A sideways look at economics
Fathom is about to embark on another major consultancy project. This isn’t a brag — or not mainly. It’s a note about what — I’ve come to realise over the years — consultancy really is, at least the way we do it. [1]
The project involves gathering and cleaning tons of data, reading piles of learned papers, defining the question as tightly as possible in the light of all that (the question being rather loosely defined at present), designing an approach to that question, and answering it as best we can. The ‘answer’ is not usually a number (42, for example): it’s a narrative. A coherent and intellectually satisfactory story that’s consistent with the data that are available.
A successful project will generate a successful narrative. Success here means: first, the client is happy and has learned something useful about the subject matter; and second, the narrative we develop is not part of the mainstream at the time we develop it and it becomes part of the mainstream thereafter. The first condition of success is a necessary condition for all consultancies. The second is nice to have, and confers on us and on our clients a sense that we saw something important ahead of the mainstream.
The second condition — the narrative starts outside the mainstream and ends up inside the mainstream — is also a description of a successful virus. Robert Shiller’s fascinating new book Narrative Economics explores this theme.
Shiller introduces a new definition of narrative economics. The old definition was more or less the description of the typical consultancy project above — a coherent story consistent with the data. Shiller’s new definition says that narrative economics is the study of how different narratives compete and which ones go viral: which ones are most ‘contagious’, drawing on the analytical tools developed in epidemiology.
Accepting this description of economics involves an important step for economists: a sort of humbleness in the face of uncertainty.
The scientific method (per Popper and others) is not a good description of how economic thought evolves. Instead, there are life cycles of successful narratives (or of ‘constellations of narratives’, as Shiller puts it), which correspond closely to the life cycle of viruses. It’s not a competing claim to truth, so to speak, so much as competing degrees of contagiousness for different narratives, that distinguish between successful and unsuccessful economic research.
Successful economic research is viral and contagious. If we contribute to a successful, contagious narrative, then we feel validated. We succeed when our narratives are contagious. This implies that we economists must be humble when it comes to making assertions about what is true. Why?
Let’s think about the consultancy project we’re about to start.
First, the data are patchy and inconsistent. They cover different concepts and different nations; and are gathered from different primary sources. In the end, we will attempt some econometrics using these data. But it’s already clear that we won’t have enough good consistent data to be fully confident, in a statistical sense, in the robustness of any conclusions that we might draw. You need a minimum of 30 independent data points net of the number of explanatory variables employed in any single equation, assuming that the data are all stationary (or integrated order zero, I(0) in the jargon),[2] the explanatory variables are independent of each other, the data are normally distributed, and there has been no structural change or shift in the underlying policy regime during the sample period.
Few of those conditions will be met. The independence requirement really implies that we need a minimum of 30 years’ worth of consistent data, plus additional years for each explanatory variable employed — and that’s for just one equation. In fact, we will be estimating systems of equations. We don’t have consistent time series that long. And regime shifts probably happen a lot. The data tend to be I(1) rather than I(0), and it’s the mean shifts that we’re interested in above all. The explanatory variables are closely related to each other. The data aren’t distributed normally. And so on. Whatever happens in this project, any claims about the ‘truth’ will necessarily be weak.
And yet.
Life is full of situations in which you need to make an assumption in which you don’t have a high degree of confidence. What you do in those circumstances is look as broadly as you can at the hints — I won’t call them evidence — that are available; you reflect on that set of information; and you conclude something, hopefully with appropriate caveats around that conclusion to reflect the uncertainty you have encountered. We do the best that we can, using as much information — statistical, theoretical, anecdotal, drawing on our cultural hinterland and our personal experience, and thinking laterally (all of which are riddled with bias and informed by the zeitgeist of the times) — as is available. And we hold in mind our uncertainty and awareness of bias while we act. We can’t do better than that.
I find that process — a loose sort of Bayesian inference, incorporating both statistical data and our prior views, however formed — fascinating. It’s the key to nearly all of our consultancy work. And the best results are obtained, in my experience, when none of the elements of the information set we deploy has a superior claim — when they are all in balance, so to speak. This is the characteristic environment in which economic research is undertaken. That’s why I love it.
One point Shiller emphasises is that a successful, contagious narrative can be a kind of ‘joke’ — like the ‘Laffer curve’ for example, drawn on a napkin by Arthur Laffer; massively influential, fundamentally misleading. Success, in the sense of contagiousness, for a narrative, is orthogonal to truth. That feels right but implies we can claim a standpoint from which we can see the ‘truth’, or at least see what is not true. How do we know what is true, or not, if all we have is competing narratives, the viral success of which has nothing to do with their claims to be true?
“The owl of Minerva starts its flight only as dusk breaks.”[3]
Minerva was the Roman goddess of wisdom. She was accompanied, and sometimes represented, by an owl. Some readings of that proverb go along the lines: wisdom comes when the day is done: with experience; with hindsight. That’s not my take, though. I prefer to think of it as implying that wisdom comes when you can’t see clearly; when everything is a bit blurry and you have to open your eyes wide; employ all the senses; call on your intuition and on half-buried memories of patterns, similar or dissimilar to those you are half sensing now. When you have to engage your whole self in interpreting the world around you.
Emma, my wife, has poor sight at the best of times, but can recognise people in poor light at a hundred paces, just from the vague shape of their bodies and the way they move. Tiny signals, massively amplified by intelligence and experience. She says it’s because “I never have any better information than that, unless I’m standing right up close”.
If we are used (as I am) to tagging fifty or so visual clues to a person’s identity, we struggle to manage with only two or three. But if two or three is all we ever get, we manage somehow — by invoking intuition and everything else. If the evidence is unclear, we have to use our wisdom, such as it is.
And it’s the part where we use our wisdom that is the most engaging. It’s what I get out of bed to do, basically. Maybe, like my son, I should take to getting out of bed at dusk.
[1] For how we work out how to cost it, see my earlier blog.
[2] Stationary series are those where both the mean and the standard deviation are constant over time. Most economic time series — GDP, consumption, CPI and some would argue even inflation and interest rates — do not possess this property. When working with non-stationary time series, it is very easy to find entirely spurious correlations. In an amusing paper (well, relatively), Robert Matthews finds strong statistical evidence that the stork population of many European countries is positively correlated with the number of human births.
[3] G.W.F. Hegel, Preface to the Philosophy of Right: “die Eule der Minerva beginnt erst mit der einbrechenden Dämmerung ihren Flug”.