Anyhow, argument number one I dislike hearing on Twitter but can't express why easily in 140 characters: The failure to predict the financial crisis (even GFC to some) presents an existential crisis to economics. E.g.:
Economists who can't predict a GFC or the value of the Aus dollar, yet attack Scientists complex models #qanda #auspol
— steve (@steveUsername1) October 21, 2013
Not true on so many levels, but the simplest is the following. It confuses economics and forecasting, and it's very easy to argue that better economics models will not necessarily forecast better.
Forecasting depends on the thing you forecast being predictable - i.e. you could throw something in (another economic variable usually) and you'd reduce the forecast errors of forecasts over just using a random walk to predict (i.e. predict for tomorrow today's value).
So, even if we assume something *is* predictable, and we then construct a wonderful model which explains what's going on perfectly well (unlikely in economics but never mind), what happens if tomorrow a structural break happens and the variable becomes totally unpredictable? Then you've failed to forecast what happened next (other than by total fluke).
Upshot: Better economic models will not forecast better in a world that is constantly changing. So therefore economists don't need to sit around worrying about their existence because their best models generally failed to spot the crisis coming.
Clarifications:
1) This argument is based on one commonly used by Sir David Hendry, which doesn't make it right by the sheer fact of the size of his brain, but just to give credit where credit is due.
2) This argument doesn't imply I think the economics profession can instead be complacent about itself, or become arrogant again about any of its models. They are all subject to this critique. As Keynes once mused, I'd be delighted if economists were thought of as humble and competent as dentists.
3) Equally I don't make this argument because I think economics has already processed everything to be learnt from the crisis and hence is immutable to criticism related to that. Many criticisms are very valid, I just have an issue with this particular line of argument. 4) This argument is very different to the kind of argument put forward by @unlearningeconomics in his point 2 about why economists haven't got a clue in that it's obviously not based on any economics models (e.g. EMH) but instead on forecasting theory and econometrics.
Later comes argument 2: Model X/Theory Y is WRONG!
You're confusing numerical predictions, such as "variable w will be x%" with conditional predictions, such as "if we don't sort out variable y, z will do this". There are a number of recurring patterns that emerge from bubbles and business cycles and identifying and alleviating them should be a top priority for economists. Economic reality may be more malleable and difficult to pin down than physical reality but that doesn't mean it is completely random and intractable. If you believe that why do economics at all?
ReplyDeleteI should also link you to this:
http://noahpinionblog.blogspot.co.uk/2012/01/seven-principles-for-arguing-with.html
which deals with both this and your 'second argument I'm tired of hearing', too.
No, I'm not making that mistake. Any forecast based on any model is subject to this critique, since even if that model is perfect, the thing being modelled may change. If you choose to be really specific and give %s and decimal places to forecasts, you're just setting yourself up for a larger fall.
ReplyDeleteSo the issue is really that of non-stationarity (which structural breaks lead to in statistical processes), and how to forecast given that.
The conclusions drawn from the forecasting literature there are fairly intuitive: Update your forecast model frequently, and often very simple ("naive") forecasting models (change today is change tomorrow, today's level will be tomorrow's level) will work better than complicated "structural" models.
Will go and read the link before responding directly to how that counters my two points.
No, you're going to have to explain exactly how it deals with my arguments. As I point out here, the argument I'm making here has nothing to do with any economic model like EMH, it's based on the instrinsic properties of economic data - i.e. they are (in general) non-stationary.
Delete