MV=PQ, and issues with GDP accounting
WARNING: this is me in full thinking-out-loud mode. Please take it in that spirit. I would, in fact, be positively delighted to be shown this isn’t something I should worry about.
MV=PQ is a tautology. Of itself, it places no constraints on reality and is merely a device for organising our thoughts. But if you start empirically defining the variables, then it becomes useful. For example, it seems to me that there is an M pretty amenable to definition – the monetary base. And whilst a fair amount of electrons have been spilled on the blogosphere about the impossibility of defining P, (almost) everyone seems to agree that PQ can be defined – as NGDP. V is then left over as whatever it is than happens to the monetary base in order to produce PQ. NGDP-targeting fans such as myself say the central bank should set a rule such that it will manipulate M (and, most importantly, our expectations about future path of M) in such a way as to produce a certain level of NGDP. Whilst we can’t observe V empirically and can only infer it from M and PQ, we know the kinds of things that affect it just be thinking how the economy works (the banking system, the payments system, anything that affects demand for money relative to goods and services etc.). So when we see changes in those things, we forecast changes in NGDP (assuming constant M), and NGDP-targeting advocates like me think the Fed should pinky swear to offset these changes by producing more or less M, if necessary. We don’t need an independent way of accurately measuring V, we just need to know the kinds of things that affect it in order to stabilize the forecast for PQ, which itself stabilizes PQ. Isn’t theory beautiful.
…But, I have learnt to always be VERY wary when dealing with tautologies. One minute you can be saying something interesting and useful, and the next minute you can be saying, well, nothing at all. So, bearing in mind this is me thinking out loud, what if it is the case that we can’t define PQ? For example, because NGDP does not factor in depreciation, whether you count something as an intermediate good or a final investment good seems to affect NGDP (because you count intermediate goods being used up, but not capital). This distinction does not seem to me to be an economically relevant one, and has big implications for what counts as ‘PQ’.
So, suppose I’m right and there is no economically relevant distinction between using up intermediate goods and depreciating the capital stock, or that there is no way of drawing a line between the two in reality (completely prepared to be proved wrong on this). What has happened to MV=PQ? Well, we can define M and… er… that’s it. To be consistent, we can either start counting (some) intermediate goods as final investment goods, or we can start factoring in depreciation of final investment goods. Whichever way we go, we have some extremely nasty conceptual issues before we even get into the question of making oranges equal to Apples for real GDP/price level disentangling purposes.
I’m inclined to say that there is no fact of the matter as to what NGDP is, because there is no fact of the matter as to what counts as ‘final goods and services’. And I’m also inclined to say that it is decidedly non-obvious that NGDP is the relevant nominal expenditure figure for macroeconomic stability, as we’re getting into the issue of what exactly it means for ‘the economy’ to be stable. When I have been saying things like ‘recessions are caused by an increase in the demand for money relative to goods and services’, I thought I had a pretty tight concept of what goods and services meant. Now I’m not so sure.
The reason this worries me is that the statistical organisations have some kind of way of producing a final number, and I haven’t got the foggiest idea how they can possibly do it. The measurement issues alone boggle the mind, but there seem to me some fairly profound questions about what it is they are trying to measure at all. Which leaves me with my biggest worry: that when trying to calculate a statistic with both deep conceptual issues and high potential for mismeasurement, even if you can arrive at an empirical definition of the concept, you will (consciously or otherwise) start making the tricky judgement calls in such a way that the result fits with your prior understanding of the world. And that, my friends, would be a problem.