2012 is starting with a lot of forecasts for the year ahead, a time-honoured and yet futile tradition. If there is one thing 2011 taught us, it’s the scope for radical uncertainty to throw any prediction off course.

Besides, correct predictions can often be unwelcome. Back in 2005-2007 I was, rather reluctantly, doing some UK macro work for a client (this is not my usual type of consultancy work). I’m a big fan of charting your data, so each quarter I prepared a book of about 30 charts, which replicated the overview I used to help prepare when I worked in the Treasury in the mid-1980s, including asset prices, house price ratios, yield curves and money and credit aggregates as well as the obvious GDP, inflation, unemployment etc. As the assignment progressed, I told them that either the benign combination of rapid global growth, low inflation and booming asset markets would continue – or we should take seriously the inverted yield curve as a signal of forthcoming recession, in which case asset prices would plunge (as they do), and by the way global asset prices including housing were highly correlated. I lacked the intellectual courage to go with the anti-consensus pessimism, because after all I don’t do macro, and the client certainly didn’t like that scenario so they picked the other one.

Anyway, all this came to mind because my eldest son is tussling with econometrics in his course and I learnt that his textbook is co-authored by my thesis adviser. Stock and Watson’s

looks a terrific text. The innovation in econometrics in the past 20 years or so has been sensational – I was just about in time for Grainger revolutionising time series work, and for the use of vector autoregressions, and had been taught a lot of that by Mark Watson anyway. Hendry and Clements are also excellent on the time series econometrics in their and earlier books.  However, although my micro-econometrics teachers could not have been bettered (Zvi Griliches, Dale Jorgenson and Jerry Hausman), I finished before the Heckman and McFadden inspired innovations became widely used.

But the trouble is that a lot of the econometrics that is done is poor quality. Not only do economics journals not publish negative results (this afflicts most sciences, I think), a lot of economists plug data into statistical packages and torture it until it delivers ‘statistical significance’. On the way, they lose sight of the underlying hypothesis, the alternative against which it is being tested, the properties of the raw data, the economic meaning of the coefficients, the effect of omitted variables and simultaneity – pervasive in macro time series data – and so on. Few have even learnt to plot their data before they start, to pick out visually the outliers and breaks in series, as I was taught.

So for any economist who takes empirical work seriously, I think McCloskey and Ziliak on

is a must, along with McCloskey’s . I’ve not read many econometrics textbooks recently. I like  Angrist and Pischke’s , which is fabulous about formulating hypotheses and counterfactuals (there’s a website too). Anyway, I’m sure my son is in good hands with his textbook.

And the moral for 2012? Think carefully before you make a prediction, and work out what to do if you’re wrong. Or, as the old joke puts it, forecast what will happen, or when, but never both at the same time.

[amazon_image id=”1408264331″ link=”true” target=”_blank” size=”medium” ]Introduction to Econometrics[/amazon_image]

[amazon_image id=”0691120358″ link=”true” target=”_blank” size=”medium” ]Mostly Harmless Econometrics: An Empiricist’s Companion[/amazon_image]