As a policy wonk, I get a sinking feeling when I'm reading a terrific non-fiction book and find there's a "policy recommendations" chapter.

I’ll be happily devouring chapter after chapter, lapping up the insights. Then I reach the second-to-last chapter; this, by some sort of mysterious ancient law of publishing, is where the author is expected to say what their insights mean for public policy.

And I usually find myself mightily disappointed. Nine times out of ten, the originality of the author’s thesis seems to evaporate like the morning dew, and I’m left plodding through 15 pages of policy recommendations I’ve read 1,000 times before.

It’s not that the policy ideas are stupid; in fact they’re usually perfectly sensible. But so often they fall short of the originality of the rest of the book, and indeed seem semi-detached from it.

(Economists seem particularly bad at this. Time after time, the most interesting general-interest economics books seem to fall back on recommendations from a sort of general economists’ sourcebook: relax planning laws, increase competition, reform schools, tackle inequality in some vague way, end occupational licensing, loosen IP rules... To be clear, I’m not saying these are bad ideas. I just feel I’ve heard them all before.)

Until recently, I just accepted this as a fact of life, and resorted to skim-reading the inevitable policy chapter and enjoying the rest of the book. But increasingly I’ve started to wonder why it is that obviously bright people write such dull chapters. And I think I have an answer.

The problem is that when a person sits down to write a set of policy recommendations, there are actually a number of subtly different things they might be trying to do.

Let’s call the first option the adviser’s approach. This involves trying to write down the best set of measures to tackle a particular problem. It’s the kind of answer a government expert would use when writing a white paper on the subject that was actually going to be implemented. Examples include Bob Gordon’s 10 recommendations for American growth (excerpted from his impressive new book), or the recommendations in Brynjolfsson & McAfee’s excellent Race Against the Machine (more entrepreneurship, school reform, R&D spending, and regulatory reform).

The advantage of the adviser’s approach is that if it’s done well, the list of recommendations actually represents a good programme of action for government to take. This is very useful if the reader actually is a government minister (though most of us of course aren’t).

The disadvantage of the adviser’s approach is that it doesn’t put a premium on originality. After all, most good policy ideas aren’t original, and most original policy ideas aren’t good. You would expect even a very good policy white paper to include lots of old ideas. This is a problem if you are expecting the chapter to tell you something new and fresh.

An alternative way to approach policy recommendations is what I’ll call the wonk’s approach. Here the point is to be original, ideally in a way that proceeds directly from the argument of the rest of the book. Some examples of this include Thomas Piketty’s plan in Capital in the Twenty-First Century for a global wealth tax, or Mariana Mazzucato’s recommendation in The Entrepreneurial State that the government should take equity in firms that commercialise state-funded technologies, or Brynjolfsson & McAfee’s recommendations for a basic income in their follow-up book, The Second Machine Age.*

The advantage of this approach is the reader gets ideas they haven’t heard before (or at least, a strong rationale for ideas that are rarely proposed). This is a big plus for me.

There are two downsides to the wonk’s approach. (a) It generally doesn’t aspire to be a complete solution to the problem – the point is to say something new and interesting rather than to list all the things that need to be said. (b) The advice is on average more likely to be wrong than if you follow the adviser’s approach (it’s harder to be original and right than simply right**).

Personally, I’ll take the wonk’s approach over the adviser’s approach any day. I’d rather hear an original or semi-original recommendation that proceeds from the author’s overall thesis than even a very sensible set of recycled policies. After all, I’m reading a book because I want to hear new ideas, not to decide whether to hire its author as a policy adviser.

Post-script: think-tanks, the wonk’s approach and the adviser’s approach

The fact that there are two different sorts of policy recommendations also has implications for think-tankers who are writing reports.

The more I think about it, the more I recognise these two different approaches in think-tank reports.

I’d like to think the wonk’s approach of making original, idiosyncratic policy recommendations is the nobler one for think-tanks. After all, one thing think-tanks should be good at is policy formulation and novelty. (And a good policy recommendation does not always have to be a practical one: sometimes a policy recommendation is useful because it sharply outlines a particular debate. But that’s a different topic.)

But I can see some places where the adviser’s approach of bundling up less original recommendations might appeal to think-tanks.

Most of the time when a think-tank writes a “manifesto for X” (for the tech sector, for manufacturing, for London, or whatever), it’s generally following the adviser’s approach. You can expect lots of worthy but not especially original recommendations.

There are two occasions when the adviser’s approach is especially tempting for think-tanks: first of all, if it wants to sign up a lot of public backers for its report. After all, it’s easier to build a movement or a campaign behind familiar proposals than behind your random new idea, even if it’s brilliant.

The adviser’s approach is also useful is if the think-tank wants to demonstrate how many of its proposals subsequently get accepted. If your aim is tick off the number of proposals that government implements (either out of the noble desire to change the world, or because you need to demonstrate “impact” to funders or trustees), you’re better off packaging up known proposals than marketing the brand new.

I’m not sure either of these are good reasons for think-tanks to reheat old policy ideas – but I’m pretty sure I see them a lot. But let me conclude with another option for think-tanks: not making policy recommendations at all.

For a long time this approach seemed totally wrong to me. (Perhaps this is my déformation professionelle as a former management consultant; always give the client actionable advice!)

But increasingly I’ve come to the opinion that (a) think-tanks add much of their value by the way they frame problems and diagnose the situation, not by the specific solutions they propose; (b) it’s often good to separate the task of diagnosis from the task of policy design.

So let me end with a plea to think-tanks: to produce fewer me-too policy recommendations, and instead to either come up with original ones, or not to bother with them at all.

* The Second Machine Age is interesting in this respect because the authors make a point that they wanted to set out some more extreme policy recommendations, alongside some explicitly near-term ones.

** I’ve pointed out my concerns about the policy recommendation in The Entrepreneurial State, for what it’s worth – but even though I don’t agree with it, it’s certainly interesting.