fedea

Fundación de Estudios de Economía Aplicada

DYPES: A Microsimulation model for the Spanish retirement pension system by 六 F. J. Fernández-Díaz ** C. Patxot *** G. Souto

Documento de Trabajo 2013-06

September 2013

This research received institutional support from the Spanish Science and Technology System (Projects Nº ECO2012-37572 and ECO2012-35054), the Catalan Government Science network (Projects Nº SGR2009-600 and SGR2009-359 as well as from XREPP – Xarxa de Referència en Economia e Polítiques Públiques). Authors would like to acknowledge Martin Spielauer’s collaboration in the model code revision..

* Instituto de Estudios Fiscales, Madrid.

** Xarxa de Referència en Economia i Polítiques Públiques (XREPP), Universitat de Barcelona (UB).

Departament d’Economia Aplicada, Universitat Autònoma de Barcelona.

Los Documentos de Trabajo se distribuyen gratuitamente a las Universidades e Instituciones de Investigación que lo solicitan. No obstante están disponibles en texto completo a través de Internet: http://www.fedea.es. These Working Paper are distributed free of charge to University Department and other Research Centres. They are also available through Internet: http://www.fedea.es. ISSN:1696-750

Jorge Juan, 46 28001 Madrid -España Tel.: +34 914 359 020 Fax: +34 915 779 575 infpub@fedea.es

DYPES: A Microsimulation model for the Spanish retirement pension system

F. J. Fernández-Díaz¹, C. Patxot², G. Souto³

Abstract

This paper presents the results of DyPeS, the first dynamic microsimulation model of the retirement pensions system applied to the Spanish case. The simulation of the reform approved in 2011 shows that only the delay in retirement age (from 65 to 67) would have a significant effect on pension expenditure, while other measures changing the computation of the initial pension for new retirees have a limited impact. Paradoxically, it is found that the consideration of more contribution years in the computation of the initial pension amount, despite fostering the Bismarckian nature of the system, has a positive impact on redistribution.

JEL Classification: H53, H55, H68.

¹Instituto de Estudios Fiscales, Madrid

² Xarxa de Referència en Economia i Polítiques Públiques (XREPP), Universitat de Barcelona (UB).

Acknowledgements

This research received institutional support from the Spanish Science and Technology System (Projects Nº ECO2012-37572 and ECO2012-35054), the Catalan Government Science network (Projects Nº SGR2009-600 and SGR2009-359 as well as from XREPP – Xarxa de Referència en Economia e Polítiques Públiques). Authors would like to acknowledge Martin Spielauer’s collaboration in the model code revision.

3 Departament d’Economia Aplicada, Universitat Autònoma de Barcelona

1. Introduction

The pension system is one of the main components of the Spanish welfare state. At present, it is composed of two different types of benefits: non-contributory (or Beveridgean) and contributory (or Bismarckian). The former can be considered a system of minimums, as it provides, under some conditions, a minimum income to those individuals who cannot access the contributory level. The latter is the main part of the system.¹ It is organized on a pay-as-you-go (PAYGO) basis and includes different kinds of pension benefits - retirement, disability and survival - for those individuals who meet the eligibility requirements regarding age and past contributions to the system.

Clearly, the contributory retirement pension is the most important social expenditure programme. In 2007, it represented over 65% of all contributory pension expenditure, and over 5% of GDP. Moreover, a huge increase in its size can be expected in the face of demographic ageing: retirement pension expenditure will be more than 75% of contributory pension expenditure and almost 12% of GDP in 2060 according to the last projection exercise of the Economic Policy Committee (2009). Therefore, the concern about the necessary reforms to make it sustainable in the long term is fully justified.

Reform proposals vary from a complete restructuring of the system – like a switch to a true or to a notional capitalization system³ – to marginal adjustments of the legal parameters of the current system. Given the expected increase in the ratio of pensioners to contributors, all of them imply raising contributions and/or reducing pensions. Alternatively, an increase in the retirement age is often proposed, as this might improve sustainability both on the expenditure and on the revenue side. Although the legal retirement age used to be 65 for most workers in most countries, the fact is that average retirement age is clearly lower. Governments in the EU, concerned about the importance of this phenomenon, are interested in incentivizing workers to delay retirement. Though average retirement is quite high in Spain by European standards – according to Eurostat it was 62.6 years in 2008, 1.1 years above the EU15 level -, Spain will be hit by one of the most pronounced ageing processes starting around 2020, a little later than in other European countries given that the baby boom in Spain occurred later than in other countries. Following the recommendations of the European Council, a reform package was debated and approved in 2011. Apart from the continuation of some of the proposals contained in the Toledo Agreement⁴ , it was mainly composed of a delay in retirement age.

¹ The civil servants have their own system.

² This projection is in line with previous projections and also with those done in academic work.

The possibility of switching to a funded system is limited by the so-called transition problem: the initial gift given to generations who did not contribute and received a pension now needs to be offset by a double burden on current workers, who need to maintain the old PAYGO system and contribute to the new one. Some countries have opted for introducing some kind of notional funding, which is somehow a sophisticated way of introducing a defined contribution system.

The design and evaluation of reform measures requires the availability of analytical tools. Simulation models have been developed, especially in the last decades thank to the availability of an increasing amount and quality of databases and computing tools. In the context we are dealing with, those simulation tools need to take both a macro and a micro perspective. The first is essential if one aims at analyzing in a consistent way the sustainability of pensions or any other welfare state transfer. The second is crucial when considering the adequacy of the benefit level in respect to income redistribution. The EU, in the joint report of the European Policy Committee (EPC) and the Social Policy Committee (SPC), has been recommending to attend to both dimensions in order to maintain the welfare state, one of the main achievements of Europe in the past century.

In this paper we present a simulation model of the contributory retirement pension system in Spain and it is used to analyze the effects of the reform enacted in 2011. To the best of our knowledge it is the first mode of this kind developed for the Spanish case, thanks to the availability of the MCVL – an administrative data set published by the Social Security Administration since 2004. The model is based on a sub-sample of the 2007 wave of this data set.

The paper is organized as follows. Section 2 is devoted to the description of the Spanish pension system. Section 3 describes the model structure and the data employed, including the projection rules implemented in the model. Section 4 presents the results and Section 5 makes some final remarks.

⁴ Toledo Agreement arose in 1995 when the main political forces in Spain approved a set of recommendations to drive the necessary reforms of the Social Security System in order to ensure its sustainability.

⁵ The data set is called continuous working life sample (Muestra Continua de Vidas Laborales) and the first wave was published for 2004 (MTAS, 2006b).

2. The Spanish pension system

The Spanish Social Security system as such was introduced in 1967. Previously, workers were insured against disability and retirement by mutual societies⁶. Some individuals who belonged to that previous system are still working and, because of that, they still keep some special privileges in acceding to early retirement within the contributory system. Due to this particularity, we will refer to them as Old System (OS) workers henceforth.

The Social Security is organized on a PAYGO basis and under a defined-benefit scheme with a partial Bismarckian orientation - the pension received is mainly determined by the past contributions of the worker. Social Security covers other contingencies besides retirement, such as disability and survival, but retirement is clearly the most important of them. Below we outline some legal features relevant to our analysis. For the moment we concentrate on the system prior to the last reform approved in 2011 (27/2011 Act), as it is starting to be applied from 2013, with a long transition period, until 2027. Section 4.b) is devoted to detailing the main changes introduced in this last reform and to simulating their effects in the future. Beneficiaries of retirement pensions are affiliated workers who meet the legally established conditions of i) having the legal age (generally 65) and ii) proving a minimum contribution period of 15 years - 2 of them should be included in the 15 years prior to retirement date.

The initial retirement benefit is determined according to a formula depending on retirement age, the number of contribution years and the amount of past contributions. Since the reform posed by the 24/97 Act, the computation is as follows. First, a basic amount (BR) is computed as a mean of the level of contributions made in the 15 years prior to retirement⁷. Second, a percentage depending on the number of years of contributions – p(n) – is applied to the BR in this way: 50% for the first 15 years, an increase of 3% for each additional year between the 16th and the 25th, and an increase of 2% for each additional year from the 26th until it reaches 100%, at 35 years of contributions. Furthermore, when retirement age deviates from the legal retirement age a correction coefficient – cc(n) – is applied, depending also on the number of years of contribution. Thus, in general terms, the legal formula to calculate the initial retirement benefit (IRB) or entry pension is:

⁶ This system was called Seguro Obligatorio de Vejez e Invalidez (SOVI).

⁷ The legal name is base reguladora. In the computation, the contributions are updated from the third year according to the evolution of the consumer price index.

\[I R B = p (n) \cdot B R \cdot c c (n)\tag{1}\]

The coefficient will be above (below) one for workers delaying (advancing) their retirement from age 65. In the first case an additional 2% will be applied for each year of delay, or 3% if the worker has 40 or more years of contributions – this premium either acts on increasing up to 100%, holding equal to one, or increasing above one if the worker already reached the 35 contribution years. In the second case, if the worker retires before age 65 an annual penalty between 6% and 8% is applied, depending on the years of contribution.⁸ Apart from this general rule, there are also other possibilities to reduce the age of retirement: i) certain professional activities, ii) disabled workers, iii) special retirement at age 64 with no penalty and iv) partial and flexible retirement. In each of them, different correction coefficients are also applied.

There are several ways to access retirement in Spain, some of them depending on the labour status. The pension formula is the same for all of them, except for the value of . Table 2.1 summarizes all those different possibilities or retirement paths. The ordinary retirement age – 65 – is not compulsorily established by law, but it is agreed in most collective wage settlements. In any case, at the moment 65 is the reference age for retirement rules and incentives. For the first time, the 2011 reform establishes a gradual increase to 67, as explained in Section 4.

8 In particular, the 8% affects those that only reach the minimum eligibility requirement of 15 contribution years. This penalty is reduced gradually for those crediting enough years to move to the next contribution years scale: 31-34 (7.5), 35-37 (7), 38-39 (6.5) and 40 plus (6).

Table 2.1 Retirement paths in Spain according to labour status

Notes: n= number of years of contribution; * Only disabled below age 65 might change state back to work; ** 7.5% (n=30-34), 7% (n=35-37), 6.5% (n=38,39), 6% n≥40; *** In 40/07 Act the minimum n was increased to 30 and 6 years of seniority in the same firm were required.

Early retirement pathways have been developing gradually with this direct purpose, but also with an indirect objective of employment policies, like intending to improve the labour market access for youth. Regarding the latter, special retirement at age 64 and partial retirement were established in 1985 and 1999, respectively.⁹ Regarding the former there are several rules. First, apart from the disabled, for some professional activities – especially those which are dangerous or which imply being away from home – there is a possibility of retiring at a fixed or at a reduced age. Second, early retirement was generally available from age 60 under certain requirements but with some penalty in the old system. Given that this would eventually disappear, in 2002 (35/2002 Act) new rules were established for new system workers, allowing early retirement from age 61 with similar but stronger penalties, and including an explicit requirement of being unemployed. In 2007, some changes in penalties were introduced.

⁹ The former was introduced in 1985 (RD 1994/1985) and the latter in 1999 (RD 144/1999).

In parallel, and also due to the need to foster sustainability in the face of demographic ageing, some rules were also introduced intending to delay retirement. On the one hand, the possibility of combining work and pension receipts was introduced in 2002 by reformulating partial retirement and introducing flexible retirement (the latter allowing retirees to come back to the labour market as part-time workers). On the other hand, incentives to continue working full-time after age 65 were introduced in the regular retirement path, as detailed above. Finally, there is a possibility to exit the labour market through pre-retirement, i.e. through a special agreement with the firm that leads to private or public subsidies until the age of eligibility.

Table 2.2 Distribution of new entries by pathways (Spain 2002-2007)

Note: Pre-retirement includes only those which can be identified as receiving public subsidies. Source: Authors’ elaboration from the MCVL.

Table 2.2 shows the distribution of new entries to retirement in Spain over six subsequent years (2002- 2007). Clearly, ordinary retirement is the most common and includes both a small share – around 1% – of early retirement at fixed age and a sizeable share – above 10% – of delayed retirement.¹⁰

Interestingly, while the share of pensioners opting for the OS early retirement rules is still 20.24%, early retirement at age 61 has not been very attractive. To some extent, this fact is explained by the requirement of being unemployed, but one can think that workers have opted for other less costly ways to access retirement before the legal age, like partial retirement. In some cases they even use the flexible retirement situation to improve their benefits once retired (in 2006 only 12.38% of pensioners in this situation were below age 65).¹¹ Indeed, in practice, some of the measures intending to delay retirement have still been used as a way to improve the pension benefit by early retirees. In 2006, 98.75% of new entries and 87.76% of pensioners in partial retirement situation were below age 65. Flexible retirement has indeed been used to stay longer in the labour market, but its impact is very limited as only 0.20% of new entries in 2007 have chosen this option.

Despite the abovementioned reform efforts, the average retirement age for males has remained quite stable in Spain in recent years. Interestingly, at present females retire later, perhaps due to a joint retirement decision or to the need to complete their shorter contribution histories. This would also explain some of the differences observed between male and female retirement probabilities. Despite the stability of the average retirement age, the evolution of the share of new entries by age and sex has undergone substantial changes which seem to be driven by cyclical movements. In fact, changes in early and delayed retirement move in different directions. The share of those retiring after age 65 is only slightly affected by incentives to delay retirement introduced in the 35/2002 Act. One should bear in mind that most collective wage settlements deny workers the possibility of delaying retirement, so that incentives to work beyond age 65 might become inoperative.

¹⁰ The table classifies new pensions using the variables “type of pension” and “situation of the pension” in the MCVL. A description of the database is provided in Appendix A.

¹¹ With respect to partial retirement, it can be extended beyond age 65. Nevertheless, in 2006 only 6.56% of pensioners were older than 65.

3. The model

This section is devoted to the description of DyPeS, the microsimulation model developed to analyze the Spanish contributory pension system. Subsection a) outlines the model structure, Subsection b) details the data employed, while Subsection c) details the way retirement decision is modelled.

a) Model structure

As said in the introduction, DyPeS is a micro-based model, thanks to the availability of the MCVL. The construction of a microsimulation model involves taking several decisions, mainly affected by the question analyzed and the data set on which the model is based (see Jinjing and O’Donoghue, 2012). In our case, the nature of the pension policy necessarily implied the use of a dynamic model – meaning that it simulates the micro units over time. In this setting, dynamic does not necessarily mean behavioural.¹² In fact, behavioural responses are kept to a minimum at this first stage, for simplicity. Nevertheless, the model presented serves as a starting point for future analysis, including more sophisticated behavioural responses on retirement and other labour market decisions.

Given that our main focus are retirement pensions, the rest of events are modelled in the simplest way given data availability. The following are the main events experienced by agents. Agents experience first birth and second entry in the labour market. Third, other labour market transitions, including changes in the qualification level and unemployment events, are derived based on transitions observed in the data set. Fourth, once agents attain the eligible retirement age (fixed from 59 to 75), they start computing their expected pensions in each of the available pathways depending on their labour market status and, eventually, retire. Finally, agents die according to exogenous age and gender specific mortality rates evolving in line with those used in the standard population projections. The projection routine of the model starts in 2008. Hence, for events occurring before – affecting agents alive in 2007 - the observed date is taken from the data set.

12 The term dynamic has been mixed up with the term behavioural for some time, especially in static microsimulation models. See Dekers et al., (2008) and Jinjing and O’Donoghue (2012) for a survey on dynamic microsimulation models.

Regarding the programming strategy, DyPeS is a case-based model – one case is simulated after another -, as opposed to time-based – where all cases are simulated in each period. It is an open model, in the sense that new agents are introduced, apart from those in the initial sample; and population-based, instead of cohortbased. It is programmed in continuous time, though some of the events happen only once a year. DyPeS has been developed using ModGen, a generic dynamic microsimulation programming language developed and maintained by Statistics Canada and widely used in social science microsimulation.¹³

b) Data employed

DyPeS starts from a subsample of individuals registered with the Social Security in 2007 extracted for the 2007 wave of the MCVL. The year 2007 is chosen as the base year and the reference point for most data. This way the data employed for transitions are not distorted by the effects of the crisis. Certainly, future waves of the MCVL will improve perspectives to measure the effect of the crisis. For the moment, we start in 2007 and simulate the effect of the crisis in a stylized way, as described in Section 4.¹⁴ At this first stage we stick to information given in the MCVL and no alignments to external data are made, unless strictly necessary. Below, the main decisions taken regarding the MCVL data set and information used from external sources are summarized.

Table 3.1 Initial sample from the 2007 MCVL

Source: Authors’ elaboration

13 Modgen supports the creation, maintenance and documentation of most dynamic microsimulation model types, including both continuous and discrete time, case and time-based models as well as interacting and non-interacting populations. It is freely available at the Statistics Canada web site.

⁴ The last complete wave – including fiscal data - available at the moment is for year 2009.

Table 3.1 describes the initial sample. Basically, individuals present in the 2007 wave of the MCVL are taken. The MCVL extracts 4% of the population who have some relation with Social Security administration. Then, all past information on their working careers and contributions is added. Nevertheless, the quality of data worsens for the past. A cleansing process was needed in order to keep those workers whose contributions were well recoded (see Appendix A for details). The sample includes both pensioners and contributors born from 1907 to 1991. Hence, in order to project future expenditure and revenue, new entries in the labour market from 2008 on and new births from 1991 need to be added to the model. In order to add newborns, we compare the number of people in the 2007 population and in the 2007 MCVL wave. Figure 3.1 below shows this relation for males and females.

Figure 3.1 Number of people in the MCVL to Total Population, by birth year (year 2007).

Source: Authors’ elaboration

The distance between males and females for cohorts born from 1922 to 1960 is explained by low participation rates of females in those cohorts. Quite surprisingly, females overcome males for some particular cohorts. To avoid this temporary bias, we fix the weight given to males and females as the average of cohorts born between 1967 and 1971. Then, using standard population projections, we add the resulting number of future newborns and entries in the labour market by gender to the 2007 sample.

The data employed to simulate each of the events are described below. The first step is to assign an education level. Decisions on education level are minimized and simplified as much as possible. The MCVL contains information about the education level of the individuals. Nevertheless, this variable is collected from a different data set that is not updated very frequently. As a result, the education level is frequently missing or underestimated. For individuals registered in the MCVL we keep the value reported and correct it up in case there is an inconsistency between the value of education and the contribution group. For “future” individuals, born from 1991 on, the final education level is assigned randomly so as to reproduce the educational distribution observed for the Spanish population by MEC (2010). According to this publication, the education level has grown substantially. In 1997 – Figure 3.2, panel a) - one can already observe that the young population was more educated than the total population. The improvement in the education level continues until 2008, as one can see by looking at panel b) in the same Figure: the education level increases for both young males and females from 1997 to 2008. Interestingly, the education level is currently higher for young females that for males, inverting the initial situation. Note that we are assigning the final educational attainment without any attention to transitions from one education level to the next one, given the data constraints and the focus of our analysis.

Figure 3.2 Distribution of population according to education level in Spain (%) a) Young (25-34) versus total (25-64) in 1997

b) Evolution of the education level of the young (25-34) from 1997 to 2008

Source: Authors’ elaboration from MEC (2010).

In a second step, once the main characteristics of the individuals are assigned and they reach the age of 16, they are exposed to the risk of entering the labour market by age, gender, education and initial qualification level. This is obtained from the observation of the entry path of the last cohort, which had completed their incorporation to the labour market – those aged 36-40 in 2007 (see Table 3.2.). As shown below, most entrances are in the contribution group (composed of part-time workers) except for the most educated – more than ¼ of them enter once they obtain their degree. The pattern is similar for females, with slight differences.

Table 3.2 Entrance in the labour market by education and qualification level (CG)

Source: Authors’ elaboration

In a third step, once individuals enter the labour market, they are exposed to the labour market transitions. The hazards observed are extracted from the MCVL. In particular, transitions between qualification levels within employment and transitions between employment and unemployment are obtained by age and gender and qualification level when necessary. To that effect, the 13 contribution groups in the general regime of the Spanish Social Security are grouped in five subgroups – those subject to the same contribution limits (thresholds). As the transition hazards among the different qualification levels are quite stable during the period observed (2002-2007), the value of the last transition observed before the economic crisis (2006 to 2007) is taken, and it is held constant for the future (see Appendix B for details). In Section 4.a), an explanation is given of how these hazards are altered during the crisis.

Regarding earnings, some decisions need to be taken. On the one hand, the initial value of wages for future workers needs to be decided. On the other hand, it is necessary to decide how wages are to be projected to the future both for 2007 workers and for those entering the labour market afterwards. First, regarding initial wages, for those working or contributing in 2007, the initial value of wages in 2007 is taken from the fiscal module of the MCVL. In case this value is missing, the contribution basis is taken. For future contributors, the final education level attained determines the contribution group, entry age and wage, as shown above. Similarly, the error term observed in each cell is also used to ensure variability of initial wage.

Second, the projection is modelled taking into account that they evolve with time and age. The literature on this field highlights the fact that it is not possible to identify the separate effect of age, cohort (or birth year) and year, as they are linked by a linear combination. To overcome this problem, Fitzenberger et al. (2001) assume that life cycle wage growth is independent of the calendar year and find that this hypothesis -called the unified wage growth hypothesis- cannot be rejected. Moral-Arce et al. (2009) apply this method to the Spanish case using the MCVL finding a similar result. The separability condition holds, which implies that an age profile independent of calendar year exists. Nevertheless, parameters still cannot be identified. Hence, in order to project wages, we opt for a simplifying approach. The growth rate of wages is decomposed in a pure time component -represented by the productivity growth rate- and a deviation from it, depending on age, sex and contribution group. The latter is obtained using past data on contribution from the MCVL. By taking the average deviation from the past, cyclical effects might be compensated. Although the cohort effect cannot be eliminated in this way –the data base is mainly observing the entry of the baby boomers in the labour market– it is the best we can do. To minimize errors, only full time wages are used to that purpose.

During the simulation, earnings (and contribution bases) are updated on a continuous time basis. For that purpose, both a current value and an accumulated value are maintained and updated in the following cases. First, earnings are updated at the beginning of the year, according to equation (2) – using age, gender and group-specific deviation from the average productivity growth rate - and adding the expected inflation rate. At the same moment, contribution bases are also updated. Second, whenever a labour status transition occurs – both among contribution groups within employment status, and between unemployment and employment status –, a change in wage is applied depending on gender and the original and final states. For that purpose, the average change in wage observed is used.

Finally, each time one of the abovementioned changes occurs, total earnings (and contribution bases) functions are updated. This also happens at the end of the year, so that the annual flow of earnings and contribution bases can be recovered and stored.

Once agents reach retirement age, the retirement module – detailed below – produces the retirement event. Finally, an exogenous age, gender and time-specific mortality rate, coherent with demographic projection, is applied.

c) The retirement decision: Handling with multiple retirement paths

In this section, we describe in more detail the way the retirement decision is modelled. We deliberately opt to focus on the changes in the level of pensions, allowing for the minimum reaction in the retirement decision. For that purpose, the retirement hazards observed are fixed by age and gender. As in the rest of transitions, the values observed for the year 2007 are kept. Nevertheless, to be able to simulate the effect of reforms on pension rights, we consider the possibility of reaching retirement through different pathways detailed in Table 3.1 above. In order to account for these possibilities, first a matrix of expected pension benefits by age and pathway is computed for each individual and year, once they reach retirement age. This is updated annually during the simulation period. Second, the best pension given labour status is chosen, so that it can be assigned to the individual once retirement age is reached. This way, a minimum behavioural reaction is permitted.

A thorough consideration of the behavioural response of agents is beyond the scope of this paper. Ideally, this should take into account a simultaneous consideration of the decision both of the firm and the individual. This is rarely possible, given the lack of data on the position of the firm, and most studies focus on the perspective of the individual. Along this line, structural models of retirement behaviour take into account utility maximization along the life cycle and, hence, the whole set of contribution and pension receipts from the pension system. Nevertheless, due to constraints in availability of longitudinal data sets, it has become common using reduced form estimations of retirement decision. This estimation strategy is not exempt of problems related to the identification of the incentives to delay retirement. These are usually measured as the increase in the so-called social security wealth – the present net value of net benefits received from the pension system. However, this variable might be endogenous and it might not be possible to separate the effect of financial incentives and the taste for work – both interacting with age. The MCVL certainly opens the way for this kind of analysis. There is some work on this field applied to Spain. This showed a poor effect of retirement incentives on retirement decision, indicating that age was the main determinant.¹⁵ This work was based on a preliminary experimental version of MCVL. Recently, this work has been extended (García-Pérez et al. 2010) using the MCVL and the results seem promising for future improvements in DyPeS.

As said above, we opt to consider only the effect of age, so the age pattern of retirement observed in 2007 is considered, except in the case of the reform delaying retirement age. Note that, when simulating reforms, the model will mainly capture changes in the level of pensions given the retirement age. Interestingly, if a reform cutting pension rights were implemented, we would capture mainly the cut in pension benefits, ignoring the possibility of agents escaping from it. Hence, we would obtain a proxy for the maximum effect of the reform.

4. Results of the simulation model

This Section is devoted to showing the first results of DyPeS. First (Subsection a), the baseline situation is characterized, explaining how the effects of the crisis are simulated. Second (Subsection b), the impact of the reform of the Spanish pension system approved in 2011 is analyzed. In order to interpret the results, it is important to bear in mind that we are keeping the retirement hazards observed by age, not allowing for a behavioural response in the retirement age decision. This implies that the burden of the adjustment will be mostly on the level of pensions. If we observe changes in the time path of the number of pensions, they will be mainly due to direct policy changes in retirement age.

a) The baseline situation and the effects of the crisis

As said above, the year 2007 is chosen as the base year to prevent the projections from being permanently

affected by the current crisis. At the same time it is necessary to take into account the effect of the crisis, although it is still early for a thorough account of it. Hence we opt for a simplifying ad hoc simulation of a temporary increase (decrease) in the job destruction (creation) rate, in line with the evolution observed in the first years of the crisis, shown in Figure 4.1. Given the uncertainty of the duration of the current crisis, we assume a slow recovery ending in 2018. The change observed in job destruction and job creation rates during the crisis (Figure 4.1) is applied to the transition hazard rates observed in the MCVL (see Figures B.1-3 in Appendix B) from 2008 to 2018.

15 See Boldrin et al. (2004) and Jimenez-Martin (2006), mainly focused on employed individuals, while the alternative pathways to retirement - long-term unemployment benefits and disability - were included using exogenous age and gender-specific probabilities. The first study is the chapter devoted to Spain in an international comparative study edited by Gruber ad Wise (2004). Although incentive measures turn out to be significant for most countries, results for Spain seemed to be an example of the identification problem. See also Argimón et al. (2007 and 2009). In the former, retirement incentives are not considered among the explanatory variables. In the latter, retirement incentives are included, but there is no clear way of controlling the alternative pathways to retirement.

Figure 4.1 Evolution of the job creation and destruction rates (2008-2011) a) Job creation rate

b) Job destruction rate

Note: Job creation rate: share of unemployed who get a job in the next trimester; Job destruction rate: share of workers who lose their job in the next trimester. Source: Observatorio laboral de la crisis, (www.fedea.es/observatorio)

As a result, the unemployment rate doubles, as panel a) in Figure 4.2 illustrates.¹⁶ This increase in unemployment probably understates the long-term impact of the current crisis, but is enough to take into account the effect of the crisis and it illustrates the potentialities of the model. Panel b) in Figure 4.2 shows the effect of the current crisis on entry pensions (both number and level). The main driving force of the changes observed is the decrease in the pension level, due to poorer working careers. It is indeed sizable, reaching 4% for the cohort retiring in 2050. There is also an impact on the time path of the number of entry pensions. The number of new entries reduces between 2013 and 2030 approximately, and grows slightly afterwards. Despite the assumption of constant retirement hazards, there is a reaction on retirement decisions due to the big movements in labour status that the current crisis produces. As a result, the total effect of the crisis in terms of expenditure to wage bill (panel c and d) is small, despite the sizable decrease in entry pensions shown in panel b. If behavioural reactions were allowed, the opposite could happen. If agents can exit the labour market through early retirement, escaping penalties, the effect of the crisis could be worse.

¹⁶ Note that the unemployment rate (Figure 4.2 panel a) increases during the retirement of baby boomers. This is a composition effect due to the fact that unemployment rates are higher for older workers might be compensated in reality by the relative scarcity of labour supply. A general equilibrium setting would be necessary to account for this.

Figure 4.2 Effect of the current crisis a) Evolution of unemployment rate

b) Changes in the entry pension

c) Ratio of total expenditure to wage bill

Source: Authors’ elaboration.

d) Change in the ratio expenditure to wage bill (%)

We take as a baseline scenario the one that incorporates the effect of the crisis. Figure 4.3 shows the evolution of retirement pension expenditure as a share of both the wage bill and the contribution bases. The low values in the first years are not representative until 2007 – the MCVL wave we are working with - and values from 2008 are projected. It is interesting to note, first, the fact that there is no discontinuity between the value observed in the sample until 2007 and the value predicted from 2008 on. This shows that the simulation model is producing valid projections. Second, the increasing distance between the two ratios. This is reflecting the role of the contribution and pension’s thresholds, showing the extent to which they help to contain expenditure.

Figure 4.3 Evolution of the ratio retirement pension expenditure to potential revenue

BC= total contribution basis Wages= total wage bill Source: Authors’ elaboration.

b) The effects of the 2011 reform

Below, the effects of the main reform measures approved in 2011 in Spain are simulated. Table 4.1 gives an overview of the kind of measures approved (see Appendix C for further details). The first two measures aim at increasing contributiveness, or proportionality between contributions and pensions, strengthening the Bismarckian nature of the system. This motivation has been present in the Spanish reform agenda since it was explicitly expressed in the Toledo Agreement in 1995. Interestingly, these measures tend to decrease pension rights in different ways, as we will see later. This way they have a positive effect on sustainability and potential effects on redistribution. The first reform measure increases the amount of past years’ contributions considered to compute the basic pension amount – BR – from 15 to 25. The second goes in the same direction by modifying the share of BR received as a pension –p(n)– to make it more linear. Nevertheless, it also increases the total number of years needed to obtain 100% of BR, implying a direct cut in pension rights.

The last and most discussed measure modifies the reference retirement age from 65 to 67. It also changes the retirement penalty for early retirement and premium for delayed retirement -cc(n). All the measures are implemented gradually from 2013 to 2027. Below, we will present the simulation results and discuss the effects on pension expenditure sustainability and redistribution.

In order to interpret the results one should bear in mind that the effects of the reforms might differ depending on a main issue: the extent to which reform affects one pathway or all of them. The latter would imply that being able to choose the pathway does not make a difference and, hence, we will not observe changes in the number of pensions, but only in the level. Most of the approved measures apply to all the pathways at once: Changes in BR and p(n). Only the change in cc(n) could have an impact on pathways, but this measure is implemented together with the change in the retirement age, which does, indeed, affect the number of pensions.

Table 4.1 Summary of the main reforms in retirement approved in 2011 (27/2011 Act)

Source: Authors’ elaboration

Figure 4.4 shows the effect of the first reform set: reforms affecting BR and p(n). The change in p(n) implies a decrease in the level of the initial pensions, reaching 3% once the reform is fully implemented. This average negative effect is probably hiding positive and negative effects for people with different working careers. The corresponding effect on the number of entry pensions is null as this measure affects all retirement paths to the same extent. In general, reforms of p(n) have potential effects on redistribution, as discussed in Appendix C. Nevertheless, in this particular case they are very small, due to the small scale of the current reform.

With respect to the changes in the number of years used to compute BR, the result of the 2011 reform, increasing the years gradually from 15 to 25 (Br25 in Figure 4.4), might seem surprising at first glance. The reform implemented has, on average, a small negative effect – around 1% depending on the year of retirement. The expected effect of this measure depends on the shape of the lifetime real earnings profile.¹⁷ If the real earnings profile is increasing, when BR takes more years in the past this implies a reduction in the level of wages considered and, hence, a cut in pension rights. Nevertheless, the earnings career does not always grow to the same extent during the life cycle. It is expected to grow more at the beginning, to stabilize around the 50s and then remain constant or possibly worsen, if the working career is interrupted because of unemployment events. Hence, the effect of this measure can be a small cut or even an increase in pension benefit if wages are not growing in real terms. The fact that we obtain this result on average is probably due to the fact that the increase from 15 to 25 years implies adding years when the real earnings profile grows very slowly, or is almost flat. Quite interestingly, from 2050 on the effect of the reform is an increase in the initial pension. This is probably reflecting the effect of the crisis on the cohorts entering the labour market precisely in this period.

In order to test the validity of this result, we simulate alternative measures along this line. First, the effect of applying this measure in 2007 is implemented (Br25_2007 on scenario). This simulation has the advantage of running on past wages observed and the effect goes in the expected direction: Entry pensions are cut for the cohorts retiring from the initial simulation year 2008. From 2027 on, the number of years is also kept constant to 25, having the same effect on initial pensions as the approved reform. The cut in pensions in this case is bigger, reaching a maximum of 14% for the cohort retiring in 2007. This figure is probably overstating the effect of the pension cut, as the measure in this case implies going back to year 1982, where the quality of the data set is worse. Note that missing data on past wages are replaced with the minimum contribution level according to Spanish regulations.

¹⁷ The relevant magnitude is real wages (and hence contributions) as the formula to compute BR updates wages to inflation two years prior to retirement.

Second, a continuation of the increase from 15 to 25 until 37 years is simulated (Br37 scenario) with the same calendar – adding one year each year. In this case, the pension cut is much higher. With respect to the Br25 scenario, pension benefits continue decreasing from 2023 on reaching a 6% cut in 2034 when the reform is completed.

c) Redistributive effect of Br25 reform starting in 2007 d) Redistributive effect of Br25 and Br37 gradual reforms Figure 4.4 Effects of changing BR and p(n) on the level of initial pension (with respect to the baseline) a) Change in years considered in BR

b) Change in the p(n) –2011 reform

Source: Authors’ elaboration

The results might also be affected by the projection rule applied to wages during the simulation period. The estimated growth rate is based on past data, mostly proceeding from the baby boom cohorts, and might be overstating the effect of age. Along this line, there is an additional explanation related to the way the pension corrects for past inflation. As seen above, the computation of BR uprates past wages to inflation but not to productivity growth. This implies that the difference between these two rates affects the replacement rates. The higher the difference, the higher the replacement rate. Note that, as seen in Figure 4.5, there was a big distance for some of the past years, which has been reduced and is kept constant during the projection years. Hence, this explains part of the lower replacement rate of younger cohorts with respect to older ones. The redistributive effect of applying this Bismarckian reform is shown in Figure 4.4 both for scenario Br25 and Br37 )– the latter being the continuation of the former. The time path of pension cut and the change in the Gini index of the initial pension is depicted together, showing that the intuition raised by Jimeno (2003) holds. The author claims that this measure has a redistributive effect. The reason is that high-income individuals have steeper lifetime earnings profiles, and hence lose more by recovering past wages in the computation of the pension formula. Again, in order to test the validity of this result, we analyze the redistributive effect of moving from 15 to 25 years right in the base year (panel c). This simulation is based on observed – not predicted - earnings profiles and gives similar results: the higher the wage, the higher the pension cut, so that redistribution improves. It is not possible to extend the simulation beyond 25 years, due to the lack of data on oldest contributions.

Figure 4.5 Past evolution of inflation and productivity

Source: INE and OECD

Figure 4.6 shows the effect of delaying the normal retirement age. In this case, the simulation needs to be done in three accumulative steps. First, the effect on the pension level is computed. Second, the age dummies are shifted. Third, the premium to delay retirement is changed. Panels a) and b) in Figure 4.6 show that the main effect is driven by the age dummies. The number of entry pensions is substantially reduced during the implementation of the reform, while pensions show both positive and negative changes on quite an erratic path. The abrupt changes observed in both magnitudes are mainly due to the fact that age dummies cannot be moved in a smooth way. Panels c) and d) show the impact on the ratio expenditure to wage bill and on the retirement age. In this case the reform has a sizable impact. The average retirement age observed increases almost two years and this implies a cut in expenditure to wage bill of slightly more than two percentage points.

Figure 4.6 Effects of the reforms of retirement age (with respect to the baseline) a) Change in the level of entry pension

b) Change in the number of entry pension

c) Changes in the ratio expenditure to wage bill

Source: Authors’ elaboration.

d) Changes in average retirement age

Finally, Figure 4.7 shows the overall effect of the reform measures approved in 2011, including the scenarios change in “p(n)”, “Br25” and “65 to 67_delayed”. The difference between the baseline and the reform scenario reaches almost 3 percentage points around 2044 (panel a). In panel b), the isolated effect of each measure is depicted. The most effective reform saves more than 2.5 points, while the others save less than 0.5 points.

Figure 4.7 Total effect of the 2011 reforms a) Ratio pension expenditure to wage bill

Source: Authors’ elaboration.

b) Change due to each individual measure (% points)

5. Final remarks

The reform of the pension system is one of the main topics on the EU policy agenda. The design and evaluation of reform measures requires the availability of analytical tools suited to that purpose.

This paper presents the results of DyPeS, the first dynamic microsimulation model of the retirement pensions system applied to the Spanish case. The model is based on a sample drawn from the 2007 wave of the MCVL - the administrative data set of the social security administration. It reproduces the main events occurring over the life cycle: birth, entrance into the labour market, unemployment events, retirement and death. DyPeS is mostly arithmetic, with no behavioural reactions except for the minimum introduced to account for retirement pathways. Nevertheless, it can serve as a starting point for future behavioural analysis on retirement and other labour market decisions. Regarding the programming strategy, it is a casebased model, population based, and it has been modelled in continuous time. DyPeS has been implemented using ModGen, a programming language developed by Statistics Canada.

In order to show the potentialities of the model, the effects of the reform approved in 2011 are simulated. The results show that the reform would have a sizable impact on pension expenditure, but it does not seem to be sufficient to ensure sustainability. In particular, once the effect of the current crisis is introduced, three main reforms are implemented: The increase in the number of years considered to compute the basic pension benefit (BR) from 15 to 25, the change in the share of BR received as a pension, p(n), – implying changing the weight given to the years of contribution - and the delay in retirement age.

The reforms fostering the Bismarckian nature of the system produce a small change, despite the fact that the simulation does not allow behavioural reactions of agents to escape from the cut in pension benefits. First, the increase in the number of years required to obtain 100% of the basic pension benefit in p(n) implies a pension cut. The second case – change in BR – is not so straightforward. As long as earnings profiles are increasing, recovering past wages reduces the pension benefit. The simulation of the effect of both measures, nevertheless, shows that the cut in pensions is reduced, due to the small extent to which they have been implemented. The change in p(n) goes in the right direction but is small. The change in BR from 15 to 25 years affects a region of the longitudinal earnings profiles when they tend to be flat in real terms. As an attempt to analyze the effect of a fully Bismarckian system, a continuation of this extension from 25 to 37 years is simulated, obtaining a substantially higher decrease in pension level.

Interestingly, the changes in BR, despite fostering the Bismarckian nature of the system, have a positive impact on redistribution. The reason is that individuals with higher wages also have steeper longitudinal earnings profiles and, hence, are more affected by this measure.

The third reform, the delay in retirement age, is clearly the most effective in terms of cutting expenditure, despite the fact that retirement age is only delayed by two years. For example, retirement delay would reduce expenditure (in terms of wage bill) by 2.5%, while reforms in BR and p(n) would reduce it by only 0.2% and 0.4% at most respectively.

Future research analysis is required on the following lines. First, it is possible to use new waves of the MCVL in order to exploit it as a panel and introduce behavioural responses in the retirement decision. This could also serve to improve the wage projection rule and investigate the way it affects results. Finally, the modular structure of the model allows for the introduction of the remaining contributory pensions and the analysis of the sustainability of the complete contributory pension system, implementing the necessary alignments.

References

Argimón, I; C.I González; and R. Vegas (2007). “Jubilación entre los 60 y 65 años. Algunas características”.

Presupuesto y gasto público, Nº 47, pp. 161-184

Argimón, I; M. Botella, C.I González; and R. Vegas (2009). Retirement behavior and retirement incentives in

Spain, Documento de Trabajo, Banco de España Nº 0913.

Benavides, A. (2006). Análisis Práctico de las Prestaciones de la Seguridad Social. Lex Nova, Valladolid.

Boldrin, M., S. Jimenez, and F. Peracchi (2004). “Micro-modeling of Retirement Behaviour in Spain”, In Social

Security and Retirement around the World (Jonathan Gruber & David Wise, eds.). NBER, University of

Chicago Press, 2004, 499-567.

Börsch-Supan, A., R. Schnabel, S. Kohnz and G. Mastrobuoni (2004): “Micro-Modeling of Retirement

Decisions in Germany”. In Social Security and Retirement around the World (Jonathan Gruber & David Wise,

eds.). NBER, University of Chicago Press, 2004, 499-567.

Dekkers, G., H. Buslei M. Cozzolino, R. Desmet, J. Geyer, D. Hofmann, M. Raitano, V. Steiner, P. Tanda, S.

Tedeschi, and F.c Verschueren (2008). What are the consequences of the AWG-projections for the adequacy

of social security pensions? An application of the dynamic micro simulation model MIDAS for Belgium, Italy

and Germany, Report of the Work Package 4 of The AIM project, Federal Planning Bureau report, September

2008.

Eklof, M.-Hallberg,D.(2006). Estimating retirement behaviour with special early retirement offers,

Department of Economics Discussion Paper, Uppsala University.

European Policy Committee (2009). 2009 Ageing Report, Economic and Budgetary Projections for the EU-27

Member States (2008-2060). European Economy 2/2009.

García-Pérez, J.I., S. Jiménez-Martín and A. R. Sánchez-Martín (2010). Retirement incentives, individual

heterogeneity and labour transitions of employed and unemployed workers, UPF WP 1239.

Gruber, J. and D. A. Wise (2004). “Introduction and Summary”, In Social Security and Retirement around the

World (Jonathan Gruber & David Wise, eds.). NBER, University of Chicago Press, 2004, 499-567.

Herce, J.A. and J. Alonso (1998). Los Efectos Económicos de la Ley de Consolidación de la Seguridad Social:

Perspectivas Financieras del Sistema de Pensiones tras su Entrada en Vigor. Documento de Trabajo de FEDEA

98-16.

Jimenez-Martín, S. (2006). “Evaluating the Labor Supply effects of Alternative Reforms of the Spanish

Pension System”, Moneda y Crédito, 222, 271-307.

Jimeno, J.F. (2003). “La Equidad Intrageneracional de los Sistemas de Pensiones”. Revista de Economía

Aplicada, nº 31, pp. 5-32.

Jinjing, L. and C. O'Donoghue (2012). A methodological survey of dynamic microsimulation models, UNU-

MERIT Working Paper Series, #2012-002, http://www.merit.unu.edu/publications/wppdf/2012/wp2012-

002.pdf

Maes, M. (2008a). Survey of empirical research on retirement behaviour and retirement incentives, Working

Paper 2008-18, Department of Economics, University of Brussels

Maes, M. (2008b). Redistributive impact of reforming the old-age pension system in Belgium, Working Paper

2008-19, Department of Economics, University of Brussels

MEC (2010). Sistema estatal de indicadores de la educación, Instituto de evaluación, Ministerio de

Educación.

Mincer, J. (1974). Schooling, Experience and Earnings. New York, Columbia University Press.

Monasterio, C., I. Sánchez, and F. Blanco (1996). Equidad y Estabilidad del Sistema Español de Pensiones.

Serie Economía Pública, Fundación BBV.

MTAS (2003). La Renovación del Pacto de Toledo. MTAS, Madrid.

MTAS (2006a). Acuerdo sobre Medidas en Materia de Seguridad Social. MTAS, Madrid.

MTAS (2006b). La Muestra Continua de Vidas Laborales, Colección Informes y Estudios, Seguridad Social,

Ministerio de Trabajo y Asuntos Sociales (MTAS), Madrid.

http://www.seg-social.es/Internet_1/Estadistica/Est/Muestra_Continua_de_Vidas_Laborales/index.htm

Appendix A: The MCVL data set and the sample extraction¹⁸

The MCVL is a sample extracted from Social Security administrative data. Four per cent of all individuals registered with the Social Security administration – both contributors and recipients of benefits – over the sampling year are selected and their entire life history in the social security records is included in the data set.¹⁹ Thus, although it is not a pure panel, the data set is rich in longitudinal data. This feature, however, complicates the structure of the information as the registration unit varies substantially ranging from the person – in the personal data file – to the contract – in the affiliation file – or to the contract and year – in the contribution file. This structure also complicates the data selection. Furthermore, the quality of data is not homogenous, deteriorating the further back in time we go as more data are missing. The data collection itself was initiated at different points in time: data on pensions were first included around 1996; data on contributions around 1980; while some data on affiliation (contract registering) are available from as early as 1970. Clearly, all these factors condition our analysis. We provide details of the data employed below. We focus primarily on the pension file whose registration unit is defined by the individual, the benefit and the year, but we also recover contributory data for those individuals in our sample.

Among the difficulties of dealing with such a large administrative data set – the sample size reaches about a million people in 2005 – the most challenging are dealing with empty contribution bases and relating contribution, affiliation and benefit data from the same individual, all defined with different registration units. In particular, in order to extract reliable data regarding contributions in a specific time unit, it is necessary to follow up all the contracts in which an individual has been involved, computing time and contribution separately so as to avoid an erroneous correspondence between working time and contribution per unit of time. Below we describe in detail how we dealt with this.

¹⁸ See MTAS (2006b) for a detailed description of the Muestra Continua de Vidas Laborales (MCVL), available upon request at www.seg-social.es/Internet_1/Lanzadera/index.htm?URL=82.

19 Both workers and pensioners are thus included and also individuals receiving unemployment benefits or benefits prior to early retirement. The latter can be identified by the type of relation they have with the Social Security.

The model presented in this paper drives mainly on administrative data, extracted from the Social Security administration – the MCVL. It seeks to examine the impact of certain reform measures on the probability of retiring. The MCVL allows for this kind of analysis because it contains data of the main factors included in the initial pension formula. Specifically, over the period covered by the database, we are able to recover the number of working years, the life-cycle contributions of the individual and the retirement age, thus determining any penalizations for early retirement. It is also possible to recover these variables – except the not yet observed retirement age – to analyze future pension rights for potential pensioners – i.e. all the individuals in the sample who can opt for retirement. For pensioners, the total number of lifetime working years considered on computing the initial pension is also registered in the MCVL. Nevertheless, it is also necessary to obtain the annual working time in order to fill the gaps in the contribution data, in line with Spanish legislation.²⁰ Hence, the annual contribution period or working time is obtained by recovering all the contracts signed by the individual for each year, taking into account part-time work as well as the possibility of contracts that ran simultaneously.

At the same time, the average hourly contribution is obtained. One of the main problems we faced was the existence of missing contribution data. This can occur either within a contract registered in the contribution file or due to a lack of correspondence between the affiliation data – starting long before 1980 – and contribution data – starting in 1980. For example, we might find, even after 1980, no recorded contribution for one specific worker, while data regarding affiliation showed the worker to be actually contributing.²¹ An imputation process is developed to provide figures for the missing contributions. By tracking affiliation and contribution data, we treat missing values differently considering whether the individual is actually working – actual missing value – or not – if he is out of the labour force. In the former case, data from the same individual during the same year are used in order to recover absent contributions. If this is not available in the same year, the individual is dropped from the sample. This implies keeping more than 60% of the 2007 sample. In the latter case, the gaps occurring in the last 15 years are filled with the minimum contribution threshold, according to Spanish legislation.

²⁰ When the individual presents a non-contributory period within the last 15 years considered for computing the RB, the minimum contribution threshold rather than zero is considered to compute the RB.

21 Information regarding contributions was first gathered in 1980, but is more reliable after 2001. The providers of the sample found that the share of contracts with missing data fell from 78% in 1984 to 94% in 1992 and to 99% in 2003.

Appendix B: Data employed for the labour market transitions.

This appendix details the data employed in the employment transitions. Abstracting from exits and reentering the labour market, they have been simplified to changes among qualification level – being employed - and changes from employment to unemployment and vice versa. Figures B.1 to B.3 show the evolution of the transition hazards observed from 2002 to 2007. Figure B.1 shows the changes in qualification level for workers re-entering employment from unemployment, both for male (panel a) and female (panel b). In all cases the transition hazards remain quite stable during the period covered. Only slight changes – more pronounced in 2004-05 - can be appreciated for males entering from groups 2 and 3 and females entering from group 3 to group 4.

In Figure B.2, the transition hazards from employment to unemployment are shown. In this case, stronger differences can be appreciated among the different years, probably because the Figures show the detail by age. Nevertheless, the trend is also quite stable both for males and for females. Finally, in Figure B.3 hazard rates for transitions between different qualification levels (contribution groups) within employment can be observed. Again, no significant changes are appreciated for the different years analyzed.

As the transition hazards among the different qualification levels are quite stable during the period observed, the value of the last transition observed before the economic crisis (2006 to 2007) is taken and it is held constant for the future. In Section 4.a), an explanation is given of how these hazards are altered during the crisis.

Figure B.1 Changes in qualification level from unemployment to employment Panel a) Male

Figure B.2 Changes in qualification level from employment to unemployment

Figure B.3 Changes in qualification level within employment

Appendix C: Past and future reforms of the pension formula

This appendix summarizes the main features of past and future reforms affecting the pension formula. Past reforms of the pension formula refer to each of the three factors in Equation [1] in the text. First, changes in RB basically relate to the number of past BC used to compute the BR. Only two years were considered when the system started. The number of years increased to 8 in 1985 (26/1985) and to 15 in 1997 (24/1997 Act). The current reform has increased it to 25, still below the recommendation of the Toledo agreement of taking into account the whole contribution history. Second, changes in the cc(n) have been varied and depend very much on the retirement path. Finally, changes affecting p(n), i.e. the weight attached to the number of contribution years, so as to compute the share of the RB received as a pension. Table I.1 summarizes the different legal changes undergone by this scale. According to the present one, fixed by the 24/1997 Act, p(n) is decreasing after the minimum, so that the weight attached to the first years is higher – which results in a redistributive effect. The former scale gave a lower weight to the initial years. The following columns show further reforms in line with the Toledo Agreement proposal, fostering the Bismarckian nature of the system: specifically, full proportionality considering the present maximum of 35 years – that is 2.86% a year - or a maximum of 40 years – that is 2.5% a year. As shown in Table I.1, for the various scales considered, the weight attached to one particular year oscillates between 5% in the first ten years prior to the 1997 reform, to 2% during the last years of the current legislation. The reform of p(n) approved in 2011 is in the same direction but has a small impact, while it complicates the computation now based on months instead of years.

Figure C.1 illustrates the potential effect of reforms on this legal parameter. On the one hand, the function p(n) is plotted for each of the legal scales considered. It is worth noting, first, that individuals who do not meet the minimum eligibility requirement would clearly benefit from a proportional rule. Second, it is clear that for those crediting between 15 and 35 years, both the previous and the present rules (26/1985 and 24/1997 Acts, respectively) are more generous than the two proportional rules.

On the other hand, the cumulative distribution of new registrations with respect to the number of contribution years, in the 2004 MCVL wave, is also shown. This highlights the share of individuals affected by each legal scale and hence its specific effect. First, note that most individuals – 72% – credit up to 35 contribution years and, as such, are affected neither by the legal changes already enacted nor by moving to a system of full proportionality with a maximum of 35 years. Second, we can see that the legal change introduced in 1997 only affected 6% of new pensions. Finally, it is interesting to note that an eventual change to full proportionality would affect almost 50% of individuals, which accounts for the highest effect obtained for this simulated legal change.

Table C.1. Weight attached to contribution years in the share of BR (several legal scenarios)

(*) 60%: according to the same previous scale, 50% from the first 10 years (5% a year) plus 10% from the next 5 years (2% a year).

Figure C.1. Average effect on pensions from fixing different functions of p(n) Source: Authors’ calculations using MCVL data and legal parameters.

Table C.2. The 2011 reform in the pension formula

References

1. 2013-05: “DYPES: A Microsimulation model for the Spanish retirement pension system”, F. J. Fernández-Díaz, C. Patxot y G. Souto.

References

1. 2013-05: “Vertical differentiation, schedule delay and entry deterrence: Low cost vs. full service airlines”, Jorge Validoa, M. Pilar Socorroa y Francesca Medda.

References

1. 2013-04: “Dropout Trends and Educational Reforms: The Role of the LOGSE in Spain”, Florentino Felgueroso, María Gutiérrez‐Domènech y Sergi Jiménez‐Martín.

References

1. 2013-03: “Understanding Different Migrant Selection Patterns in Rural and Urban Mexico”, Simone Bertoli, Herbert Brücker y Jesús Fernández-Huertas Moraga.

References

1. 2013-02: “Understanding Different Migrant Selection Patterns in Rural and Urban Mexico”, Jesús Fernández-Huertas Moraga.

References

1. 2013-01: “Publicizing the results of standardized external tests: Does it have an effect on school outcomes?, Brindusa Anghel, Antonio Cabrales, Jorge Sainz y Ismael Sanz.

References

1. 2012-12: “Visa Policies, Networks and the Cliff at the Border”, Simone Bertoli, Jesús Fernández-Huertas Moraga.

References

1. 2012-11: “Intergenerational and Socioeconomic Gradients of Child Obesity”, Joan Costa-Fonta y Joan Gil.

References

1. 2012-10: “Subsidies for resident passengers in air transport markets”, Jorge Valido, M. Pilar Socorro, Aday Hernández y Ofelia Betancor.

References

1. 2012-09: “Dual Labour Markets and the Tenure Distribution: Reducing Severance Pay or Introducing a Single Contract?”, J. Ignacio García Pérez y Victoria Osuna.

References

1. 2012-08: “The Influence of BMI, Obesity and Overweight on Medical Costs: A Panel Data Approach”, Toni Mora, Joan Gil y Antoni Sicras-Mainar.

References

1. 2012-07: “Strategic behavior in regressions: an experimental”, Javier Perote, Juan Perote-Peña y Marc Vorsatz.

References

1. 2012-06: “Access pricing, infrastructure investment and intermodal competition”, Ginés de Rus y M. Pilar Socorro.

References

1. 2012-05: “Trade-offs between environmental regulation and market competition: airlines, emission trading systems and entry deterrence”, Cristina Barbot, Ofelia Betancor, M. Pilar Socorro y M. Fernanda Viecens.

References

1. 2012-04: “Labor Income and the Design of Default Portfolios in Mandatory Pension Systems: An Application to Chile”, A. Sánchez Martín, S. Jiménez Martín, D. Robalino y F. Todeschini.

References

1. 2012-03: “Spain 2011 Pension Reform”, J. Ignacio Conde-Ruiz y Clara I. Gonzalez.

References

1. 2012-02: “Study Time and Scholarly Achievement in PISA”, Zöe Kuehn y Pedro Landeras.

References

1. 2012-01: “Reforming an Insider-Outsider Labor Market: The Spanish Experience”, Samuel Bentolila, Juan J. Dolado y Juan F. Jimeno.

References

1. 2011-13: “Infrastructure investment and incentives with supranational funding”, Ginés de Rus y M. Pilar Socorro.

References

1. 2011-12: “The BCA of HSR. Should the Government Invest in High Speed Rail Infrastructure?”, Ginés de Rus.

References

1. 2011-11: “La rentabilidad privada y fiscal de la educación en España y sus regiones”, Angel de la Fuente y Juan Francisco Jimeno.

References

1. 2011-10: “Tradable Immigration Quotas”, Jesús Fernández-Huertas Moraga y Hillel Rapoport.

References

1. 2011-09: “The Effects of Employment Uncertainty and Wealth Shocks on the Labor Supply and Claiming Behavior of Older American Workers”, Hugo Benítez-Silva, J. Ignacio García-Pérez y Sergi Jiménez-Martín.

References

1. 2011-08: “The Effect of Public Sector Employment on Women’s Labour Martket Outcomes”, Brindusa Anghel, Sara de la Rica y Juan J. Dolado.

References

1. 2011-07: “The peer group effect and the optimality properties of head and income taxes”, Francisco Martínez-Mora.

References

1. 2011-06: “Public Preferences for Climate Change Policies: Evidence from Spain”, Michael Hanemann, Xavier Labandeira y María L. Loureiro.

References

1. 2011-05: “A Matter of Weight? Hours of Work of Married Men and Women and Their Relative Physical Attractiveness”, Sonia Oreffice y Climent Quintana-Domeque.

References

1. 2011-04: “Multilateral Resistance to Migration”, Simone Bertoli y Jesús Fernández-Huertas Moraga.

References

1. 2011-03: “On the Utility Representation of Asymmetric Single-Peaked Preferences”, Francisco Martínez Mora y M. Socorro Puy.

References

1. 2011-02: “Strategic Behaviour of Exporting and Importing Countries of a Non-Renewable Natural Resource: Taxation and Capturing Rents”, Emilio Cerdá y Xiral López-Otero.

References

1. 2011-01: “Politicians' Luck of the Draw: Evidence from the Spanish Christmas Lottery”, Manuel F. Bagues y Berta Esteve-Volart.

References

1. 2010-31: “The Effect of Family Background on Student Effort”, Pedro Landeras.

References

1. 2010-29: “Random–Walk–Based Segregation Measures”, Coralio Ballester y Marc Vorsatz.

References

1. 2010-28: “Incentives, resources and the organization of the school system”, Facundo Albornoz, Samuel Berlinski y Antonio Cabrales.