ISBN 92-2-111797-9
ISSN 1020-5322
First published 1999

Foreword

The present study is part of the ILO's follow-up to Commitment 3 of the Declaration and Programme of Action of the World Summit on Social Development in Copenhagen in 1995. Commitment 3 reiterates the importance of full, productive and freely chosen employment, as a basic condition for social progress among UN organizations. The ILO has a specific responsibility for monitoring the progress made by countries in the fulfilment of Commitment 3.

In addition to the preparation of Country Employment Policy Reviews (CEPRs) in the developing and transition countries the ILO also decided to review progress of some OECD countries towards full employment. Among the countries experiencing improvements in their labour markets in recent years, four smaller European countries (Austria, Denmark, the Netherlands and Ireland) were selected. Contrary to some of the bigger European countries, these four countries have been experiencing an impressive labour market recovery, or have maintained a low level of unemployment over the long term. For each of these countries, a national CEPR was prepared in consultation with the social partners and the governments.

The present CEPR shows the achievements of the Netherlands in curbing unemployment and in increasing employment and employment rates. Especially unemployment has decreased continuously since 1993 reaching a low of around 3,4 per cent today.

Reasons for the success on the Dutch labour market are the social partner's commitment to wage moderation, labour market and social security reform and a stability oriented macroeconomic policy. Other reasons include the increase in part-time and other "flexi" jobs and a depreciation of wage costs relative to the country's most important trading partners.

However, some problems remain: while long-term unemployment is decreasing quickly, it remains high and a large number of people is still dependent on social security transfer payments ( e.g. invalidity pensions). Gender gaps in employment are also important, especially when measured in full-time equivalent employment rates.

This report, together with those on Austria, Denmark and Ireland, form part of the project on Country Employment Policy Reviews in selected OECD countries. Two other publications are planned under this activity: Peter Auer "Employment revival in Europe: Labour market success in Austria, Denmark, Ireland and the Netherlands (ILO, Geneva, forthcoming) and Peter Auer (ed.) "Labour market institutions for decent work (ILO, Geneva, forthcoming).

Gek-Boo Ng
Chief
Employment and Labour Market Policies branch
Employment and Training Department

1. Introduction

Nations are just like individuals. In consumer theory, the reference group effect has long been recognized as an important factor in determining household consumption behaviour, with consumption decisions in one household geared to observed behaviour in other households. Similarly, political debates often have a setting like a densely populated street, where everyone is watching everyone else, with little spy-mirrors on the window for fear of missing out on important moves: what are the neighbours doing, how can they afford it?

If we step behind, and look from a distance, it is like watching a merry-go-round. We see a country take the lead in economic performance on some dimension and all the spectators are impressed. But the carrousel makes its eternal turns, both in the long run and in the short run. The eye-catching leader disappears around the corner, and the runner-up proudly rises to prominence. Long ago, the historian Herbert Heaton (Heaton, 1963, 724) noted how countries imitated social protection measures from each other, with England copying health insurance schemes from Germany and Germany copying the English unemployment insurance system. This process still goes on, in a continuing attempt to learn from the good and the bad experiences in other countries.

A striking recent illustration is given in Van Ark and De Haan (1997). They analyse GDP growth performances in north-west Europe since 1960. North-west Europe consist of 11 countries.⁽¹⁾ For every country except Sweden, they can identify a period of above-average growth (five years or more of growth above the NW Europe average). In 12 out of 14 cases, the five year period after the interval of success had below-average GDP growth. Strikingly, the average growth surplus across countries and periods was 1 per cent the average shortfall in the subsequent five years was also 1 per cent! Assuming the reverse also holds, this suggests an intriguing hypothesis. A period of lagging behind in the reference group then calls forth a period of catching up. There may be responses in the economic system that prevents permanent, ever stronger lagging. A key response may be the adjustment through economic policies.

This report is not meant to test such a general hypothesis. It only attempts to offer a little spy-mirror on the window for watching The Netherlands. The performance of the Dutch economy in the last decade caught a lot of attention, and we will outline that performance. But it turns out that the Dutch case does seem to fit in the general pattern, of lagging behind from an initially strong relative position and then catching up again. We will offer a review of Dutch policies and behaviour of key players in the last two decades. We will do that against a brief sketch of the historical development of the Dutch economy, outlining some long-lasting structural features and a brief overview of main postwar developments. Obeying the laws of perspective that also reign in little mirrors, the present will be larger than the past, but of course cannot be disconnected from it. In the end, evaluating our results, we will face the question to what extent there exists something like a "Dutch model".

2. Setting the stage

2.1 Old traits

History does not determine the present in every minor detail. But throughout history, a nation develops some traits and behavioural rules that become essential elements in the way it operates. And the Dutch economy has some features that are so old as to look like a genetic structure of our economic bones: rich, fertile, low female participation, consensus oriented.

The Dutch nation has been a rich nation for centuries. In fact, De Vries and Van der Woude (1995) argue, in a fascinating study, that modern economic growth started in the low countries, preceding the industrial revolution in England, and perfectly obeying Rostow's "stages of economic growth", fitting even better than the English case. Population growth has been extremely high. In the five centuries after 1500, the Dutch population increased fifteenfold, whereas the French increased only fourfold and the Belgian sixfold (De Vries and Van der Woude, 67). Married women's participation was very low by international standards in the prosperous seventeenth century (De Vries and Van der Woude, 1995; Schama, 1988), easily interpreted as a strong income effect on female choices.

High income and wealth, low female participation and strong population growth are still characteristic for the twentieth century. For most of the century, the Dutch fertility rate was the highest among the seven countries compared by Pott-Buter (1993, 181) with some swings more marked than elsewhere. The baby boom after the Second World War was more excessive, and so was the drop after 1970. In fact, after 1970, the Dutch fertility rate sank from the top to the bottom of the distribution. Still, all during the period 1913-90, the rate of population growth in The Netherlands was substantially above that in north-west Europe, easily reaching double that rate in many sub-periods (Van Ark et al., 1994, table 1). In the seven-country comparison, the labour force participation rate of Dutch married women was the lowest for most of the century (Pott-Buter, 1993, 200-201). The growth rate of GDP in the period 1913-79 was higher than in north-west Europe, and higher than in the United States.⁽²⁾ Only in the period 1979-90 was it lower compared to both standards. Growth rates per capita were not always the highest, because of high population growth, low participation rates and reductions in working hours. But GDP per hour worked was always above the NW European average and climbed from 71 per cent of the United States value in 1913 to 97 per cent of that value in 1990 (Van Ark et al., 1994).

Another structural Dutch feature is the consensus oriented mode of operating in politics and social relations. Some observers trace this even to the earliest history of Dutch society, where protecting (and winning) land from water and floods created the oldest cooperative public bodies of government (the waterboards, founded in the twelfth and thirteenth century). Others link it to the particular modus vivendi that emerged in a religiously divided country where no political party has a majority and coalition governments are the only viable option. Under that modus, the late nineteenth century brought the development of a system of political management where severe ideological separatism at the bottom combines with willingness to compromise at the top, to produce a fair degree of social and political stability. The consensus orientation is clearly present in the institutions of the labour market that will be extensively discussed below, and that exhibit a remarkable continuity throughout the century. The postwar institutional structure is clearly rooted in the prewar institutions (Windmuller and De Galan, 1979). The cooperative spirit of labour relations is often traced to deliberate choices made to that effect in the beginning of the century by one of the leading union federations, after bitterly loosing a major strike (Windmuller and De Galan, 1979). The consensus approach often seems all pervasive. Macroeconomic forecasts by CPB, the government advisory institute have always played a prominent role in policy discussions. D'Alcantara, commenting on forty years of CPB model building, even senses the consensus tradition in research decisions, characterizing Dutch macroeconomic modelling "as a mechanism which builds up consensus, which selects from the results those that are common to all models" (d'Alcantara, 1988, 198).

2.2 Highlights of postwar economic performance

Table 2.1 shows the very high population growth rates relative to NW Europe during the postwar period. GDP per capita grows less than in NW Europe in the early decades. In the period 1960-73 the growth rate is about equal to the average, and then it starts to lag behind. Van Ark and De Haan (1997) extend the data and use 1973 and 1987 as period boundaries. GDP growth per capita is relatively sluggish in 1973-87 (1.2 versus 1.8 per cent) and relatively strong in 1987-1996 (2.1 versus 1.4 per cent). The growth rate in GDP per hour does not recover in the later years. GDP per hour grows faster until the end of the 1970s and then starts to lag behind. It was clearly above the NW European average in 1973-87, but it was clearly below in 1987-96 (Van Ark and De Haan, 1997, figure 4). Still, the level of GDP per hour worked was always above the average for north-west Europe and in 1990 was only 3 per cent below the United States level. In that year, value added per hour worked in manufacturing was higher than in Germany, the United Kingdom and the United States (Van Ark et al., 1994). In 1996, GDP per hour worked was about the same as in the United States, while the average for NW Europe was 89 per cent of the United States level. GDP per capita in the Netherlands was then 75 per cent of the United States level, while the average for NW Europe was 78 per cent of the United States level (Van Ark and De Haan, 1997, table 6). The growth in income per capita may be related to the growth in labour force participation, from below the EU average in 1983 to above the average in 1996.⁽³⁾

Some key statistics for the labour market are presented in table 2.2. Labour force growth has always been very strong. For men, the growth rate was never below the EEC average, for women it was always substantially above that average, and in the eighties even substantially above the United States growth rate. Employment growth matched this, and was never below the EEC average. As a result, the unemployment performance was relatively favourable most of the time. The exceptions are the period right after the first oil crisis and the recession period in the early eighties. Figure 2.1 illustrates. It also illustrates the faster than average recovery after the deep dips. This is the eye-catching performance: an unemployment rate markedly below the EU average during the 1990s. Whereas the EU average was 11 per cent in 1995, it was 6.5 per cent in the Netherlands. Similarly, annual employment growth was clearly above the EU average (see the Appendix tables). Moreover, GDP growth was strong, with an average of 2.8 per cent per annum in 1983-89 and 2.6 per cent during 1990-96. Clearly, a good performance by international standards.

2.3 A brief postwar economic history

Recent economic policies build on a history of policies and institutions. As ever, understanding the present is enhanced by knowing the past. We will give a brief summary here, and refer to Hartog and Theeuwes (1993) for a fuller account and further references.

The first postwar decades, until 1964, were characterized by reconstruction and catching up. Right from the beginning, wages were a key variable of economic policy, because of their pivotal role as a cost factor for an open economy that had to sell much of its product in competitive world markets. From 1945 until 1954 there was government wage control, until 1959 there were national wage guidelines, between 1959 and 1962 the guidelines permitted differentiation by industry, in 1963-64 the system exploded. From 1945 to 1954, collective agreements were only binding after approval of the Committee of Government Negotiators (with independent members, but appointed by the government); the Committee always first asked advice from the Foundation of Labour. The Foundation of Labour is a private organization, with half the seats for union federations, the other half for employer organizations, and the chair rotating between them. Between 1954 and 1959, the wage guidelines were no longer binding: they only set an upper limit. In practice, the upper limit became the actual increase. Between 1959 and 1962, the formal system was maintained (advice by the Foundation, approval by the Negotiators), but rather than setting uniform national wage increases, differentiation by industry became accepted: wages should follow the increase in the industry's productivity. In 1963, the rules were changed. Now, the Foundation became the authority to approve (or reject) collective agreements. Agreements should be compared with norms set by the Foundation, based on assessment of the general economic outlook by the Social Economic Council. The Social Economic Council is a public organization, with the seats equally divided between union federations, employer associations and independent members, appointed by "the Crown" (i.e. the government, with a signature of the Queen). The chairman is independent. In the decision-making cycle of the initial postwar decades, the Central Planning Bureau would analyse the economic situation and give annual economic forecasts. The Social Economic Council would outline desirable policies, and produce a semi-annual advice. Until recently, the government was even under legal obligation to ask advice of the Council on all important socio-economic policies and legislation. In 1968, the system was formally abandoned. The Committee of Government Negotiators lost its authority, contracts were formally approved by the Foundation without testing against any norms: wage determination was free. Effectively, the system of centralized wage determination already broke down in 1963, when the rank-and-file translated the excessive excess demand in the labour market (unemployment below 1 per cent) into much higher wage demands than centrally agreed, and individual employers agreed. The important point is that there is an early tradition of consultation and bargaining at the central level, with wage cost as a key variable, and with peak level organizations of workers and employers dealing with each other and with the government. And there is a tradition of government intervention, interwoven in a process of consultation and advice, not as bluntly imposed rulings. In the Foundation, social partners had been jointly involved in setting the norm for the wage increases, which would then be imposed or sanctioned by the Government.

The decade of the 1960s, for the present purpose running from 1963 to 1973, was the period of building up the welfare state at high speed: 1963, Law on Child Allowances; 1965, General Social Assistance Law; 1967, Disability Law; 1966, Law on Sickness Benefits. The share of taxes and social security premiums increased by 1 per cent point per year. Revenue from natural gas started to flow towards the government. It was also the period of a very tight labour market; in the period 1960-66 unemployment on average was below 1 per cent. Yet, the first signs of structural changes were also visible. After 1967, unemployment increased and there are clear indications that the UV curve shifted outwards. Government concern over wage cost never waned, and in 1970, the Law on Wage Formation was passed. It allowed the Minister of Social Affairs to nullify wage agreements and to impose a temporary freeze on wage increases (the right is removed in 1976). Resistance against that law from social partners has been quite strong.

The oil price shock in 1973 definitely marked the end of the "golden era". Unemployment rapidly increased. The problems were initially tackled with Keynesian expansion policies, job creation programmes, investment subsidies, all aided by the Government's share in the gas revenue. The large export of gas, however, also made for a strong external position, a high exchange rate, and large losses of employment in the exposed sector. The government revenue from gas increased in a few years from 2.4 to 4.9 per cent of the national income (it is now down to 0.5 per cent). After 1975, the policy goals changed towards budget deficit reduction and government retrenchment. However, the policy utterly failed. In a period of five years, the government share in national income went up by 9 per cent points. This is the period of "Dutch disease": export curtailed by a strong guilder, government (transfer) programmes fuelled by high gas revenues. Wages continued to increase (stimulated by the shifting of taxes and social security premiums), profits were squeezed. The wage share in market sector income increased to 92 per cent. The government kept hammering out the necessity of wage restraint, and actually intervened in 1974, 1976, 1979, 1980, 1981 and 1982 with mandates on wages: wage freezes, outlawing automatic cost-of-living adjustments, etc. The period 1973-82 was a period of muddling through with the government's fiscal policies, rapidly increasing unemployment and intervention in wage formation that increasingly irritated and frustrated social partners.

2.4 Turning point 1982

The year 1982, in hindsight, is usually marked as the turning point. It is the year of the Central Agreement between the leading labour federation and the employers' federation, struck in Wassenaar, and hence dubbed the "Wassenaar Agreement": wage restraint, reduction in working hours, intended to restore profit levels and create employment growth. At the same time, the government became more determined to control its expenditures and its deficit, and to reform social security. In particular, the inflow into the disability pension had been incredibly high. More generally, and following the mood in the rest of the world, the policy orientation definitely changed from Keynesian to a neoclassical, market based orientation. After 1982, the great Dutch consensus machine became effective again, as it had been before.

Why could this new start be made in 1982? There are several reasons. Perhaps the key reason is that unemployment had really skyrocketed in the early 1980s, and there was a strong sense of urgency. At some point in time, unemployment increased by 10,000 persons per month (Visser and Hemerijck, 1997, Chapter 1): the boat was sinking dramatically fast. Second, unions and employers "were not amused" by the repeated government interventions in wage formation. The government had intervened with six binding wage regulations of various sorts in the period 1974-82. They really had a strong desire to settle their affairs among themselves. And third, the whole institutional structure had always been there to allow a return to the consensual nature of industrial relations that is typical of the Dutch situation. A social agreement fits in the history and in the institutional and socio-cultural environment, it is by no means a social innovation. Centralized, corporatist decision making could resume its course again. The Wassenaar Agreement has been updated and confirmed on several occasions, as bi- or tripartite agreements, in 1989, in 1993 and in 1997. Other agreements concern early retirement and labour market flexibility.

The programme that has been developed after 1982 has several components and several actors. It has not been formulated as some "Manifesto 1982", but the goals were certainly around at the time (they were not suddenly set 1982), and so was the sense of interrelatedness of the elements. It consisted of the following components:

-- The government was to get its expenditures under control: government retrenchment, reduction in the burden of taxes and social security premium.

-- The social partners were to implement the Wassenaar Agreement: wage restraint, working time reduction, more labour market flexibility.

-- Social security was to be reformed.

Before considering this programme in greater detail, in the next section we will sketch the Dutch institutional setting, and the way it works. The particular institutional environment (and history) has been essential to allow the policy mix that has been implemented.⁽⁴⁾

2.5 The system of labour relations

The Dutch labour market has strong corporatist features. It is not easy to give a sharp definition of corporatism, let alone measure it unequivocally, but there is wide agreement on core elements.⁽⁵⁾ Essentially, corporatism is a form of organization where government and organized interests (mainly trade unions, employer associations) jointly develop and implement social-economic policies. The government does not operate at a distance, organized interest does not have to lobby, they are welcome partners in the conference room. There is tripartite consultation at all stages of legislation and policy making. Often this is institutionalized in advisory bodies, but much coordination takes place in an informal way. Organized interests are usually united in large federations or peak organizations. Bargaining over collective labour agreements is usually highly centralized. Quite often there is a strong concern for equity issues and incomes policies. Highly corporatist countries are Austria and the Nordic countries. Germany is also rated as rather corporatist. Non-corporatist countries have very decentralized, firm level bargaining and no institutionalized consultation between government and organized interest. Examples of non-corporatist countries are the United States and Canada, and in Europe, the United Kingdom. Teulings and Hartog (1998) document that the higher the degree of corporatism in a labour market, the smaller the magnitude of non-competitive wage differentials, such as differentials associated with industry affiliation, firm size and collective bargaining regime. We also argue that there is certainly no evidence that wage flexibility (a complicated concept to deal with, both theoretically and empirically) is higher in decentralized than in corporatist settings.

The Dutch institutional setting is governed by two key laws on collective agreements. The 1927 Law on the Collective Agreement makes such an agreement binding for all workers in the firm, not just members of the union who signed the agreement. Formally, it is possible to negotiate in the agreement that contract terms only apply to union members, but if so, this cannot be exclusively for a particular union: they should apply to members of any union. Exclusive bargaining for union members only is never possible.

Formally, the legal basis of labour relations allows free entry of new unions. Anyone can start a union and ask for negotiations with an employer. The employer is free to honour or to ignore such a request. Formally, he may choose to negotiate with one union and not with another, or he may use different channels, like works councils. There are no formal representation rules, no compulsory elections, no rules on bargaining or bargaining structures. Because of the 1927 Law on the Collective Agreement, when an employer signs a contract with a single union, this contract automatically applies to all its employees. Other unions are then left out. This happens infrequently, but when employers demand concessions in harsh times, it sometimes occurs that one union agrees, and that the others can only follow suit, their protests notwithstanding.⁽⁶⁾

The 1937 Law on Mandatory Extension states that if a collective agreement covers a substantial majority of the industry, the Minister of Social Affairs can extend the agreement to the entire industry. By now, a substantial majority means 55 per cent of the workers employed by employers that are directly bound by the agreement. To get extension, at least one of the parties that signed the agreement must send a request to the Ministry of Social Affairs. The Ministry checks some formal criteria and publishes the request to allow other parties to raise objections. Formally, the Ministry is obliged to ask the Wage Committee of the Foundation of Labour ("Stichting van de Arbeid", a private organization introduced above) for advice. In practice, advice is only asked when objections are raised. After receiving the advice, the Ministry comes to a decision on the request. The maximum duration of an extension is two years.

In the Foundation of Labour, bargaining parties meet and consult each other, give joint recommendations to their members regarding wage restraint, training and additional employment policies. In some years in the past, a Central Agreement was negotiated within the Foundation of Labour. The Wage Committee of the Foundation of Labour plays an important role in the system of labour relations.

Mandatory extension enables contracting partners to extend contracts in a large number of industries. Only in retail trade and wholesale, where there are so many firms and their size is so small, is it difficult to satisfy the requirement of a substantial majority of the workers to be bound directly. Instead, joint public bodies (PBOs) of employers and trade unions have been installed by law in most of these branches. Negotiations on industry agreements take place within these bodies, which then have the legal right to extend these agreements to the whole industry without interference from the Ministry. These agreements are called rulings. The character of these rulings differs from the usual industry contracts, in the sense that they are really imposed upon the firms in the sector, as most firms are too small to affect the outcome. The most important ruling is in retail trade.

In 1996, only 28 per cent of the workers is member of a union. Most unions belong to one of three federations that are member of the Foundation of Labour: FNV, with socialist/catholic roots,⁽⁷⁾ the protestant CNV, and the middle and higher personnel federation MHP. FNV is by far the largest federation. In 1985, 58 per cent of the union members were affiliated with FNV, 20 per cent with CNV, 7 per cent with MHP and 15 per cent belonged to other unions. At the bargaining table, the most common situation was the joint presence of FNV and CNV unions: 52 per cent of the collective agreements. 29 per cent of the agreements were entered jointly by FNV, CNV and MHP, 11 per cent were entered only by a FNV union, and 4 per cent by unions belonging to the two main federations (FNV, CNV) and some other union (like a company union, such as for the railways).⁽⁸⁾ The organizational structure is now highly stable.⁽⁹⁾ The largest federation, the FNV, consisted until recently of 20 industry unions. However, their size distribution was heavily skewed, the two largest unions (civil servants and manufacturing) covering about 50 per cent of its total membership. In the Spring of 1998, four large unions merged, further adding to the skewness. A similar structure (but somewhat more concentrated in the public sector) can be found in the CNV.

On the employers' side, there were until recently two national federations (VNO and NCW) which have now merged into a single federation (VNO-NCW). Membership is made up of both firms and lower level employer organizations, often organized by industry. These lower level organizations are the ones involved in bargaining. Their involvement differs across sectors. There is a separate organization for small and medium size firms, KNOV. Sometimes an employer organization bargains directly with the unions (in the metal and banking sector for instance), sometimes the employer organization provides mainly support to employers who do the negotiations themselves.

The institutional setting of the Dutch labour market generates four relevant bargaining regimes in the private sector.⁽¹⁰⁾ First, a company can negotiate its own collective agreement with one or more unions; second, a company can be part of an industry agreement; third, it can be roped in by an extension of an industry agreement; finally, a firm can be uncovered and have no collective bargaining.

Whenever there is a collective agreement in the firm, this does not necessarily imply that all employees are covered. Top level management is always excluded from the agreement. Furthermore, other categories of workers may be excluded. Mostly, these are low-paid workers on temporary employment. The formal rules for exclusion are not clear and we have no good information on exclusion practices. Table 2.3 gives an overview of the coverage situation in 1990 and 1997(showing the joint effect of firm and individual coverage). Most collective agreements are industry agreements, and the share is even rising. In 1990, 82 per cent of covered workers were covered by an industry agreement, in 1997 it was 88 per cent. The share of the company agreement was 14 per cent in 1975, it climbed to 18 per cent in 1990 and then dropped to 12 per cent in 1997. The number of workers covered by a non-extended industry contract has been increasing, from 3 per cent of covered workers to 10 per cent Casual evidence suggests that non-extension occurs mainly because either almost every firm in the industry joins the contract "voluntarily" or because requesting extension has a disadvantage for the signing parties.⁽¹¹⁾ Coverage through mandatory extension has been declining. As the table indicates, by far the majority of covered workers is covered by industry level bargaining. But focussing just on the formal bargaining level would not do justice to the intricate system of labour relations. The system has a remarkable balance with unions and union federations playing an important role, precisely because their formal position is rather weak. Union density is less than a quarter of the labour force and multiple unionism combined with the non-discriminatory status of collective agreements weakens the union side (an agreement signed with any union is binding for all workers in the firm and possibly in the industry).

Mandatory extension seems to give organized interest a strong grip on their playing field, but the grip is different from a tight monopoly. The system appears to operate in a very flexible way. Flexibility comes through different channels. The first channel is through flexible boundaries. The collective agreement itself defines the domain of application, the firms it intends to cover. Coverage is subject to bargaining and may move over time. Firms may seek coverage by other agreements if they become dissatisfied. Peak organizations advise on boundary conflicts (The Wage Committee of the Foundation of Labour) and they try to solve conflicts as much as possible by inviting the parties themselves to work out a solution. Shifting boundaries prevents rigid imposition of collective bargaining results by mandatory extension. Sometimes employers seek reduction of wage cost by moving particular jobs from one collective agreement to another. An example is the transfer of jobs covered by an agreement in construction to an agreement in agriculture, where pay is lower. There are also some ways to evade the effect of mandatory extension. Several agreements explicitly allow for permission to deviate from the terms of the agreement under "special circumstances". Also, some agreements have a clause calling for adjustment of the agreement under "exceptional and grave changes in the general socio-economic situation". However, these clauses only serve as emergency exits when the continuity of the firm is at stake. But existence of a company agreement is sufficient argument for exemption from mandatory extension. And after a recent, tough battle it has been accepted that a company agreement does not have to be concluded with a union that belongs to any of the established federations. Firms may also opt for non-compliance with an industry agreement. This may occur in sectors where the union is too weak to enforce compliance. It is known to happen in construction, retail trade and wholesale trade.

The existence of escape routes and the potential vulnerability of established positions lead to quite moderate results. There is no evidence that actual wages are substantially above market wages. Econometric analysis shows that wage differentials between bargaining regimes are quite small, no more than a few percentages (Hartog, Leuven and Teulings, 1997). The regime with mandatory extension does not generate the highest wage levels: they are realized under company agreement. Collective bargaining mostly sets minimum wage levels. If these were binding constraints we should observe spikes in the wage distribution at the level of the contract wage, but this does not happen. We only observe spikes in cleaning and retail trade, but they occur at the level of the legal minimum wage. Many wages seem to follow the pressure from market forces. For example, in construction wages are determined "at the gate", when employers recruit for a going construction project. In wholesale trade, wages in the Eastern part of the Netherlands are paid quite close to the contract level, but in the Western part wages are easily 10-15 per cent higher. And representatives of unions and employers do not deny that payment below agreed scales occurs if the market situation requires it.

Freeman, Hartog and Teulings (1995) analyse the Dutch system of wage formation and draw the following conclusion:

"In summary, the Dutch system of corporatism as it stands now is characterized by weak unions which, however, have got leverage by a system of mandatory extension of contracts that is based on legal support by the Ministry of Social Affairs. Mandatory extension provides a barrier to entry for new unions because they are not represented in the Wage Committee of the Foundation of Labour. It helps therefore to keep together the federation of unions. A similar argument applies to employer federations. Employers gain from this system by the fact that unions do not care too much about having a strong organization within the firm (like the shop steward system in the United States and the United Kingdom), so that distributional conflicts are resolved outside the firm. The essential characteristic of the system is the balancing act that the organizations of workers and employers have to perform. They get a fair amount of background support from the government, endorsing, extending and influencing the outcome of bargaining. But the organizations have no monopoly, no legal exclusivity and they can only play their role as long as workers, employers and member organizations of the federations are satisfied. Potentially, there are escape routes for the players in the game. The escape routes are seldom used however, suggesting the system does not function as a straightjacket on labour market participants."

There is more evidence that the Dutch system of wage formation is flexible and sensitive to the interest of outsiders. Kniesner and Goldsmith (1987) discuss aggregate models for the United States labour market and they start with some regressions to assess the cyclical sensitivity of unemployment and wages. They regress unemployment, nominal and real wages on the logarithm of GNP, for the period 1948-85. Replicating that relation for the Netherlands points to smaller unemployment fluctuation and greater wage flexibility in the Netherlands.⁽¹²⁾ Layard, Nickell and Jackman (1991, page 58, table 7) find that real wage rigidity in the United States and the Netherlands rank about equal, while nominal rigidity is ranked much higher in the United States than in the Netherlands. However, it is important to repeat the warning on the complexities of assessing aggregate wage flexibility (see the elaboration in Teulings and Hartog,1998, Chapter 7).

In Teulings and Hartog (1998), interviews with key players are used to picture how the game of collective bargaining is played. These interviews clearly establish the influence and the power of labour federations over member unions, with federations always emphasizing the need for wage restraint. In fact, union federations are the essential institutions for imposing this restraint. Van den Toren (1996, Chapter 4) tells the same story. Negotiators indicate that their claims in collective bargaining are heavily influenced by internal coordination within the union movement. Central agreements carry not much direct weight, but the translation of such an agreement into the formulation of the initial bargaining claims by the union federation is important. Our conclusion is also supported by results reported in Hartog, van Opstal and Teulings (1996). For the United States, Beaudry and DiNardo (1991) established asymmetry in wage adjustment to labour market conditions. Wages appear to respond to the minimum unemployment rate during a worker's tenure: wages are increased in booming times, but a slack labour market does not affect incumbent's wages. The burden of adjustment thus falls on workers who are new to the firm. Such asymmetry cannot be established for the Netherlands: market conditions affect incumbents and the firm's new hires alike.

As mentioned above, our findings for the Netherlands are part of a broader result on wage formation under corporatism (Teulings and Hartog, 1998). Empirical research shows that non-competitive wage differentials are smaller in more corporatist countries. In that sense, corporatist countries get closer to the competitive rule of one price for a given quality of labour: deviations associated with industry affiliation, firm size, tenure and bargaining regime (e.g industry versus company bargaining, or union-nonunion in decentralized countries) are smaller. Such a non-competitive wage differential serves no allocative goal. It can even be harmful, as is demonstrated by Hibbs and Locking (1995) for Sweden. Augmenting a production function with a term for wage dispersion, they find that increased wage dispersion within the firm is beneficial to labour productivity, while increased dispersion between firms and between industries is harmful.

Freeman, Hartog and Teulings (1995) present evidence that structural adjustment of labour does not appear to be prevented by the system of wage formation. Their report examines the degree to which employment moved across industries in the period 1975-89. The measure of reallocation is the Index of Structural Change (ISC), defined as half the sum of the absolute value of the change in the share of employment in the relevant grouping. It gives the percentage point change in the distribution of workers that would equate the distribution of employment in two periods. Large values indicate that the labour market is subject to large shocks and that the labour market is adjusting well to those shocks. Low values are more difficult to interpret. They could mean that large shocks are taking place but the labour market is adjusting sluggishly or that the shocks are small and the labour market is making the appropriate adjustment.

European countries, like the United States have redeployed their work force away from manufacturing into service sectors. From 1975 to 1989, the ISC is calculated for the Netherlands, the United States and various European economies, based on one-digit industry classifications. See table 2.4. It turns out that the ISC for the Netherlands and countries with mandatory extension (ME) was essentially no different than those for the United States and various economies without mandatory extension.⁽¹³⁾ In these data, the United Kingdom stands out with a high index value, pointing to large shocks and quick adjustment. The data do not support the notion that the Netherlands and other countries with ME had smaller quantitative flexibility over this period. The same conclusion emerges from a more disaggregate grouping of 3 and 4 digit ISIC code industries using data for 35 manufacturing industries from 1977 to 1990. Again, the resultant statistics show no difference in ISCs by use of ME in wage-setting. The countries with the most structural change are the Scandinavian countries without ME but with high collective bargaining. The Netherlands has a roughly comparable index of structural change as the United States and the United Kingdom There is no evidence that the reallocation of workers across industry lines was greatly affected by the system of industry wage-setting and industry bargaining.

For several years, there has been a public debate about the system of mandatory extension and the employment losses it would entail. Freeman, Hartog and Teulings (1995) summarize the debate and argue that moving towards deregulation by abolishing mandatory extension is unlikely to bring benefits. Rather, it may threaten the entire system of labour relations. In some sectors unions and collective bargaining may disappear, in other sectors unions may be forced to increase membership by costly unionization drives. Instability may emerge as a more confrontational system of labour relations takes hold with employers opposing unions, possibly taking the form of "organization wars" and resulting in an increased dispersion in collective bargaining coverage across industrial sectors. And as the empirical literature indicates, it is not very likely that the new system, with more wage variation, will bring any benefits in terms of additional employment.⁽¹⁴⁾

Table 2.1 Population and DGP: annual growth rates

Table 2.2 Labour force growth, employment growth and unemployment

Table 2.3 Employees in private and subsidized sectors bound by collective agreements in 1990 (in millions and per cents)

Table 2.4 Country indices of structural change, 1975-1989

Figure 2.1A Standardized employment rates, 1969-1997

Figure 2.1B Real GDP growth rate, 1969-1997

Figure 2.1C Employment growth rate, 1969-1997

3. Impact effects: Government budget, wages, working hours

3.1 Government budget control⁽¹⁵⁾

The central government budget deficit has been reduced from 8.9 per cent of GDP in 1983 to 1.4 per cent in 1996, the deficit of the total public sector (central government, local government, social insurance) from 8.6 per cent to 2.0 per cent. The share of taxes and social security premiums was reduced from 47.4 per cent to 44.4 per cent. The share of public expenditures in GDP fell from 58.0 per cent to 46.9 per cent. Employment volume in the public sector dropped from 698,000 person years to 650,000, which means a drop from 15 per cent of employment volume to 11.7 per cent. The national debt, at 63 per cent of national income in 1983, initially increased to about 81 per cent in 1993 but since then has fallen to 78 per cent in 1996. The "wedge", i.e. the difference between gross and net wages, as a percentage of gross wages has fallen from 34 per cent at the level of the minimum wage to 21 per cent in 1996, and at the wage level for the "modal employee" (a specific Dutch definition, i.e. an employee with a wage at the lowest upper boundary for compulsory social insurance) from 48 to 41 per cent. At the lowest wage levels, there are substantial rebates on social insurance contributions (see section 6).

To see what caused the reduction in the burden of the public sector, consider the following decomposition:

In a period of 13 years, the share of public expenditures has been reduced by 13.5 per cent gross and by 11.1 per cent net. There are three main contributors: income transfers, civil servant wages, and loans.

The drop in transfers and in civil servant wages should be decomposed in a volume and a wage component. Wage restraint in the public sector has been substantial. The index for contractual wages, base 1980 = 100, stood at 155 for private sector workers in 1996, while for public sector workers it stood at 123 (contract wage increases refer to the changes in scale wages: they are net of individual wage changes related to tenure, experience and other career effects).⁽¹⁶⁾ Whereas in earlier days public sector wages had been indexed to the private sector wage index, a substantial gap was created in the first half of the 1980s. By 1983, public sector wages had dropped 2 per cent, while private sector wages had increased by 15 per cent. By 1985, private sector wages further climbed to a 21 per cent gain, while the public sector wage stabilized. After 1985, contractual wages developed almost parallel: the relative wage cut for civil servants has been accomplished within just a few years.⁽¹⁷⁾ As noted above, the share of employment in the public sector fell substantially. Note that this substantial decline in employment and relative wage has taken place without major disruptive labour conflicts. This requires accommodating labour union behaviour. As everywhere, union density is relatively high in the public sector: 46 per cent in 1985, relative to 17 per cent in the private sector (Visser, 1990, table 11, page 50). Only 1982 had a relatively high incidence of strikes, with 21/100,000 workdays lost; in other years, it is usually below ten (Statistisch Zakboek). Public sector unions are affiliated with the large federations, and share the generally restrictive policies.

The reduction in expenditures on social security (transfers) is a pure price effect. From table 4.3 below, we can calculate that between 1980 and 1995, the number of benefit years increased from 2.2 million to 3.4 million, an increase by 52 per cent. The average benefit per benefit year increased from Dfl 14,040 to Dfl 15,116, i.e. an increase by less than 8 per cent, while over the same period national income per inhabitant increased by 79 per cent.⁽¹⁸⁾

So, the conclusion on the relative reduction in government expenditures can be clear. The bulk of the cuts came from reduction in transfers and in the wage bill. For transfers, it is overwhelmingly a price effect, as the number of recipients increased substantially. For the public sector wage bill, it is both a price and a volume effect. Over the period 1983-96, public sector employment fell by 7 per cent. Over the period 1980-96, nominal public sector wages increased 30 per cent less than private sector wages. The relative wage reduction was mostly established in the early 1980s. This is illustrated in figure 3.1, copied from Salverda (1997).

3.2 Wage restraint

Nominal wage increases after 1983 have been modest, with a peak of 4 per cent in 1992; in most years, the increase was less than 2 per cent. Figure 3.1 gives the index of contract wages, in the private and in the public sector. Note that real market wages were stable between 1975 and 1979, then decreased, and stabilized again after 1981. Hence, the Wassenaar Agreement was not the beginning of wage restraint: rapidly increasing unemployment during th 1970s already had triggered moderation. As figure 3.2 shows, unit wage cost in 1996 were at the same level as they were in 1981. Over the same period, they increased by 40 per cent in Germany and by 15 per cent on average in the EU. The wage share in market sector income dropped from 93 per cent in 1987 to 82 per cent in 1997.

At the low end of the wage distribution, restraint was even stronger. The adult legal minimum wage, in real terms, has been reduced by 20 per cent in less than 20 years. The legal minimum wage is differentiated by age. In 1979, half of the minimum wage population was on the minimum youth wage, in 1994 it had dropped to a third (Salverda, 1997). The minimum youth wage has declined even more than the adult minimum; this is reflected in the graph for the weighted minimum, i.e. weighted by the age category's employment share in 1979. The share of employment yielding no more than the legal minimum wage has dropped from 7.9 per cent in 1979 to 2.6 per cent in 1994 (in full-time equivalents). In 1979, the legal adult minimum wage was 70 per cent of the median wage. The share of employment at or below that relative wage level was 12.4 per cent in 1979 and had climbed to 14.0 in 1994 (Salverda, 1997). In other words, the share of workers at the minimum wage has declined substantially. But since the minimum wage itself has dropped below its former relative level, the share of workers in the low wage tail defined by the initial relative level has increased.

3.3 Working time reduction

Actual hours worked per year have been reduced to the lowest level in Europe: 1452 hours in 1993, while the average in the EU stood at 1669, and Germans worked 1592 hours. This is mainly due to part-time work. Average annual hours for full-time workers are about equal: 1788 in the Netherlands, 1797 in the European Union. Great Britain is the exception here, with an average of 1953 annual hours. Dutch part-timers work less (817 annual hours versus 921 in the EU), and there are relatively more of them. The reduction fits in a downward trend over the postwar period. Contractual hours took their first substantial dive in 1961-62, with the introduction of the five-day working week. There clearly has been a substitution of persons for hours. Total hours worked per annum by the total labour force in 1990 were equal to the total in 1970 (Den Butter en Van Vuuren, 1997). See figure 3.3. In 1975, all full-time workers had a standard 40 hours working week. In 1998, the average working week for a full-timer is about 36 hours (Plantenga and Dur, 1998).

Table 3.2 portrays the shift in employment composition for 1987 and 1995, two cyclically comparable years (the unemployment rates in both years are virtually equal). Both for men and women, annual contractual hours have been reduced. Overtime hours have increased, as could be anticipated from the relative increase in fixed cost when contractual hours decline. Part-time and flexible contract work have increased. The share of full-time male work has diminished from 66 to 60 per cent of the annual employment volume. The total of full-time work, for men and women together, decreased from 83 to 75 per cent. The share of women increased from 29 to just under 33 per cent, with a big increase in the share of part-time female work. Part-time work increased from 13 to 19 per cent, flexible contract work increased only modestly from 4 to 5 per cent.

Part-time work covers all jobs in which the agreed hours of work are less than the full-time contractual hours of the firm. Over three quarters of the volume of part-time work is done by women. Mostly is in line with preferences. The labour force participation rate for women has increased steadily (from 40 per cent in 1987 to 49.9 per cent in 1996), and many women like to work part-time: given the social and institutional environment, there are no signs that those working part-time want to supply more hours. The relation between women's labour supply behaviour and child care facilities in the Netherlands is not well established. Day care facility provision by employers is not very frequent. Use of informal child care (friends, relatives, neighbours) is more common than use of formal facilities, both among women with and without paid employment. In fact, among women using child care facilities, the differences by types of care between women with and without paid employment are modest. The effect of the price of the child care facility on leisure demand has been estimated as zero (Maassen van den Brink, 1993; Maassen van den Brink and Groot, 1994). Perhaps, availability of facilities is relevant for labour supply, but given the high probability that at least informal facilities are available, it is hard to conclude that lack of facilities is the key factor to explain the high incidence of part-time work among women.

The reduction in contractual hours after 1982 did not stem from demands by the workers. The earlier reductions had been increases in demand for leisure induced by increases in the wage rate: an income effect dominating the substitution effect. After 1982, the stimulus came from union leadership and workers only reluctantly accepted. With constant or even declining real wages they were not interested in buying more leisure. More recently, however, surveys indicate an interest in a shorter working week, both among men and women (Plantenga and Dur, 1998).

Labour supply has also been reduced by a strong reduction in the participation of older men. They got a disability pension, benefitted from early retirement schemes or, when long-term unemployed, got excempt from the obligation to search for a job. In 1996, only 39 per cent of the men aged 55 to 65 had paid work. Among women, this was 13 per cent. (Kroniek van de Sociale Verzekeringen, 1997). The financial cost to the individual pensioner was usually very low. There were private sector supplements for workers older than 57.5 to their unemployment benefit. In 1975, unemployed over 60 became eligible for unemployment benefit until age 65 without the requirement of active job search and involuntary lay-off (which implied benefit entitlement from the age of 57.5 until the age of national pension entitlement 65). Optional early retirement schemes started to emerge after 1976, organized at the level of firms and industries (Teulings, Van der Veen and Trommel, 1997, page 31).

3.4 Effectiveness

Government policy consisted of reducing expenditures by reducing the public sector wage bill, reducing transfers and lowering the legal minimum wage. The initial impact of the policy is negative: it directly reduces public sector employment and it reduces consumption demand from civil servants and transfer recipients. The benefits should come from wage cost reduction, stimulated by a lower tax burden, a lower legal minimum wage and lower benefit levels (lowering reservation wages). Wage cost reduction should be reinforced by deliberate wage restraint by the labour unions. At given output, lower wage cost stimulates more labour intensive, less capital intensive production. The lower wage will reduce consumption demand from employees, but higher profits may stimulate investment demand and lower export prices may stimulate foreign demand. The lower minimum wage may lead to a more extended lower tail in the wage distribution, stimulating growth of low-wage jobs. The lower wage rate may reduce labour productivity, from reduction of the capital intensity of production and from less pressure to seek and implement process innovations. Reduced consumer demand may reduce the stimulus to product innovations. While the policy may benefit employment, it might come with a reduced labour productivity and with demand composition effects from consumption towards investment and exports.

To assess the balance of potential positive and negative effects on employment, the Central Planning Bureau has compared the actual situation in 1990 with the predicted outcome under alternative policies (CPB, 1991). The econometric model has a wage equation in which wage growth responds to prices, unemployment, the burden of taxes and social security premiums and the net replacement rate. Investment responds to the share of wages in national income. Before the policy interventions, public sector wages, social security benefits and the legal minimum wage were indexed to the private sector wage. Had this indexation continued (at budget deficit as actually realized, hence at higher taxes), private sector employment in years would have been lower by 125,000 (by 150,000 in persons) and unemployment would have been higher by 105,000 persons. Between 1979 and 1990, the share of labour income in private sector value added decreased by 10 percentage points. In the CPB calculations, three quarters of this decrease was due to the endogenous effect of higher unemployment. Had wages increased to maintain the private sector labour share at the level of 1979, private sector employment in years would have been lower by 220,000 and unemployment in persons would have been higher by 275,000 (actual unemployment in 1990 was 419,000). In both simulations, consumption would have been higher and investment and exports would have been lower. Had both indexation and constant labour share wage development occurred, predicted private sector employment in 1990 would have been 8 per cent lower than actually observed. Private sector labour productivity would have been higher. Had indexation continued, real national income would have been lower. Had wage development maintained the labour share in income, real national income would have been higher. Between 1982 and 1990, relevant world trade annual growth was 1 to 1.5 per cent higher than in the previous eight year period. According to simulations that suppress this higher growth rate, the effect on employment and unemployment was very modest, although of course exports would have grown much faster.

How effective has work sharing been for creating additional employment? A simplistic, strict proportionality between contractual working hours and unemployment is unrealistic for many reasons, and in fact not many people adhere to such a naive view. But many believe the positive relation sufficiently strong to have policy relevance. Calmfors (1985) is probably the first to analyse the potential effects of work sharing. Work sharing will tend to increase the cost of labour, which by itself will reduce demand for labour. With fixed cost relatively more important, labour demand will shift in favour of hours per person over persons, hence stimulating working overtime. This has indeed been observed, as noted above (see table 3.2). In bargaining models, the monopoly union model and the right-to-manage model yield ambiguous predictions on the effect on wages (and hence employment). Efficient bargaining models predict a decrease in annual earnings and an increase in the hourly wage. Layard, Nickell and Jackman (1991) argue that working hours have no effect on the unemployment rate. They cite an OECD wage equation for 19 countries (and 30 years) in which the effect of average weekly hours on wages is insignificant. With wage pressure unaffected, equilibrium unemployment is also unaffected in their model. They offer further evidence ("casually", as they call it) by plotting the decrease in working hours against the increase in unemployment in 11 countries for the period 1975-88. The highest increase in unemployment appears to have occurred in countries with the strongest reduction in working hours. However, De Regt (1997) points out that 1975 and 1988 mark different stages in the business cycle. Drawing up a figure for cyclically comparable years 1983 and 1993 exactly reverses the relation! Plotting changes over the years 1979-89, both more or less cyclical peaks, restores the relation as found by LNJ.

Dur (1997) reports an estimated wage equation for The Netherlands for the period 1966-93, and finds that annual wage cost decreases by 0.6 per cent if contractual hours are reduced by 1 per cent, implying an increase in the hourly wage of 0.4 per cent. De Regt (1997) refers to nonreported estimation results in which hourly wages are constant if working hours are reduced, a result similar to the LNJ finding above.

Dur (1996) estimates a four-equation model: employment, unemployment share of long-term unemployed, real hourly wages and labour supply, for the period 1969-94. Employment in hours (annual employment in labour years multiplied by contractual annual hours) is not significantly affected by contractual hours (coefficient 0.14, t-value 0.56). This suggests proportionality for work sharing, at given hourly wage cost and at given total employment, i.e. equivalence of persons and hours. Work sharing thus has the potential of reducing unemployment. But real hourly wages (including employer premiums for social security) respond significantly to contractual hours, with an elasticity of -0.6: a reduction of contractual hours by 1 per cent increases hourly wage cost by 0.6 per cent. Since wages affect employment with an elasticity of -0.5, there is a negative effect of work sharing on employment: through cost increases. Dur claims his results for the wage equation to be similar to results for Germany, Denmark, Finland and Australia. Labour supply responds significantly to wages and unemployment, positive to the former, negative tot the latter. Wage effects and labour supply effects take away much of the proportionality effect of work sharing: less than 20 per cent of the proportional effect on unemployment remains. This is due to a primary effect of wages on employment, cutting directly into the proportionality, and to secondary effects, as reduced unemployment increases wages and labour supply. Work sharing indeed reduces unemployment, but at a high cost in terms of employment and output. In his decomposition of the unemployment history between 1970 and 1994, unemployment was strongly pushed upward by population growth. Reduction in contractual hours had a substantial mitigating effect between 1970 and 1989. Negotiated reduction in working time in collective agreements mostly took place in two periods: by 8 per cent in 1970-75 and by 6 per cent in 1980-85. During the nineties, the effect has been negligible (Plantenga and Dur, 1998).

Kapteyn, Zaidi and Kalwij (1997) draw on the international literature and new econometric evidence for a sceptical (or perhaps agnostic) view on the effects of worksharing. Among the empirical regularities that can cautiously be deduced from the modest existing evidence are the following. Actual hours worked per employee tend to follow standard hours worked rather closely, with a coefficient (or elastity) just below 1. The partial elasticity of employment with respect to hours, holding wages fixed is about -0.5 to -0.6. Perhaps the effect of hours worked on wages is most disputed. Some claim no effect (Layard et al., see above). Other studies suggest that a reduction in hours worked increases the hourly wage; this would hold for Germany, Sweden and Norway, and according to Dur's results, for the Netherlands. Kapteyn et al. present estimation results for 13 OECD countries for the period 1971-94. It is a highly aggregate analysis, estimating relations between employment in persons, gross hourly wages (national averages), average actual weekly hours, the unemployment rate, GDP and some other variables. The results are mostly dominated by absence of significant effects. In particular, working hours have no effect on employment, unemployment, wages, labour supply or GDP per capita. This suggests that work sharing is both useless and harmless: no effect on the target variables (employment, unemployment), no cost effect (GDP per capita). The authors also offer a simulation analysis of policies for the Netherlands (country specific effects emerge from country dummies in the relations), indicating that reduced working hours reduce employment and increase unemployment. But they rightly point out that given the low significance levels of the estimates, these predictions are very uncertain.

Clearly, there is no strong empirical support for a policy of worksharing as a method to reduce unemployment. CPB (1991) also concludes to a very modest effect. It may have had some effect in the Netherlands, but at a cost to the volume of employment. An important issue, not studied in the econometric analyses, is the decision on the timing of working hours. It will matter a great deal whether standard weekly hours are rigidly reduced, or only reduced on average, and it will also matter whether workers can take time off at will or whether employers can decide when the worker should take additional days off. A survey covering the period 1982-85 indicates that in half of the firms reduced working hours per employee was met by reduced operating hours of the firm, while in the other half, operating hours were maintained (CPB, 1991, page 10). Several recent collective agreements attempt to combine reduced standard hours with increased flexibility, with a say for employers on the timing of work and leisure. In banking, reducing the standard week to 36 hours (in 1995) is combined with extension of opening hours to the public. A large retail firm moved to a 35 hours working week with a 60 hours operational week, also extending opening hours (restrictions on opening hours of retail stores have recently been lifted). A large chemical company introduced an average working week of 36 hours, with the company deciding over the timing of 50 per cent of 23 days of annual leave and the possibility for the worker to sell some of these days for working (Plantenga and Dur, 1998). If additional leisure can be taken during slack times, the effect on wage cost is reduced, but so is the effect on labour demand. Indeed, authority over the spacing of working time seems quite important for the effects.

Table 3.1 Public expenditure categories, share in GDP

Table 3.2 Shifts in employment composition by type of contract

Figure 3.1 Real contract wages, market sector and public sector, 1970-1996 (1979=100)

Figure 3.2 Real wages in the Netherlands, 1979-1997 (1979=100)

Figure 3.3A Annual contractual hours 1950-1995

Figure 3.3B Total employment volume 1950-1990

Figure 3.3C Hours worked per head of total population

4. Social security reform

4.1 The reforms

The period of construction and expansion of the welfare state lasted until the late 1960s. By the end of that decade, the legal structure of insurances and provisions was more or less complete. Later periods added streamlining and refinement. In the period from the late 1960s until the mid-1980s, the period for which the term "Dutch Disease" was coined, the welfare state provided a comfortable safety net, with generous benefit levels for sickness (100 per cent or more!), unemployment and disability, and rather easy access. But halfway during the 1980s reforms got under way, with 1987 marking a watershed.⁽¹⁹⁾ Let us consider this history in some detail.

The system of social security consists of social insurance and social provision. Social insurance is financed by premiums paid by the insured, and is divided into employee insurance, covering employees only (unemployment, sickness, medical expenses, disability) and national insurance, covering the entire population (national pension, widowers, child allowance and some other insurances). Social provisions, financed from general tax revenue, aim at guaranteeing the social minimum income. Social provisions also include the Income Supplement Act, which guarantees minimum income levels for recipients of benefits from social insurance. The big adjustments took place in the employee insurances.

Until the reforms in 1995, the organization of social security was distinctly corporatist. Eligibility, benefit level and benefit duration, i.e. the terms of the contract, are specified by law. But implementation, administration, control, was in the hands of social partners through their position in Industry Associations. Industry Associations were created by law in 1953 (Law on Organization of Social Security) precisely for this purpose. Initially there were 26 such Associations, governed by the organizations of employers and employees (as a result of mergers only 18 were left in 1995). The social partners had a strong grip on operating the insurances for unemployment, sickness and disability. For example, in the disability insurance they jointly decided on disputed benefit claims in a Disputed Claims Commission. Up until 1994, social partners even controlled supervision of the Associations, because they had a joint majority in the supervisory Social Insurance Council. The Social Insurance Council consisted of union and employers' representatives and independent members appointed by the Minister of Social Affairs (eight members for each category), under a chairman appointed by the Crown (i.e. the Queen). The same applies to national insurance (old age and widow(er) pensions), where the Social Insurance Bank with tripartite governance administers the insurance and sets the premiums.

In all cases, social partners had a strong hand in setting premiums. The premiums for the Sickness Insurance were set by the Industry Associations, and hence, differed by Association. The premium for Disability Insurance was set by the Social Insurance Council, where partners had a (two-thirds) majority. The premium did not fluctuate across Associations, but was uniform across the labour force. Unemployment Insurance has a mixed structure. Part of the premium is uniform across the labour force, set by the Unemployment Fund, with tripartite governance. Another part (on average, roughly half the premium) was set by the Industry Association, and hence, industry specific. The industry specific part relates to the first stages of unemployment, i.e. the first eight weeks of unemployment (later extended to 26 weeks). Considered from the perspective of the decision makers at the industry level, there was no scope for shifting the burden of generous policies in the Sickness Insurance to outsiders (premiums were specific for the Industry Association).⁽²⁰⁾ There was however full scope in case of Disability Insurance (because of the undifferentiated, national, premium) and some scope in case of Unemployment Insurance.

During the 1980s, when unemployment had risen dramatically and a large share of the population was covered by social security, attempts to reduce expenditures have been launched along three routes: along the employee side, the employer side and at the level of the controlling and supervising organizations. By making benefit regimes less attractive and less accessible for employees and by improving cost-reducing incentives for employers, the price and volume of benefit recipiency should be reduced, to reduce premiums and wage cost. Also, policies to increase the outflow from unemployment and disability were intensified (for example, allowing schooling while on unemployment benefit and subsidies to hire disabled). Changing the management organization of social insurances should generate a self-governing system more conducive to sober and efficient utilization.

The reform on the employee side meant first of all a reduction of benefit ratios: in the period 1985-87, the basic replacement ratio for unemployment, sickness and disability was lowered from 80 to 70 per cent. But for each of these insurances, eligibility rules were also tightened.

Eligibility for unemployment benefits required 13 weeks of work out of the last 52 in the period 1949 to 1987. In 1987, this was increased to 26 weeks of work out of the last 52, in 1995 it was raised to 26 out of the last 39 weeks, and in addition four years of work (for a minimum of 52 days) out of the last five years. Benefits stood at 80 per cent, for a maximum of six months in the period 1965-85, then were reduced to 70 per cent, and in 1987 were "kinked": 70 per cent of last wages for a maximum of six months (with extension up to four and a half years as a function of work experience) and a so called extended benefit of 70 per cent of the legal minimum wage for a maximum of one year. The unemployed not covered by unemployment insurance (for example, for lack of sufficient work experience, were covered by a national benefit plan financed from tax revenue (with benefits at the level of the social minimum); this was abolished in 1996. After exhausting the benefits, the unemployed are entitled to General Welfare Assistance, means tested, with partners' incomes pooled, and limited to the social minimum.

As a crude test on the effectiveness of tighter eligibility rules, we may relate the number of unemployment beneficiaries to the number of "real" unemployed. The real unemployed are those who qualify for the ILO definition: without a job, actively searching and immediately available for work. We may also look at the average unemployment benefit per "true" unemployed (see table 4.1). With tighter eligibility rules, one would expect the number of beneficiaries to be closer to the number of "real" unemployed and hence a lower ratio.

The first and perhaps surprising result is the fact that we have 25 to 40 per cent more unemployment beneficiaries than unemployed. This is a consequence of the fact that several categories of unemployment beneficiaries are exempt from the obligation to seek a job, such as elderly unemployed. The pick-up rate has certainly not declined structurally over time, although there is a slight dip after 1987, when eligibility rules were tightened up. The total amount of benefits paid, related to the "real" unemployed, has increased by 11 per cent between 1983 and 1995. Over the same period national income per capita increased by 56 per cent. Thus, the benefit level has indeed been lagging the general income development. This reflects the decline in benefit levels and may also reflect some composition effect towards legal regimes with lower benefit level.

The incidence of sickness covered under Sickness Insurance is shown in figure 4.1. It has increased strongly from the early fifties to the mid-seventies. Until 1985, the national Sickness Insurance paid 80 per cent of the wage during the sickness, starting at day three of a sickness interval, for a maximum of one year (after which transition to the Disability programme was considered). Premiums were differentiated by industry (for 18 Industry Associations). Through collective agreements, the benefits were raised to 100 per cent, from the first day of sickness, at the expense of the employer. In 1985, the benefit level was lowered to 75 per cent, in 1986 to 70 per cent, but again collective agreements safeguarded the workers by simply maintaining the employer supplements (Teulings, Van der Veen and Trommel, 1997, page 287). There was thus no financial loss for sick employees and until 1985 often even gains in net income, as sickness benefits were exempt from social insurance premiums. Moral hazard will lead to excess "consumption" (high levels of sickness absenteeism) which can only be countered by intense monitoring. It is quite likely that the strong increase in sickness incidence was due to the combination of full income coverage and a very lenient system of monitoring. In the reform of 1994, employers were mandated to pay at least 70 per cent of the wages for the first six weeks of sickness (two weeks for small employers), with of course the option for employers to seek private insurance. In 1996, this was extended to full employer liability for 70 per cent of the wages for one year: essentially, the Sickness Insurance programme for employees has been privatized.

To evaluate the effect of the reforms, there is the disrupted data series of CTSV and there is a new dataseries initiated by CBS, the Sickness Absence Survey. The CTSV series is disrupted when employers became liable for sickness compensation and no longer needed to report to the Industry Association to make sure the employee got the benefit.⁽²¹⁾ Absence percentages from the latter survey, for the private sector, excluding maternity leave are given in table 4.2. Between 1993 and 1995, the sickness absence rate fell markedly. Since then, it appears to have stabilized. Just at the face of it, without conditioning on external factors (cold winter? influenza epidemic?), the extension of employer liability from six weeks to one year does not seem to have had much of an impact, while the first step (liable for the first six weeks) is associated with a substantial drop in sickness rates. However, the sickness absence rate in the public sector also fell: a drop from 6.3 per cent in 1993 to 5.6 per cent in 1994. In 1995, it increased again to 5.8 per cent (Sociaal Economische Maandstatistiek, Mei 1996, page 14). If we use the public sector sickness rate as a crude index for external conditions, the indexed private sector rate would have been 5.5 per cent in 1994 and 6.3 per cent in 1995. That would imply a drop of sickness incidence by 11 per cent in 1994 and another drop by 11 per cent in 1995, giving a total reduction due to privatization of 22 per cent.

Privatization of sickness leave insurance is not a sudden policy disrupture. Sickness absenteeism policies had already been on the agenda for collective bargaining before privatization was enacted.⁽²²⁾ Some of the reduction of sickness incidence might be due to the increase in monitoring that took place in recent years, often on the basis of the bargaining results in collective agreements. Recently, sickness incidence has been an important topic on the bargaining agenda (Van den Toren, 1996, Chapter 5) and employers and employees have jointly adopted measures to reduce it. It has resulted in more attention for the quality of working conditions, but also in more frequent and earlier checks by physicians. As the data in figure 4.1 indicate, sickness incidence had already fallen from its peak in the mid-1970s. Sickness leave compensation as a percentage of the wage bill increased from 2 per cent in the early 1950s to almost 7 per cent in the late 1970s. It declined towards 5 per cent in the late 1980s and then rose again, before the privatization dip took place (Kroniek van de sociale verzekeringen, 1997,Grafiek 4.2.b, page 67). It has always been possible to opt out of the collective insurance system at the industry level and to either as a firm fully absorb the risk or to set up an insurance under joint governance of employer and employees. In terms of total benefits paid, the latter two groups initially were about about half the size of the collective insurance system, but the share gradually dropped to about a quarter (before returning to one half when the privatization reforms started). The developments of sickness incidence (measured by benefits paid relative to the wage bill) appear very similar for the different systems (Kroniek, 1997, Grafiek 4.1.b, page 65). This might suggest that the type of organization of the insurance is rather immaterial as long as the insurance contract (level of benefits paid, intensity of monitoring etc.) is not affected. However, a more inquisitive comparison might reveal otherwise.

The privatization of sickness leave compensation (and the attention already paid to the high absence rates before the privatization took hold) does appear to have changed the policies of employers and insurance associations. There is more intensive monitoring and there are more active prevention policies. In 1996, under full privatization, five out of ten firms paid less than 100 per cent as sickness benefit, and four out of ten paid 70 per cent. Before privatization, compensation at 100 per cent was the rule. There are also indications that firms are more selective in hiring and that they evaluate absenteeism indicators in their hiring decision (CTSV, Augustus rapportage arbeidsongeschiktheids verzekeringen 1997, pages 53-60).

In conclusion, the sickness incidence rate, after a long climb to a high plateau maintained between the mid-seventies and the mid-nineties, has come down in reaction to privatization and a set of policies leading to lower financial compensation, more intense monitoring and more active prevention policies. The down side however, is more selective hiring decisions by firms.

The most worrisome development in the Netherlands, and a painful claim to international fame, has been the explosion in the number of disability benefit recipients. This has been caused by a combination of a generous benefit scheme, generous and fuzzy eligibility criteria and a governance structure that was not conducive to restraint. The Disability Insurance was enacted in 1967. Individuals were placed in disability intervals, with a disability of 80-100 per cent giving entitlement to the full benefit of 80 per cent of lost earnings. Disability was determined with reference to the wage that could be earned for the incapacitated's education and experience, taking into account the situation in the labour market. This condition implies that disability became related to the level of unemployment: if the state of the labour market would make it hard to find a job for the remaining capacity to work, the disabled would be placed in a higher disability interval. The inflow into the disability was massive and dramatic, and clearly responsive to economic incentives (Aarts and De Jong, 1992).

In 1987, the benefit level was lowered to 70 per cent However, it stimulated the provision of supplementary benefits through collective agreements, from applying to 68 per cent of employees in 1984 to 90 per cent in 1989 (Teulings, Van der Veen and Willems, 1997, page 286). Also, the "labour market criterion" was abolished. In 1993, eligibility conditions were tightened. The disability had to be a direct consequence of the impairment, by objective standards, and the remaining earnings capacity came to be estimated from a selection of the best paying jobs that would be open to the individual. Also in 1993, the principle of constant benefit levels until age 65 was abandoned. The benefit level came to be related to age and work experience. Recipients younger than 50 were subjected to new medical examinations, in successive cohorts. Half of the new examinations for "cohort 94" (recipients younger than 35) led to lower disability levels and hence to lower benefit levels; in 37 per cent of the cases, the benefit was even ended altogether. For "cohort 1995" (aged between 35 and 40), 35 per cent got a lower estimate (of which 18 per cent saw their benefit ended)(CTSV, In en uit de WAO, pages 50-51).

Until 1993, disability benefits were constant until retirement (allowing for purchasing power indexing). This made disability a far more attractive exit route than unemployment, where benefits are reduced to the social minimum after some time. An element of generosity, no doubt, was the permission to take the labour market situation into account when determining the benefit level. Fuzziness surrounded the very concept of disability. This is to some extent always inevitable, but in combination with the lenient control and supervision culture, it opened the gates widely and let many pass through (Aarts en De Jong, 1992).

The percentage of workers on disability increased to a high of 13.5 per cent in 1984 and then slowly declined to 9.7 per cent in 1996. In equivalence years, correcting for the degree of disability, there was a peak of 568,000 years (for employees) in 1992 and 1993, and then a decline to 509,000 years in 1996. Figure 4.2 depicts the annual inflow and outflow. For the first time since the introduction of the law, the outflow surpassed the inflow in 1994. However, in 1996, the tide turned again. For employees, the inflow rate in 1996 was 1.4 per cent, down from a high of 2.4 per cent in 1978; the outflow rate was 9.6 per cent, up from a low of 8.2 per cent in 1989-90.

It has extensively and convincingly been argued that the disabled population contains a large component of unemployment, and this is clearly reflected in its time path: it parallels the development of the unemployment rate quite closely. During the 1990s the inflow rate seems to have responded to the policy measures with a structural decrease. The outflow rate has clearly gone up, from 8.2 per cent in 1990 to 12.2 per cent in 1994. The reform of 1987, in particular the elimination of the labour market situation from the disability criterion, appears to have no effect. Econometric analysis of inflow into disability before 1987 showed a considerable impact of the situation on the labour market, one study using post-reform data shows a much smaller impact, suggesting the change may have had an impact (Stigter, 1997, page 15). However, Teulings, Van der Veen and Willems (1997, page 80) report that after 1987, the labour market situation remained influential for examiners' decisions, in particular for older men. And the time series in figure 4.2 does not show signs of a marked effect. What did change, however, is the share of partial disability: the gradual increase of this share after 1977 steepened after 1986 and after 1990.

Among the three lines of attack on the expanded welfare state distinguished above, the second is targeted on employers. The changes have already been mentioned above. In 1994, employers became directly liable for wages of sick employees in the first six weeks (or first two weeks for small employers). In 1996 this six week period was extended to a full year. The employer is held responsable for a policy to reduce sickness absenteeism (which includes the obligation to consult expert advice) and is under obligation to monitor the worker's inability to work. Between 1992 and 1994, the disability insurance had a bonus/malus system: a penalty for every employee of the firm that becomes disabled, a premium for hiring a disabled person. Since January 1998, there is experience rating in the firm's premium for disability insurance. There is an obligation for firms to consult a Working Conditions service ("Arbo dienst") to improve working conditions and prevent the incidence of disease and disability.

The third line of attack is focussed on the organizational structure. Unemployment, sickness and disability insurance used to be administered by the Industry Associations, supervised by the Social Insurance Council, and social partners dominated administration and operative control. In 1995 the Council was discontinued and supervision was handed over to the Commission for Supervizing Social Insurances, CTSV. This is an independent body, without participation from social partners. CTSV has three independent members (and a staff of some 200 employees). It oversees the Industry Associations and initially the temporary committee for coordinating the Associations' policies, TICA. In the same year, the Joint Medical Service GMD was abolished. It used to do the medical inspections for the disability claims. In 1997, both TICA and all the Industry Associations are abolished. Administration of employee insurances, disability insurance and the Income Supplement Act is now the responsibility of the National Institute for Social Insurances, LISV, which contracts out to five Executive Institutes, UVIs, organized by (clusters of) industries. LISV is supervised by CTSV. Thus, social partners have lost their operational control over social insurances.

With respect to disability and unemployment, the new coalition government that resulted from the May 1998 elections intends to change the organization. The discussion centers on the extent of privatization. Social partners (and independent members) in the Social Economic Council have presented a unanimous advice to let private organizations evaluate claims for disability compensation, provided that unions and employer jointly select the organization. The government intends to maintain claim evaluation as a public sector activity, while agreeing on private sector administration of the disability payments. Social partners were shocked⁽²³⁾ by the planned refusal to accept the Social Economic Council advice.⁽²⁴⁾ The discussion between government and social partners is still going on.

There is a widespread view that social partners were responsible for the dramatic expansion of the population on disability benefits, a view underscored by an extensive Parliamentary Investigation. Social partners have been punished for their mismanagement by the drastic administrative reform. In the old system, until 1995, social partners administered disability and part of the unemployment benefits through their role in the Industry Associations. In the supervisory body, they had a joint majority over the independent members. After 1997, when the Industry Associations were abolished, they are no longer involved in administration or supervision. This is an important move away from the former intentionally corporatist organization set up in the early 1950s.⁽²⁵⁾

4.2 Effectiveness of the reforms

So, what is the upshot of all these reforms?

Let us take a gross picture of what happened over a decade and a half of discussions and policy changes. We compare 1980, 1986 and 1995. 1986 is the last year before serious reforms in social security got under way. We consider four categories: sickness benefits, disability benefits and unemployment benefits, with the latter grouped in two subcategories: the standard unemployment benefits from entitlement under the WW, the unemployment benefit act, which covers the first six months of unemployment, and other unemployment. Other unemployment covers all these special provisions, under headings as unemployment provision, general assistance, etc..

Even admitting the crudity of the picture generated by table 4.3, it certainly does not strike as a splendid succes story. Indeed, the size of the labour force increased, from 5.344 million in 1980 to 5.955 million in 1986 and to 6.596 million in 1996. The incidence of sickness is at the same level of 1980 and hence, relative to the labour force has clearly diminished.⁽²⁶⁾ The other categories have increased substantially, in particular unemployment. The cost per recipient year has increased for sickness and disability, and decreased for the unemployed. Cost and incidence of sickness have developed favourably, disability and unemployment recipiency have increased strongly.

Figure 4.3 pictures the development of the relative social minimum over 25 years: from two-thirds of the average wage in the early seventies to over 80 per cent between 1975 and 1983, and then gradually back to two-thirds again. The social minimum is the guaranteed minimum income, either as minimum wage, or minimum benefit in combination with Income Supplements. The increase in the relative benefit level has made benefit status more attractive relative to work. This can be illustrated by a study of Wolfe et al. (1984). They estimate a labour supply equation (weekly hours worked) on predicted earnings, predicted benefits and a vector of demographic characteristics in 1974 and in 1980. They find substantial effects of increased generosity of the benefits between the two years. For the earnings elasticity they find (for the Netherlands) 0.77 in 1974 and 1.35 in 1980, for the benefit elasticity they find -1.51 and -.82. With the transfer elastic