De-standardisation and differentiation of retirement trajectories in the context of extended working lives in the Netherlands

Abstract

During recent decades, labour market participation among older workers in the Netherlands has increased significantly. Postponing workers’ labour market withdrawal potentially makes their retirement patterns more uncertain and less predictable. This article uses Dutch register data to analyse de-standardisation and differentiation of retirement trajectories of men and women born between 1940 and 1946 for the age bracket of 59–65 (N = 12,843). The results indicate that retirement trajectories of men have become more homogeneous, whereas those of women somewhat more heterogeneous. Simultaneously, retirement patterns of both men and women became more complex from one birth year to another.

Keywords

De-standardisation differentiation life course policy reforms retirement sequence analysis

Introduction

After decades of declining labour market participation among older workers, the trend towards earlier exit has largely been reversed in many European countries since the late 1990s. Even during the economic crisis of recent years, employment levels of the older segments have continued to rise or remained relatively stable (OECD, 2013). Part of this increasing labour market participation can be attributed to cohort effects: older workers increasingly enjoy better health, have a higher life expectancy and are better educated (OECD, 2006). Some of this increase in activity can also be attributed to trends in reforming pension and social security systems. Early exit has been made less attractive or even impossible, while statutory pension ages are being raised. In addition, ‘active ageing’ strategies have been introduced, aiming at extending people’s working lives (European Commission, 2012; Walker and Maltby, 2012).

Growing employment of the age group of 55+ is generally perceived as a positive development. Using the full potential of the whole population is considered beneficial for all. Older employees can remain active and lead a purposeful existence as part of the labour force. Employers can gain from the experience and knowledge of their older employees. Moreover, higher employment rates result in people being less dependent on benefits and paying more in taxes and contributions, thereby reinforcing the fiscal sustainability of the pension system as well as the welfare state as a whole (Hofäcker, 2010).

There are, however, also potentially unintended effects of extending people’s working lives, resulting in an increase in social risks and inequalities (Anxo et al., 2012; Buchholz et al., 2009). First of all, inequalities exist in the extent to which people have the opportunities and freedom to determine their own timing and mode of retirement (Schils, 2008). More privileged groups may continue enjoying possibilities of early exit with a comfortable pension, whereas others have no choice but to continue working until the statutory basic old-age pension comes within reach. This leads to a growing heterogeneity between individual retirement patterns, or in other words: a de-standardisation of retirement trajectories (Fasang, 2012).

Second, early exit is not always a privilege but can be involuntary, especially when non-employment is the result of poor health or displacement (Dorn and Sousa-Poza, 2010; Van Solinge and Henkens, 2007). Limiting the options of early exit for these older workers may result in insecure late career patterns, where they experience an increase in the number of labour market transitions (between different jobs and social security schemes) in the years preceding retirement, especially as secure jobs are less available after displacement or exiting a career job (Dingemans and Henkens, 2014; Tatsiramos, 2010). This results in a growing complexity within individual retirement patterns, or in other words: a differentiation of retirement trajectories (Fasang, 2012).

The Netherlands constitutes an important study case, as it has witnessed an impressive growth of labour market participation among both men and women over the past two decades. Subsequent governments have introduced a series of policy reforms in early retirement and benefit schemes, aimed – directly and indirectly – at increasing labour market participation among older workers (Visser et al., 2016). In that regard, people’s working lives have effectively been extended in the Netherlands. However, little is known about how these changes have affected the heterogeneity and complexity of retirement trajectories.

The aim of this article is to analyse to what extent an overall increase in labour market participation among older workers in the 2000s has been accompanied by a de-standardisation and differentiation of retirement trajectories of Dutch men and women between the ages of 59 and 65. In order to identify patterns in these retirement trajectories, sequence analysis is applied to municipal register data for a group of men and women born between 1940 and 1946 from the age of 59 until 65. This article is novel in its approach by using measures for de-standardisation (change in state distribution entropy) and differentiation (change in turbulence) to analyse the diversification of retirement trajectories between periods and cohorts. Moreover, this study seeks to disentangle to what extent these changes can be attributed to reforms, cohort effects, or other factors.

De-standardisation and differentiation of (institutionalised) retirement trajectories

In this study, the concept of ‘retirement trajectory’ as a life course pattern is used. Aisenbrey and Fasang (2010) argued that ‘the trajectory is a theoretically superior concept [compared to the discrete transition] because it emphasises that single events should not be isolated from each other but have to be understood in their continuity’ (Aisenbrey and Fasang, 2010: 421). Under this definition, retirement is more than a single transition and rather a sequence of movements in and out of the labour market (Fasang, 2012; Hayward and Grady, 1990). Identifying patterns in these movements enables studying the de-standardisation and differentiation of trajectories.

Fasang (2012) defines a ‘retirement trajectory’ as the ‘sequence of primary income sources within the age bracket during which old-age pension entrance is theoretically possible’ (Fasang, 2012: 687). Retirement trajectories differ from one another in terms of timing, order and complexity of income statuses. For example, the retirement trajectory of someone who retires on an old-age pension at the age of 65 differs, in terms of timing, from a trajectory within which that transition takes place at the age of 63. The trajectory of someone who transitions from employment into retirement differs from the trajectory of someone who ‘unretires’ and re-enters the labour market from retirement. Even though both trajectories consist of the same two statuses (employment and retirement), the order of these statuses vary. The complexity of a trajectory increases with the number of transitions between various statuses. In that sense, a trajectory of someone who is unemployed between being employed and retired (two transitions) is more complex than a trajectory of someone who retires directly after employment (one transition only).

Aisenbrey and Fasang (2010: 448) distinguished two main dimensions in the diversification of trajectories over time. First, de-standardisation occurs when the heterogeneity between individual statuses at each point in time increases (Brückner and Mayer, 2005). Retirement patterns de-standardise when there is growing diversity in the timing or order of transitions in each individual trajectory. Second, differentiation refers to the ‘process where the number of distinct states or stages across the life time increases’ (Brückner and Mayer, 2005: 33). It indicates an increase of the complexity within individual retirement trajectories.

Each individual retirement trajectory can be regarded as unique and the result of a set of individual characteristics (e.g. health, education, marital status) and personal histories (e.g. unemployment spells, child birth). However, the institutional and policy context conditions the life course and to a large extent determines the possibilities for retirement (Aisenbrey and Fasang, 2010; Kohli, 2007). Comparative research has shown that welfare state arrangements and social security schemes are a strong determinant of exit behaviour, even when controlling for individual characteristics (Engelhardt, 2012; Schmitt and Starke, 2016: 2). As a result, welfare state arrangements can lead to the standardisation and institutionalisation of certain types of retirement trajectories (Brückner and Mayer, 2005; Hofäcker, 2010; Kohli and Rein, 1991).

In many European countries, starting in the late 1970s, early retirement schemes, disability benefits and unemployment benefits transformed into institutionalised pathways out from the labour market by forming a bridge for the period between exit from a career job and entry into the normal old-age pension system (Ebbinghaus, 2006; Kohli and Rein, 1991: 6). Whereas a single transition from work to pension at a statutory pension age was previously the norm, the creation of these exit pathways institutionalised a more de-standardised late life course in many countries, (Hardy, 2011; Henretta, 2003). Simultaneously, they led trajectories to differentiate, as the exit pathways added at least one extra transition to trajectories that previously consisted of a single transition between work and retirement (Ebbinghaus, 2006).

Reversing the trend of early exit in the Netherlands

In the 1980s and 1990s the Netherlands was a prime example of a country with a culture of early exit (Henkens and Kalmijn, 2006; Hofäcker, 2010: 242). Large numbers of older workers were exiting the labour market before the official retirement age of 65, mainly through early retirement schemes and disability benefits, as well as to some extent through unemployment benefits. By 1995, 42.5% of the people between ages 55 and 65 were non-employed through one of the exit pathways, with 23.5% in disability benefits, 13.3% in pre-retirement (VUT) schemes, and 5.7% in unemployment benefits (Ebbinghaus, 2006: 138). Furthermore, in the Dutch ‘male-breadwinner society’, labour market participation among women had always been low in all age groups.

By the mid-1990s, early exit pathways became less of a labour market solution and more of a problem due to the increasing pressure on public finances and the squandering of potential in the labour force (Hartlapp and Kemmerling, 2008; Schmitt and Starke, 2016, Van Oorschot, 2007). Subsequent Dutch governments therefore initiated a series of reforms to close alternative exit pathways and increase labour market participation. Table 1 summarises the most important reforms that were introduced since the mid-1990s to prevent early exit through the alternative pathways.

Table 1.

Major reforms in the Dutch social security system that affected early exit pathways in the 2000s.

Year of reform	Exit route	Type of reform
1997	Early retirement	Replacement of more generous VUT early exit arrangements by more actuarially neutral pre-pension schemes
2002	Sickness and disability benefits	‘Law on the improvement of gatekeepers’, enhanced screening in order to reduce the influx into sickness and disability benefits
2003	Unemployment benefits	Abolishment of ‘follow-up benefits’ that extended the period of entitlement for those over 57½
2004	Sickness and disability benefits	Adoption of stricter set of criteria for assessing disability
	Sickness and disability benefits	Shift of responsibility for an employee’s sickness absence from state to the employer for the first two years of sickness
	Unemployment benefits, social assistance	Abolishment of exemption from job-search requirements for those over 57½
2006	Sickness and disability benefits	Comprehensive reform of disability benefits system, introduction of new law on work and income according to labour potential
2006	Early retirement	Abolishment of tax exemption for pre-pension schemes
2008	Unemployment benefits	Tightening of requirements to take up work that mismatches one’s qualifications

Sources: Based on Ebbinghaus (2006), Van Gestel et al. (2009), Van Oorschot (2007), Visser et al. (2016) and Vrooman (2010).

The most comprehensive reforms took place in early retirement schemes and disability benefits. In 1997, the VUT was abolished and replaced by less generous and actuarially more neutral pre-pension schemes. VUT funds, which are tied to sectors and governed by trade unions and employer representatives, did not disappear immediately, but their numbers have since then been declining (Visser et al., 2016). The abolishment of the tax exemption for pre-pension schemes in 2006 further reduced the attractiveness of early retirement.

Sickness and disability benefits were reformed in several rounds. Medical screening for the take-up of disability benefits was made stricter by the reforms of 2002 and 2004. The main responsibility for return from sickness to work of the employee was transferred to employers in 2004, increasing the costs for employers in the case of long-term sickness absence (Van Gestel et al., 2009; Vrooman, 2010). In 2006, the whole disability benefit system was overhauled and the new law on work and income according to labour potential was introduced. This meant that only those fully incapacitated are entitled to full disability benefits. Otherwise, the employer has the responsibility to find work according to the employee’s capabilities.

Until their abolishment in 2003, so-called ‘follow-up benefits’ offered older workers a flat-rate unemployment benefit of three and a half years after the maximum period of earnings-related benefits of five years expired, functioning as a bridge between standard unemployment and (early) retirement (Van Oorschot, 2007). Since 1 January 2004, older workers who end up in unemployment benefits or social assistance are subjected to the same job-search requirements as those younger than 57.5 years old (De Graaf-Zijl and Hop, 2007; Lammers et al., 2013). Finally, in 2008 requirements for unemployed of all ages were tightened to accept work that is not necessary in line with one’s qualifications.

Overall, there has been a significant increase in labour market participation of older workers in the 2000s. The net labour market participation rate of men aged 60–65 has grown from 17.5% in 1995 to 54.2% in 2013. For women in the same age group, this increase was from 5.1% to 29.2% (Statistics Netherlands, 2015). The latter can be much less attributed to the reforms of exit pathways than to the steady inflow of women into the labour market since the 1980s, especially into part-time employment (Arts and Otten, 2013).

Studies have found that these changes can to a certain extent be attributed to reforms of exit pathways. The reform of VUT into pre-pensions has induced workers to postpone retirement (Euwals et al., 2010). Visser et al. (2016) found a relation between the steady decrease of early retirement funds and increasing labour market participation for the 55–65 age group. Incidence of disability benefits among 55–65 years old has started to decrease since they reaching its peak level in December 2003, especially among men (Arts and Otten, 2013). Euwals et al. (2012), in their study on the Dutch health care sector, found a relation between the closing off of disability as an exit pathway and an increase in labour market participation. Incidence of unemployment benefits for the same age group shows more cyclical movements in reaction to changing unemployment rates. Still, Lammers et al. (2013) found that after the introduction of stricter job-search requirements, a higher outflow from unemployment benefits took place.

Scenarios of de-standardisation and differentiation of Dutch retirement trajectories

Little is known about how these changes have affected the heterogeneity and complexity of retirement patterns. The aim of the replacement of the VUT with pre-pension schemes, the stricter rules applying to disability and the abolishment of follow-up unemployment benefits has been to increase the share of the population that is employed until the age of 65 and only then enters into old-age retirement. All other things equal, this should have led to a standardisation of retirement trajectories, i.e. a synchronisation of employment and retirement statuses, as well as of the timing of a single transition into retirement. In this scenario, patterns also should have become less complex. As intermediary statuses of disability and unemployment have been removed, two transitions (e.g. employment → disability → retirement) have been reduced to one only (employment → retirement).

However, several other scenarios were likely to happen. These scenarios are not necessarily mutually exclusive and are even possible all at the same time. First of all, in the literature, exit pathways have been found to function as substitutes. Closing off one pathway potentially opens up another through which workers continue to find a way to exit the labour market early (Ebbinghaus, 2006: 148; Guillemard and Van Gunsteren, 1991: 364). In the Netherlands, for example, there has been some indication that unemployment rates have been higher as a result of more restricted access to early retirement and disability benefits (Visser et al., 2016). In this scenario, no changes in the heterogeneity and complexity of retirement patterns can be expected to have taken place, as one status is simply replaced by another.

Second, the Dutch reforms were not necessarily aimed at preventing the inflow to sickness, disability and unemployment benefits, but rather at shortening spells and incentivising outflow. Therefore, it can be expected that the number of transitions and, as a result, the complexity of trajectories has remained the same, even though the spell durations have changed. However, as the timing of the transitions in this scenario has become more scattered, trajectories can be expected to have de-standardised.

Third, shortening spells of benefit entitlements does not necessarily mean that by the time of their expiration people have jobs or are suddenly capable of working. As a result, they end up in inactivity or enter another benefit scheme. Lammers et al. (2013), for example, found that after the introduction of stricter job-search requirements in 2004, a higher outflow from unemployment to sickness and disability took place. Moreover, displacement among older workers has been found to decrease the likelihood of re-employment (Tatsiramos, 2010) and to shorten employment spells after re-employment (Chan and Stevens, 1999). Hence, in this scenario, patterns are expected not only to have de-standardised, but also differentiated, as more transitions between a wider variety of status types have taken place.

Fourth, even when remaining employed until the statutory retirement age without experiencing unemployment or sickness spells, the number of people working full-time in their career jobs until the statutory retirement age has been decreasing. Instead, various types of gradual retirement have been on the rise. These can include phased retirement, i.e. working fewer hours for the same employer, or partial retirement, i.e. moving to a less demanding ‘bridge job’ or becoming self-employed (Bloemen et al., 2016). Moreover, people might ‘unretire’ and return to employment after already having been retired. Under these scenarios, it can be hypothesised that retirement trajectories have de-standardised, especially in the case of unretirement, and differentiated, as the variety of status types has increased.

Finally, the increase in labour market participation of women during the past decades has most likely contributed to a standardisation of retirement patterns between men and women. At the same time, within the female population it is expected to have led to greater de-standardisation and differentiation. Instead of consisting of a single homogeneous inactive group, women are now to a larger extent employed, although with an increased risk of unemployment, sickness and disability. Moreover, even though employment rates among women have increased, they still work predominantly part-time and tend to experience more and longer breaks during their careers than men, due to an unequal division of family care responsibilities (Fasang et al., 2013; Widmer and Ritschard, 2009). This makes their labour market position more vulnerable and causes a larger variation in their retirement possibilities and entitlements (Radl, 2013).

Data and methods

Data

To analyse (de-)standardisation and differentiation of retirement trajectories during the 2000s, the SECMBUS register data from Statistics Netherlands on monthly primary sources of income was used. These data were available for the years 1999–2011 at the time of the research and consist of spells of each type of main income source for the population included in municipal registers, tax authorities and social insurance institutions. There were in total 11 types of income in the data, which have been recoded into six main types: employment (including self-employment), sickness and disability, unemployment, social assistance, retirement and other. For the 59–65 age group in this study, the last category mainly includes those who have no source of income of their own.

One limitation of analysing primary sources of income was that it was impossible to distinguish between cases where there is more than one source of income. Moreover, the data did not contain any information about whether employment was full-time or part-time, whereas the difference between the two might be relevant, especially for women. It should also be noted that, unfortunately, transitions between different jobs or forms of employment were not visible in these data. In spite of such limitations, the data offered the possibility to measure the heterogeneity between income statuses at each point in time, as well as the complexity of spell sequences for each individual.

The study focused on the Dutch population born between 1940 and 1946, which is followed from the year of turning 59 (i.e. taking place in the seven consecutive years from 1999 until 2005) until the year of turning 65 (i.e. 2005–2011). Within the limitations of data availability, this was considered an optimal solution for the aims of this study. First, there were seven birth years which should allow taking into account cohort and period effects. Second, each individual was followed for a period of seven years before reaching the statutory retirement age of 65. A longer follow-up period, e.g. 10 years, would have been more ideal, but would have decreased the number of birth years that can be taken into account. Third, having the oldest cohort followed for the period before the major reforms were completed (1999–2005) and the youngest for the period when most major reforms had been fully implemented (2005–2011), offered the opportunity to take into account the changes in the policy context over time.

With approximately 1.3 million individuals in the population for the period under investigation, a random sample of 1% was drawn, yielding a total of 12,843 individuals, among which were 6474 men and 6369 women. Most research on retirement behaviour either focuses on men only or takes men and women as two strictly separate groups, as retirement timing and patterns have been found to be different for men and women (Han and Moen, 1999; Radl, 2013). As it was hypothesised that de-standardisation and differentiation might occur at different rates for men and women, changes in their retirement trajectories were analysed separately.

Measuring de-standardisation and differentiation

To analyse the retirement trajectories of each cohort of men and women, sequence analysis was applied. Sequences can be defined as ordered lists of states (Abbott, 1995: 94). Those states were, in the case of this study, main sources of income on a monthly basis. Sequence analysis is a method within the algorithmic statistical tradition. It allows to detect patterns in data and identify the processes that produce them, without making prior assumption about the processes that generate the data (Aisenbrey and Fasang, 2010: 425). Sequence analysis is closely related to the theoretical trajectory concept, as it is a tool to identify typologies within longitudinal categorical data. Moreover, the possibilities for visualisation offer an insightful way of presenting the data.

To measure the extent of de-standardisation and differentiation of trajectories, the changes over time in respectively heterogeneity (variation between trajectories) and complexity (variation within trajectories) were analysed. Figure 1 depicts how the concepts of heterogeneity and complexity are operationalised. The number of different individual states is situated on the y-axis and time is indicated on the x-axis. For each point in time t_j, each individual has a certain state (employed, retired, etc.). Each individual has a unique sequence of states x_i that stretches out over time.

Figure 1.

Measuring trajectory heterogeneity and complexity.

Heterogeneity is measured at each t (in this case: month) by calculating the state distribution entropy indicator. State distribution entropy at any point in time is calculated by the following equation:

h (p_{1}, \dots, p_{s}) = - \sum_{i = 1}^{s} p_{i} l o g (p_{i})

where p_i is the proportion of cases in state i at the considered time point and s is the number of possible states (Gabadinho et al., 2010: 67). Entropy equals 0 when all respondents are in the same state and is 1 when all cases are equally distributed among all state combinations (Fussell, 2005). Using the TraMineR package in R, the state distribution entropy indicator was calculated for each month between the ages 59 and 65 for each of the seven birth cohorts (Gabadinho et al., 2010).

The concept of trajectory complexity is operationalised by Elzinga’s turbulence indicator (Elzinga, 2010; Elzinga and Liefbroer, 2007). Turbulence for any sequence x is calculated by

T (x) = \log_{2} (φ (x) \frac{s_{t, m a x}^{2} (x) + 1}{s_{t}^{2} (x) + 1})

where φ(x) is the number of distinct sub-sequences, $s_{t}^{2}$ is the variance of the state duration and $s_{t, m a x}^{2}$ is the maximum value that this variation can take given the length of the sequence. The maximum is calculated by

s_{t, m a x}^{2} = (n - 1) {(1 - \bar{t})}^{2}

with $\bar{t}$ being the mean consecutive time spent in the distinct states (Elzinga, 2010; Gabadinho et al., 2010: 85).

In the context of this study the turbulence indicator is superior to other measures of sequence complexity, such as counting the number of transitions or estimating the within-sequence Shannon entropy (Gabadinho et al., 2010: 77). The main asset of the turbulence indicator is that it measures complexity within sequences by not only taking into account the variety of states, but also the time spent in various states and the variation of these durations. Sequence turbulence increases when longer spells are spent in different states, but the more time is spent in one particular state, the less turbulent the sequence is (Elzinga and Liefbroer, 2007: 233).

The lengths of the spells matter, as the reforms as they were implemented in the Netherlands might not necessarily be effective in preventing the transition into sickness or unemployment benefits. To a large extent, reforms have been aimed rather at improving outflow, i.e. shortening spells. Therefore, if exit pathways have been effectively closed off, it is expected that turbulence increases not only because of an increase in the number of transitions, but also because of a shortening of the spells of residing in unemployment, disability and retirement. Finally, the turbulence indicator captures whether someone returns to a previous state (e.g. employment after a period of unemployment) after a certain spell or transitions into yet another state (e.g. employment → unemployment → disability). This enables identifying the second case as more turbulent than the first.

The average turbulence of each birth cohort for each follow-up period of seven years was calculated, while splitting the results for gender, education and labour market status at the time of turning 59. Education is of importance, as it was expected that lower skills will lead to more turbulent trajectories due to a higher likelihood of being employed in declining industries and precarious jobs. Level of education is divided into three categories: lower, middle and higher. Unfortunately, data on education were only available for a smaller segment of the sample (n = 2155). Income status at the age of turning 59 was used to indicate different labour market statuses at the beginning of the trajectory. It can be expected, for example, that someone who has already retired at the age of 59 will have a much less turbulent trajectory than someone who is still employed.

Apart from several plots displaying the turbulence averages, a series of two-way ANOVA models were run to analyse the differences in turbulence by cohort, gender, income status and education. The aim of these models was to check whether there are significant differences in turbulence between the consecutive cohort years, when controlled for other factors. Interactions were included for cohort year with gender and cohort year with labour market status, in order to analyse whether within certain groups the change in turbulence over time has been greater or smaller.

Level of education was used in order to control for cohort effects. The average level of education tends to increase with each cohort (Herd, 2006), whereas higher levels of education have been found to decrease the risks of unemployment and poor health, while increasing the likelihood of re-employment and longer working lives (Järnefelt, 2010; Schuring et al., 2013). Controlling for the level of education allows estimating whether changes in turbulence between cohort years were the effect of reforms or rather the result of a cohort being more highly educated and therefore having longer employment spells with lower risks of experiencing spells of unemployment, sickness or disability.

Findings

Increase in labour market participation

Figures 2a –d show the state distribution plots for the male and female populations and the oldest (born in 1940) and youngest (born 1946) cohorts. On the x-axis, time in months from the beginning of the year of turning 59 (t = 1) until the end of the year of turning 65 (t = 84) is represented. The y-axis indicates the percentage of individuals receiving their primary income from a particular source in each particular month. The difference between men and women is obvious for both birth years: men had higher employment rates and were more often recipients of disability benefits. Women, on the other hand, were more often without income (other) and more likely to receive social assistance.

Figure 2a.

State distribution plot of men from the age of 59 (month = 1) until 65 (month = 84), birth year 1940.

Figure 2b.

State distribution plot of men from the age of 59 (month = 1) until 65 (month = 84), birth year 1946.

Figure 2c.

State distribution plot of women from the age of 59 (month = 1) until 65 (month = 84), birth year 1940.

Figure 2d.

State distribution plot of women from the age of 59 (month = 1) until 65 (month = 84), birth year 1946.

Analysing the changes between cohorts, the graphs confirm that employment for both men and women has increased. This change was especially due to a decrease in disability benefit take-up for men and decreasing levels of inactivity and, to a lesser extent, social assistance among women. Simultaneously, however, the proportion of early retirees before the age of 64 has increased somewhat among both men and women. Among women, there has been a slight increase in unemployment. Nevertheless, the graphs confirm that the average working life among both genders has been extended in the Netherlands, even within this relatively short period of seven cohort years.

De-standardisation?

Figures 3a and b show the state distribution entropy indicator as a measure of heterogeneity for every six months between ages 59 and 65 for each of the cohort years, as well as for men (Figure 3a) and women (Figure 3b) separately. The higher the bars, the higher the levels of heterogeneity. First of all, it is shown that levels of heterogeneity were on average higher for women than for men. This indicates that at almost any point in time there was greater dissimilarity of income statuses among women than among men. Moreover, this heterogeneity remained rather constant for women until the age of 64, whereas for men there has been a slight rise until month 30 (age 61.5), after which it started to decline. This suggests that the income statuses of men at the age of 59–61 were more homogeneous, which can be explained by the majority of them still being employed at that point in time. After the age of 61.5 heterogeneity increased, as larger sections of men entered into especially disability benefits and (early) retirement. This was followed by a period in which the share of men in (early) retirement took the upper hand and heterogeneity again decreased somewhat. There was no such fluctuation in heterogeneity among women until 64 years of age, after which heterogeneity among them rapidly decreased, as a large proportion among them entered into retirement.

Figure 3a.

State distribution entropy per every six months, men.

Figure 3b.

State distribution entropy per every six months, women.

Figure 3a shows that for men status distributions have become more standardised over time, especially in the age bracket 59–62. After the age of 62, the standardisation trend levelled out somewhat, but heterogeneity remained at a lower level than among the older cohorts. For women, there has been no such clear change. There are differences between the cohorts, but no clear trends towards either de-standardisation or standardisation. There are some signs of de-standardisation between ages 62 and 64, but it is unclear whether the differences between the youngest and oldest cohorts are significant. The lack of a clear trend for women, in spite of the large movement between cohorts from inactivity into employment, has likely been the result of the shift out from inactivity being accompanied with a simultaneous and equal shift into employment, leaving the state distribution entropy indicator close to being unaltered.

Differentiation?

Figures 4a –c show the average turbulence for each cohort year, grouped for various individual characteristics. Figure 4a indicates that for the total population, as well as both men and women separately, there was a slight differentiation (i.e. an increase of complexity of trajectories) over time. A one-way ANOVA test indicated that there was a difference between the birth years, F = 5.24 (d.f. = 12,842), p < .01, and a posthoc test indicated that the average turbulence of birth year 1940 differed significantly from 1945 and 1946, while 1941 differed significantly from the average of those born in 1946 (not reported here). The gradual increase in turbulence across birth years could indicate that it is the result of the phased implementation of reforms throughout the early 2000s, but that it could also be the result of cohort and period effects. However, the finding that the turbulence within the 1946 cohort, which experienced the full range of reforms, was significantly higher than that within the predominantly pre-reform 1940 and 1941 cohorts, strongly suggests that the total package of reforms did have an effect.

Figure 4a.

Average turbulence by gender, cohorts 1940–1946.

Figure 4b.

Average turbulence by level of education, cohorts 1940–1946.

Figure 4c.

Average turbulence by income status at t = 1, cohorts 1940–1946.

Men’s trajectories were overall more turbulent than those of women. There is no evidence that those with lower levels of education would have had more complex trajectories at any point or experienced a higher degree of differentiation over time (Figure 4b). There were differences in turbulence according to income status at the time of turning 59 (Figure 4c). Those who were unemployed and employed were more likely to have more complex trajectories towards retirement. Those who were already retired obviously had the least complex trajectories, whereas those who were receiving social assistance, disability benefits or no income were somewhere in between. Only in the cases of employment and unemployment, might one detect a trend of differentiation between the older and younger cohorts.

A series of two-way ANOVA tests were run to analyse whether there were differences in trajectory complexity between the genders and the income status at the age of 59, as well as whether differentiation over time can be explained by belonging to any of these groups. Moreover, it was checked whether it could be assumed that differentiation is explained by the reforms of early exit or whether this was the result of a cohort effect of an increasingly more skilful workforce. Model 1 in Table 2 confirms that there were significant differences between birth years, gender and income status at the beginning of the trajectories. However, Model 2 incorporating interaction terms between birth year and gender as well as birth year and income status, shows that there was no difference in the pace of differentiation between these groups. In other words, there was no evidence either that women experienced a more rapid increase in turbulence than men over time, or that the pace of change in turbulence was different for each of the groups defined by labour market status at the beginning of the trajectory.

Table 2.

Two-way ANOVA tests for levels of turbulence.

Model 1	SS	d.f.	MS	F
Year of birth	66.23	6	11.04	2.28*
Gender	56.74	1	56.74	11.71**
Income status	12,601.85	5	2520.37	520.29**
Corrected model	14,054.77	12	1171.23	241.78**
n = 12,709, R² = .19
Model 2
Year of birth	75.21	6	12.54	2.59*
Gender	53.21	1	53.21	10.99**
Income status	12,155.06	5	2431.01	501.98**
Year of birth × gender	12.58	6	2.10	0.43
Year of birth × income status	179.59	30	5.99	1.24
Corrected model	14,245.33	48	296.78	61.28**
n = 12,709, R² = .19
Model 3
Year of birth	47.16	6	7.86	1.39
Gender	95.86	1	95.86	16.98**
Income status	1480.70	5	296.14	52.47**
Year of birth × gender	52.16	6	8.69	1.54
Year of birth × income status	141.58	30	4.72	0.84
Level of education	0.43	2	0.216	0.04
Corrected model	2260.18	50	45.20	8.01**
n = 2155, R² = .16

p < .05, ** p < .01.

Finally, the level of education was included as a control variable for the possibility that (a lack of) differentiation is related to longer working lives as a result of increasing skill levels. The average level of education of the cohorts increased significantly over time (not shown), but, as Model 3 shows, this had no effect whatsoever on the turbulence levels, F = 0.04 (2). In this Model the effect of birth year became insignificant, but this was more likely due to a decrease in sample size and degrees of freedom than to the inclusion of the education variable, which had an F-value that was nowhere close to indicating a significant difference between groups.

Discussion

Since the late 1990s, the Netherlands has seen an impressive growth of labour market participation among older workers. Part of this can be ascribed to cohort effects: a healthier and more educated population is able to work longer. Moreover, coming from very low participation rates in the early 1980s, women have been catching up and have delivered a strong contribution to the growth of labour market participation, with the inclusion of the 55–65 age segment. At the same time, a set of comprehensive reforms in pensions, early retirement schemes, disability benefits and unemployment insurance were aimed at closing off alternative early exit pathways and extending working lives. Studies have found that these single reforms have to some extent been effective in postponing labour market exit and have contributed to re-employment of older workers (Euwals et al., 2010, 2012; Lammers et al., 2013; Visser et al., 2016).

The findings of the present study confirmed that, when comparing the 1940 cohort with the 1946 cohort, increases in employment among men largely came at the expense of sickness and disability. However, there potentially was some substitution effect of disability benefits with early retirement: the incidence of retirement between the ages of 62 and 64 had increased somewhat. This also suggests that by 2005–2006 the reforms in disability and sickness had come into full effect for the youngest cohort, whereas the phasing out of the VUT schemes and restricting of pre-pensions were still in progress.

Women in the 1946 cohort were especially less likely to be inactive or on social assistance and more likely to be employed than those born in 1940. However, at the same time, their incidence of unemployment had increased somewhat. This suggests that the increase in labour market participation among women has also increased their risk of becoming unemployed, while their longer careers might have expanded their eligibility under the earnings-related unemployment benefit scheme (Schils, 2008).

The effects of these changes on the heterogeneity between individual trajectories were somewhat mixed and did not show an unambiguous trend towards either standardisation or de-standardisation. Among men up to the age of 62, individual trajectories towards retirement have standardised. This suggests that the use of various alternative exit pathways, and in particular sickness and disability benefits, has decreased between cohorts for this age group. However, after the age of 62 almost no difference between cohorts was noticeable. Possibly, early retirement schemes acted as a substitute for the closed-off disability pathway, leaving existing levels of heterogeneity intact. Towards the age of 65, no trend towards de-standardisation was discernible, suggesting that, if there had been a trend towards ‘unretirement’, this had no effect on heterogeneity levels.

Women’s retirement patterns showed to be more heterogeneous than men’s across ages and cohorts. Among women there has been no major (de-)standardisation from the older to the younger cohorts, although between the ages of 62 and 64 a cautious trend towards de-standardisation is visible. This might be due to the fact that, as women’s labour market participation has risen since the 1980s, they have enjoyed longer careers and have been accumulating entitlements to unemployment, disability and early retirement schemes. Considering that a substantial part of the influx of women into the labour market has been in the form of part-time work, a stronger trend towards de-standardisation might have been detected if data on part-time employment as a separate state had been available.

Retirement trajectories have become more complex and unpredictable, as the results for the analysis of differentiation show. For the whole population, as well as for both men and women separately, trajectories have differentiated. This means that during the period of being 59–65 years old, the younger cohorts had more complex trajectories than the older cohorts, indicating that they made more transitions between employment and various social security schemes, as well as having longer spells in a greater variety of statuses. The results suggested that differentiation was due to reforms, as the youngest cohort born in 1946 and exposed to the full range of new regulations showed significant differences in turbulence with the oldest 1940 and 1941 cohorts.

Especially the trajectories of men and those who were employed or unemployed at the age of 59 were more and increasingly complex, which indicates that labour market participation has become accompanied with more labour market risks. Those who had already been in disability or retirement were less affected, as reforms were mainly concentrated on new cases and not on the re-employment of those already outside the labour market. The results also suggested that, if reforms caused spells of unemployment and disability to shorten, these episodes were not necessarily followed by a return to employment or, if they were, longer spells of employment. However, more research will be needed to analyse if certain new types of status sequences have become common as a result of the reforms.

The Dutch reforms eliminated some of the ‘pull factors’ of early exit, namely by making early exit through early retirement, disability and unemployment programmes less attractive or even impossible. Hence, although successful from an economistic or productivist perspective, reforms did not necessarily address the ‘push factors’ of early exit, such as sectoral decline, obsoleteness of skills, deteriorating health or age discrimination (De Preter et al., 2013; Ebbinghaus, 2006; Van Oorschot and Jensen, 2009). Younger cohorts did attain on average higher levels of education. Higher skill levels have the potential of increasing older workers’ resilience in the face of labour market risks and have been found to be associated with better health (Leopold and Engelhardt, 2011; Schuring et al., 2013). However, no effects of education on the complexity of retirement trajectories were found. Further research will be needed to investigate how health, education and their possible interaction relate to de-standardisation and differentiation, and how more comprehensive active ageing strategies might affect these (Boudiny, 2013; Walker and Maltby, 2012).

The Netherlands provides an instructive case of how increasing labour market participation coincides with changes in the heterogeneity and complexity of retirement trajectories. Still, it has to be borne in mind that the findings in this study are not automatically generalisable to other countries. Certain features of the Dutch welfare state and labour market are regime- or even country-specific and retirement needs to be studied in this context (Buchholz et al., 2009; Ebbinghaus, 2006; Van Oorschot and Jensen, 2009). More comparative research is needed on how the presence of other factors in the institutional setting, including for example job security or the presence of active labour market policies, affects changing trajectories. It would be especially interesting to compare countries that reformed early exit to different extents: from closing off single pathways, like in France, Germany or Italy in the 1980s (Ebbinghaus, 2006: 148), to more recent comprehensive reforms in, for example, Sweden in the late 1990s and early 2000s (Karlström et al., 2008) or Finland in the mid-2000s (Kyyrä, 2015).

Recent policy developments in early retirement in the Netherlands also require further investigation. While in 2012 the short-lived Life Course Savings Scheme (Levensloopregeling) was abolished, its foreseen substitute, the Vitality Scheme (Vitaliteitsregeling), did not come into existence for budgetary reasons. With no policies in place that allow early retirement for the younger generations, transitions into disability and unemployment benefits might be on the rise again in the near future. Moreover, one of the limitations of this study is that its subjects were born in the years 1940–1946, hence predominantly during the deprived years of the Second World War and German occupation. It is not unlikely that this has had specific and remaining effects on their health, education and working lives. Finally, the real test for the reforms of early exit began when the baby-boom generation started to approach retirement in more recent years. The availability of new data should allow for analysing whether the detected trends in de-standardisation and differentiation have continued for this generation.

Working longer is a double-edged sword. While the benefits for the financing of pension systems and social security are highlighted in the public discourse, the risks are often under-exposed. Increasing labour market participation among older workers does not have the same outcomes for everyone. Some might still have the private or collective means to retire early and voluntarily, whereas others are forced to continue work in low-quality and insecure jobs or work in spite of health problems. Although researching the changing heterogeneity and complexity of retirement trajectories does not directly provide evidence for these inequities, it can give an indication of the uncertainties and insecurities that longer working lives may entail. Previous research has shown, for example, that the loss of a career job in one’s late career can negatively affect income as well well-being in case of re-employment or taking up a ‘bridge job’ (Chan and Stevens, 1999; Dingemans and Henkens, 2014). Policy-makers might want to consider the broader spectrum of outcomes of their reforms. It is one thing to restrict early exit, but yet another thing to equip all older workers with skills and opportunities to perform meaningful work until the statutory retirement age comes within reach.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research leading to these results has received funding from the European Union’s 7th Framework Programme (FP7/2007-2013) under grant agreement n° 262608, DwB – Data without Boundaries.

Author biography

Aart-Jan Riekhoff is a PhD student in Social Policy at the University of Tampere, Finland. He is writing his thesis on changing retirement patterns in the Netherlands and Finland with the use of longitudinal register-based data. His main research interests are related to labour market policies and socio-economic inequalities.

References

Abbott

(1995) Sequence analysis: New methods for old ideas. Annual Review of Sociology 21: 93–113.

Aisenbrey

Fasang

(2010) New life for old ideas: The ‘second wave’ of sequence analysis bringing the ‘course’ back into the life course. Sociological Methods and Research 38(3): 420–462.

Anxo

Ericson

Jolivet

(2012) Working longer in European countries: Underestimated and unexpected effects. International Journal of Manpower 33(6): 612–628.

Arts

Otten

(2013) Stijgende arbeidsparticipatie en minder uittreding bij ouderen. Sociaaleconomische Trends No. 2013/4. The Hague: Centraal Bureau voor de Statistiek.

Bloemen

Hochguertel

Zweerink

(2016) Gradual retirement in the Netherlands: An analysis using administrative data. Research on Aging 38(2): 202–233.

Boudiny

(2013) ‘Active ageing’: From empty rhetoric to effective policy tool. Ageing and Society 33(6): 1077–1098.

Brückner

Mayer

(2005) De-standardization of the life course: What might it mean? And if it means anything, whether it actually took place? In: MacMillan

(ed.) The Structure of the Life Course: Standardized? Individualized? Differentiated? Amsterdam: Elsevier, pp. 27–53.

Buchholz

Hofäcker

Mills

, et al. (2009) Life courses in the globalization process: The development of social inequalities in modern societies. European Sociological Review 25(1): 53–71.

Chan

Stevens

(1999) Employment and retirement following a late-career job loss. The American Economic Review 89(2): 211–216.

10.

De Graaf-Zijl

Hop

(2007) 45-Plus en 55-Plus in de SUWI-Keten. Hoe Vergaat het Oudere Werklozen op Zoek naar Werk? Amsterdam: SEO Economisch Onderzoek.

11.

De Preter

Van Looy

Mortelmans

(2013) Individual and institutional push and pull factors as predictors of retirement timing: A multilevel analysis. Journal of Aging Studies 27(4): 299–307.

12.

Dingemans

Henkens

(2014) Involuntary retirement, bridge employment, and satisfaction with life: A longitudinal investigation. Journal of Organizational Behavior 35(4): 575–591.

13.

Dorn

Sousa-Poza

(2010) ‘Voluntary’ and ‘involuntary’ early retirement: An international analysis. Applied Economics 42(4): 427–438.

14.

Ebbinghaus

(2006) Reforming Early Retirement in Europe, Japan and the USA. Oxford: Oxford University Press.

15.

Elzinga

(2010) Complexity of categorical time series. Sociological Methods and Research 38(3): 463–481.

16.

Elzinga

Liefbroer

(2007) De-standardization of family-life trajectories of young adults: A cross-national comparison using sequence analysis. European Journal of Population 23(3–4): 225–250.

17.

Engelhardt

(2012) Late careers in Europe: Effects of individual and institutional factors. European Sociological Review 28(4): 550–563.

18.

European Commission (2012) EEO Review: Employment Policies to Promote Active Ageing 2012. Luxembourg: Publications Office of the European Union.

19.

Euwals

Van Vuren

Van Vuuren

(2012) The decline of substitute pathways into retirement: Empirical evidence from the Dutch health care sector. International Social Security Review 65(3): 101–122.

20.

Euwals

Van Vuuren

Wolthoff

(2010) Early retirement behaviour in the Netherlands: Evidence from a policy reform. De Economist 158(3): 209–236.

21.

Fasang

(2012) Retirement patterns and income inequality. Social Forces 90(3): 685–711.

22.

Fasang

Aisenbrey

Schömann

(2013) Women’s retirement income in Germany and Britain. European Sociological Review 29(5): 968–980.

23.

Fussell

(2005) Measuring the early adult life course in Mexico: An application of the entropy index. In: MacMillan

(ed.) The Structure of the Life Course: Standardized? Individualized? Differentiated? Amsterdam: Elsevier, pp. 91–122.

24.

Gabadinho

Ritschard

Studer

, et al. (2010) Mining Sequence Data in R with the TraMineR Package: A User’s Guide. Geneva: University of Geneva.

25.

Guillemard

A-M

Van Gunsteren

(1991) Pathways and their prospects: A comparative interpretation of the meaning of early exit. In: Kohli

Rein

Guillemard

M-A

, et al. (eds) Time for Retirement: Comparative Studies of Early Exit from the Labor Force. Cambridge: Cambridge University Press, pp. 362–387.

26.

Han

S-K

Moen

(1999) Clocking out: Temporal patterns of retirement. American Journal of Sociology 105(1): 191–236.

27.

Hardy

(2011) Rethinking retirement. In: Settersten

Jr Angel

(eds) Handbook of Sociology of Aging. New York, Dordrecht, Heidelberg and London: Springer, pp. 213–227.

28.

Hartlapp

Kemmerling

(2008) When a solution becomes the problem: The causes of policy reversal on early exit from the labour force. Journal of European Social Policy 18(4): 366–379.

29.

Hayward

Grady

(1990) Work and retirement among a cohort of older men in the United States, 1966–1983. Demography 27(3): 337–356.

30.

Henkens

Kalmijn

(2006) Labour market exits of older men in the Netherlands, an analysis of survey data 1979–1999. In: Blossfeld

Bucholz

Hofäcker

(eds) Globalization, Uncertainty and Late Careers in Society. London and New York: Routledge, pp. 79–99.

31.

Henretta

(2003) The life-course perspective on work and retirement. In: Settersten

RA Jr

(ed.) Invitation to the Life Course: Toward New Understandings of Later Life. Amityville, NY: Baywood Publishing Company, pp. 85–105.

32.

Herd

(2006) Do functional health inequalities decrease in old age? Educational status and functional decline among the 1931–1941 birth cohort. Research on Aging 28(3): 375–392.

33.

Hofäcker

(2010) Older Workers in a Globalizing World: An International Comparison of Retirement and Late-Career Patterns in Western Industrialized Countries. Cheltenham and Northampton, MA: Edward Elgar.

34.

Järnefelt

(2010) Education and Longer Working Lives: A Longitudinal Study on Education Differences in the Late Exit from Working Life of Older Employees in Finland. Helsinki: Finnish Centre for Pensions.

35.

Karlström

Palme

Svensson

(2008) The employment effect of stricter rules for eligibility for DI: Evidence from a natural experiment in Sweden. Journal of Public Economics 92(10): 2071–2082.

36.

Kohli

(2007) The institutionalisation of the life course: Looking back to look ahead. Research in Human Development 4(3–4): 253–271.

37.

Kohli

Rein

(1991) The changing balance of work and retirement. In: Kohli

Rein

Guillemard

M-A

, et al. (eds) Time for Retirement: Comparative Studies of Early Exit from the Labor Force. Cambridge: Cambridge University Press, pp. 1–35.

38.

Kyyrä

(2015) Early retirement policy in the presence of competing exit pathways: Evidence from pension reforms in Finland. Economica 82(325): 46–78.

39.

Lammers

Bloemen

Hochguertel

(2013) Job search requirement for older unemployed: Transitions to employment, early retirement and disability benefits. European Economic Review 58: 31–57.

40.

Leopold

Engelhardt

(2011) Bildung und Gesundheitsungleichheit im Alter: Divergenz, Konvergenz oder Kontinuität? Eine Längschnittuntersuchung mit SHARE. Kölner Zeitschrift für Soziologie und Sozialpsychologie 63: 207–236.

41.

OECD (2006) Ageing and Employment Policies: Live Longer, Work Longer. Paris: OECD Publishing.

42.

OECD (2013) All in it together? The experience of different labour market groups following the crisis. In: OECD Employment Outlook 2013. Paris: OECD Publishing.

43.

Radl

(2013) Labour market exit and social stratification in Western Europe: The effects of social class and gender on the timing of retirement. European Sociological Review 29(3): 654–668.

44.

Schils

(2008) Early retirement in Germany, the Netherlands, and the United Kingdom: A longitudinal analysis of individual factors and institutional regimes. European Sociological Review 24(3): 315–329.

45.

Schmitt

Starke

(2016) The political economy of early exit: The politics of cost-shifting. European Journal of Industrial Relations. Epub ahead of print 27 December 2015. DOI: 10.1177/0959680115621137

46.

Schuring

Robroek

SJW

Otten

FJW

, et al. (2013) The effect of ill health and socioeconomic status on labor force exit and re-employment: A prospective study with ten years follow-up in the Netherlands. Scandinavian Journal of Work, Environment and Health 39(2): 134–143.

47.

Statistics Netherlands (2015) CBS Statline database. Available at: statline.cbs.nl.

48.

Tatsiramos

(2010) Job displacement and the transitions to re-employment and early retirement for non-employed older workers. European Economic Review 54: 517–535.

49.

Van Gestel

De Beer

Van Der Meer

(2009) Het Hervormingsmoeras van de Verzorgingsstaat. Veranderingen in de Organisatie van de Sociale Zekerheid. Amsterdam: Amsterdam University Press.

50.

Van Oorschot

(2007) Narrowing pathways to early retirement in the Netherlands. Benefits 15(3): 247–255.

51.

Van Oorschot

Jensen

(2009) Early retirement differences between Denmark and the Netherlands: A cross-national comparison of push and pull factors in two small European welfare states. Journal of Aging Studies 23: 267–278.

52.

Van Solinge

Henkens

(2007) Involuntary retirement: The role of restrictive circumstances, timing, and social embeddedness. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences 62B(5): S295–S303.

53.

Visser

Gesthuizen

Kraaykamp

, et al. (2016) Trends in labour force participation of older men: Examining the influence of policy reforms, normative change and deindustrialization in the Netherlands, 1992–2009. Economic and Industrial Democracy 37(3): 425–447.

54.

Vrooman

(2010) Een succesvolle gedaantewisseling? De hervorming van de sociale zekerheid, 1985–2010. Tijdschrift voor Arbeidsvraagstukken 26(4): 358–377.

55.

Walker

Maltby

(2012) Active ageing: A strategic policy solution to demographic ageing in the European Union. International Journal of Social Welfare 21: S117–S130.

56.

Widmer

Ritschard

(2009) The de-standardization of the life course: Are men and women equal? Advances in Life Course Research 12(1–2): 28–39.