by Richard G. Lipsey and Kenneth Carlaw, Simon Fraser University, with a contribution by Davit D. Akman, Research Associate, October 1998
Executive Summary
This monograph evaluates a selection of government policies designed to influence technological change. Public policy can seek to accelerate or retard technological change, using incentives or disincentives that are either generally applied or narrowly focussed. Policy can also seek to exert its influence indirectly by altering such structural variables as the concentration of industry, the amount of foreign investment, the location of firms and the education system.
At the level of abstraction at which economists commonly operate, the distinctions between policies, programs and projects are ignored, and the general term policy is used. Our analysis requires that the distinctions be highlighted in assessing policies and also in evaluating programs. Furthermore, within policies we distinguish three types. Framework policies, such as research and development (R&D) tax credits, provide general support for specific activities across the entire economy; they are usually single-instrument policies that do not discriminate between firms, industries or technologies. Focussed policies are designed to encourage the development of specific technologies such as nuclear power, and are usually so narrowly focussed that falling within the focus is both a necessary and sufficient condition for obtaining benefits under the policy. Blanket policies, such as the Industrial and Regional Development Program (IRDP), incorporate elements of both framework and focussed policies: like framework policies, they have broad-based objectives; like focussed policies, they use multiple instruments and have some form of assessment mechanism enabling administrators to tailor the assistance they provide.
We study the policy advice that follows from two distinct theoretical paradigms. Neoclassical theory provides a precise set of objectives for policies and programs, and a means of assessing whether or not they have contributed gains. What we call structuralist-evolutionary theory starts from different assumptions about the behaviour of the economy, and it reaches different conclusions about both the role of policy and the substance of policy/program evaluation. We find that these differences account for striking divergences in views about the efficacy of various policies and programs.
The two types of theory also suggest different criteria for assessing incrementality. Because they treat structure and institutions as "black boxes," neoclassical theories assess incrementality solely by the effects on technological change, usually measured by changes in R&D expenditure. Because the structuralist decomposition stresses the relation between technology and the underlying structure through which it operates, the incrementality criterion allows for policies that alter structural relations without necessarily affecting the level of R&D expenditure or inducing specific technological changes.
Neoclassical theory recommends policies that give generalized support to all R&D over policies that are more selective, while structuralist theories suggest a place for both generalized and selective policies.
Neoclassical theory provides criteria for optimal policies while the structuralist model suggests that irreducible elements of judgment are required for all policy decisions and judgments over which reasonable people may simply have to agree to disagree.
The direct approach of measuring the outputs of blanket policies is seldom easy and often impossible. As an alternative assessment procedure, we turn to the criteria derived from Lipsey and Carlaw's 1996 study of the conditions that favour success or failure in focussed programs, and we use these criteria in our current assessment of the design and operation of blanket policies and programs. First, we examine the available assessments made by others. These concentrate mainly, but not exclusively, on outputs. Second, we compare the design of the policy or program in question with Lipsey's and Carlaw's design and operation criteria, using them to judge the potential for success or failure. Third, where there is agreement between the judgments reached under both of the above procedures, we conclude that there is a strong case for either success or failure. Fourth, if the judgments resulting from the two procedures disagree, we seek to reconcile the differences by comparing the theoretical perspectives adopted by ourselves and the other assessors. The differing judgments often arise from the different assumptions that characterize the theories employed.
In the cases of the Defence Industry Productivity Program (DIPP) in Chapter 2 and of the Industrial Research Assistance Program (IRAP) in Chapter 4, our assessments are favourable. Where the assessments of other evaluators differ from ours, we find two main causes. The first is that different incrementality criteria are used, with our assessments allowing for a broader set of objectives. The second is that neoclassical theorists prefer framework policies and reject policies with elements of focus suggested by structuralist theory.
In Chapter 3, we agree with the other assessors in rating the IRDP and its predecessors as failures. Our criteria allow us, however, to focus on specific structural interactions and design characteristics as causes of failure. These causes are either ignored or downplayed by those using other theoretical perspectives.
We conclude that, in contrast to neoclassical theory's strong and exclusive recommendation in favour of framework policies, the ideal structuralist policy set must have the single aim of encouraging technological advances but should use multiple policies and programs to achieve that aim. Framework policies provide the general push. Focussed policies cover particular spots where market failures are large and specific. A few blanket policies can be cautiously applied when a relatively broad-based single need is identified and clearly communicated to the administrators. But before such middle-range policies are used, very careful study is needed and much fuller and more careful study than has typically gone into the design of past policies and programs that were often hastily put together in response to political pressures. Before spending millions of dollars on any new blanket policy, a few tens of thousands should be spent on clearly defining its goals, selection criteria and administrative structure. In principle, this advice is easier to follow than the neoclassical advice of searching out the optimum level of R&D and instituting neutral policies to achieve it. It may, however, be no less difficult to follow in practice.