Over the years I have spent some time in policy events where economists make recommendations to a crowd of civil servants, journalists and politicians. I always considered that as a good intellectual exercise, but it was clear to me that such events had very little influence on the actual course of policy. For an academic economist who spends most of his time doing research, i.e. exploring the unknown, intervening in the policy debate, i.e. turning knowledge into action, is very difficult. In this note I discuss four pitfalls which, if you fall in one of them, make it quite likely that your intervention is useless. And I don't claim I know how to avoid them.

1. The Academic attitude.
The Academic pitfall consists in being unable to get out of academic ways of thinking and expressing oneself. The result is a discourse which is uninteresting or difficult to understand from the point of view of policy makers. Our discipline is increasingly specialized and technical. Even when we think we're making an effort to be pedagogical when adressing bureaucrats and policymakers, we are still off-mark. Being pedagogical for such an audience requires more effort than being pedagogical for students. Students have already being introduced to basic notions such as elasticity, production function, consumer's surplus, or liquidity trap. In addressing policymakers, one has to assume that they don't know any of these things. Thus 'elasticity' may sound cristal clear to us, but it may be absconse to many non economists. When talking to physicists I was amazed to discover that what they refer to as a parameter, a variable, or robustness, are quite different things from what these words mean to us. The same is probably true for bureaucrats and politicians. Instead of such jargon we should make an effort to use common-language words. 'Responsiveness' is as good as elasticity, etc.
One should under no circumstances use the word 'model', even though models structure our thinking and we are constantly using them, even implicitly. Policymakers do not want you to talk about models, they want you to talk about the real world. To address policymakers you should supplement your knowledge of models with judgement, i.e. an assessment of which models, assumptions, and results are more relevant. Example:  result A holds if skilled labor is more mobile than unskilled labor. Is that a reasonable assumption? Data and experience suggest that it is the case, so we can confidently use result A. Result B, in contrast, holds in a two-period OLG model with strong nonlinear income effects. While it enhances our understanding of the model, it is probably not relevant for the real world. We should abstain from making much out of it, or even refer to it, when talking to policymakers.
'Two-handedness' ("X has shown that, on the other hand, Y, using different data and specification, finds no effect. Therefore, one should be cautious about...") should be ruled out. It is a virtue for an academic, but a vice for somebody who adresses policy makers. If the literature is inconclusive, why should busy decisionmakers lose their time hearing about it? Either you have made your own opinion, based on own judgement, introspection, and personal experience, and then you should have made up your mind, even though the literature has not. Or you just don't know, in which case don't talk about it. Similarly, criticizing other economists for failing to control for this or that should be avoided. When you are attending a policy event where a fellow economist is speaking, refrain from intervening if it is only to add provisos and qualifications to what he said. Many people do that just to call attention on their own work.

2. Banalities and vagueness.
One tempting attitude is to protect oneself by only expressing banal recommendations on which everybody agrees. This is especially tempting as the public policy debate is getting less controversial, and more hypocritical, in many areas. People assertively claim to be against war, AIDS, or racism, as if there was something surprising or shocking about it.
Thus, you should refrain from saying banalities such as "we need more education ...", "we need to invest in social cohesion..." Nobody would contradict you and people would perhaps look very happy and express agreement. But are they learning anything or just having a good time? We need more education ... but at what cost? Social cohesion...what does it mean?
A complementary attitude is vagueness. By being vague one is less likely to be proved wrong. People can infer what they like from your statements. Vagueness is again useless. Avoid statements such as "it may well be that..." or "adequate policy measures should be taken in order to ..." Which policy measures? A tax? Of how much? What if there is a representative of the producers of the good you want to be taxed in the room? "Specific social groups should be compensated ..." Who are we talking about exactly? have we made the computations of gains and losses for these people? What if we discovered that they lose and they did not find out? Isn't some politician going to get angry at us if we reveal it?
Sentences with "appropriate" or "adequate" are tautologies. Their contrary could never conceivably be true in any state of the world (this is how a tautology is defined).

3. The eclectic dirigiste
Another defensive attitude is to advocate complex policies which use many instruments. This is a comfortable stance for several reasons. First, it makes it difficult to find out whether what you propose improves or worsens things. Second, bureaucrats like complexity because it increases the demand for their product: complicated regulations which increase their power and ability to distribute rents and property rights. Hence they are very happy with economists who tell them that all this mess is actually on the right track. Third, these proposals allow to bring together all we have read and written in a grand policy scheme, and to avoid using judgement to select what's really important. As Nietzche said: "being open minded means being unable to say neither yes nor no".
In many cases complexity hides contradiction. Your proposal draws from different models which are actually incompatible with one another. Or, you want to neutralize several distorsions that are in fact the result of your policies. Example: Z, a smart economist, has found out that employment is too low at the bottom of the distribution of income. So he first advocates a reduction in marginal tax rates for the poor. Then he realizes that this may create budget deficits. So he advocates an increase in marginal tax rates for the rich. But this reduces incentives to invest in higher education. So let us subsidize higher education. But then we need to raise taxes further. So let us introduce a tax on capital...however, this is bad for capital accumulation in the long run, which may hurt wages...wait...how about an investment tax credit, then?
Just as with the vague/banal option, you won't make enemies. "Vous caressez les bureaucrates dans le sens du poil". Out of the 35 or so measures that you advocate, the odds are that some will be implemented. So you could even think you had an influence. But the truth is that you did not make a difference.

4. The smart economist's idea
At the other end of the spectrum, there is the risk of proposing policies that are rejected simply because they are miles away from the basic culture of bureaucrats and politicians. Politicians have a constituency of voters of limited intelligence, who typically resist radical changes. Bureaucrats do not want to be made useless, and resist competition from economists. Finally, ideology imposes further constraints.
Unfortunately, these proposals are often those which make most sense from an economic viewpoint, like introducing new markets, opening trade, creating tradeable property rights, and so on.
I am myself fond of  such smart, simple ideas. But if you propose them to policymakers, you often sound like an idiot. First if it sounds too simplistic, you don't look like an expert, which reduces your credibility. Second, these people constantly meet each other, and eventually develop a common language, and a common approach to policy. They play safe: who's going to blame you for a mistake if everybody has made it? Whatever these people do and say is irrelevant provided others do and say the same.
The economist who comes with a smart simple idea is threatening to destroy this consensus capital. If, as a bureaucrat, I deviate and endorse the economist's view, I may well fail alone, and this would be damaging. If I stay with the herd and ridicule the economist, not much can happen to me. So the Nash equilibrium is to reject the smart economist's idea. This would not be the case if all these policymakers were newcomers and had not spent so much time and effort building consensus.
Hence coming with new provocative ideas is also useless, except in circumstances (revolutions, severe crises) when a spirit of "table rase" prevails.

After this rather depressing list the reader may ask whether there is anything to be done. Since I have never contributed to a policy change myself, I am probably not the best person to answer that question. But I have learned a little bit from colleagues who were more involved in policies than I was. In particular, I got from them the following possible strategy: Introduce a change which makes no difference in the short run but paves the way for gradual change in the long run. This sounds vague (pitfall 2), but here is an example.

An example: The CSG in France

How do you increase the tax base in a country where only 50 % of households pay income taxes? Well, you introduce a new tax (the CSG), but its rate is almost zero. In order to make it even less painful, you collect it at the source. Voters don't scream: it is a negligible event. Bureaucrats are happy: a new scheme to be administered. Then, as time passes, you loudly reduce the income tax (which 50 % of the people pay) and silently increase the CSG, which 100 % of the people pay. Hence you have increased the tax base, and reduced the distortionary impact of taxation (it takes 15 years).
Note that economics play a role here: it teaches us that the tax base should be broad, so as to have lower tax rates. However, all the skills involved are in making people swallow the move, not in actually computing the optimal tax system, and then brutally impose it. Computing the actual optimal CSG tax is a second-order problem; most of the benefits come from actually being able to increase the tax base, and most of the effort is in making it politically acceptable.
Note also how the reform is deeply consistent with dominant state culture: a new acronym, a new tax, a small change, followed by many other small changes. (Ironically, a subsequent prime minister, rather than increase it by 1 % preferred to introduce yet another tax). Policymakers are in a familiar environment. No crazy economist who wants to suppress the income tax overnight and replace it with a broad-based tax with one or two  brackets.