Once a fairly exhaustive inventory has been created, the next major step in theory construction is to move towards taxonomy. That is, examining the list of the phenomena under consideration and attempting to classify or categorize the subject matter into a smaller number of mutually exclusive categories which, together, retain the exhaustive character of the original lists. Many such schemes were developed in the policy instruments literature of the 1960s to 1980s.

Kirschen and his fellow authors (1964), for example, utilized a resource based taxonomy of governing instruments to group instruments into five general ‘families’ according to the ‘governing resource’ they used: public finance, money and credit, exchange rates, direct control, and changes in the institutional framework (16–17). However, this scheme was very sectorally specific and focused on the specific problem of achieving economic development goals. More generic schemes were developed such as that put forward by Theodore Lowi (1966; 1972), which heavily influenced later thinking on the subject. Lowi developed the insight first put forward by students of public administration in the USA like Cushman (see Figure 1), that governments had only a small number of alternative choices in any given regulatory situation, depending on the amount of coercion they wished to employ in that situation – in Cushman’s case, choosing either to regulate or not and, if so, to regulate either by the use of public enterprises or regulatory commissions. This analysis, among other things, introduced the idea that instrument choices were multi-level and nested, an insight which would be further developed in the years to come.

Figure 1: Cushman’s three types of policy tools (Cushman, 1941)

In his own work, however, Lowi argued that a four-cell matrix based on the specificity of the target of coercion and the likelihood of its actual application would suffice to distinguish all the major types of government implementation activity. The original three policy types he developed included the weakly sanctioned and individually targeted ‘distributive’ policies; the individually targeted and strongly sanctioned ‘regulatory’ policy; and the strongly sanctioned and generally targeted ‘redistributive’ policy. To these three Lowi later added the weakly sanctioned and generally targeted category of ‘constituent’ policy (see Table 1).

Table 1: Lowi’s matrix of policy implementation activities

This was a significant advance, since it attempted to reduce the complexity of instrument choice to a single two-dimensional framework. However, Lowi’s categories of tools – distributive, redistributive, constituent and regulatory – did not fit well with existing tool inventories and hence were difficult to operationalize and test (Roberts and Dean 1994). As a result, many other classification schemes emerged in the literature in the mid-to-late 1980s.

Many of these efforts, but not all, followed Lowi and Cushman’s lead in focusing on some aspect of coercion as the key element differentiating policy instrument types. In a key development, however, some also introduced a greater number of differentiating criteria. Balch (1980), for example, talked about both ‘carrots’ (inducements or incentives) and ‘sticks’ (coercion or disincentives), while Bardach (1980) argued that government had three ‘technologies’ which they could utilize in any given choice situation: enforcement, inducement, and benefaction; and that these strategies required different combinations of four critical governmental resources: money, political support, administrative competency, and creative leadership. Elmore (1987) identified four major classes of instruments: mandates, inducements, capacity-building, and system-changing.

Further studies moving in the design direction refined this idea of only a limited number of ‘governing resources’ lying behind each tool. Christopher Hood (1983; 1986) generated a major work on the subject in 1986 which was heavily influenced by detailed studies of the British and German policy implementation processes undertaken previously by Dunsire (1978) and Mayntz (1975). It involved the elaboration of a fourfold resource-based categorization scheme for policy instruments which served as an admirable synthesis of the other, earlier, models.

Hood argued that governments have essentially four resources at their disposal which they can use to either effect changes in their environment or detect them: nodality, meaning the resource that existed simply by nature of the fact that government’s existed at the ‘centre’ of social and political networks but which can also be thought of as ‘information’ or ‘knowledge’; authority; treasure; and organization (or ‘NATO’ in Hood’s terminology). In Hood’s scheme, implementation instruments are grouped together according to which of the NATO resources they most or primarily relied upon for their effectiveness, fully recognizing that most used some combination of these resources in practice (Anderson 1977; Hood 1986).

This taxonomy proved useful in providing a limited number of eight clearly differentiated categories of instruments (see Table 2).

Table 2: Hood’s 1986 taxonomy of substantive policy instruments