the conversion factor from currently 6.8% to 5.2% would reduce pension income by 23.5%),¹² two different scenarios are simulated: 1) the conversion rate remains at today’s level, 2) the conversion rate will be lowered in the future. As a reduction in the conversion rate lowers the relative share of 2^nd pillar pension income in total pension income, we also expect that this affects the simulated GPG, since the GPG is largely based on the gender differences in the 2^nd pension pillar.

Life events, such as becoming a parent, reducing work to care for relatives, unemployment and divorce, all affect pension upbuilding. For example, periods spent upraising and caring for children are credited in the 1^st pillar pension scheme, but not in the other pillars. Other life events, such as unemployment or interruptions and reductions in employment, affect pension income through their impact on earnings. As is standard in a discrete time model, the likelihood that a given event will take place is simulated for every individual in each period.

The simulation of demographic events in MIDAS_CH uses the SDBL method (see section 2.1). To ensure that the simulated population structure stays as close as possible to the population structure projected by the FSO (FSO, 2017; FSO, 2020c), the simulation results are aligned to the official projections. But also other relevant official projections on household formation and dissolution; labour market participation -, unemployment -, and care-giving rates; part-time work rates; earnings growth rates, etc. are being taken into account in the simulations of MIDAS_CH. Table 1 provides an overview of the data sources used for the alignment.

The next sections provide an overview of the modelling of demographic events as well as labour market participation and labour supply.

3.3 Disability and informal care

The disability rate is closely related to age. While in 2020 less than 2% of the total resident population under the age of 30 received a disability pension, this share was 11.4% for men and 9.6% for women shortly before they reached the statutory retirement age (FSO, 2022b). When modelling disability, the probability of becoming disabled is estimated for each individual each year, depending on age, civil state and an interaction variable of age and education, before the alignment by sorting is applied according to the projected prevalence of disability by gender and age group. In the absence of official projections of disability prevalence, projected disability prevalence is assumed to follow the trend of recent years, but to begin to stabilize after 2040.

The next step is to simulate the need for informal care. In 2017, 13% of the population received help from relatives, acquaintances, or neighbours for health reasons. This percentage, which is higher among women, increases sharply from the age of 75. The proportion of people providing unpaid care is highest among those aged 45 to 64 (FSO, 2020a). Hence, the probability of need for care is simulated for each person each year, depending on age, an interaction between age and education, gender and disability as arguments. This logit is combined by an alignment by sorting that is applied according to the projected need for care by gender and age group. In the absence of official projections about the need for care, projected need for care is assumed to follow the trend of recent years, but to begin to stabilize after 2040.

Based on the need for informal care by family members, the provision of informal care is simulated as individuals providing informal care to relatives are granted care credits. When modelling the provision of informal care, the probability of provision is estimated for each person each year, depending on the need for informal care from relatives (spouse, children, parents, and in-laws), before adjustment through sorting by projected prevalence of informal care by sex and age group. In the absence of official projections of informal caregiving, the projected incidence of informal caregiving is assumed to follow the trend of recent years but to begin to stabilize after 2040.

Figure 3 shows the need (patterned bars) and provision (solid bars) of informal care as a fraction of the total population for the simulated year 2025 by gender. The comparison across age groups shows that for men, the informal care provided as well as the informal care needed is relatively stable over the life cycle. For women, on the other hand, there are clear fluctuations over the life course. The need for informal care is higher in younger and older age groups, and the amount of informal care provided is highest in midlife.

Figure 3

Download asset Open asset

3.4. Labour market participation and labour supply

The modelling of labour market participation is a key element in MIDAS_CH, as interruptions and reductions in employment significantly determine pension income after retirement. Since the goal of our model is to simulate the GPG, the employment- and part-time work rate, especially of women is of particular importance. As in most European countries, the majority of men in Switzerland work full time. However, for women, the picture is a bit more complex. In the EU28, the employment rate of women on averaged 64.1% in 2019. In Switzerland in the same year, 76.3% of women were employed, making it the second highest employment rate for women compared to the European countries – only surpassed by Iceland (81.9%; FSO, 2020b). However, the labour market participation of mothers is much lower than that of childless women (Ravazzini, 2018). Switzerland also ranks second in terms of part-time employment among women. In 2019, 62.7% of employed women in Switzerland worked part-time according to the European definition¹⁸ (2010: 60.8%) – a percentage only surpassed by the Netherlands (FSO, 2020b).

The conclusion is that although the overall employment rate of women in Switzerland is high, it is often at a reduced employment level. To fully capture the heterogeneity of work intensity, we model it in three steps. First, whether a man or woman works or not is modelled by combining the aforementioned alignment information (see Table 1) with the results of a logistic regression model, with arguments age, civil state and previous work status (employed, unemployed, other inactive). Next, given that a person works, we use the results of a logistic regression model (referring to gender, number of children below age 11 and age 15, civil status, age, and education level) and the alignment information to simulate whether the person works part-time or full-time. The third and final step then models the extent of part-time work for those that are selected to work part-time in the first step. Here we distinguish three groups i) less than 50% of full time, ii) between 50% and 69% of full-time, and, iii) between 70% and 89% of full time. These three groups match with the dimensions of the projections from the FSO (FSO, 2020c) that we use in our alignment tables.

With regard to labour market participation at the extensive and intensive margins, we extended the projections of the FSO from 2050-2070. Figure 4 illustrates the actual and projected labour market participation of men and women in Switzerland. The activity rate of men is declining mainly because of demographic ageing, and partially due to increasing educational attainment levels that delay labour market entry. These developments are also at play with women, but their impacts are more than compensated by the strong increase in the activity rates of women, especially in the middle aged groups, which is driven by a rising level of education and other factors, such as measures to better reconcile work and family life, interest in careers and the need to contribute to household income (FSO, 2020c). It is assumed that the labour market participation of men and women will further converge and stabilise (see Dekkers et al., 2022, for a further discussion).

Figure 4

Download asset Open asset

In 2019, around 18% of men worked part-time, compared with 60% of women. For this reason, women have significantly lower FTE employment rates than men, especially those of family age. Assuming a better reconciliation of family and work, the FTE employment rate of women will increase significantly by 2050 and a trend towards more part-time employment among men is expected. However, the FTE employment rate of women will still be significantly lower than that of men (FSO, 2020c).

3.6 Immigration and emigration

Switzerland is a migration country, and the MIDAS_CH model therefore has to allow for immigration and emigration. Each year, around 145.000 people immigrate, while around 126.000 people emigrate. The population is around 8 million. In 2019, Switzerland had approximately 17 immigrations and 15 emigrations per 1000 inhabitants, a rate higher than that of other Western European countries of similar sizes, such as Austria (12 / 8 per 1.000), Belgium (13 / 9 per 1.000), and the Netherlands (12 / 6 per 1.000).¹⁹ In 2018, the share of immigrant population in Switzerland equals 29.7%. Apart from micro-states, such as Andorra and Liechtenstein, this is one of the highest percentages in Europe, only surpassed by Luxembourg.

Moreover, the profile of immigrants in Switzerland is somewhat different than in EU countries. According to the Swiss Mobility Migration Survey, immigration to Switzerland is mainly motivated by professional reasons (Steiner and Wanner, 2019). 73% of the immigrants benefited from the “Agreement on the Free Movement of Persons" (AFMP)²⁰ with the EU. The contribution of other types of migrants (1.7% labour migrants from outside the EU, 17% accompanying family members; and 5.5% humanitarian migrants (OECD, 2020) is limited.

According to the projections of the Swiss FSO (FSO, 2020c) both immigration and emigration will increase further in the coming years. As people born during the baby boom in the 1960s gradually reach retirement age, Swiss companies will rely increasingly on foreign labour to fill the gaps (Beerli et al., 2021). Due to the greater mobility of people with a high level of education, who will make up the majority of migrants in the coming years, not only the number of immigrants but also the number of emigrants will increase. In the long term, however, migration would decrease as the European population ages and several countries will experience population decline (FSO, 2020c).

This high proportion of migrants, and the projected developments of immigration and emigration make it necessary for MIDAS_CH to include migration. Furthermore, we decided to model emigration and immigration separately as the age and gender distribution of immigrants and emigrants differs greatly. Finally, we aimed to align our modelling based on the projections of the Swiss FSO (2020c).²¹

3.6.1. Modelling of immigration

With regard to the modelling of immigration, two challenges arise. First, the characteristics of new immigrants are by definition unknown. This includes the panel history and weighting information needed for the shared weights approach (Schonlau et al., 2013). This can be resolved by the donor approach (Duleep and Dowhan, 2008) by which existing individuals in the dataset are selected to represent prospective immigrants. This donor sample is then cloned and added to the existing dataset. An additional problem is that immigrants often enter the country as households while the alignment tables are expressed in numbers of individuals. We therefore follow Chènard (2000). See Dekkers (2015), O’Donoghue et al. (2010) for a discussion on modelling this in LIAM2.

Since the sociodemographic characteristics of immigrant differ by country of origin (FSO, 2019), we have grouped all individuals in the model into three areas, based on country of birth: Switzerland (returning emigrants), the European Economic Area, and other countries. The alignment procedure is based on the projections of the FSO (FSO, 2020c, Scenario A-00-2020), which are made for these three groups. Figure 5 illustrates the age distribution of immigrants by the three regions of origin. Figure 5 compares the projections (FSO, 2020c) and the results of the simulation.

Figure 5

Download asset Open asset

3.6.2. Modelling of emigration

Modelling of emigration is broadly analogous to the modelling of immigration. However, in the case of emigration, households are not cloned but removed from the dataset. Nevertheless, the households that emigrate must be selected, and the total number of emigrating persons must correspond to the projections of the Swiss Statistical office (FSO, 2020c). To identify the households with the highest probability of emigration, we use the results from Wanner (2021), model 3) to simulate the individual probabilities of emigration. In a second step, we compute the estimated average probability of emigration at household level. Finally, we use the Pageant algorithm again to select households for emigration, and these are removed. Because individuals with different citizenships may live together in one household, households were categorized according to the most common citizenship within the household.²² The alignment is performed, as in the case of immigration, according to projections of the Swiss Statistical office (FSO, 2020c, Scenario A-00-2020), by gender and age. Analogous to our modelling of immigration, we divided the alignment matrix into three areas (Switzerland, European Economic Area, and other countries), as proposed in the MOSPI Report (2020).

Figure 6 shows the projected (FSO, 2020c) and the simulated number of immigrants and emigrants. The graphs show that the steadier the predicted figures are, the more accurate the results of the alignment process. This is because any deviations from the selection process are carried over to the next period.

Figure 6

Download asset Open asset

The deviation of the simulations from the projection remains however for most of the simulated years within a range of (+/- 5% points), as shown in Figure 7, which depicts the projected number of immigrants (dashed line, +/- 5% confidence interval) and the simulated number of immigrants (solid line).

Figure 7

Download asset Open asset

The Swiss pension system consists of three pillars. The 1^st pillar, which consists of the old-age and survivors' insurance (OASI), the disability insurance (DI) and means-tested supplementary benefits (SB), is a publicly financed pay-as-you-go system. The 2^nd pillar is an employer-based, fully funded occupational pension system while the 3^rd pillar is a voluntary, individual, tax-privileged private provision in addition to the first and second pillars (see Kirn and Baumann, 2021).

Figure 8 illustrates the relationship between annual retirement benefit and the average annual lifetime income, assuming an uninterrupted working career of 44 years and applying a conversion rate equal to 6.8% for the mandatory part of the 2^nd pillar pension income. The 1^st pillar, represented as a dark grey area, aims at providing a basic retirement income. The mandatory part of the 2^nd pillar (light grey area) and the supplementary part of the 2^nd pillar and 3^rd pillar (hatched area) should enable the “continuation of the accustomed way of life”. The three pillars are coordinated in such a way that a replacement rate equal to 60% of the average annual lifetime income is achieved (blue line). By contrast, for wages below CHF 30,072, which is the risk-of-poverty threshold,²³ the benefit target of 60% is already met by the 1^st pillar pension alone. However, the maximum pension benefit one can receive from the 1^st pillar alone (which is CHF 28,680) is itself below the at-risk-of-poverty threshold. In 2019, 16% of the elderly population received a pension income (excluding notional rent) below the risk-of-poverty threshold, which is slightly below the European average²⁴ (FSO, 2021a).

Figure 8

Download asset Open asset

Next we discuss the simulation of 1^st and 2^nd pillar pension income in more detail. The model does not cover the 3^rd pillar, so we will not discuss this further.

4.1 Simulation of 1^st pillar pension income

The 1^st pillar includes old age, survivors and disability benefits (OASI/AHV & DI/IV) as well as supplementary benefits (SB/EL). In old age or in case of death and injury, the OASI/DI takes over the insured person’s livelihood. If the income from pensions is not sufficient to ensure subsistence, means-tested supplementary benefits (SB) cover the necessary living requirements. OASI insurance is compulsory for all residents of Switzerland, including non-working persons, students, people receiving public transfers (such as unemployed and disabled workers), and the self-employed. The system is financed by a proportional contribution on earned income, without a cap.

The OASI pension income depends on the number of eligible contribution years and the ‘relevant’ annual income, which is the sum of revalued incomes (employment incomes, non-employment contributions, split income²⁵ are uprated according to the mixed index)²⁶ and the sum childcare and care credits. Childcare and care credits are warranted to those persons who care for children under 16 years of age²⁷ or relatives in need of care during their contribution years.²⁸ Both credits correspond to three times the minimum annual pension (CHF 3,585 per month, which is about 60% of the median income of women (CHF 6,067 per month) (FSO, 2020b).

Figure 9 illustrates the conversion of earned, split income and credits to ‘relevant’ pension income, which is then used as the basis of the 1^st pillar pension benefit. The conversion of the annual average income to pension income is based on a pension formula, which is linked to a parameter table and indexed by the mixed index.

Figure 9

Download asset Open asset

A full pension is paid to anyone who has paid AHV contributions without interruption every year from the age of 20 up to the statutory retirement age (SRA);²⁹ missing contributions years therefore result in partial pensions. However, since the entire population is compulsorily insured, incomplete contribution careers are rare, except for those immigrants that have not been in Switzerland for their entire working life.

Figure 9 shows that the transfer from annual average income to pension income is highly degressive. Through the use of redistributive parameters such as care credits and income splitting, the benefits depend weakly on income and in a linear way on the number of contribution years.

Besides being highly degressive, the 1^st pillar pension income is subject to both a minimum and a maximum benefit. But here too the Beveridgean character of the pillar reveals itself: the former (CHF 14,340 or €14,704 per year) is half of the latter (CHF 28,680 per year), while the maximum benefit remains below the risk-of-poverty threshold (CHF 30,072 per year) (see Figure 8). Finally, the pension of a married couple is capped at 150% of the maximum individual pension. If a married couple decides to divorce, all income earned during the spouses' marriage is divided equally between both parties. In case of death of a beneficiary, the surviving partner gets a benefit equal to at most 80 percent of the deceased spouse’s 1^st pillar pension. If someone can claim a 1^st pillar or disability pension at the same time as the widow’s or widower’s pension, only the higher pension will be paid.

As an application of the new MIDAS_CH model, we simulate the income trajectory of individuals over their entire lifespan and the resulting pension income and analyse the Gender Pension Gap (GPG). The gender pension gap is usually much higher than the gender pay gap, since the income differences between the sexes and gendered behaviour (prevalence of part-time work and leaves due to care obligations) add up. As however the redistributive elements within pension systems are designed to compensate for losses in earnings due to non-standard working careers and/or non-standard family arrangements (Möhring, 2021), the relationship between the earnings gap and differences in labour market participation on the GPG is not linear (Kirn and Dekkers, 2022).

According to the Eurostat definition of the GPG, pension income includes gross retirement pensions, gross survival pensions as well as the means-tested supplementary benefits for the elderly. People with zero pensions, as well as those below age 65 are excluded from the calculation. In addition to this general definition of the GPG, there are numerous variants that can be distinguished based on Dekkers’ four dimensions (Dekkers et al., 2022). First, the underlying definition of pension income can be restricted on retirement pensions (covering all pillars), both old-age and survivors’ pensions, and exclude the means-tested “guaranteed minimum income” (denoted SB) for the elderly. As we focus on pension income, we report the GPG based solely on retirement income, excluding SB.

Second, the standard definition of the GPG excludes the 65+ without a pension and is based on the assumption that people who do not have a retirement benefit are not retired. However, it may in some cases be relevant to construct a variation of the GPG that extends the standard definition of the GPG by including zero pensions. The latter GPG can then be interpreted as a combination of the standard GPG and the gender pension coverage gap, which measures the extent to which women have their own independent access to pension system benefits (European Commission, 2018). It is of course not very relevant for the pension benefit of the 1^st pillar only, where the coverage gap is close to zero because persons who are not gainfully employed must also contribute to the OASI. But this is in contrast to the 2^nd pillar, where the overall coverage rate is 49% in 2019 (Kuhn, 2020), and we therefore can expect important gender differences. Hence, comparing the standard GPG with that including zero-values is still very relevant for the pension income from both pillars together. In doing so, we include zero 2^nd pillar pensions in our analysis to compute the GPG.

Third, the standard GPG compares the mean gross benefits of men and women. However, the GPG can also be calculated by using other measures of location, such as percentiles or quartiles. In this article, we will present the GPG at the mean, the 25th and 75^th percentile. Furthermore, we measure pension inequality by the P75/P25 ratio. To analyse the development of the GPG, we compare the simulated P75/P25 ratios to the 2019 statistics.

Fourth, the GPG can be derived separately for age groups. As in other studies (Bonnet et al. (2006); Halvorsen and Pedersen (2019)), we focus on one birth-cohort of individuals born in 2000. This cohort will reach the statutory retirement age of 65 in 2065. The dynamic simulation covers the entire span of the contribution obligation, i.e. 44 years for the 1^st pillar and 40 years for the 2^nd pillar. At the simulation horizon in 2070, we simulate pensions of those that retire at the SRA as well as early retirement (two years before SRA) and retirement five years later, since retirement at a later point in time is also possible. In the standard measure of the GPG we compute the GPG at the SRA.

Although the selected simulation horizon for the 2000 birth cohort covers the entire contribution period, it does not automatically follow that the life histories analysed have no contribution gaps. As explained earlier, migration at an age older than 21 leads to gaps in contributions, even in the 1^st pillar of the Swiss pension system. For this reason, we analyse the pension income of immigrants separately.

In the next section, we analyse the current GPG at SRA. To analyse the impact of the most recent pension reform – the increase in the SRA of women from 64 to 65 – we simulate the pension incomes and compare the GPGs in 2070 (policy scenario I and II). Next, we simulate a reduction of the conversion rate of the mandatory part of the 2^nd pillar – which is a likely future policy scenario – and analyse its potential impact on the GPG in 2070 (policy scenario III).

5.2 Policy scenario I: SRA of women remains at 64 – GPG in 2070

The AHV 21 reform, which among other things provides for an increase in the retirement age for women from 64 to 65, was adopted in September 2022. To analyse the impact of an increase in retirement age, we compare the simulated 1^st pillar pension with a retirement age of 64 for women (scenario I) to the pension income with a retirement age of 65 for women (scenario II).

The simulated 1^st pillar pension incomes in year 2070 are illustrated by the scatterplot (Figure 10), whereas it is assumed that the SRA is at the age of 65 for both men and women.³⁴ There is a positive correlation between average lifetime income (x-axis) and 1^st pillar pension income (y-axis), but the variance is greater for women, which is due to the redistributive elements that predominantly benefit women. In addition, the upper limit of the pension income of the 1^st pillar (which is twice the minimum pension income) is expected to be 43,174 CHF in 2070, as well as the maximum pension income of married couples, which is expected to be 64,761 CHF.

Figure 10

Download asset Open asset

Table 5 reports the percentiles of the simulated 1^st pillar pension income at SRA in the year 2070. According to the projections, women would continue to receive a lower pension income from the 1^st pillar than men across the entire income distribution in 2070. However, compared to 2019, the gender gap would narrow significantly in 2070 and become marginal at the top 50% income distribution. This effect is mainly due an increase in the number of women reaching a maximum pension. The 1^st pillar pension income at the 25^th percentile of women is 43% above the minimum pension income, for men it is 66%, which indicates that the relative income distribution at the bottom 25% of the income distribution will not change significantly.³⁵ However, with regard to the top 75% of income distribution, we expect that the 1^st pillar pension income at the 75^th percentile of women (men) is 94% (98%) of the maximum pension income, which means that almost all men and women in the top quartile would receive the maximum 1^st pillar pension income, which would however, be below the risk-of-poverty threshold. Compared to 2019, we expect women in the top 75% of income distribution to catch up and the GPG in the top income distribution to become smaller.

Table 5

	mean	p25	p50	p75	p75/p25
total population	37,391	33,614	38,722	42,044	1.25
women	35,682	30,781	37,283	40,672	1.32
men	39,027	35,806	40,070	42,376	1.18
GPG	8.6%	14.0%	7.0%	4.0%
non-immigrants	37,297	33,463	38,680	41,905	1.25
women	35,642	30,711	37,179	40,642	1.32
men	39,131	36,098	40,304	42,470	1.18
GPG	8.9%	14.9%	7.8%	4.3%
immigrants	38,138	34,821	39,055	42,193	1.21
women	36,526	.	.	.	.
men	38,534	35,057	39,276	42,193	1.20
GPG	5.2%	.	.	.

1. Source: own computations. SRA for men 65 years and women 64 years. Quantile incomes of female immigrants are not reported due to the relatively small number of cases in the simulated dataset.

This is also reflected in pension inequality, where the P75/P25 ratio of women (men) increased from 1.02 (1.03). to nearly 1.32 (1.18). Hence we see a greater inequality of 1^st pillar pension income. This effect may be explained by a higher share of women having tertiary educational attrition, which is associated with higher wages, leading to a reduction of the GPG in the upper income range. However, since labour market participation as well as the extent of part-time employment are positively correlated with the level of education, this leads to a larger variance in women’s lifetime earnings. This finding is in line with Kuhn and Ravazzini (2017) who observe that women in couples are more likely to work if they earn higher wages, however, working women with higher wages tend to work slightly less than working women with lower wages.

The simulation results suggest that pension income of male immigrants would still be slightly below the pension income of male non-immigrants across the entire income distribution. However, the pension income of female immigrants would be at the mean slightly above the pension income of female non-immigrants. This result may be driven by the on average higher labour market participation of female immigrants in comparison with female non-immigrants.

5.4 Policy scenario III: Reduction of conversion rate – GPG in 2070

Up to now, the simulation results have described the reference scenario, which assumes a constant conversion rate at 6.8%. However, the conversion rate in practice is an ad hoc function of life expectancy among other things, so we can expect that it will decrease over time. Hence, we developed a first alternative scenario where we assume that the conversion rate would gradually decline from 6.8 to 5.2%.

Figure 11 illustrates the simulated the simulated 2^nd pillar pension income (mandatory part) at SRA, where a conversion rate of 5.2% is assumed, given the increased life expectancy and low interest rate expectations. When interpreting the results, it should be taken into account, that a reduction of the conversion factor from 6.8% (status quo) to 5.2% (as assumed for 2070) will lead to a reduction of pension income by 23.5% (1-5.2/6.8=0.235).

Figure 11

Download asset Open asset

Table 9 reports the simulated 2^nd pillar pension income (mandatory part) at SRA, where a conversion rate of 5.2% is assumed, given the increased life expectancy and low interest rate expectations. Since the lower conversion rate reduces the pension income of both men and women equally, the GPG does not change. On average, 2^nd pillar pension income would decrease by 23.5% if the conversion rate were reduced from 6.8% to 5.2%.

Table 9

	mean	p25	p50	p75	p75/p25
total population	23,196	15,380	22,653	31,654	2.06
women	18,102	11,234	18,634	24,552	2.19
men	28,016	20,687	29,847	34,549	1.67
GPG	35.4%	45.7%	37.6%	28.9%
non-immigrants	22,926	15,217	22,316	31,342	2.06
women	18,139	11,177	18,703	24,538	2.20
men	28,180	20,822	29,932	34,631	1.66
GPG	35.6%	46.3%	37.5%	29.1%
immigrants	25,220	16,587	26,866	32,637	1.97
women	17,414	.	.	.	.
men	27,271	18,875	29,417	33,188	1.76
GPG	36.1%	.	.	.

1. Source: own computations. SRA for men and women: 65 years. Second pillar: Mandatory pension, conversion rate: 5.2%.

Table 10 reports the percentiles of the simulated 1^st and 2^nd pillar pension income (mandatory part). According to the simulation results the GPG is expected to decline at the mean from 21.6% to 19.8% if the conversion rate were to be reduced. A reduction in the conversion rate would reduce the GPG more for higher incomes, since the relative share of pension income from the 2^nd pillar is larger for higher incomes. Pension inequality, measured by the P75/25 would also decline.

Table 10

	mean	p25	p50	p75	p75/p25
total population	60,696	50,463	61,552	73,592	1.46
women	53,854	44,037	55,854	64,218	1.46
men	67,169	58,774	70,312	76,465	1.30
GPG	19.8%	25.1%	20.6%	16.0%
non-immigrants	60,303	50,166	61,309	73,124	1.46
women	53,843	43,754	55,931	64,213	1.47
men	67,393	59,131	70,643	76,683	1.30
GPG	20.1%	26.0%	20.8%	16.3%
immigrants	63,643	53,860	65,319	74,649	1.39
women	54,063	.	.	.	.
men	66,159	56,152	68,558	75,304	1.34
GPG	18.3%	.	.	.

1. Source: own computations. SRA for men and women: 65 years. Second pillar: Mandatory pension, conversion rate: 5.2%.

As far as we are aware, MIDAS_CH is the first dynamic cross-sectional microsimulation model for Switzerland. It has been developed to assess the pension adequacy of the 1^st and mandatory part of the 2^nd pillar pension income of the Swiss Pension system. It simulates the life spans of the cohort born in 2018 and migrating individuals with all its underlying demographic and labour market processes, such as labour supply, marriage and divorce, family formation and resolution as well as the need for informal care.

The microsimulation model runs on LIAM2 and has taken the MIDAS_BE model as a source of inspiration and example. As such, it belongs to a class of older -and further developed- models, such as MOSART, SESIM, DESTINIE, T-DYMM. MIDAS_CH is based on the Swiss SILC dataset and aims to ensure consistency with forecasted target proportions. It has an extensive alignment procedure, as well as a dynamic weighting procedure.

With regard to the 1^st pillar pension income at SRA in the year 2070, the simulations suggest that the GPG would decline from 9.7% to 8.7%. At the upper half of the income distribution, the GPG would almost disappear. Furthermore, a higher proportion of women would receive a maximum pension. This trend is also reflected in the GPG of the 1^st and the mandatory part of the 2^nd pillar pension income at SRA. At mean, the GPG would decline only marginal from 23.6% to 21.6%. However, at the top 25% of the income distribution, the gap would narrow significantly and decline from 23.3% to 17.6%. At the bottom 25% of the income distribution, however, the GPG would increase slightly from 23.6% to 27.5%. The catching-up process of women may be driven by two trends: first, compared to earlier cohorts, the share of women with tertiary educational attainment levels would increase, which means that their average wage would increase. This leads to a reduction of the GPG in the upper income range. However, since labour market participation and part-time employment are positively correlated with the level of education, we expect that the variance in women’s lifetime earnings will also increase. Here our research confirms for Switzerland the trends found in other countries (Bonnet et al., 2006; Dekkers et al., 2022; Halvorsen and Pedersen, 2019).

In contrast to the 1^st pillar pension system, which has strong retributive elements, the gender-specific differences are significantly larger in the mandatory 2^nd pillar pension system, which has the character of a DC/DB hybrid. According to our simulations, the coverage rate of the 2^nd pillar for women would increase from about 49% today (Kuhn, 2020) to almost full coverage in 2070. Thus, almost all female pensioners would receive a pension from the 2^nd pillar, although in the first quartile, the pension income of women would remain significantly lower than that of men. However, at the median and in the 75^th percentile, the GPG would decline considerably between 2020 and 2070. The simulated 2^nd pillar pension income of women at the 75^th percentile would be 71% that of men, coming up from 63.6% in 2020.

Since pension income from the 1^st pillar, even if the maximum pension is reached, does not cover the subsistence minimum, pension income from the 2^nd pillar is of key importance for securing subsistence in old age today, and might be even more so in the future. Thus, policy measures should be attempted to reduce the GPG in the lower income range. Mothers, in particular, have a low labour market intensity due to a lack of childcare options and the high cost of childcare (Budig and England, 2001; Gash, 2009; Gutiérrez-Domènech, 2005). This is in line with Ravazzini (2018), who finds that an expansion in childcare places increases the working hours of mothers of 3-year-olds, whereas the impact is stronger for mothers with upper-secondary education. However, an increase in the supply of childcare places does not lead to an increase in the labour supply of low-educated or single mothers, who are the most vulnerable group because of the high costs.

With regard to 2^nd pillar pension income we also analyse the impact of an increasing life-expectancy, which would result in a lower conversion factor within the 2^nd pillar pension framework. We confirm the expectation that this reduction of the conversion factor from currently 6.8% to 5.2% would result in a decline in 2^nd pillar median pension income of around 24%.

The simulation of migration allows us to simulate the future pension income of immigrants. Our simulations suggest that the pension income of immigrants is slightly above that of non-immigrants across the entire income distribution. However, at the bottom of the income distribution the differences are higher than at the top, where they are marginal. This result may be explained by noting, first, that immigrants with low pension incomes are more likely to return to their country of birth, since their comparatively lower retirement income may not cover the high cost of living in Switzerland (Wanner, 2021). As pensions drawn abroad are not included in the distribution of pension income in Switzerland, this may lead to a positive bias at the lower end of the distribution of pension income of immigrants. Second, the higher labour market participation of immigrants in comparison with native-borns, and their high levels of education (e.g. driven by demand in the R&D and financial sectors) result in higher earnings and labour market activity rates (Hercog and Sandoz, 2018; OECD, 2015). These, in turn, increase their pension accumulations and hence their pension benefits.

This first version of MIDAS_CH can be used to project pensions in Switzerland for the 2018 cohort of native-borns as well as immigrants. Further research will be needed to project the intertemporal path of the GPG and identify the impacting factors. Among other things by taking a look back to analyse the employment histories of today’s employed or today’s new retirees. However, this requires the linking e.g. of SILC data with administrative records, as used by Kuhn (2020) or else the use of statistical methods to impute retrospective data for the respondents in the 2018 SILC. In addition, the modelling of the labour supply, educational choices, and health aspects could be refined. Finally, it would be interesting to analyse in more detail the interpersonal as well as the intertemporal redistributive effects of the Swiss pension insurance system.

1. 1
   The dynamic cross-sectional microsimulation model MOSART

   1. L Andreassen
   2. D Fredriksen
   3. HM Gjefsen
   4. E Halvorsen
   5. NM Stølen

   (2020)

   International Journal of Microsimulation 13:97–119.

   https://doi.org/10.34196/ijm.00214

   • Google Scholar
2. 2
   EXPERT FOCUS (Der Schweizer Treuhänder)

   1. R Baumann

   (2016)

   307–315, Die Solidarität in der Beruflichen Vorsorge, EXPERT FOCUS (Der Schweizer Treuhänder), Vol, 5, p.

   • Google Scholar
3. 3
   The Abolition of Immigration Restrictions and the Performance of Firms and Workers: Evidence from Switzerland

   1. A Beerli
   2. J Ruffner
   3. M Siegenthaler
   4. G Peri

   (2021)

   American Economic Review 111:976–1012.

   https://doi.org/10.1257/aer.20181779

   • Google Scholar
4. 4
   The Gender Gap in Pensions in the EU

   1. F Bettio
   2. P Tinios
   3. G Betti

   (2013)
5. 5
   Effects of Pension Reforms on Gender Inequality in France

   1. C Bonnet
   2. S Buffeteau
   3. P Godefroy
   4. D Tash

   (2006)

   Population (English Edition) 61:41–70.

   https://doi.org/10.3917/pope.601.0041

   • Google Scholar
6. 6
   Matchmaker, Matchmaker, Make Me a Match

   1. N Bouffard
   2. R Easther
   3. T Johnson
   4. RJ Morrison
   5. J Vink

   (2001)

   Brazilian Electronic Journal of Economics 4:135–144.

   • Google Scholar
7. 7
   LIAM2 User Guide Release 0.12.0

   1. G Bryon
   2. G Dekkers
   3. G de

   (2018)

   Brussels: Belgian Federal Planning Bureau.

   • Google Scholar
8. 8
   LIAM2 User Guide, version 0.12.0

   1. G Bryon
   2. G Dekkers
   3. G De Menten

   (2018)
9. 9
   The Wage Penalty for Motherhood

   1. MJ Budig
   2. P England

   (2001)

   American Sociological Review 66:204–225.

   https://doi.org/10.2307/2657415

   • Google Scholar
10. 10
    Gender Segregation, Occupational Sorting, and Growth of Wage Disparities Between Women

    1. F Busch

    (2020)

    Demography 57:1063–1088.

    https://doi.org/10.1007/s13524-020-00887-3

    • PubMed
    • Google Scholar
11. 11
    Microsimulation Modelling for Policy Analysis

    1. D Chènard

    (2000)

    Individual alignment and group processing: an application to migration processes in DYNACAN, Microsimulation Modelling for Policy Analysis.

    • Google Scholar
12. 12
    Informal Caregiving, Employment Status and Work Hours of the 50+ Population in Europe

    1. N Ciccarelli
    2. A Van Soest

    (2018)

    De Economist 166:363–396.

    https://doi.org/10.1007/s10645-018-9323-1

    • PubMed
    • Google Scholar
13. 13
    On the Modelling of Immigration and Emigration Using LIAM2

    1. G Dekkers

    (2015)

    Brussels: Belgian Federal Planning Bureau.

    https://doi.org/10.13140/RG.2.1.4373.8967

    • Google Scholar
14. 14
    What Are the Consequences of the AWG-Projections for the Adequacy of Social Security Pensions?

    1. G Dekkers
    2. H Buslei
    3. M Cozzolino
    4. R Desmet
    5. J Geyer
    6. D Hofmann
    7. M Raitano
    8. V Steiner
    9. P Tanda
    10. S Tedeschi
    11. F Verschueren

    (2009)

    CEPS.

    https://doi.org/10.2139/ssrn.2033662

    • Google Scholar
15. 15
    On weights in dynamic-ageing microsimulation models

    1. G Dekkers
    2. R Cumpston

    (2012)

    International Journal of Microsimulation 5:59–65.

    https://doi.org/10.34196/ijm.00072

    • Google Scholar
16. 16
    Applications of Microsimulation Modelling

    1. G Dekkers
    2. K Van Den Bosch

    (2016)

    13–33, Prospective microsimulation of pensions in European Member States, Applications of Microsimulation Modelling, Budapest, Társadalombiztosítási Könyvtár, p.

    • Google Scholar
17. 17
    Projections of the Gender Pension Gap in Belgium Using MIDAS

    1. G Dekkers
    2. K Van den Bosch

    (2021)

    Brussels: Belgian Federal Planning Bureau.

    • Google Scholar
18. 18
    How Do Gendered Labour Market Trends and the Pay Gap Translate into the Projected Gender Pension Gap? A Comparative Analysis of Five Countries with Low, Middle and High GPGs

    1. G Dekkers
    2. K Van den Bosch
    3. M Barslund
    4. T Kirn
    5. N Baumann
    6. N Kump
    7. P Liégeois
    8. A Moreira
    9. N Stropnik

    (2022)

    Social Sciences 11:304.

    https://doi.org/10.3390/socsci11070304

    • Google Scholar
19. 19
    LIAM2: a New Open Source Development Tool for Discrete-Time Dynamic Microsimulation Models

    1. G De Menten
    2. G Dekkers
    3. G Bryon
    4. P Liégeois
    5. C O’Donoghue

    (2014)

    Journal of Artificial Societies and Social Simulation 17:9.

    https://doi.org/10.18564/jasss.2574

    • Google Scholar
20. 20
    Adding immigrants to microsimulation models

    1. HO Duleep
    2. DJ Dowhan

    (2008)

    Social Security Bulletin 68:51–65.

    • Google Scholar
21. 21
    The cost of caregiving: wage loss among caregivers of elderly and disabled adults and children with special needs

    1. A Earle
    2. J Heymann

    (2012)

    Community, Work & Family 15:357–375.

    https://doi.org/10.1080/13668803.2012.674408

    • Google Scholar
22. 22
    What Comes after Caring? The Impact of Family Care on Women’s Employment

    1. U Ehrlich
    2. K Möhring
    3. S Drobnič

    (2020)

    Journal of Family Issues 41:1387–1419.

    https://doi.org/10.1177/0192513X19880934

    • Google Scholar
23. 23
    Do Highly Paid, Highly Skilled Women Experience the Largest Motherhood Penalty?

    1. P England
    2. J Bearak
    3. MJ Budig
    4. MJ Hodges

    (2016)

    American Sociological Review 81:1161–1189.

    https://doi.org/10.1177/0003122416673598

    • Google Scholar
24. 24
    Panel Surveys

    1. LR Ernst

    (1989)

    135–159, Weighting issues for longitudinal household and family estimates, Panel Surveys, New York, Wiley & Sons, p.

    • Google Scholar
25. 25
    Long-Term Care Workforce: Employment and Working Conditions

    1. Eurofound

    (2020)

    Luxembourg: Publications Office of the European Union.

    • Google Scholar
26. 26
    The 2018 pension adequacy report

    1. European Commission

    (2018)

    Current and future income adequacy in old age in the EU.

    • Google Scholar
27. 27
    Long-Term Care Report

    1. European Commission

    (2021)

    Luxembourg: Publications Office of the European Union.

    • Google Scholar
28. 28
    Eurostat Gender Pay Gap; online data code: ILC_PNP13

    1. European Commission

    (2021)

    https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Gender_pay_gap_statistics, Accessed, 3 December 2021.

    • Google Scholar
29. 29
    Combining work and family life: the pension penalty of caring

    1. M Evandrou
    2. K Glaser

    (2003)

    Ageing and Society 23:583–601.

    https://doi.org/10.1017/S0144686X03001314

    • Google Scholar
30. 30
    Gender Pension Gap in Der Schweiz: Geschlechtsspezifische Unterschiede Bei Den Altersrenten: Schlussbericht

    1. R Fluder
    2. R Salzgeber
    3. L von Gunten
    4. D Kessler
    5. R Fankhauser

    (2016)

    BSV.

    https://doi.org/10.24451/arbor.9208

    • Google Scholar
31. 31
    https://www.bfs.admin.ch/bfs/de/home/statistiken/bevoelkerung/geburten-todesfaelle/lebenserwartung.assetdetail.2103067.html

    1. FSO

    (2017)

    Sterbetafeln für die Schweiz 2008/2013.
32. 32
    MikrodatenSILC 2018, Codebook

    1. FSO

    (2018)

    Neuchâtel: Swiss Federal Statistical Office.

    • Google Scholar
33. 33
    Wie geht es den Personen mit Migrationshintergrund in Der Schweiz? Analysen Zur Lebensqualität der Bevölkerung mit Migrationshintergrund 2017

    1. FSO

    (2019)

    Neuchâtel: Swiss Federal Statistical Office.

    • Google Scholar
34. 34
    Schweizerische Gesundheitsbefragung 2017

    1. FSO

    (2020a)

    Neuchâtel: Swiss Federal Statistical Office.

    • Google Scholar
35. 35
    Erhebung Über Die Einkommen Und Lebensbedingungen (SILC)

    1. FSO

    (2020b)

    Swiss Federal Statistical Office.

    • Google Scholar
36. 36
    Szenarien Zur Bevölkerungsentwicklung. Ergebnisse Des Referenzszenarios

    1. FSO

    (2020c)

    Swiss Federal Statistical Office.

    • Google Scholar
37. 37
    Fruchtbarkeit

    1. FSO

    (2020d)

    Neuchâtel: Swiss Federal Statistical Office.

    • Google Scholar
38. 38
    Armutsgefährdung

    1. FSO

    (2021a)

    Neuchâtel: Swiss Federal Statistical Office.

    • Google Scholar
39. 39
    Neurentenstatistik

    1. FSO

    (2021b)

    Neuchâtel: Swiss Federal Statistical Office.

    • Google Scholar
40. 40
    Erwerbsbeteiligung

    1. FSO

    (2021c)

    Neuchâtel: Swiss Federal Statistical Office.

    • Google Scholar
41. 41
    Lohnentwicklung im Jahr 2021: Schweizerischer Lohnindex

    1. FSO

    (2022)

    Lohnentwicklung im Jahr 2021: Schweizerischer Lohnindex, https://www.bfs.admin.ch/bfs/de/home/statistiken/arbeit-erwerb/loehne-erwerbseinkommen-arbeitskosten/lohnindex.gnpdetail.2022-0102.html.

    • Google Scholar
42. 42
    Periodensterbetafeln Für Die Schweiz (1876-2150) Nach Jahr, Geschlecht Und Alter

    1. FSO

    (2022)

    Periodensterbetafeln Für Die Schweiz (1876-2150) Nach Jahr, Geschlecht Und Alter, https://www.pxweb.bfs.admin.ch/pxweb/de/px-x-0102020300_102/-/px-x-0102020300_102.px/.

    • Google Scholar
43. 43
    Lebenslagen Im Wandel: Analysen

    1. HP Galler

    (1987)

    295–317, Zur Längsschnittgewichtung des Sozio-oekonomischen Panels, Lebenslagen Im Wandel: Analysen, Vol, 2, Campus, Frankfurt, p.

    • Google Scholar
44. 44
    Sacrificing Their Careers for Their Families? An Analysis of the Penalty to Motherhood in Europe

    1. V Gash

    (2009)

    Social Indicators Research 93:569–586.

    https://doi.org/10.1007/s11205-008-9429-y

    • Google Scholar
45. 45
    First and second births among immigrants and their descendants in Switzerland

    1. EA Guarin Rojas
    2. L Bernardi
    3. F Schmid

    (2018)

    Demographic Research 38:247–286.

    https://doi.org/10.4054/DemRes.2018.38.11

    • Google Scholar
46. 46
    Optimal age at motherhood. Theoretical and empirical considerations on postponement of maternity in Europe

    1. S Gustafsson

    (2001)

    Journal of Population Economics 14:225–247.

    https://doi.org/10.1007/s001480000051

    • Google Scholar
47. 47
    Employment after motherhood: a European comparison

    1. M Gutiérrez-Domènech

    (2005)

    Labour Economics 12:99–123.

    https://doi.org/10.1016/j.labeco.2004.04.004

    • Google Scholar
48. 48
    Desktop Companion to the German socio-economic panel study (SOEP) Technical report

    1. JP Haisken-DeNew
    2. JR Frick

    (2005)

    German Institute for Economic Research (DIW).

    • Google Scholar
49. 49
    Closing the gender gap in pensions: A microsimulation analysis of the Norwegian NDC pension system

    1. E Halvorsen
    2. AW Pedersen

    (2019)

    Journal of European Social Policy 29:130–143.

    https://doi.org/10.1177/0958928717754296

    • Google Scholar
50. 50
    PSID cross-sectional individual weights, 1997–2009 Technical report

    1. SG Heeringa
    2. PA Berglund
    3. A Khan
    4. S Lee
    5. E Gouskova

    (2011)

    Institute for Social Research, University of Michigan.

    • Google Scholar
51. 51
    The chicken or the egg? Endogeneity in labour market participation of informal carers in England

    1. A Heitmueller

    (2007)

    Journal of Health Economics 26:536–559.

    https://doi.org/10.1016/j.jhealeco.2006.10.005

    • PubMed
    • Google Scholar
52. 52
    The effects of informal care on paid-work participation in Great Britain: a lifecourse perspective

    1. U Henz

    (2004)

    Ageing and Society 24:851–880.

    https://doi.org/10.1017/S0144686X04002351

    • Google Scholar
53. 53
    Selecting the highly skilled: norms and practices of the Swiss admission system for non-EU immigrants

    1. M Hercog
    2. L Sandoz

    (2018)

    Migration Letters 15:503–515.

    https://doi.org/10.33182/ml.v15i4.2

    • Google Scholar
54. 54
    Inequality And Ageing. In Facing the Facts: The Truth about Ageing and Development

    1. R Jolly

    (2014)

    London: Age International.

    • Google Scholar
55. 55
    Project MIGAPE: Work Package 3: Results of the Dynamic Simulations for Switzerland

    1. T Kirn
    2. N Baumann

    (2021)

    Mimeo University of Liechtenstein. Deliverable D3.1/D3.2 of the MIGAPE project.

    • Google Scholar
56. 56
    Paper presented at the 3rd workshop of the Swiss Network on Public Economics (SNOPE)

    1. T Kirn

    (2022)

    KOF ETH Zurich.

    • Google Scholar
57. 57
    Project MIGAPE: Work Package 2: Results of the Standard Simulations for Liechtenstein

    1. T Kirn
    2. K Thierbach

    (2020)

    Mimeo University of Liechtenstein. Deliverable D2.1/D2.2 of the MIGAPE project.

    • Google Scholar
58. 58
    Was Schweizer Wirtschaftsforschende Über Inflation, Geldpolitik Und Den Franken Denken

    1. KOF

    (2022)

    KOF Bulletin.

    • Google Scholar
59. 59
    Augmented wealth in Switzerland: the influence of pension wealth on wealth inequality

    1. U Kuhn

    (2020)

    Swiss Journal of Economics and Statistics 156:19.

    https://doi.org/10.1186/s41937-020-00063-9

    • PubMed
    • Google Scholar
60. 60
    Early retirement as A privilege for the rich? A comparative analysis of Germany and Switzerland

    1. U Kuhn
    2. MM Grabka
    3. C Suter

    (2021)

    Advances in Life Course Research 47:100392.

    • Google Scholar
61. 61
    The Impact of Female Labour Force Participation on Household Income Inequality in Switzerland

    1. U Kuhn
    2. L Ravazzini

    (2017)

    Swiss Journal of Sociology 43:115–136.

    https://doi.org/10.1515/sjs-2017-0006

    • Google Scholar
62. 62
    Longitudinal and cross-sectional weighting of the survey of labour and income dynamics Technical report

    1. S LaRoche

    (2003)

    Statistics Canada.

    • Google Scholar
63. 63
    The Impact of Parental Leave Duration on Later Wages

    1. L Lequien

    (2012)

    Annals of Economics and Statistics 107:267.

    https://doi.org/10.2307/23646579

    • Google Scholar
64. 64
    Projections of the Gender Pension Gap in Luxembourg Using LU-MIDAS 2020

    1. P Liégeois

    (2021)

    Luxembourg Institute for Socio-Economic Research (LISER), Esch-sur-Alzette.

    • Google Scholar
65. 65
    Evaluating Binary Alignment Methods in Microsimulation Models

    1. J Li
    2. C O’Donoghue

    (2014)

    Journal of Artificial Societies and Social Simulation 17:15.

    https://doi.org/10.18564/jasss.2334

    • PubMed
    • Google Scholar
66. 66
    Handbook of Microsimulation Modelling

    1. J Li
    2. C O’Donoghue
    3. G Dekkers

    (2014)

    305–343, Dynamic Microsimulation, Handbook of Microsimulation Modelling, 293. edn.

    • Google Scholar
67. 67
    The consequences of non‐standard working and marital biographies for old age income in Europe: Contrasting the individual and the household perspective

    1. K Möhring

    (2021)

    Social Policy & Administration 55:456–484.

    https://doi.org/10.1111/spol.12720

    • Google Scholar
68. 68
    MIGAPE, Work Package 3

    1. A Moreira
    2. K Wall

    (2021)

    Projections of the Gender Pension Gap in Portugal using DYNAPOR, MIGAPE, Work Package 3, Institute of Social Sciences, Lisbon, University of Lisbon.

    • Google Scholar
69. 69
    Make It so: Event Alignment in Dynamic Microsimulation

    1. R Morrison

    (2006)

    Ottawa: DYNACAN Team.

    • Google Scholar
70. 70
    T-DYMM 3.0 Forecast Model Report, Istituto Nazionale per l’Analisi delle Politiche Pubbliche (INAPP)

    1. MOSPI Report

    (2020)
71. 71
    Bericht Finanzielle Lage Der Vorsorgeeinrichtungen 2020

    1. BV OAK

    (2021)

    Bern: Oberaufsichtskommission Berufliche Vorsorge.

    • Google Scholar
72. 72
    Simulating Migration in the Pensim2 Dynamic Microsimulation Model

    1. C O’Donoghue
    2. H Redway
    3. J Lennon

    (2010)

    International Journal of Microsimulation 3:65–79.

    https://doi.org/10.34196/ijm.00039

    • Google Scholar
73. 73
    Les Indicateurs de l’intégration Des Immigrés 2015

    1. OECD

    (2015)

    Paris: OECD Publishing.

    https://doi.org/10.1787/9789264233799-fr

    • Google Scholar
74. 74
    International Migration Outlook 2020

    1. OECD

    (2020)

    Paris: OECD Publishing.

    https://doi.org/10.1787/ec98f531-en

    • Google Scholar
75. 75
    Excluding Institutionalized Elderly from Surveys: Consequences for Income and Poverty Statistics

    1. H Peeters
    2. A Debels
    3. R Verpoorten

    (2013)

    Social Indicators Research 110:751–769.

    https://doi.org/10.1007/s11205-011-9957-8

    • Google Scholar
76. 76
    Childcare and maternal part-time employment: a natural experiment using Swiss cantons

    1. L Ravazzini

    (2018)

    Swiss Journal of Economics and Statistics 154:15.

    https://doi.org/10.1186/s41937-017-0003-x

    • PubMed
    • Google Scholar
77. 77
    Public-private interactions: Mandatory pensions in Australia, the Netherlands and Switzerland

    1. M Rein
    2. J Turner

    (2001)

    Review of Population and Social Policy 10:107–153.

    • Google Scholar
78. 78
    The implementation of cross-sectional weights in household panel surveys

    1. M Schonlau
    2. M Kroh
    3. N Watson

    (2013)

    Statistics Surveys 7:37–57.

    https://doi.org/10.1214/13-SS104

    • Google Scholar
79. 79
    Konjunkturtendenzen SECO

    1. SECO

    (2022)

    Sommer 2022.
80. 80
    The relationship between women’s work histories and incomes in later life in the UK, US and West Germany

    1. T Sefton
    2. M Evandrou
    3. J Falkingham
    4. A Vlachantoni

    (2011)

    Journal of European Social Policy 21:20–36.

    https://doi.org/10.1177/0958928710380475

    • Google Scholar
81. 81
    Modelling foreign labour inflows using a dynamic microsimulation model of an ageing country-Slovakia

    1. M Štefánik
    2. T Miklošovič

    (2020)

    International Journal of Microsimulation 13:102–113.

    • Google Scholar
82. 82
    Migrants and Expats: The Swiss Migration and Mobility Nexus

    1. I Steiner
    2. P Wanner

    (2019)

    328, Migrants and Expats: The Swiss Migration and Mobility Nexus, Cham, Springer Nature, p, 10.1007/978-3-030-05671-1.

    • Google Scholar
83. 83
    In Migrants and Expats: The Swiss Migration and Mobility Nexus

    1. I Steiner
    2. P Wanner

    (2019)

    103–122, In Migrants and Expats: The Swiss Migration and Mobility Nexus, Cham, Springer, p, 10.1007/978-3-030-05671-1.

    • Google Scholar
84. 84
    Modelling private wealth accumulation and spend-down in the Italian microsimulation model CAPP_DYN: A life-cycle approach

    1. S Tedeschi
    2. E Pisano
    3. C Mazzaferro
    4. M Morciano

    (2013)

    International Journal of Microsimulation 6:76–122.

    https://doi.org/10.34196/ijm.00084

    • Google Scholar
85. 85
    Labour Market Effects of Parental Leave Policies in OECD Countries

    1. O Thévenon
    2. A Solaz

    (2013)

    OECD Social, Employment and Migration Working Papers 141.

    • Google Scholar
86. 86
    Family resources and male– female educational attainment: Sex specific trends for Dutch cohorts (1930– 1984)

    1. M Van
    2. G Kraaykamp
    3. MHJ Wolbers

    (2015)

    Research in Social Stratification and Mobility 40:29–38.

    https://doi.org/10.1016/j.rssm.2015.02.001

    • Google Scholar
87. 87
    Cross-sectional and longitudinal weighting in a rotational household panel: applications to EU-SILC

    1. V Verma
    2. G Betti
    3. G Ghellini

    (2007)

    Statistics in Transition 8:5–50.

    • Google Scholar
88. 88
    Can Migrants’ Emigration Intentions Predict Their Actual Behaviors? Evidence from a Swiss Survey

    1. P Wanner

    (2021)

    Journal of International Migration and Integration 22:1151–1179.

    https://doi.org/10.1007/s12134-020-00798-7

    • Google Scholar
89. 89
    Longitudinal and cross-sectional weighting methodology for the HILDA survey Technical report

    1. N Watson

    (2012)

    Longitudinal and cross-sectional weighting methodology for the HILDA survey Technical report, University of Melbourne.

    • Google Scholar
90. 90
    The World Bank pension conceptual framework. World Bank pension reform primer No 45728

    1. World Bank

    (2010)

    https://documents.worldbank.org/en/publication/documents-reports/documentdetail/389011468314712045/the-world-bank-pension-conceptual-framework, Accessed, 27 July 2022.

    • Google Scholar
91. 91
    1 CHF to Eur: convert Swiss Francs to Euros

    1. Xe Currency Converter

    (2022)

    https://www.xe.com/currencyconverter/convert/?Amount=1&From=CHF&To=EUR, Accessed, 27 July 2022.

    • Google Scholar

Author details

1. Tanja Kirn

   University of Liechtenstein, Liechtenstein, and Albert-Ludwigs-Universität Freiburg, Freiburg, Germany

   For correspondence

   tanja.kirn@uni.li

2. Gijs Dekkers

   Federal Planning Bureau and CESO, KU Leuven, Leuven, Belgium