As Linder and Peters (1990: 307) noted, once the tools of government have been inventoried and classified, ‘the need to do something more becomes irresistible’. The next logical step was ‘to explore functional connections’ involving ‘matching instruments to goals, policy problems, social impact and organizations’. Taking this next step towards the idea of policy design, as we have seen, requires clarifying the nature of the criteria by which experts assess policy tools and the nature of the contexts in which they can reasonably be anticipated to perform as expected.

Kirschen et al., in their early 1964 work, had already gone some distance towards that goal by arguing that the key determinants of policy choice in the case of the economic instruments they had identified were the economic objective or goal pursued and the structural and conjunctural context of the choice. The economic objectives, they argued, were determined by the interaction of political parties and their representatives in government, administrators, and interest groups (1964: 224–36), while the structural and conjunctural context, in turn, was affected by the influence of long-term economic processes and structures, and current economic conditions (236–38). They argued that the actual choice of instrument from within the set that fit these epistemic and contextual constraints should be made on essentially technical grounds, according to efficiency and cost criteria; although the political preferences of interest groups and governments – including sociological and ideological constraints – and the institutional limitations of the political system itself had to be taken into account as factors influencing key decision-makers (238–44).

This was a prescient analysis of the overall set of factors affecting instrument choices (Majone 1976; 1989), combining as it did both technical and political factors. However, it was also one which was not adequately grounded in a classification of instruments so as to be able to produce specific recommendations or hypotheses concerning appropriate instrument selections and policy designs in different circumstances or times.

The first models of instrument choices which were so grounded attempted to identify a limited number of criteria upon which policy tools varied; creating single or multiple ‘spectrums’ or ‘continuums’ of instrument characteristics which it was hoped could then be associated with specific government preferences among these criteria. Dahl and Lindblom, for example, as early as 1953 had argued that the number of alternative politico-economic instruments is virtually infinite, and proposed five long continua as a method of assessing tool preferability in specific contextual situations (Dahl and Lindblom 1953). Their first continuum ranged instruments according to whether they involved public or private enterprises or agencies; the second according to whether they were persuasive or compulsory; the third according to whether they involved direct or indirect controls over expenditures; the fourth according to whether they involved organizations with voluntary or compulsory membership; and the fifth according to whether government agencies were autonomous or directly responsible to legislators or executive members. Although Dahl and Lindblom did not pursue any further the question of whether and to what extent governments actually used these criteria in order to choose a particular instrument, their idea of arranging instruments on a continuum in order to better clarify the reasons behind their choice was adopted by many authors.

A simplified version of this model was put forward by a group of Canadian scholars including Bruce Doern, Richard Phidd, Seymour Wilson and others, who published a series of articles and monographs in the late 1970s and early 1980s that turned Lowi’s two-dimensional matrix of policy choices into a single continuum of policy instruments based on the ‘degree of government coercion’ each instrument choice entailed. They first placed only self-regulation, exhortation, subsidies, and regulation on this scale (Doern 1981) but later added in categories for ‘taxation’ and public enterprise (Tupper and Doern 1981) and  finally, an entire series of finer ‘gradiations’ within each general category (Phidd and Doern 1983).

For Doern and his colleagues, the development of their coercion spectrum model led to the hypothesis of a twofold rationale of instrument choice; one that fitted very well with the notion of a ‘continuum’ of choices, and which offered a great deal of explanatory power in the context of liberal-democratic states. This rationale was based on an appreciation of the ideological preferences of liberal-democratic governments for limited state activity and on the difficulties posed to this principle by the relative political ‘strength’ of the societal actors in resisting government efforts to shape their behaviour.

Assuming that all instruments were more or less technically ‘substitutable’ – or could perform any task although not necessarily as easily or at the same cost – they argued that in a liberal-democratic society, governments, for ideological reasons, would prefer to use the least coercive instruments available and would only ‘move up the scale’ of coercion as far as was necessary in order to overcome societal resistance to attaining their goal. As Doern and Wilson put it:

politicians have a strong tendency to respond to policy issues, (any issue) by moving successively from the least coercive governing instrument to the most coercive. Thus they tend to respond first in the least coercive fashion by creating a study, or by creating a new or reorganized unit of government, or merely by uttering a broad statement of intent. The next least coercive governing instrument would be to use a distributive spending approach in which the resources could be handed out to constituencies in such a way that the least attention is given as to which taxpayers’ pockets the resources are being drawn from. At the more coercive end of the continuum of governing instruments would be a larger redistributive programme, in which resources would be more visibly extracted from the more advantaged classes and redistributed to the less advantaged classes. Also at the more coercive end of the governing continuum would be direct regulation in which the sanctions or threat of sanctions would have to be directly applied. (Doern and Wilson 1974: 339)

This model was lauded for its simplicity and elegance but, as critics pointed out, was still problematical in its application to policy design decisions since ‘coercion’ appears to be indivisible, or at best, still very difficult to operationalize with the degree of precision required by the model (Trebilcock et al. 1982: 22–23).