Professional Documents
Culture Documents
TORENVLIED AKKERMAN. Soft' Policy Implementation in
TORENVLIED AKKERMAN. Soft' Policy Implementation in
net/publication/233700806
CITATIONS READS
57 523
2 authors:
Some of the authors of this publication are also working on these related projects:
Linking the discontented employee and the discontented citizen View project
The role of external experts in crisis situations: a research synthesis of 114 post-crisis evaluation reports View project
All content following this page was uploaded by A. Akkerman on 23 October 2017.
Introduction
Processes of policy implementation are highly relevant to political scientists.1
The implementation of public policies determines whether elected officials
deliver the policies they promised to their citizens. Policy implementation is the
domain of the executive branch of government but has an important impact on
the political survival of elected officials. If the executive drifts away from the
democratic decisions of elected officials, this could generate political
instability, and ultimately harm democracy.
René Torenvlied and Agnes Akkerman
‘Soft’ Policy Implementation in Multilevel Systems in the Netherlands
32
case. The final section discusses the results and puts them in a broader
perspective.
assumptions follow from the spatial theory of voting, as well as the state-
centric notion of policy-making. First, unitary actors rationally strive to realize
the substantive policy alternative they prefer most: their policy position. The
extent to which the realization of a policy position contributes to the actors’
organizational goals translates into the salience they attach to different policy
issues. Actors are fully informed about policy issues, alternatives, positions,
salience, and outcomes of collective decision-making. Second, decision-makers
and implementation agencies have single-peaked preference-loss functions,
defined over the policy scale, monotonically increasing with increasing distance
from the policy position. Additionally, implementation agencies have single
peaked reputation-loss functions monotonically increasing with increasing
distance from the outcome of collective decision-making: the political decision.
Third, decision-makers have the following behavioural option: they are
involved in influencing, apply coercion or political exchange in order to
change the (voting) position of other decision makers more closely towards
their own position. Their issue salience determines the level of effort a
negotiator will put in this process. The explanatory variable is (in)formal
bargaining power. Subsequently, decision-makers vote in order to arrive at a
political decision, explained by the formal voting power.3 Theoretically,
lobbying activities of implementation agencies could result in a political
decision that reflects the opinions of implementation agencies. However,
implementation models do not focus on these lobbying processes. Fourth,
implementation agencies have the behavioural option to comply with or
deviate from the political decision. If the policy position of an implementation
agency differs from the political decision, the agency has an incentive to deviate
from the outcome. The agency applies its discretion to realize a policy
performance, which is as close to their policy position as possible — given their
restrictions. Together with the incentive for deviation, agency discretion
determines the amount of deviation.
The third range of assumptions concerns the structure of the implementation
process, and the interaction mechanisms between decision-makers and
implementation agencies. These assumptions all follow from the state-centric
notion of the policy-making process. First, all models, implicitly or explicitly,
assume the existence of an implementation structure, which defines the
principal–agent relation between decision-makers and agencies. Second, the
outcome of collective decision-making is binding for all actors involved in
implementation. This restriction is absent in collective decision-making. Since
all actors are assumed to have full information about the outcome of collective
decision-making, any policy deviation is a voluntary defection.
Many of the above assumptions reflect the traditional ‘state-centric’ policy
making system. Below, we describe the features of yet another type of policy-
making system — one that has increasingly received attention over the last
Acta Politica 2004 39
René Torenvlied and Agnes Akkerman
‘Soft’ Policy Implementation in Multilevel Systems in the Netherlands
36
specific policy). The second type is most different from the traditional ‘state-
centric’ system of policy-making, because in this type there is no hierarchical
relationship between the levels of collective decision-making. For example,
memberships of decision-makers at different levels are sometimes overlapping,
(representatives of) organizations meet again at the different levels. This type
of multilevel policy-making system concerns us here in the present paper.
The second feature of multilevel governance is the relative autonomy of the
different levels of collective decision-making. Collective decision-making at one
level does not necessarily restrict collective decision-making at another level.
For example, different administrative bodies in regional areas often operate
autonomously from each other and from the traditional counties or (national)
states. This feature clearly distinguishes multilevel policy-making from the
more ‘traditional’ state-centric system with its top-down framework of
regulatory decision-making.
A third feature, increasingly observed in multilevel policy-making systems of
the second type, is the presence of ‘soft’ policy instruments: non-binding policy
recommendations, guidelines, informational devices, or voluntary agreements.
Despite the autonomy of the different levels of decision-making, non-binding
soft policies are directed from one level towards another level in the system.
The emergence of soft policy instruments could be explained from increasing
needs for coordination — for example when ‘policies of one jurisdiction have
spillovers (y) for other jurisdictions’ (Hooghe and Marks, 2003, 239). In
multilevel policy-making, the costs of coordination are high because of the
large number of actors involved and interactions between them (Scharpf,
1997). For this reason, soft instruments may be an efficient means to
coordinate policies across different levels.
The literature on binding decisions is overwhelming compared with literature
on soft policies. Nevertheless, recently an increasing number of examples are
reported of ‘soft’ policy instruments to coordinate policies across different
autonomous levels of decision-making. A major example is European Union
policy-making. In the European Union, we find at one extreme treaty-based,
legally binding agreements that carry the threat of sanction, such as directives
and regulations. The effect of using these hard policy instruments in European
policy-making is to constrain effectively the choices available to policy makers
at the implementation phase in member states and at the regional level. At the
other end of the continuum, we increasingly observe the use of ‘soft’
instruments, associated with the ‘open method of coordination’ in the
European Union (European Commission, 2000a; Ahonen, 2001). This method
refers to a variety of instruments and procedures that cannot be enforced in a
legal sense and do not carry the threat of sanction. We find these instruments in
a variety of EU policy domains: the European Employment Strategy
formulates ‘Broad Economic Policy Guidelines’, EU environmental policy
Acta Politica 2004 39
René Torenvlied and Agnes Akkerman
‘Soft’ Policy Implementation in Multilevel Systems in the Netherlands
38
Logic of ratification
Domestic ratification of international treaties is an example of multilevel
policy-making. It displays two important features of such systems. First,
ratification of a treaty is not a performance by an agency, but a collective
decision made at another level in a two-level system with nested cascades of
collective decision-making. Second, the levels of diplomacy and domestic
ratification are autonomous because domestic constituencies need not
necessarily ratify an international treaty, and a non-ratified treaty is non-
binding.
The central actor in the theory of ‘two-level games’ is the level-I negotiator,
who bargains with other level-I negotiators. The agreement at level-I (as well as
its feasibility and attainability) depends upon the relative bargaining strength,
the distribution of policy positions of the level-I negotiators, as well as the
institutional voting rules.4 The relative bargaining strength and distribution of
policy positions of level-I negotiators furthermore depends upon the policy
positions of the level-II constituencies of the negotiators, who must ratify the
agreement at level-I.
The theory of ‘two-level games’ explains cross-level policy coherence
through the mechanism of strategic anticipation to the ‘logic of ratification’.
Negotiators play at two tables: the level-I table, and the level-II table.
Horizontal decision-making processes determine the level-I game with other
negotiators. Vertical coordination processes determine the level-II game with
the constituency. Dependent upon the interaction between these policies,
agreements succeed or fail. Some agreements fail because they will never
survive the ratification stage. This happens if the win-set of policy positions of
different constituencies is empty. Consequently, the theory of ‘two-level games’
predicts full policy coherence between a level-I agreement and the agenda of
level-II collective decision-making. The reason is that level-I negotiations take
place in the shadow of subsequent ratification. Negotiators strategically
anticipate on the chances of ratification and a projected rejection must be
credible in order to put pressure on the negotiations. In the theory of ‘two-level
games’, there simply would be no agreement to sign if this agreement is likely to
be rejected later. This leads to the ratification hypothesis:
Ratification hypothesis: There is full cross-level policy coherence between the
level-I agreements and the level-II agenda and the outcomes of level-II
collective decision-making.
negotiators and their level-II representatives to base this agenda upon strong
collective support at level-I. The reason is that agreements with weak collective
support would require difficult renegotiations at high costs at level-II. Weakly
supported agreements would be put on the agenda both by their proponents
(who strive for coherent implementation) and adversaries (who seek to
renegotiate the agreement). As the relative bargaining power of representatives
varies on level-II, it is uncertain to what extent the weakly supported
agreement will be reflected in all level-II outcomes. In other words, cross-level
policy coherence in the agenda is not guaranteed for agreements with weak
collective support. For level-I agreements with strong collective support, the
prospects for cross-level policy coherence in the agenda are far better than for
agreements with weak support. Unlike the case of the weakly supported
agreements, it is rational for only the representatives of supportive negotiators
to put the agreement on the level-II agenda. We now formulate the following
two hypotheses:
Agenda coordination hypotheses: (1) Representatives of the level-I negotiator
who initiated an agreement will put the agreements on the level-II agenda; (2)
Cross-level policy coherence in the level-II agenda is positively associated with
collective support for the level-I recommendation.
It makes little sense to put further pressure on level-II and coordinate the
process and outcomes of collective decision-making at this level, as, in the
context of multilevel policy-making, level-II decision-making is an autonomous
process. Consequently, we do not expect level-I negotiators to affect the
outcomes of level-II collective decision-making. In absence of ‘vertical’
coordination pressures, outcomes at level-II can only be explained by
characteristics of the ‘horizontal’ collective decision-making process. We can
apply existing collective decision-making models (Thomson et al., 2003) to
explain the level-II outcomes (and consequently cross-level policy coherence in
outcomes) from distributions of policy positions, issue salience, and relative
bargaining power. We formulate the following hypotheses:
Outcome coordination hypotheses: (1) Cross-level policy coherence in the
level-II outcomes is not associated with collective support for the level-I
recommendation. (2) Cross-level policy coherence in the level-II outcomes is
explained by a weighted5 distribution of policy positions of the level-II
representatives involved in collective decision-making.
Level I
Labour Foundation:
Negotiations
Level II
Agenda setting
Level II
Collective bargaining
Data collection
In 2000, we retrospectively collected data about the negotiation process in the
Labour Foundation in 1997, and about the collective bargaining in the metal
industry in 1998. We interviewed the most important negotiators in the Labour
Foundation (one informant from employer peak organization VNO-NCW,
and the other from the largest trade union confederation FNV) and in the
metal industry (one informant from the organization of employers in the Dutch
metal industry FME-CWM, and the other from the trade union FNV-
Bondgenoten). The four negotiators provided information about themselves
and all other negotiators operating at their bargaining level. The data we
obtained from these highly structured interviews were supplemented by a
careful study of documents. Additional interviews were taken from officials of
the Dutch Ministry of Social Affairs and Employment. From the data, we
Acta Politica 2004 39
René Torenvlied and Agnes Akkerman
‘Soft’ Policy Implementation in Multilevel Systems in the Netherlands
47
Subsequently, interviews were held with the two most important negotiators
in the metal industry. These negotiators provided information on the substance
of all bargaining issues in their sector in 1998. They indicated that 12 issues lay
on the bargaining table: three concerning wages, four regarding training and
human development, two on saving time for sabbatical or parental leave, and
three about child care. In order to map the policy positions of all the
negotiators involved in collective bargaining on one-dimensional issue scales,
we used the same highly structured questionnaires as in the previous interviews
(Rojer, 1999; Torenvlied, 2000; Achterkamp and Akkerman, 2003). The
informants also provided information about which of the issues resulted in
what final agreement. Just like during the interviews conducted on decision-
making within the Dutch Labour Foundation, the informants also estimated
the salience, that is, the relative importance of the bargaining issues for the
different negotiators. Informal power resources often play a role, not only
within the Labour Foundation, but also behind the scenes of industry-wide
negotiations. Hence, formal indicators for bargaining power do not adequately
reflect the actual distribution of power. Therefore, the negotiators were asked
to assess also the informal bargaining power of all the negotiators in the metal
industry. Finally, the informants in the metal industry provided the mapping of
collective bargaining issues with the recommendations in the Labour
Foundation.
Results
Degree of cross-level policy coherence
To what extent do central policy recommendations in ‘Agenda 2002’ reappear
in the agenda of the 1998 collective bargaining in the metal industry? Table 1
provides an answer to this question from the perspective of the two largest
organizations involved in the bargaining processes. The first column lists the 11
selected policy recommendations that were subject to negotiations in the
Labour Foundation. The second column provides information about the
organization that took the initiative in formulating the recommendation.
Together, the first two columns provide information about the central
negotiations in the Labour Foundation. Five recommendations were initiated
by the employers’ peak organization VNO-NCW, while seven were initiated by
the trade union confederation FNV. In one case, both negotiators initiated a
recommendation (about wage moderation).
Level I. Labor Foundation: ‘Agenda 2002’ Level II. Collective bargaining in metal industry
Recommendation in ‘Agenda 2002’ Initiator of recommendation Initiator of agenda proposal Collective agreement
11. Stimulate provisions for child-care Trade union confederation FNV FNV trade union ‘Bondgenoten’ Arrangement in
collective bargaining
agreement
49
René Torenvlied and Agnes Akkerman
‘Soft’ Policy Implementation in Multilevel Systems in the Netherlands
50
tions reappear in the 1998 metal industry collective bargaining agenda, a score
of 64%. That is, more than one-third of the central policy recommendations
disappears in the ‘vertical’ process between recommendation and agenda
proposal. The third column also lists which of the two key organizations is
responsible for the agenda proposal, if any proposal is made: the employers’
organization in the metal industry FME-CWM, or trade union FNV-
Bondgenoten. FNV Bondgenoten puts more recommendations (six) on the
agenda of collective bargaining than the employer organization FME-CWM
(three).
Test of hypotheses
We formulated two sets of hypotheses, each reflecting a different mechanism
behind cross-level policy coherence in multilevel policy-making systems: (1) the
ratification hypothesis and (2) the multilevel coordination hypotheses.
The ratification hypothesis states that full cross-level policy coherence can be
expected between the level-II agenda and the level-II outcomes of collective
decision-making. Table 1 demonstrates that hypothesis 1 must be rejected for
the data in our case study. Neither the agenda of collective bargaining (64%),
nor the final agreement (36%) in the metal industry displays full coherence
with the recommendations from the Labour Foundation.
The agenda coordination hypothesis consists of two parts: (1) representatives
of the level-I negotiator who initiated an agreement put the agreements on the
level-II agenda; and (2) cross-level policy coherence in the level-II agenda is
positively associated with collective support for the level-I recommendation.
Table 1 provides the necessary data for a test of the first part of the agenda
coordination hypothesis. It only partly corroborates the first part of the agenda
coordination hypothesis. It shows that from the three recommendations that
Acta Politica 2004 39
René Torenvlied and Agnes Akkerman
‘Soft’ Policy Implementation in Multilevel Systems in the Netherlands
51
60
50
40
Collective Support (S)
30
20
10
0
−10
−20
−30
0 1 2
Elaboration on Agenda
Note. 0 = No proposal; 1 = Simple proposal; 2 = Elaborate proposal.
Figure 2 Association between collective support of policy recommendations in ‘Agenda 2002’ and
their elaboration on the agenda in the Dutch metal industry. Note: 0 ¼ No proposal; 1 ¼ Simple
proposal; 2 ¼ Elaborate proposal.
support in the Labour Foundation, but did not appear on the agenda of
collective bargaining in the metal industry.
Table 2 provides further clues. It shows the average level of collective
support for three different ‘categories’ of recommendations: first all
recommendations in ‘Agenda 2002’; second, those on the agenda in the
metal industry, and those that were left out of it; third, those reflected in
the final agreement, and those that were lost in collective bargaining in the
metal industry. Table 2 makes clear that the level of collective support for
recommendations that disappeared from the agenda of bargaining in the
metal industry were the recommendations with on average weakest
collective support. A t-test is the standard test to compare differences
in average between two groups. However, because we have only few
cases that have not been selected on the basis of a random sample, a non-
parametric Mann–Whitney U-test is more appropriate. This test compares rank
orders rather than interval scores. The one-tailed test reveals that the difference
is not significant (Z ¼ 0.945; P ¼ 0.12), and that we must reject the second
part of the agenda coordination hypothesis on the basis of this, rougher
comparison.
The outcome coordination hypothesis states that cross-level policy coherence
in the level-II outcomes is not associated with collective support for level-I
recommendation and that, alternatively, the distribution of policy positions,
issue salience, and relative bargaining power of the level-II representatives
better explain the outcomes. Table 2 shows no difference in average collective
support between the recommendations that were lost during bargaining and
those that reached the final agreement. The average collective support for
recommendations reflected in the bargaining agreement of the metal industry is
23.5 (SD ¼ 24.9). We can compare this average with the average collective
support for recommendations that were not reflected in the agreement (either
not put on the agenda or disappeared during bargaining). The collective
support for these recommendations is 15.7 (SD ¼ 27.9). A one-tailed non-
parametric Mann–Whitney U-test shows that no significant differences exist in
References
Marks, G., Hooghe, L. and Blank, K. (1995) ‘European integration since the 1980s. State centric
versus multilevel governance’, Journal of Common Market Studies 33(3): 341–378.
Mayer, F.W. (1992) ‘Managing domestic differences in internbational negotiations: the strategic
use of internal side-payments’, International Organization 46(4): 793–818.
McCubbins, M., Noll, R. and Weingast, B.R. (1989) ‘Structure an process as solution to the
politician’s principal agent problem’, Virginia Law Review 75(2): 431–482.
Mo, J. (1995) ‘Domestic institutions and international bargaining: the role of Agent Veto in two-
level games’, The American Political Science Review 89(4): 914–924.
Moe, T. (1997) ‘The Positive Theory of Public Bureaucracy’, in D.C. Mueller (ed.) Perspectives on
Public Choice, Cambridge: Cambridge University Press, pp. 455–480.
Putnam, R. (1988) ‘Diplomacy and domestic politics: the logic of two-level games’, International
Organization 42(3): 427–460.
Rojer, M.F.P. (1999) ‘Collective decision-making models applied to labor negotiations in The
Netherlands: a comparison between an exchange model and a conflict model’, Rationality and
Society 11(2): 207–236.
Scharpf, F. (1997) Games Real Actors Play. Actor-centered Institutionalism in Policy Research,
Boulder, CO: Westview Press.
Scharpf, F. (1999) Governing in Europe. Effective and Democratic, Oxford: Oxford University
Press.
Schelling, T.C. (1960) The Strategy of Conflict, New York: Oxford University Press.
Schilstra, K. (1998) ‘Industrial relations and human resource management’, Ph.D. dissertation,
Tinbergen Institute, Erasmus University Rotterdam.
Thomson, R., Stokman, F.N. and Torenvlied, R. (2003) ‘Models of collective decision-making:
introduction’, Rationality and Society 15(1): 5–14.
Torenvlied, R. (1996) ‘Political control of implementation agencies’, Rationality and Society 8(1):
25–56.
Torenvlied, R. (2000) Political Decisions and Agency Performance, Dordrecht/Boston/London:
Kluwer Academic Publishers.
Torenvlied, R. and Thomson, R. (2003) ‘Is implementation distinct from political bargaining? A
micro-level test’, Rationality and Society 15(1): 64–84.
Van Assen, M., Stokman, F. and Van Oosten, R. (2003) ‘Conflict measures in
cooperative exchange models of collective decision-making’, Rationality and Society 15(1):
85–112.
Van den Toren, J.P. (1996) Achter Gesloten Deuren? CAO Overleg in de Jaren Negentig,
Amsterdam: Welboom.
Walton, R.E. and McKersie, R.B. (1965) A Behavioral Theory of Labor Negotiations, New York:
McGraw-Hill.
Weingast, B.R. (1984) ‘The congressional bureaucratic system: a principal agent perspective (with
applications to the SEC)’, Public Choice 44: 147–191.
Appendix A
The data on negotiations in the Labour Foundation are summarized in
Table A1.
Recommendations
1 50 50 50 50 50 50 13
2 100 100 100 100 100 100 25
3 100 100 100 100 100 70 8
4 60 80 80 100 60 20 44
5 80 100 100 100 100 100 39
6 80 80 80 40 40 40 32
7 80 50 50 100 80 80 50.5
8 100 100 100 50 50 50 40
9 80 80 80 80 60 60 14
10 100 100 100 80 100 80 3
11 60 60 60 70 70 70 28
Salience-weighted positions, relative bargaining power of the six negotiating peak organizations,
and collective support (S).
Notes
1 We benefited from discussions with several Dutch policy makers and negotiators in industrial
relations. We thank two anonymous reviewers for their detailed comments on an earlier draft.
Torenvlied acknowledges the Dutch Ministry of Social Affairs and Employment for financial
support.
2 We will not reiterate the discussion about relative strengths and weaknesses of a rational choice
approach here. We refer to the debate Green and Shapiro (1994) evoked in the American
political science community.
3 The outcome of collective decision-making on each policy issue is a weight of the (shifted) policy
positions — for example, the median or average position (cf. Thomson et al., 2003). The
weighting procedure and the shifts in positions reflect different modes of decision-making, such
as conflict, logrolling, or social influence. Differences in opinion among decision-makers (their
distance in positions on the policy scale) are assumed to affect positively agency discretion, and
consequently trigger policy deviations. The presumed mechanism is that a difference in opinion
reduces the collective support for a political decision, which results in a political decision open to
a broader interpretation. In such situations, decision-makers have less capacity for monitoring
and sanctioning.
4 A series of further extensions of the model were proposed, for example focusing on effects of
side payments (Mayer, 1992; Friman, 1993) or effects of incomplete information (Mo, 1995).
5 Where the weights refer to issue salience, and relative bargaining power.
6 During the time that lapsed between these social pacts, the Labour Foundation produced a few
‘affirmations’ of existing social pacts, sometimes because a new social pact appeared to be
unattainable.
7 More levels in the system can be observed. From bottom to the top-level of negotiations we
distinguish: (a) meetings between trade union members and their representatives, or between
managers and their interest organization, (b) collective bargaining between employer and
employee representatives in industries and enterprises, (c) national platforms for negotiation: the
‘Labour Foundation’ comprises representatives of the peak organizations of employers and
employees, the ‘Socio-Economic Council’ comprises negotiators in the Labour Foundation and
representatives from government, during spring and fall there is consultation with the Dutch
cabinet. Negotiations are diverse, fragmented, sometimes general and sometimes very specific.
8 The figure is quite simplified. For example, policy-making within peak organizations is not
conceived of as a sub-process of the process of implementation of central recommendations of
the Labour Foundation. This does not deny the importance of co-ordination and decision-
making within peak organizations.
9 The selection of a large industry makes it also more likely that the social pact is built upon
previous collective bargaining in the large industry, thus resulting in policy coherence.
Mechanisms of previous negotiations or previous lobbying are strictly kept outside the
implementation models, and are not the subject of the present study. However, effects of these
mechanisms can be tested in a more elaborated, longitudinal design. Such a design would require
more data that are difficult to obtain.
10 For details, we refer to Rojer (1999), who reports an average model accuracy of 65% of the
outcomes of a large number of collective bargaining agreements (for a more general
introduction, see Thomson et al., 2003; Van Assen et al., 2003).