8. The Netherlands: From convergence to divergence in Europe? Social dialogue and industrial relations in the face of household labour supply*

One important – if not the most important – aim of the European Union (EU) is the upward convergence of socio-economic outcomes between the member states. However, doubts have been growing that this is actually happening, fed by diverging trends resulting from the financial crisis and related policy-making. The recent initiative of the European Pillar of Social Rights (EPSR) is one response to this situation, illustrating that policy action is necessary to change such trends and to work towards achieving that aim. In this chapter I consider how the Dutch outcomes stand with regard to convergence or divergence, and how that relates to social dialogue and industrial relations in the country by international comparison. In line with the aim of this volume the main focus is on wages, working time and non-standard forms of employment.1 In EU policymaking (for example, the Employment Strategy or EU2020) the Netherlands is often one of the countries taken as an example for others to follow. This implies that the country may deem convergence more of an issue for other countries to pursue than for itself to contemplate and account for with respect to the converging or diverging effects of its own policy-making and corresponding social dialogue. That convergence is largely taken for granted makes instances of divergence all the more important. The starting question, in section 2, is how the Netherlands compares with other EU countries in terms of realising socio-economic convergence, making use of comparative indicators derived from the EPSR and other European sources. The focus is on the situation of employees, to whom the social dialogue and relevant Dutch laws and European directives apply.2 Working-time patterns, wage formation (for which read, moderation) and a relentless rise of flexible labour contracts – the main form of non-standard employment – are found to be the most important deviations. I am brief on wages and working time as I have dealt with those extensively before (Salverda 2018b). By contrast, I pay more attention to contractual flexibility as, interestingly, it is a field in which the Netherlands was a forerunner of EU policy-making and in which policy-making at both levels (Netherlands and EU) comes together; furthermore, core responsibility rests here with social dialogue as both the Dutch social partners and the European social partners literally wrote the legal stipulations that were adopted in the Dutch Flexicurity Law and


INTRODUCTION
One important -if not the most important -aim of the European Union (EU) is the upward convergence of socio-economic outcomes between the member states. However, doubts have been growing that this is actually happening, fed by diverging trends resulting from the financial crisis and related policy-making. The recent initiative of the European Pillar of Social Rights (EPSR) is one response to this situation, illustrating that policy action is necessary to change such trends and to work towards achieving that aim. In this chapter I consider how the Dutch outcomes stand with regard to convergence or divergence, and how that relates to social dialogue and industrial relations in the country by international comparison. In line with the aim of this volume the main focus is on wages, working time and non-standard forms of employment. 1 In EU policymaking (for example, the Employment Strategy or EU2020) the Netherlands is often one of the countries taken as an example for others to follow. This implies that the country may deem convergence more of an issue for other countries to pursue than for itself to contemplate and account for with respect to the converging or diverging effects of its own policy-making and corresponding social dialogue. That convergence is largely taken for granted makes instances of divergence all the more important.
The starting question, in section 2, is how the Netherlands compares with other EU countries in terms of realising socio-economic convergence, making use of comparative indicators derived from the EPSR and other European sources. The focus is on the situation of employees, to whom the social dialogue and relevant Dutch laws and European directives apply. 2 Working-time patterns, wage formation (for which read, moderation) and a relentless rise of flexible labour contracts -the main form of non-standard employment -are found to be the most important deviations. I am brief on wages and working time as I have dealt with those extensively before (Salverda 2018b). By contrast, I pay more attention to contractual flexibility as, interestingly, it is a field in which the Netherlands was a forerunner of EU policy-making and in which policy-making at both levels (Netherlands and EU) comes together; furthermore, core responsibility rests here with social dialogue as both the Dutch social partners and the European social partners literally wrote the legal stipulations that were adopted in the Dutch Flexicurity Law and Wiemer Salverda -9781788978071 Downloaded from Elgar Online at 04/28/2021 08: 19:55AM via free access the two EU Directives, on temporary work in 1999 and on temporary work agencies (TWAs) in 2008. Even this three-pronged policy action has been unable to prevent the strong growth of flexibility, which has become a bone of contention in current Dutch policy debates. This scrutiny offers us the chance to learn about the effectiveness of policy-making, social dialogue and industrial relations at both national and EU levels.
In section 3, I analyse the Dutch situation regarding flexible contracts in further detail and in relation to household labour-market behaviour, which may affect not only the socio-economic outcomes, reducing the effects of social dialogue, but also the nature of industrial relations and the corresponding social dialogue. In section 4, I scrutinise the role of Dutch social dialogue, underlying industrial relations, and government policy-making with regard to divergence, touching upon wages, working time and, in particular, contract flexibility. Section 5 sheds more light on this, especially industrial relations and household context, with a brief internationally comparative case study of the possible explanation that Belgian-Dutch differences in industrial relations and social dialogue may offer for the noticeable divergence of contract flexibilisation in these largely similar, neighbouring countries. Section 6 concludes with a summary and a discussion of whether, after the disappointing effects of the Directives, the EPSR might be of help in changing divergence into convergence.
Before turning to the Dutch-EU comparison in section 2, I emphasise the relevance of the relationship between individual earnings from labour-market activity and household incomes for understanding labour-market outcomes in the face of social dialogue and industrial relations. This relationship has changed radically and become much more complex in recent decades owing to the massive shift away from the single-breadwinner to the dual-earner household and a simultaneous rise in part-time jobs. Paradoxically, with the growing individualisation -mirrored in women's large and rapid gains in employment participation and educational attainment -the role of the household vis-à-vis the labour market has become even more significant. The current relationship between the earnings of individuals in the labour market, on the one hand, and household income distribution, on the other, is highlighted in Box 8.1. It differs radically from the traditional singleearner world in which each household fielded one worker and workers' annual earnings largely determined the income distribution. The labour-market effects on incomes were more transparent then and made the role of social partners and the instruments they could use much more obvious. The upshot is that now the highest household incomes can go together with relatively low individual earnings and different labour-market positions, not only between households but especially within them (Salverda 2015a(Salverda , 2015b(Salverda , 2016a(Salverda , 2016b, and that consequently most low-paid jobs are held by non-poor households (Salverda 2018a). Thus, in the current dual-earner world, policies aimed at improving the income position of poor and low-wage households by general means -such as a higher minimum wage or tax credits -may incur deadweight losses to the extent that these policies benefit households receiving higher incomes and comprising low-paid employees, and will no longer automatically reduce income inequality.
Accounting for this changed relationship, first, helps us to understand how the supply of labour can affect socio-economic outcomes and, consequently, the industrial relations that underlie the social dialogue. Second, it suggests possible weakening limitations on what industrial relations and social partners' dialogue on their own may be able to achieve without a broader involvement of the state. This works both ways: diverging   labour-market behaviours combine in households, but, equally, those household combinations incite or enable diverging behaviour. Such behaviour may come with different individual choices of labour conditions -regarding, for example, pay, occupation or working time -that deviate from and potentially bias against labour supply coming from other types of households, in particular single earners. Second earners who are content to work part-time and flock massively into easy-to-combine, usually low-paying occupations provide a good example. This mechanism is further examined here for people on temporary contracts. Note that the charts in Figure 8.1 also portray the EU Panel A shows MLHH shares among all households within the ten deciles of the household income distribution. They increase from 20 per cent in the bottom decile to 85 per cent in the ninth and tenth deciles. The distribution of MLHH is strongly skewed towards the top of the income distribution. The split by earner numbers in the household shows their essential contribution and how this differs from a world of single earners only. Single-earner and dual-earner households each comprise 44 per cent of all MLHH, while multiple-earner households provide the remaining 13 per cent. At the bottom, virtually all MLHH are single earners, while at the top they make up only one-tenth. This contrasts with the large shares at the top for dual earners, almost 60 per cent, and multiple earners, 30 per cent. Panel B presents the same for incomes instead of numbers. Understandably, this looks very similar to panel A across the deciles as households fit the same decile cut-off incomes. Potentially, only the top decile could be different, as it has no upper cut-off, but actually it does not -MLHH have the same 86 per cent share for both numbers and incomes. Top earnings largely match enterprise and capital incomes, which commonly concentrate in that same top.
Panel C distributes the corresponding wage-earning individuals over the same ten deciles. This amplifies the importance of dual-and multiple-earner households as they have more earners. Single earners comprise only 25 per cent of all employees, while the remaining three-quarters majority all share a household with at least one other employee. Individual earners concentrate even more strongly towards the top of the income distribution. Only 2 per cent are found in the bottom decile, while the top decile holds a ten times larger share, of 20 per cent. At the bottom, 95 per cent concern single earners, while at the top this is only 5 per cent. The remaining 95 per cent at the top is divided almost equally between dual earners (50 per cent) and multiple earners (45 per cent). Finally, panel D indicates where in the income distribution their wage earnings are found. The skewed numbers combine with increasing earnings levels into an extremely skewed household earnings distribution. The bottom decile's share is a miniscule 0.3 per cent of all MLHH earnings, while the top decile's share amounts to 32 per cent, 100 times more. This far exceeds the difference in average individual earnings at the two ends. Unsurprisingly, this difference is larger for single earners (26 times), who need much higher individual earnings to reach the top compared with dual earners (16 times) and multiple earners (9 times). Obviously, the household incomes at the top of the latter two categories are related to significantly lower positions in the individual earnings distribution which they bring together in the household, while single earners with top incomes are undoubtedly found also at the top of the wage distribution -at levels two and three-and-a-half times higher compared with dual and multiple earners, respectively. Dual and multiple earners often combine lower and higher earnings from different members within the same household, but nonetheless their main earners still lag considerably behind single earners at the top. Single earners obtain 28 per cent of total earnings, dual earners 54 per cent and multiple earners 19 per cent.
Note: * Main labour households' wage earnings contribute 80 per cent (EU27, 77 per cent) to their total income. The remainder derives from other sources, including self-employment, capital ownership and social transfers. Main labour households' numbers make up 88 per cent (EU27, 89 per cent) of all households receiving any earnings, be it a minority or majority contribution to the income. Finally, the comparison of the Dutch performance in the perspective of the EU as a whole leads me to conclude that the country's secular rise in flexible labour contracts is the most striking feature of a fast-growing divergence. This motivates the special attention this chapter pays to that issue.

Convergence Indicators of the EPSR
The EPSR encompasses 20 principles and rights that are divided over three broad chapters: equal opportunities and access to the labour market; fair working conditions; and social protection and inclusion. Most aims have a broad labour-market relevance, while a few seem too general for the purpose of this study of industrial relations and social dialogue (numbers 11 children, 16 health care, 18 long-term care, 19 housing and 20 access to essential services). All indicators are provided by Eurostat from its international databases, while previously European data sources were still insufficient and indicators were partly nationally provided (for example, the Laeken poverty indicators; Beil et al. 2011). However, it may still be that adequate statistical information for certain issues is missing, at Eurostat itself or not exploited from other sources, such as Eurofound. This may explain that not all EPSR aims are covered by indicators. The available indicators do not line up precisely with the 20 aims of the EPSR, nor with those most relevant for our purpose of industrial relations and social dialogue. In particular, there are none available for aims numbers Wiemer Salverda -9781788978071 Downloaded from Elgar Online at 04/28/2021 08:19:55AM via free access 8 social dialogue and workers' involvement, 9 work-life balance, 10 safety at work, 13 unemployment benefits, 14 minimum income or 6b minimum wage, and 17 inclusion of disabled in employment. In principle, the European provision of data is more efficient and also more comparable internationally. However, there may still be a risk that national situations are not adequately captured owing to the uniformity of the approach or the quality of the data gathered, and that the period covered by the data is too short to be meaningful. Unfortunately, this is particularly the case for labour contracts -the focus of this contribution -as we see below.
Eurostat retrieves the necessary information from its databases -mainly the European Labour Force Survey (ELFS), the European Structure of Earnings Survey (ESES), European Union Statistics on Income and Living Conditions (EU-SILC) and national accounts 3 -and provides a set of series that pretty comprehensively covers the period from the mid-2000s up to 2016.  (2.2), labour-market activation policy (2.3), and trends in income (2.4) and wages (2.5) for the second chapter; and income redistribution (3.1) for the third chapter. The nature of the indicators differs as to unit of measurement (column 1) and policy aim (column 2) of achieving either a low level L (for example, 1.1 early school-leaving) or a high level H (for example, 2.1 employment rate). Table 8.1 enables us to compare the Dutch situation with the average outcomes for the 28 EU countries (EU28). 4 Note that the EU averages presented in this chapter are simple, unweighted country averages that allow direct country-level comparisons. 5 The first part (columns 4 to 9) shows the most recent levels for 2016, except a few where the latest data concern 2015. It displays the EU28 average and the Netherlands for mutual comparison, and indicates also the width of the range, from the best-to the worst-performing member state. In the second part (columns 10-13) the preceding evolution since the mid-2000s is shown, broken down into two sub-periods: up to and since 2008, when the financial crisis started.
From 2016, the Netherlands occupies similar-to-average positions (unshaded) for the gender employment gap (1.2) and active labour market policy (2.3). The Dutch level is below average (darkly shaded), however, for disposable income (2.4), which fell by more than 7 per cent after 2008, at a higher speed than elsewhere, on average, though, obviously, Greece did much worse, with a 26 per cent decline. All remaining Dutch positions are above average (lightly shaded). In particular, the Netherlands is the EU's best performer for youth not in education, employment or training (NEET) (1.5), at half the average level. The best performing member states are found in the north and the east, especially the Czech Republic, for income quintile (1.3), poverty (and social exclusion) (1.4) and unemployment (2.2), and the worst in the south (for example, Greece) and the south-east (Bulgaria and Romania).
Looking back to the mid-2000s (2005 and a few for 2006), the evolution in the Netherlands initially (2005-08) conforms largely to these relative positions: above-average increases go together with above-average levels. One serious exception is that the impact of redistribution on poverty had declined (-3.5 (2.4). The mitigating effect of social transfers in the Netherlands diminishes substantially in both periods but its level is still above average in 2016.
The period since the mid-2000s may be too short to draw meaningful conclusions about convergence, all the more so because of the deep financial and economic crisis, which generated a lot of cyclical movement. Cyclical effects may be partly structural but that is hard to tell on the basis of a single cycle. As we interpret convergence as a process of structural change, drawing conclusions from these indicators is complicated by the fact that, based on year-to-year changes, most indicators actually appear not to be structural but cyclical (column 3 in Table 8.1), be it procyclical (P: for example, employment rate grows when the economy grows) or countercyclical (C: for example, unemployment rate increases when economic growth falters), while they are largely continuous for early school-leaving (1.1) and poverty reduction (3.1) only. There is no single headline indicator for which the coefficient of variation keeps declining over the years, which would indicate that countries are coming closer together; instead the coefficients shift between upward and downward movements during these years. Dutch levels consistently deviate more than one standard deviation from the EU28 average for youth NEET, employment rate and worker compensation. The Dutch unemployment rate has moved up closer to the average since 2013, while the income quintile share has moved away for the better. For early school-leaving the Netherlands is at the head of a general reduction for the EU; however, this may have come at the cost of NEET incidence, which on balance has declined very little, with a clear cyclical pattern of decline up to 2008 and increase thereafter. Success in convincing young people to stay in school until obtaining qualifications may have ended up in non-employment or very small jobs after school.
The European Commission maintains a range of 23 secondary indicators in addition to the headline indicators. Table 8.1, panel B, presents a selection of eight, based on their direct relation to the labour market, which makes them relevant for considering the effects of industrial relations and social dialogue for the Netherlands. The effects can be either direct or indirect, via the great importance of employment and wage earnings for household incomes, as explained above. The eight selected indicators range from adult learning (1.A) via gender gaps for part-time work and for pay (1.B and 1.C) and the Wiemer Salverda -9781788978071 Downloaded from Elgar Online at 04/28/2021 08:19:55AM via free access household relationship to work (1.D), to employment dynamics (2.A and 2.B) and to labour-market effects on incomes (2.C and 2.D). 6 Their numbering retains the distinction between the EPSR's chapters 1 and 2. Again the comparison is drawn to the EU28 average 7 and looks at the current situation, as well as the preceding evolution.
In 2016 Dutch levels were worse than average (darkly shaded in Table 8.1) for four of the indicators: the gender part-time employment gap (1.B) and the unadjusted gender pay gap (1.C), the transition from temporary to permanent contracts (2.B), and the pension replacement rate relative to preceding income (2.D). The country occupied similarto-average positions (unshaded) for low work intensity (1.D) and job duration (2.A), and was doing better than average for adult learning (1.A) and in-work poverty (2.C). The Netherlands was the EU's worst performer for the gender part-time employment gap (1.B). Strikingly, some of the best performers were found in the south-east -Italy for gender pay gap (1.C) and Romania for gender part-time gap (1.B) -and some of the worst in the north-west -Ireland for low work intensity (1.D) and pension replacement rate (2.D), France for transitions to permanent contracts (2.B). Looking back to the mid-2000s, the evolution for the Netherlands initially (2005-08) largely conformed to these relative positions: as before, above-average increases and levels often went together. After 2008 it deteriorated for pensions replacement (2.D) only. The Dutch gender gap in part-time employment (1.B) evolved better than average, shrinking more than elsewhere in both periods, but it was so large that it still showed the worst performance of all EU countries in 2016, remaining substantially more than two standard deviations above the EU average. Finally, also here there is not a single indicator for which the coefficient of variation shows a steady trend which would point to growing convergence; instead it increases and decreases over the years.
Thus, the secondary EPSR indicators appear to shed an important, different light on Dutch underperformance compared with the EU average.

Additional Convergence Indicators
It is important to ask how adequate this set of EPSR indicators is for the Netherlands and for examining the effects of industrial relations and social dialogue. Social rights are intended to reach each and every person to the far ends of the ranges covered by each aim. However, the indicators offer insufficient insight into the distribution of persons over these ranges, particularly with regard to incomes, employment and unemployment, and earnings. Table 8.2 assembles several additional indicators for the two EPSR chapters. This table follows the same format as, and links to the numbering of, Table  8.1. First, three distributional measures are added for incomes: ninth to first decile ratio, shares in total income of the top 10 per cent and of the bottom 40 per cent of households (1.3.a-1.3c), while other measures shine light on the earnings distribution: relative minimum wage, incidence of low pay and the three common decile ratios for earnings (2.5.a-2.5e). Second, for employment and unemployment the internationally strongly diverging effects of part-time employment are accounted for. An estimated full-time equivalent (FTE) employment:population ratio (EPOP) for all ages as well as for young people separately (2.1.a-2.1c) indicates the volume of work that is actually performed, and shared, by the population. The unemployment-over-population rate compensates for the distortionary effect that employment numbers inflated by part-time jobs have on  Wiemer Salverda -9781788978071 Downloaded from Elgar Online at 04/28/2021 08:19:55AM via free access the common unemployment rate (over the labour force), again for all ages and for young people (2.2a-2.2c). Finally, employment performance is further qualified by introducing the rates of involuntary part-time employment over part-time employment (1.B.a), on the one hand, and over the working-age population as a whole (1.B.b), on the other hand, as well as the share of temporary contracts among employees (2.B.a) and the share of TWA employees, who are a special group within temporary workers (2.B.b). Dutch income distribution is less unequal than average for all three measures, but the differences are relatively small. The same applies to the decile ratios of the earnings distribution -decile shares or averages would add important insights but are not available. 8 However, as in Germany, the Dutch incidence of low pay is above average and six times higher than the Swedish minimum. The relative minimum wage is average but has declined since 2008, in contrast to the EU. The effects of accounting for the country's many part-time jobs with the help of the FTE volume are highly significant. While the EPOP in Table 8.2 (2.1) (77 per cent), based on a headcount of employees, is well above average (71 per cent), the FTE-based rate is well below (63 per cent versus 69 per cent). The relatively even higher youth EPOP level -the Netherlands is the best performer at 61 per cent as against 35 per cent for the EU28 -falls to an average level (31 per cent) when accounting for the hours worked. Note that this throws another light on the Dutch NEET performance; for example, if educational participation were counted together with young people in full-time jobs only -which is de facto the common situation in most countries -the NEET figure would be two to three times higher than the 5 per cent mentioned in Table 8.1 (1.5) and well above the EU average. In the same fashion, exactly because of the high incidence of part-time jobs, part-time employment considered on its own shows a level of involuntariness of only 10 per cent, well below average. However, if we account for the high incidence of part-time employment involuntariness affects 3 per cent of the Dutch working-age population, which is above average. The traditionally low level of involuntariness concerns those only who are actually working part-time, while for the working-age population as a whole involuntary part-time employment is as much of a problem for the Dutch as elsewhere. Last, but not least, the incidence of temporary contracts is far above the EU28 average: 20 per cent as against 12 per cent. The Dutch incidence has doubled the gap with the EU28 average from four to eight percentage points. The incidence of temporary agency workers is also high, again two standard deviations above the EU average. This combines with the strikingly low transition rate from temporary to permanent contracts found in Table 8.1 (2.B). People risk becoming stuck in a morass of temporary jobs.
In summary, the Netherlands is in many respects an above-average performer for the EPSR, though some of that may be overestimated owing to the extensive role of part-time jobs -hours worked are actually below average. However, the financial crisis has pushed many Dutch trends downward, closer to the EU28 average. Wage and income trends diverged though and Dutch working-time patterns remain very different from the EU average and show a very different outcome in terms of persons involved (headcount) and work effort expended (hours count). The most important exception, however, concerns temporary employment contracts. These show rapid growth, especially during the period covered by the indicators above, which has led the country on a swiftly diverging path away from the EU average. Although the Netherlands shares such a high incidence with several other countries, these show considerable decline (Spain) or stagnation instead of growth (Poland and Portugal). The Dutch debate often speaks of this as the 'flexible skin' of the labour market and labour organisations. That seems rather euphemistic compared with the thin and protective skin of a fruit from which it derives. Contractual flexibility to the extent of a quarter or a third of all jobs is not a thin layer and it does not offer protection to the people involved. The Netherlands has become increasingly uncomfortable with this development, thinking that job flexibility has gone too far at the expense of security (for example, WRR 2017). However, it has not been able to really change this trend.
Taken together with the strong divergence of the Netherlands from the EU trends that it represents, this makes it worth adding first, in section 3, some detail on the explosive growth of temporary work and contractual flexibility and on its household context, which may weaken the strength of industrial relations and the effectiveness of social dialogue, and considering after that, in section 4, the linkages of convergence and divergence to social dialogue and underlying industrial relations for wages, working time and, especially, flexible contracts.

EXPANDING FLEXIBLE CONTRACTS AND THEIR HOUSEHOLD CONTEXT
First, I present how the incidence of temporary contracts among employees has evolved since the early 1980s, including the period since 2005 covered in the previous section. Figure 8.2 shows that, since the late 1980s, temporary contracts at the EU level have changed very little, while, after initial ups and downs, there has been a straightforward increase in the Dutch level over the past 25 years. Over the period since 2005 alone, the gap with the EU average has doubled. The growing incidence of contractual flexibility has reached escape velocity and brought the Netherlands into orbit at more than one standard deviation above the EU average in the 2010s. The highest grey line depicts how Statistics Netherlands (Centraal Bureau voor de Statistiek, CBS) captures the recent evolution differently from Eurostat, with a higher incidence and a steeper rise. The CBS employs a broader definition that includes people who formally have a permanent contract but no fixed hours of work. Paradoxically, people on such contracts may be worse off as legally their contracts oblige them to come to work at the behest of the employer even for minimal hours or several times a day (De Beer and Verhulp 2017). In the rest of this section I exploit these CBS data, where possible. The Dutch evolution suggests that the growth was less during the dotcom crisis 2001-04 and the first stage of the financial crisis . This could mirror the greater ease of dismissing those on temporary contracts in times of economic decline. However, the CBS data show continued growth, albeit a bit more slowly between 2007 and 2011 but clearly picking up again in the most recent years. Figure 8.3 specifies how this works out for the headcount employment rate. At the start, in 1987, the permanent part of the Dutch employment rate alone was as large as the EU's permanent and temporary parts put together. Currently, it is only the higher incidence of temporary work that lifts the Netherlands above the EU average (comparison with EU15 or EA19 averages makes little difference). Since 2001, well before the financial crisis, the permanent EPOP has declined from a level ten percentage points above the EU average all the way down to the same level, while in the meantime the temporary EPOP has grown from 9 to 13 per cent. The employment volume (FTE) cannot be pinned down precisely, but it may be estimated at roughly three-quarters of the headcount rate (estimated from Figure 8.5, panel B, by weighting numbers in the hours band with the help of the means for the three lower bands and 40 hours for the upper, full-time band), implying a growth of the temporary EPOP-FTE from a level of 6-7 per cent to 9-10 per cent of the working-age population. Consequently, the permanent EPOP-FTE may lag even more substantially behind the EU average than suggested in Table 8.2 (indicator 2.1.a). Evidently, the headcount is important in terms of the personal effects of flexibility, while the volume can help us to evaluate whether it makes economic sense.

A Relentless Rise Fanning Out in All Possible Directions
Notably, the rise of contractual flexibility diverges significantly by type of temporary contract and implies a strong shift towards ever more flexible types. Figure 8

Notes:
In 1987, the hours worked by temporary agency workers were equivalent to 2.1 per cent of all employees; total persons with a temporary contract amounted to 9 per cent. The long series for TWAs concerns the volume of hours worked, the other data concern persons. The hours worked by temporary agency workers are invoiced by the agencies. Their FTE as a percentage of total employee hours indicates the incidence. This may be slightly greater than for the number of persons, which before 2003 was known only from a minimum of 12 hours per week, according to the definition of the labour force then used.  and those that are most flexible. Temporary work agencies jobs are the best-known type and were the focus of the flexicurity deal concluded by the social partners in 1996, discussed in section 4. Their incidence doubled up to the mid-1990s from 2 to 4 per cent, and has risen more gradually since, at one-quarter of the previous speed, to 6 per cent most recently. The most flexible, indicated as 'on-call' (but including fixed and temporary contracts without fixed hours) and shown stacked to TWAs, however, bridge most of the gap between TWAs and the total as their incidence has risen from 9 to 17 per cent. 9 Behind the aggregates, the incidence of flexibility is skewed and has risen for pretty much every dimension one might look at, with the exception of gender. The incidence is always larger for women than for men but the rise has been faster for men. Women have experienced an increase from 14 to 22 per cent since 1992, men from 7 to 19 per cent, narrowing the gender gap from seven to three percentage points. However, five other important dimensions -age, educational attainment, working time, occupation and industry -show a strong fanning out of temporary contracts across the population and the economy.

Source
In two panels, Figure 8.5 provides the detail for the two dimensions showing the strongest gradients: age and working time. The highest incidence is reached by young people, at 54 per cent currently, or 70 per cent according to the CBS data (panel A). However, their rise -a doubling from 27 per cent in 1995 -is not faster than for the next two age groups. Those between 25 and 34 years of age experienced more than a doubling, from 10 to 24 per cent, and those between 35 and 44 years of age a doubling, from 6 to 12 per cent. Similar to an oil spill, flexible contracts are spreading towards older ages, albeit not yet to people aged 55 or more. The evolution of temporary work for four different bands of weekly working hours (panel B) is available only from 2003 and concerns CBS data. What transpires is a very significant overlap between flexible contracts and parttime jobs. Employees with the smallest jobs of less than 12 hours per week attain the same high incidence of 70 per cent as young people -unsurprisingly, as most employees in this category are young. Although the level is stunning, the increase seems more modest, from a start at 50 per cent; but notably this concerns a much shorter period than for the other dimensions, where starting also in 2003 would often halve the increase. Growth in the other working-time bands is at a comparable speed, extending also to full-time workers (35+ hours: 9 to 16 per cent). It suggests that the growth in temporary work has been comparable between headcount and FTE volume. Nonetheless, the gradient over working hours is very strong and the minimum-maximum gap grows from 41 to 57 percentage points, the largest directly observable difference. The low educated and middle educated (upper secondary education) both faced a doubling (15 to 30 per cent, and 10 to 20 per cent, respectively). For the highest educated the incidence is clearly less but they also experienced a considerable increase, 11 to 15 per cent, albeit less than a doubling. The low-high gap increased from four to 15 percentage points. The CBS data indicate a substantially larger gap -21 percentage points in 2016. The rapid increase has touched virtually all main occupational categories, often clearly more than doubling the incidence: elementary jobs (15 to 39 per cent), sales and service occupations (12 to 29 per cent), administrative work (8 to 20 per cent), but also professionals and technicians (5 to 14 per cent) and even managers, albeit at by far the lowest levels (3 to 7 per cent). However, this speed was easily overtaken by craft occupations, rising from 3 to 21 per cent, and machine operators (ISCO 8), from 5 to 24 per cent. The widest gap, between elementary jobs and managers, increased from 12 to 32 percentage points. Finally, across industries, 10 employees in hotels and catering reach the second-highest level of the six dimensions (45 per cent), between young people above and elementary occupations below. This industry clearly veers away from the rest as they combine the highest level at the start in 1992 (22 per cent) with the average doubling over time. The speed of the increase, however, is higher in most other industries, such as trade (9 to 25 per cent), construction (4 to 16 per cent), transportation (5 to 21 per cent), business services (10 to 23 per cent), financial services (3 to 11 per cent) and manufacturing (6 to 16 per cent). For health care (8 to 15 per cent), education (11 to 16 per cent) and, surprisingly again, agriculture (19 to 28 per cent) the increase is clearly there but at a slower pace. The widest gap, which is found between hotels and financial services, grows from 19 to 34 percentage points.
An outstanding role is played, also in international comparison, by young people interlinked with part-time jobs in sales and services occupations in (retail) trade and hotels and cafés. Under the age of 25 permanent and full-time jobs are in decline. This offers food for thought about the household context, as well as the possible role of industrial relations and social dialogue, which seems minimal among young people (Salverda 2018b). However, this cannot be the full story, as contractual flexibility is rapidly spread- ing beyond young people to older ages and higher levels of educational attainment, and across most occupations and industries, revealing a process of unravelling.

Households and Temporary Labour Supply
Persons working on temporary contracts are not necessarily happy or unhappy with it. Some like it, others do not. Asked whether they are satisfied with their work 'on the whole' they appear to be less satisfied (75 per cent) than employees with a permanent contract (78 per cent) but the difference is modest (CBS 2017). Satisfaction is lower, though, for those who prefer a permanent job and have tried in vain to find one. In particular, temporary agency workers are the least satisfied for that reason (66 per cent). Asked whether they appreciate the flexibility or have no need for security, two-thirds of flexible workers disagreed. Eurostat's admittedly imperfect data for the longer run indicate how many are unsuccessfully looking for a permanent job (Figure 8.6, light grey line linked to the right axis): between one-third and half of all temporary employees (if we disregard the probationary period), depending on the economic situation. This amounts to 3 to 8 per cent of all employees, permanent and flexible taken together. However, a substantial part may be hidden in the non-response.
We may still surmise that a fair share do not directly object to a flexible job, especially Note: In 1992, 3.9 per cent of all employees had a temporary contract because they could not find a permanent contract, while 4.9 per cent did not object, but at the time that included those during probation for a permanent job.
Source: Eurostat, ELFS (lfsa_etgar). if it is on a part-time basis and can be combined with other activities. High school pupils and students working a few hours at a time in a supermarket or a café come to mind. It underlines the importance of examining the household context for them, but also for others. A household with sufficient earnings from other sources can -voluntarily or forcibly, we do not know -provide an accommodating environment for those who are willing and some kind of safety net for the others who would prefer a permanent job. That

Figure 8.7 Temporary employees as a percentage of the total of all employees in main labour households over the distribution of household gross incomes, Netherlands and EU27, 2012
Wiemer Salverda -9781788978071 Downloaded from Elgar Online at 04/28/2021 08:19:55AM via free access is, where in the household distribution of Figure 8.1 do we find the temporary workers? 11 Figure 8.7 presents the aggregate picture in four panels, the upper two (A and B) with a breakdown by types of earner-households and the lower two (C and D) by females and young people, the two demographic groups affected more strongly by temporary work and probably also by the household environment. The left-hand and right-hand panels correspond to panels C and D, respectively, of Figure 8.1 -understandably with a much smaller scale on the y-axis: 2.5 per cent compared with 35 per cent. The new figure spreads the numbers, 16 per cent of all employees, and earnings, 11 per cent of the total, in a much flatter pattern over the income distribution compared with Figure 8.1. Particularly, single-earner temporary workers cluster much more strongly towards the bottom (panel A), for very small earnings, as implied by the considerably more strongly skewed distribution of the earnings (panel B). Dual-earner and multipleearner households 'produce' large shares of temporary workers (44 per cent and 12 per cent, respectively), not much different from the situation of all employees, and they skew numbers and especially earnings towards the higher end of the income distribution. We can imagine that people in such a situation would answer that they are not necessarily seeking security in their jobs. External effects of their behaviour on those who do seek and need security may be considerable, however. It is a highly significant feature to take into account for labour market policy. The situation is difficult to address for the social partners on their own, without involvement of the state, as they focus on individual workers and ignore -and should ignore -the household context of their employees and union members. Diverging behaviour of labour supply would potentially weaken industrial relations. The two panels compare as in Figure 8.1 with the EU averages. These comprise 12 per cent of the total number and 7 per cent of total earnings, which are exceeded by the Dutch aggregates (16 and 11 per cent), as might be expected. Importantly, Dutch aggregate numbers are more equally spread from the bottom to the top than are European aggregate numbers, probably reflecting a concentration of small jobs at the Dutch bottom. By contrast, aggregate earnings at the top are equally important for the Netherlands and EU, though the Dutch aggregate earnings excess concentrates in the middle. In comparison, the Netherlands confers a more substantial role on single-earner households and a smaller role on multiple-earner households for both numbers and earnings, as indicated by the differences between corresponding Dutch areas and EU lines in the graphs.
Panels C and D identify women and young people within these totals, two groups experiencing higher rates of temporary employment. They are shown in the graphs independently of each other, meaning that the female share, be it area or line, includes that of young females, and also overlapping each other, meaning that part of the female share hides behind that of young people and part of the total share behind both other shares. The total shares are identical to those of panels A and B, respectively. Again, the most important observation is that the groups are found across the entire distribution, with an overrepresentation of youth numbers at the bottom and the top, and of women's numbers at the bottom and the upper middle. Earnings at the bottom are negligible, as in panel B. Young temporary workers account for 3 per cent of all employees and 1 per cent of all earnings, implying very low earnings, at one-third of the average. Being on your own as a household is then a problem, which brings young persons to the bottom of the distribution, while being a member of a better-off household brings you higher up the distribution. Among temporary employees alone, young people make up 20 per cent of the numbers and 9 per cent of the wages, respectively, implying that they receive less than half of average temporary worker earnings. Female temporary workers comprise 9 per cent of the total number and receive 5 per cent of total earnings, and thus earn half the average earnings. Of temporary employees 55 per cent are female; receiving 45 per cent of all temporary earnings they are still somewhat below the average of all temps, including men.
The above concerns gross incomes, as derived from the labour market. The importance of the household context finds confirmation in disposable incomes equivalised for household size and economies of scale (Figure 8.8). The earnings of people on temporary contracts are far below those of permanent workers, at 40 per cent. By contrast, household incomes for the same flexible persons are close to 90 per cent of the permanent level. Apparently, many flexible workers are members of households that are close to average in terms of income. Although we have no information about the implied redistribution of income within the household towards temporary workers, that the great majority (75 per cent) are not happy when asked to assess their situation 'on the whole' seems to imply that there are limits to the desirability and sustainability of such redistribution (CBS: Statistics Netherlands 2016).
In summary, flexibility has risen rapidly in every possible direction, especially over the past ten years. It shows particularly strong gradients by age (young people), occupation (sales and service jobs), industry (hotels and catering, and trade) and working time ( Earnings concern primary income, which may comprise some income from sources other than earnings. Disposable income refers to disposable household income equivalised for household size and economies of scale.

Figure 8.8 Average annual individual earnings and household equivalised incomes of temporary workers as a percentage of permanent average, Netherlands, 2005-15
jobs). These characteristics hang together and are also deeply rooted in a complex pattern of household labour supply across the entire household income distribution, which makes it acceptable for a significant part of the persons involved. Consequently, it may also reinforce the level and growth of flexibility and diminish wilful action for change. This complicates finding the proper instruments and limits of what social dialogue can accomplish, certainly without clear state involvement.

CONTRIBUTIONS OF SOCIAL DIALOGUE AND GOVERNMENT POLICY
We now turn to the contributions that social dialogue and government policy have made to this stunning divergence between the Netherlands and the EU with regard to contractual flexibility and the deviating patterns for wages and working time.
The institutional set up of the Dutch Polder Model with its abundance of social dialogue -since its launch in 1950 the Social and Economic Council (SER) has produced 1200 Recommendations to the government, no less -has hardly changed over the past 70 years. I will not give a detailed picture of its main framework, actors and modus operandi (see Salverda 2018b for an extensive discussion that also covers policy debates and outcomes of inequality). Suffice to say that globalisation and European monetary union have altered Dutch industrial relations, shifting power away from unions to employers, and have diminished the government's budgetary room for policy manoeuvre. Social dialogue may be unchanged in terms of form and procedure, but this does not necessarily hold for its material content. The distribution of power in the underlying industrial relations affects its outcomes. As already mentioned, the Netherlands is one of those country models that are often put forward by the EU in its policy-making as an example for other countries to follow and to pursue with their own policy-making. That may have as a consequence that relatively little policy action, including social dialogue, is actually undertaken within the Netherlands itself, based on the contention that goals set by the EU have already been met or are within easy reach -as the Dutch above-average scores for 2016 in Tables 8.1 and 8.2 can still demonstrate, despite the many declines we have found since 2008. Finally, as the social dialogue is commonly preoccupied with the national situation, it may also pay insufficient attention to the country's contribution to convergence or divergence at the EU level. Nevertheless, the deviations from the path towards convergence warrant all the more attention.
The strong divergence with regard to contract flexibility found above seems a case in point regarding the effects of social dialogue, depending on the strength of underlying industrial relations in the context of household labour supply. I examine this in more detail in section 4.2 and later in section 5. Flexible contracts are key to non-standard forms of employment. However, I consider first, in section 4.1, the two other topics selected for this project: wages and working time. These are the prime responsibility of the social partners. I address these briefly, however, as I have discussed them previously in much greater detail than is possible here (Salverda 2018b).

Social Dialogue on Wages and Working Time
The EPSR indicator for wage developments concerns the running average of actual hourly compensation received across the economy, in euros (Table 8. 1, 2.5). Unsurprisingly, given the intra-EU differences in economic development, the Dutch level is far above the EU average (+74 per cent in 2016, +78 per cent in 2005). However, this measure is not a convincing indicator of the role of social dialogue vis-à-vis wage evolution. Employers and unions determine negotiated wages as laid down in collective agreements in terms of both wage structure and regular general adaptations of their wage levels. By contrast, actual wages depend on a double composition effect: how the workforce spreads over and within collective agreements, depending on the nature of jobs and applicable agreements in different parts of the economy. Naturally, actual wages may depend also on individual employees' wage-negotiating potential to move higher up a labour agreement's wage structure, depending on the general situation of the economy and the labour marketboom or bust. The two -negotiated and actual wages -may and do diverge significantly, as shown Figure 8.9, a selective update from Salverda (2018b, fig. 9.4a). It compares the evolution since 1990 of negotiated wages, excluding employers' social contributions made on top of employees' gross earnings, and actual employee wages and salaries, for consistency also excluding employer contributions, and both deflated by consumer prices. Although in recent years negotiated wages have again made some progress, their advance over and above the level of the early 1990s is minimal, at no more than 1 per cent in purchasing power. By contrast, average earned wages have increased by 11 per cent. The gap between the two has opened up rapidly over the 2000s and become much wider than before. This signifies a reduced grip on the part of the unions -and therefore the social dialogue -on wage formation. The steadily growing incidence of low pay (see   Netherlands, 1990Netherlands, -2017Netherlands, (1990 Chapter 3, on Belgium, in this volume) points in the same direction. Unfortunately, a comparison with the EU average is impossible owing to the lack of any systematic data on negotiated wages internationally. As I have discussed extensively elsewhere (Salverda 2018b), the Dutch labour market is characterised by a very broad and still growing incidence of part-time jobs, including a large mass of very small jobs. In 2017 more than half (51 per cent) of all employees worked part-time, that is, less than 35 hours per week, including more than 11 per cent who worked fewer than 12 hours per week. This implies a major fragmentation of low-skill employment: 75 per cent part-time, including 38 per cent fewer than 12 hours, which affects those on flexible contracts (70 per cent and 30 per cent, respectively) and low-educated people (58 per cent and 25 per cent, respectively) given their lower employment rate (47 per cent against 67 per cent on average) which may reflect the lack of substantial hours. A revised implementation of wage subsidies for low-paid jobs requires minimum employment of 1248 hours over the year (amounting to at least 24 hours per week), except for people under the age of 22. This revision came into force very recently and it is too early to say anything about the effects. Potentially, this can counteract some of the fragmentation on the job side, but there are no requirements for the persons filling the jobs. Therefore, there is no special reason to expect a more favourable employment outcome for low-educated persons. The initiative for the measure was taken by the government without a positive contribution from the social dialogue -to the contrary, a change in the youth minimum wage was agreed by the social partners on condition (demanded by the employers) that young people would be eligible for the subsidy but exempted from the 1,248 hours rule. 12

Dutch Flexicurity Law and the EU Directives as a Result of Social Dialogue
In the Netherlands a deep recession in the early 1980s was followed by rapid employment growth and another recession in the early 1990s, albeit relatively benign compared with the 1980s and other countries. This unleashed rapid growth in temporary agency work, the most important and visible type of flexible labour contract at the time. Its volume doubled in the course of a few years (Figure 8.4). The growing preference for reduced working time, as demonstrated by the shortening of full-time hours and the very rapid rise of part-time jobs -the main form of Dutch employment growth since the mid-1980s -was used as an argument by the social partners and the government for a flexibilisation of working time and labour contracts. Aimed at mainstreaming temporary agency work, the national union and employer federations concluded the deal 'Flexibility and Security' in 1996, as a direct response to the government's request for advice on contractual flexibility. The deal traded off contractual security and job growth, and made the Netherlands a forerunner of flexicurity policy-making -with an incidence of temporary agency work at two standard deviations above the EU average (Table 8. 2, 2.B.b). Unions obtained some improved contractual security and the establishment of a collective agreement for TWAs, while employers gained a route towards the consecutive easing of dismissal procedures and costs, and of unemployment benefits -a route for the long run that may be compared with the equally long path of unchanging wage moderation that has been followed since the Wassenaar Accord of 1982 (Salverda 2018b).
In 1999 the Dutch government turned this deal almost literally into law -in the same year the EU copied a similar framework agreement on fixed-term work concluded by the social partners at the EU(15) level, equally literally into Directive 1999/70. Thus, contractual flexibilisation is the full responsibility of the social partners at both levels; there is no hiding behind the government as policy-makers dutifully copied what employers and unions had agreed. The aim of the Dutch deal and the new law was to improve employment security and not the equal treatment of temporary and permanent workers, as in the Directive. Employers remained legally uncommitted to short-term temp agency workers. However, for those on repeated TWA jobs for longer terms, a stronger commitment was introduced. The law established the '3 × 3 × 3' rule for all temporary contracts (including TWA workers) and employers (including temporary agencies), meaning that after a sequence of three temporary contracts or a total of three years of temporary employment the contract would be changed by law to a permanent contract. Notably, however, it also mentioned a three-month break between contracts as an escape route that could restart the whole sequence. This corresponded, in principle, to the second aim of the Directive -to prevent abuse in the form of chaining one temporary contract after another -but it missed out on its primary objective: non-discrimination or equal treatment of temporary and permanent workers with regard to labour conditions, unless for objective reasons. The parties to the agreement at the national level left such inequality to be dealt with within the framework of local social dialogue, between the bargaining parties of collective contracts at industry level, who could try to equalise labour conditions between temporary and incumbent, permanent workers, including equal pay for TWA hires. However, collective agreements enjoy legal wriggle room here, following Dutch stipulations of 'three-quarters' law (Hoek 2003), but also following the EU Directives (for example, clause 8-2 in the European Trade Union Confederation-Union of Industrial and Employers' Confederations of Europe-European Centre of Enterprises with Public Participation, ETUC-UNICE-CEEP, agreement of 18 March 1999; although their general principle is that deviations should be more favourable to workers). Local partners are free to deviate to some extent and can even agree the opposite and actually increase inequality between temporary and permanent workers, for example, by enhancing the number of temporary contracts or the duration needed for the transition to a permanent contract or by restricting the recognition of preceding temporary contracts. Collective bargaining results as examined by the Labour Inspectorate have gone in both directions (SZW 2002(SZW , 2006(SZW , 2007(SZW , 2012(SZW , 2013, unfortunately without naming agreements. The two collective agreements of temporary work agencies themselves, Algemene Bond Uitzendondernemingen (General Bond Employment Agencies, ABU) and NBBU, also deviated from the legal provisions: the short-term period without any dismissal protection was increased from six months to twelve, while the duration before transition to a permanent contract was shortened by six months to three years. 13 The Nederlandse Bond van Bemiddelings-en Uitzendondernemingen (NBBU) collective agreement stipulated equal pay from day one, while the ABU collective agreement initially prescribed this after six months and only in March 2015 changed to day one.
The Netherlands integrated the Directive into law in 2002 by a very general stipulation on non-discrimination but without defining a comparable permanent worker (European Commission 2006). At the European level it took until 2008 before the social partners could reach agreement on the treatment of TWA workers, and the EU Wiemer Salverda -9781788978071 Downloaded from Elgar Online at 04/28/2021 08:19:55AM via free access then enacted Directive 2008/104 on temporary agency work that filled the gap left in the 1999 Directive. Again, the prime aim was to ensure the equal treatment of TWA workers by user undertakings as if they were hired directly instead of via the TWA as an intermediary. The Directive brought no change to Dutch rules and practices. Note that the Directive of 1999 exempted TWAs, while the Dutch debate mainly focused on them and fitted them into the general provisions on fixed-term contracts (or the other way around) in the deal of 1996 and the new law. This already provided the answers to the new Directive, which itself was weak on implementation (Contouris and Horton 2009;Sartori 2016). In its evaluation of the transposition of the Directive into national law the European Commission observes that the Netherlands -similar to several other countries -allows collective agreements to deviate from equal treatment and lay down prohibitions and restrictions without formally requesting that they do not prejudice the overall protection of temporary agency workers (European Commission 2014). Research by the ABU Association of temporary worker agencies into the conditions laid down in 745 collective agreements of TWA-user industries in 2011 again found stipulations going in both directions, enhancing or reducing equal treatment. For example, 164 collective agreements imposed equal pay from day one, while 113 reduced transition chances towards a permanent contract by counting all previous TWA contracts as one. In their own evaluation of the Directive, the Dutch social partners could not agree (SvdA 2011). Trade union federations advocated the application of the user collective agreement to TWA hires and equal pay, some maximisation of the use of TWA workers by enterprises, a maximisation of the individual duration of TWA contracts and prior consultation of unions/work councils on TWA hiring by enterprises. These were all opposed by the employers' associations, implying the absence of a level playing field for social dialogue on the topic. As a result, no recommendations were made to the local bargaining parties and it was left entirely up to them to evaluate the existing stipulations or not. This suggests that the diverging outcomes of applying three-quarters law to collective agreements reflect the weak position of the unions in the social dialogue, at least at the national level, but does not imply a lack of interest in the issue from their side. The Directive of 1999 has hardly changed the European incidence of temporary contracts, while the Dutch law unleashed an unstoppable growth that, 20 years on, seems to have overwhelmed the country (Figure 8.2). Figure 8.4 showed a modest increase of temporary agency work since 1999, from 4.1 to 6.0 per cent, and revealed that the unstoppable rise rests on a strong growth of other, more flexible types of temporary contract. The incidence of short-term temping has hardly changed, mainly following the ups and downs of the economy, with three consecutive peaks between 3.4 and 3.6 per cent. 14 By contrast, the longer-term TWA category with enhanced employer commitments has expanded very sharply, from 0.6 to 2.6 per cent of total hours worked in the economy and among TWA hours from 16 to 43 per cent. However, it seems that this has not led to an increase in permanent contracts with the agencies. 15 This short-term stability, together with the shift towards the two longer-term phases, is an encouraging outcome for the Dutch social dialogue of 1996-99, on its own terms, with regard to temporary agency work and leaving inequality concerns aside for a moment, which taint the success. Apparently, legal obligations, collective agreements and public debate have cemented organisational rules and procedures concerning temporary agency work. The dialogue seems to have managed to control and improve the role of temping in the economy, prob-ably helped by the fact that, organisationally, temporary agency work is easier to address because of the involvement of user organisations and agencies as clearly identifiable actors, in contrast to individual firms, which hire individuals on temporary contracts without further interference. Unfortunately, this mixed blessing is overshadowed by the unbridled growth of the weakest forms of flexibility: on-call contracts and fixed-term contracts without fixed hours, which largely fill the gap between the rapidly rising number of temporary contracts and the much flatter line of TWA hours (Figure 8.4). Apparently, employers have in no way felt morally bound by the flexicurity deal but have instead massively exploited other, perhaps more convenient and cheaper forms of temporary hiring. Despite this expansion and the heated public debate that has been going on for some years there has effectively been no response to the doubling of super-flexible contracts that compares to that of 1996 when temporary agency work had doubled. Note that the '3 × 3 × 3' chain rule of successive temporary contracts leading to a permanent contract applies only to certain categories of contracts, and not to on-call jobs or to pure temporary agency work during the initial period, when immediate dismissal is possible.
The relentless growth of flexible contacts, and particularly the increasing duration that people spend on temporary contracts, led then Minister of Social Affairs and Employment, Lodewijk Asscher, to renew law-making efforts to get things under control and improve the protection of the people involved, partly based on observations that collective agreements often changed the rules to the disadvantage of temporary workers, without good reason. In 2013, he concluded a new pact with the social partners which addressed contractual flexibility, among other issues (SvdA 2013). The three partiesgovernment, unions and employers -agreed that flexibility had gone too far in certain respects and become economically suboptimal. In the Pact improvements for temporary workers were traded off against changes in dismissal rules and the abolition of sheltered workplaces. This resulted in a new law, WWZ, that enacted a series of small adaptations to the legal rules on flexible contracts and restricted the adaptations that social partners were allowed to make in collective agreements. Collective agreement extensions to the 26-week maximum period of temporary agency work before an employment contract was needed were now capped, but at a high level of 78 weeks. The '3 × 3 × 3' rule was replaced with a '3 × 2 × 6' rule, stipulating a maximum of three temporary contracts or a total two-year duration before transition to a permanent contract, and a minimum period of six months to break the sequence. It was hoped that this would speed up transitions to a permanent contract. However, with a '6 × 4 × 6' maximum rule collective agreements were still allowed to increase the number of successive contracts to six or their cumulative duration to four years, but changing the six-month break period was now forbidden. In addition, collective agreement adaptations could no longer just be declared but had to be objectively justified. The entitlements of people with on-call contracts were enhanced by introducing a rule that a minimum of three hours should be paid for any work done, including for every separate occasion on a single day.
However, as Figures 8.2 and 8.4 show for the two and a half years since the WWZ law was introduced on 1 July 2015, the changes have had no noticeable effects on the growth of flexible contracts; fixed-term contracts without fixed hours grew even faster in 2017. A ministerial report on collective agreements in 2016 found many stipulations raising the number or duration of temporary contracts to the '6 × 4 × 6' maximum and reducing the recognition of preceding temporary contracts (SZW 2017). Evidently, no magic solution has been found for weaning employing organisations off unjustified contractual flexibility. On the contrary, employers are currently pondering the introduction of fiveyear contracts with the promise of training during those years (Volkskrant 10/2/2018) -notably an existing obligation of equal treatment according to the Directive. After the elections in early 2017, the new government concluded a coalition agreement that is supposed to lift the maximum duration of temporary contracts back to the just abolished three years and allow longer probation periods for permanent contracts, but also aims to forbid the use of the legal status of short-term temporary agency work by the payrolling institutions. This is all in the future, however, so we have to wait and see.
To sum up, if ever there was a labour-market phenomenon owing entirely to social dialogue, this is it. Flexibility rules were written by the social partners and copied by the government in 1999. The prime aim seems to be orderly contracting, not the equal treatment of temporary and permanent workers. This has created a legal framework that is massively used by individual actors in the labour market -on both sides, demand and supply or employers and employees -and no end of expansion is in sight. Starting from one end of the economy -the upper end of the strong gradients characterising most dimensions mentioned in section 3 -the conclusion of permanent contracts is progressively unravelling. Only in 2014, 15 years after the legal launch, did the then government take action towards addressing that pitiful state of affairs, but with no detectable effect on the explosive growth so far. What is more, the newly elected government is currently pondering whether to retract some of the changes. Unions, who were party to this evolution, have been unable to change the direction through renewed social dialogue. It is probable that the evolution of Dutch industrial relations has weakened them vis-à-vis the employers, something we look at in a comparative way in the next section.
Thus, explosive flexibility seems to have been added to the endless wage moderation, while both may hang together with the strong changes that have been wrought in the distribution of working time due to the great expansion of part-time employment and the corresponding distribution of employment and earnings over households.

CASE STUDY: DUTCH AND BELGIAN TRAJECTORIES OF FLEXIBILITY -A MATTER OF INDUSTRIAL RELATIONS AND HOUSEHOLD CONTEXT?
This section provides a companion case study to that in Chapter 3 in this volume, on Belgium, which compares wage developments in the two countries. Elaborating on the main divergence between the Netherlands and the EU found above -the explosive growth of contractual flexibility -this comparison focuses on temporary contracts, whose evolution also diverges considerably between the two countries, as we will see. This makes the question asked even more pertinent: to what extent is this Belgo-Dutch divergence related to differences in social dialogue and industrial relations, and the corresponding labour-market behaviour of households? I venture as a possible explanation that stronger industrial relations in Belgium lead to other social dialogue outcomes, but the argument needs -and will hopefully inspire -further research. Generally, the economic outcomes for the two countries are much the same. According to the National Accounts, the current level of GDP per capita is 11 per cent higher in the Netherlands. We may question, however, what that difference really means. Gross domestic product evolution since 1995 has been virtually identical for the two countries. The Netherlands did better in the 1990s, while Belgium has done so since the financial crisis. Belgian hourly labour productivity is 9 per cent higher. 16 The role of exports is the same (82-83 per cent of GDP), but Dutch imports (71 per cent) lag Belgian imports (82 per cent), which can explain the 10 percentage-points larger (im)balance surplus for the Netherlands. The share of individual consumption in GDP is larger in Belgium (66 to 60 per cent) but the amounts actually consumed per capita are exactly the same (€25 000). 17 In addition, the GDP shares of employee compensation (50 per cent) and the operating surplus of the economy (40 per cent) are nearly identical. Finally, government spending, excluding what is individually consumed, is identical as a share of GDP (8 per cent).
By contrast, clear differences arise when comparing labour-market outcomes. The Dutch (employee) EPOP ratio significantly exceeds that for Belgium, if measured in terms of employed persons (Figure 8.10). The gap varies between eight and 15 percentage points (in 2014 and 2001, respectively). However, the situation looks radically different if measured in terms of hours worked, estimated here as an FTE employment rate normalised on a 35-hour working week. 18 For Belgium the effect is negligible but for the Netherlands it is highly significant. While up to the financial crisis the Netherlands enjoyed a declining advantage, this has disappeared since, and has turned into a deficit in recent years. The role of part-time jobs -30 per cent of employees in Belgium, 52 per cent in the Netherlands -and the frequency of working longer hours than 35 per week explains the difference between the two metrics. Note that the lower Belgian incidence of part-time jobs goes together with a lower rate of involuntary part-time employment to population, 1.2 per cent as against 3.2 per cent for the Netherlands. That is, the two countries maintain roughly the same employment volume, but spread it differently over the population, particularly with regard to young people and women, and to a lesser extent older workers.
If, by way of thought experiment, Belgium were to adopt the Dutch employment demographics there could be 700 000 more people in work (Table 8.3), an 18 per cent expansion of current numbers. Taken together they would not deliver a larger economic effort, but just spread this over more persons, on the assumption that the broader spread would not affect the level of productivity. Two-thirds of the addition would concern young people aged below 25 years old. However, it might seem a dubious policy to adopt, as currently Belgian young people attain better educational outcomes than do the Dutch. 19 Most of the remaining addition concerns older workers, but here the two countries are already on an enduring, parallel trajectory of growth, albeit at different levels. 20

The Divide with Regard to Temporary Contracts
The incidence of temporary contracts shows another important divide between Belgian and Dutch employees besides the part-time gap. Figure 8.11 presents the Belgian incidence alongside the Dutch and the European average levels presented in Figure 8.2. Apart from a short-lived hiccup between 1998 and 2001, the Belgian level has hardly changed. It remains as flat as the European average, and well below this. The share among Dutch employees has increased to 20 per cent, while the Belgian share stayed at 9 per cent. The gap between the Netherlands and neighbouring, similarly developed Belgium exceeds that with most other EU countries and it has diverged as much, growing from a minimum of two percentage points to 11. Over the relatively short period since 2003 when the Dutch rise really took off, the gap has more than doubled.
The comparative evolution of temporary and permanent headcount employment rates is almost identical to the EU comparison drawn in Figure 8.3 (not shown). The Dutch permanent employment rate has declined to the level of Belgium and temporary jobs are now responsible for the total headcount EPOP difference of Figure 8.10. By implication  Table 8.3, adopting Dutch levels of flexibility would imply a contractual transition for 940 000 Belgian employees, significantly more than the 700 000 addition that the Dutch headcount levels would bring to the total number. Finally, the share of longer-term applications among all flexible contracts was identical in both countries, and gradually declined from 70 per cent in the early 1990s to 55 per cent on the eve of the financial crisis. Since then, the decline has continued, in Belgium, to 46 per cent in 2016 but turned into an increase in the Netherlands, up to 78 per cent (not shown). We may conclude that the broader spread of employment effort over the population in the Netherlands has come at the cost of greater job insecurity. Belgium manages to achieve the same thing without such uncertainties, at least for people in employment.

Household Context, Industrial Relations and Social Dialogue
The case of the Netherlands indicates that there are significant forces driving the labour market towards increasing flexibility and job insecurity. Section 3 suggests that, in the context of the household, individual labour supply makes an important contribution in the Netherlands and section 4 indicates the weakness of Dutch social dialogue. I briefly address these same issues for their comparative effects.
With regard to whether people are happy with their temporary jobs the Belgian situation looks quite different from the Dutch (Figure 8.12). The reasons given by employees for working on a temporary contract are negative -that is, they could not find a permanent job -to a much greater extent (80 per cent+) than in the Netherlands (50 per cent, Figure 8.6). Despite a substantially lower incidence of temporary contracts the share of all workers who complain about not having found a permanent contract is almost identical in the two countries currently (7 per cent). Figure 8.13, which shows for Belgium what Figure 8.7 shows for the Netherlands, presents where employees with temporary contracts are found across the household income distribution. Belgian numbers and earnings cover smaller shares of the total (11 and 6 per cent, respectively) than for the Netherlands (16 and 11 per cent), as might be expected from the smaller incidence. 21 Interestingly, relatively few single earners are found to have temporary contracts: less than one-third of the number compared with almost half in the Netherlands, and similarly for earnings. In particular, fewer single-earner temporary workers are found at the bottom of the distribution, while they make a striking contribution in the Netherlands. This largely reflects that country's concentration of young single earners, who have a much smaller presence in Belgium, and goes together with relatively larger contributions from dual-earner and multiple-earner households and from higher levels of the Belgian income distribution. This seems to suggest that, in principle, household effects in Belgium could have been the same, and that other factors must be constraining the use of temporary contracts. To the extent that young people have temporary contracts they are found higher up the income distribution in Belgium, while women are spread more evenly.
Most important may be the much greater strength of Belgian industrial relations, which determine how social dialogue works out, from the bottom (shop-floor) to the national top (Interprofessional Agreement). The formal institutions of the social dialogue (see Chapter 3, on Belgium, in this volume) look much the same as in the Netherlands,  Equally important is the broad spread of Belgian trade union membership over the population, which enhances the unions' societal influence. Belgian women have a three times higher union density than do the Dutch, at almost 50 per cent, and make up 40 per cent of total membership. 22 That shows up in the very low current (full-time) gender pay gap of 3.3 per cent, which is only one-quarter of the Dutch gap of 14.1 per cent (OECD 2017, fig. 1.3). This female union participation reflects a considerably stronger position of female employees in the Belgian distribution of household incomes than in the Netherlands (Figure 8.15). Their aggregate numbers differ little, but their spread does. Belgian women receive 39 per cent of all earnings compared with 34 per cent in the Netherlands; they have a particularly strong role as main earners (22 per cent compared with 17 per cent in the Netherlands) who are found higher up the income distribution. This may help to explain their interest in unionisation. Note also that the diverging incidence of low pay is rising and five times higher in the Netherlands compared with Belgium (see Chapter 3, on Belgium, in this volume).
This demographic spread in Belgium implies an equally important spread of union membership over the economy. Certain sectors have very substantial union density (60 per cent or more). 23 In the Netherlands such density would suffice to enable the Minister of Social Affairs and Employment to declare a collective agreement generally binding, but in practice only the high membership rate of employer organisations brings that about (Salverda 2018b). This economic spread is important for preventing industrial relations and social dialogue from unravelling at one end of the economy -in certain sectors and occupations -and ultimately affecting the whole economy -witness the rapid declines in union density in many sectors in the Netherlands. The difference shows up in a much flatter Belgian profile of temporary work across occupations and industries than in the Netherlands (not shown). While in the Netherlands temporary contracts among elementary jobs and hotels have mushroomed and far exceed those of other occupations and industries, they are in the middle of the pack in Belgium.

Conclusion
Taking the economy and the labour market together, Belgium seems to be weathering globalisation better than the Netherlands is: higher productivity, much more even trade balance, same consumption, same labour effort, much smaller incidence of low pay, much smaller gender pay gap, less involuntary employment per capita and much smaller and not increasing contractual insecurity. These outcomes combine with highly comparable institutions of social dialogue. In all probability, industrial relations make the difference. Its strong, broad and deeply rooted basis in society reinforces Belgian trade union power in the social dialogue, which has led to very different outcomes of social dialogue. Contrary to the Netherlands, certain subjects have simply been left out from the dialogue, which seems to have prevented the kind of rampant rule changes that are observed in the Netherlands (Berghman 1997). However, as argued in Chapter 3 on Belgium in this volume, this has not prevented the introduction of adaptive flexibility in many areas -the higher productivity of the Belgian economy speaks volumes. However, this flexibility is then pursued through local negotiations; it has not opened up a Pandora's box encouraging individual employers or employees to try to grab from it, unleashing processes of change that get out of control, as Dutch flexible contracts illustrate. This power position is consistent with the workings of the Belgian labour market in terms of its household  labour supply, with a strong role for full-time working women -who are often main earners -and with little or no interest taken in voluntary temporary work.

CONCLUSIONS
A detailed international comparison over the period since 2005 with the help of indicators for the EPSR, supplemented with additional indicators derived from Eurostat's databases, has led to the conclusion that, relative to the EU average, socio-economic outcomes for the Netherlands improved up to the financial crisis, but subsequently worsened in many respects. Even so, they are mostly still above average compared with the EU. Nonetheless, working-time patterns deviate strongly and, although wage outcomes remained relatively stable, household incomes declined more than the average. Most surprising, however, was the relentless rise of flexible contracts, which have doubled their incidence among Dutch employees. This is a clear case of divergence, as it has led the Netherlands away from the EU average, which remained largely unchanged, and has brought the country outside the range of one standard deviation around that average. Currently, the Dutch EPOP ratio for persons exceeds the EU average only because of a larger incidence of flexible contracts than elsewhere, while that ratio based on total hours worked by Dutch employees on permanent contracts is substantially below the EU average. Importantly, this has occurred despite the fact that the Netherlands was a forerunner of flexicurity, on which the social partners struck a deal in 1996, which was turned into law in 1999. In the same year the EU Directive on temporary work was enacted, which tended in the same direction, and that was later followed by the Directive on temporary work agencies in 2008. Taken together, the three are the textbook examples of social dialogue and its importance for policy-making as in each of the three cases, law and directives, the social partners agreed the new rules, which then passed literally into the legal rules. Unfortunately, in the Netherlands the three together have failed to contain contractual flexibility and keep it within acceptable limits. That is, the Dutch deal of 1996, which strongly focused on TWAs and established collective agreements for the TWA sector, has been relatively successful, as the subsequent growth of TWA employment was significantly reduced. The strong rise of flexibility rests almost entirely on that of other ever more flexible types of contracts, categorised here as 'on call'. Employers have massively exploited the options provided by the Flexicurity Law, making direct individual hires instead of through TWAs as intermediaries. Such direct hiring is generally beyond the influence of unions, which have little presence on the shop-floor in the Netherlands. Despite adjustments made at the initiative of the government with the new WWZ Law introduced in 2015 to replace the Flexicurity Law of 1999, there is no end in sight. The growth of flexibility continues as fast as before and it is spreading rapidly in almost all directions -men compared with women, older age workers compared with younger, longer working weeks compared with small jobs, higher educational attainment compared with lower, most occupations and many industries -leading to a progressive unravelling of permanent contracts across the economy. The Polder Model has opened the sluices to a flood of contract flexibilisation and taking temporary emergence action is no longer a solution. Massive flexibility is infusing a level of insecurity in the labour market, which has become a bone of contention in the Dutch public debate.
To understand this outcome, I proposed looking at the radical change in the household context of labour supply following the demise of the single-earner world and its replacement with dual-earner and multiple-earner households. Currently, three-quarters of all employees share a household with another employee, in the Netherlands as in virtually all other EU countries, albeit with different structures of working time. This change not only affects the nature of socio-economic outcomes, making it harder for the social dialogue to have an effect, but also upsets the underlying industrial relations, altering the effectiveness of the social dialogue itself.
With the Dutch case I have illustrated the former effect regarding outcomes. First, most people on flexible contracts are unhappy with that situation, but certainly not all, and, secondly, though people on flexible contracts achieve much lower earnings in the labour market compared with those on permanent contracts the disposable income of their households, on average, differs remarkably little from that of permanent workers. Both point to the contribution that the presence of another earner in the household can make, which is missing in the single-earner world (and now puts remaining single earners in a difficult position). That effect may be accommodating as well as inhibiting: encouraging those who like the flexibility, softening the unhappiness with it of the others, and possibly preventing job changes because of the within-household division of labour.
With a comparative case study of the Netherlands and Belgium, I shed light on the latter effect regarding industrial relations and the social dialogue itself. Belgium shows higher productivity in combination with similar economic performance and labour effort. It maintains a comparable format for social dialogue, but manages to fill it with a rather different content. In particular, the Belgian incidence of flexibility has remained unchanged, as much as the EU average. The main country differences concern the industrial relations and the household context. Belgian union density has not declined and is three times higher than in the Netherlands, and it covers women equally, delivering a four times smaller gender pay gap along the way. It extends strongly across industries, thus diminishing the risks of an unravelling starting at one end. The union side is much stronger in Belgium and makes a difference to the dialogue, leaving contract flexibility out of the discussion. The difference in industrial relations is consistent with that of the household context. Although the picture of single, dual and multiple earners is generally the same in the two countries, it goes together with a significantly different working-time structure. Belgian women work more often full-time and are main earners in the household, and they receive a larger share of the total wage bill. Their much stronger union membership can be no surprise and offers essential support to the industrial relations side of the social dialogue. This consistency is certainly not a full explanation, which must include the Ghent system, though the attraction of that system will likely be less for persons in smaller jobs.
The Belgo-Dutch comparison has a broader significance, since it underlines that the social dialogue is as weak as the weakest partner and tends to go the way of the strongest. This is shown most clearly where the social dialogue has been given full responsibility, as in the two Directives for flexicurity. Social dialogue characterised by weaknesses and strengths that differ importantly between countries, can offer no guarantees for European convergence unless proper attention is paid to assuring a level playing field between the participants.
Can the EPSR, still young and fresh, offer assistance? The relevant social right is Wiemer Salverda -9781788978071 Downloaded from Elgar Online at 04/28/2021 08:19:55AM via free access number 5, the first in the second chapter of the EPSR, on fair working conditions: 'Regardless of the type and duration of the employment relationship, workers have the right to fair and equal treatment regarding working conditions, access to social protection and training.
The transition towards open-ended forms of employment shall be fostered.' It encompasses both equal treatment and job security, as the two Directives did. Declaring rights is not the issue, implementing them is. The Dutch lesson is that implementation requires much stricter verification and would be greatly helped by establishing individual legal entitlements, supporting social rights as an expression of political will. This could help to level the playing field and change the content of the social dialogue for unions confronted with employers' associations that are unwilling to engage in dialogue and individual employers that make massive resort to flexibility options not covered by the social dialogue, which focuses on temporary work agencies.