6.1. Generic versus specific policy
There is a trade-off between generic policy measures and policy aimed at specific sectors. Generic policy
does not create selection problems, but contains some other drawbacks, the most important of which is
dead weight loss. Sectoral policy creates more risks of government failure, though potential gains are
higher. A policy aimed at specific sectors may then sometimes lead to a great success, such as the focus
of Finland in the 1990s on telecommunication and the rise of Nokia. On the other hand, it can also lead
to failures that are easily forgotten.
Specific policy is more appropriate if we are able to observe what the actual and crucial market failures
are, what their size is and in which sectors they occur. In particular in case of specific policy, a careful
assessment should be made if the potential benefits outweigh the higher risks of government failure.
Given our current information restrictions it is more difficult is to design a tailor-made policy mix.
6.2. A plea for experiments
Ineffective policy measures can be very costly and more information about the effectiveness would be
more than welcome. The most appealing way would be to rely on well designed experiments before
introducing new large programs. Most preferably, experiments have a control group so the impact can be
compared to a group that did not receive the subsidy. This can be achieved by randomly assigning
subsidies if the demand from companies exceeds the available budget or to build in discontinuities in
subsidy schemes. In the Netherlands, just as in other countries, these experiments are hardly used in
innovation policy. However, a recent exception is that of innovation vouchers. Approximately 800 small
businesses have signed up for a voucher which reduces the cost of a research assignment to a public
knowledge institute. Out of these 800 applications a notary has randomly drawn 100 applications which
will actually get the voucher. By looking at the behavior of all initial subscription the effectiveness of
this program is measured. This is a well-designed and controlled experiment which can make a huge
contribution to our knowledge of the effectiveness of this instrument. Unfortunately, for the purpose of
evaluating (new) policy proposals there is a lot of resistance against experiments, because of the unequal
treatment of groups, despite the large information gains. An alternative would be to rely on natural
experiments, where coincidentally a comparable control group exists in the real world, for instance due
to changing rules, population of subsidy schemes. These kinds of experiments are quite cheap and the
results are quickly available.
6.3. A plea for more accountability
Experiments would thus be a welcome supplement to evaluate policy initiatives. However, experiments
have their limitations as well. In practice, it is difficult to evaluate complete policy areas or institutional
changes. There are other ways to learn from experiences to improve the effectiveness of government
policy without a full impact analysis. The most important is that the budgetary process should contain a
framework that creates a transparent and accountable evaluation process with respect to input,
instruments and result. This means more attention should be paid to the rationale and justification of
problem identification and policy design. Such a framework should for example consist of a profound
systematic answering of the following questions (Ministerie van Financiën, 2004):
- What is or has been the reason for policy intervention and is the problem still actual?
- What’s the cause of the problem?
- Why is their a role for the government in solving the problem?
- Which level of governments is most suited to solve the problem and how is the accountability
organized? - Which objectives are formulated?