Consistency versus legitimacy

For researchers, inconsistency is often considered the ultimate sin. Inconsistent arguments in papers lead to rejection, and there is a strong instinct to avoid policy proposals that seem to be inconsistent.

Here is an example: When we do cost-benefit analysis of a new pharmaceutical, some argue that consistency demands that we should include all the costs of future treatments. For instance, a pharmaceutical that reduces a person’s probability of heart problems may increase the future costs related to cancer. Should this increase in cancer costs be included when we evaluate whether to reimburse the pharmaceutical and how much to pay? Consistency seems to demand a «yes» (1). On the other hand, many people object to the notion that «future costs» should count against a treatment today.

Fresh out of the Harkness experience, I was faced with this dilemma when I was invited to be part of a government commission to give advice about formal rules for priority setting in the health care sector. The conclusion was that the primary aim of giving advice was not to create a consistent system, but one that was legitimate. If a large share of the population believed it was ethically wrong to include future costs in this way, the system would not be good, even if it was logically consistent.

The argument that legitimacy matters is not the same as accepting that policy advice always should be constrained by majority opinion or, even worse, prejudice. Sometimes good policy advice challenges views built on poor information or unethical preferences. However, sometimes people seem to have informed and true preferences that at least at first glance appear to be inconsistent. Instead of dismissing these, one may take them as food for further reflection (2). And in any case, good policy advice needs to consider people’s actual preferences and not what we would like these preferences to be.

Optimal versus feasible

As part of a government commission on priority setting in public health, a key topic is how high the bar for evidence should be before we adopt a public health project. Researchers tend to aim for the optimal solution to problems. This involves setting up a process where evidence is collected, and the expected costs and benefits are calculated. The perceived best evidence in this case is often information from randomized controlled trials. Both the experiment and the process are often time consuming and expensive. Still, the demand that we make an optimal choice easily leads us in this direction.

While the instinct of seeking good evidence is often sensible, it may sometimes lead to hyperrationality and inferior policy advice. Hyperrationality occurs when we ignore the costs associated with the process itself, transactions costs or human costs associated with the collection of evidence and delaying the decision (3). And even if it was theoretically possible to collect the required information, it may not be politically feasible to establish the required bureaucracy. In this case, one may end up with the appearance of a rational process, but based on costly and weak information. Good policy advice takes this into account and propose solutions that that are easier to implement and «better than the current system» without necessarily being perceived as «the optimal» solution.

In the specific example about public health, one may still demand evidence but be more open to evidence from register based data, accept the use of proxy end points and implement reforms in a way that allows us to learn and adjust as we go.

Complexity versus simplicity

A hot policy question is the use of so-called sin-tax. These are taxes on unhealthy products or habits. Tobacco and alcohol are obvious examples, but also soda and products containing sugar have been included. The question in various government commissions is whether and how these products should be taxed.

From an economic perspective, there is a theoretical solution. The products often carry an externality, i.e. a cost that is borne by parties other than those who buy the product. Society contributes to health care, social security for the poor, pays to prevent crime and many other expenses that are related to some of these goods. One may calculate the size of these external costs and propose that the tax should be large enough so that the price reflects the true cost.

A good economic researcher, however, should not stop with this proposal. It is not enough to simply calculate the external costs. We also need to calculate all the other consequences (4). The consumption of one good may be linked to the consumption of others. Less sugar in chocolate may be substituted with unhealthy fat or people eating more chips. All these effects must also be taken into account, and quite quickly the solution becomes complex.

The complexity, combined with a desire to create an optimal solution, may lead to systems with many parameters. Such a solution is likely very sensitive to changes in these parameters. In addition to the technical instability, it may also be politically unstable in the sense that a complex system opens many more access points for lobbying and political pressure. This means that good policy advice should focus on simplicity and robustness, and less on optimal solutions that may be complex and unstable.

As an illustration, consider the case of how much information to collect before making a decision. In some cases, there are sophisticated rules that tell you exactly when it would be optimal to stop collecting more information, but these rules are often complex. Instead, rules of thumb, like «collect five prices and then choose the place with the lowest price» are much simpler and often not far from the optimal solution (5).

An inconclusive conclusion

The example of taxing sugar illustrates the lessons discussed so far. A practical proposal might be inconsistent, as it is technically and politically impossible to create a tax that covers all products containing sugar. People may accept the rationale for a soda tax but perceive orange juice and foods as falling in a different category. It may also be very difficult and costly to calculate the optimal tax on all products, at the same time we know that at least some tax is probably better than no tax – and that this is likely more true for some products than others. This means that a tax on some obviously unhealthy products, may be a better policy advice than a complex system trying to cover all unhealthy products in an optimal system.

The role of the researcher is to identify all the challenges, but the task of providing policy advice is to propose feasible solutions that may be simple, sometimes inconsistent, and not optimal, but still represent improvements over the current system.