One of my favourite journal paper titles is Xavier Sala-i-Martin’s AER paper ‘I just ran two million regressions’. The problem that paper tries to deal with is that there are too many potential variables that you could conceivably put in an equation explaining differences in economic growth rates among countries. There is then a serious danger of (intentional or otherwise) data mining. A researcher may want to establish that their pet new variable is important in determining growth, so they try lots of different regressions. When one set of additional variables are included the pet new variable is significant, but when another set is used it is not. Only the first group of regressions are published. Sala-i-Martin’s paper uses techniques that involve looking at all possible permutations of variables, in order to try and assess which are robust, in the sense of tending to be significant whatever else is in the regression.
A recent ECB working paper by Ca’Zorzi, Chudik and Dieppe does something similar with models of the medium term current account. Why is this important? In my view it’s a key ingredient in being able to say something about exchange rate misalignments. This idea is associated in particular with the work of John Williamson, who christened the approach Fundamental Equilibrium Exchange Rates, or FEER for short. (That led to probably the best title of any of the papers I have co-authored – ‘Are Our FEERs justified’ – where we test the FEER approach against PPP.) John’s most recent analysis, co-authored with William Cline, can be found here. This or very similar approaches often go by different names: in Peter Isard’s nice survey it is called the macroeconomic balance approach, and it continues to be used (along with other methods) by the IMF.
The idea behind the FEER approach is to model trade flows as a function of the real exchange rate and activity levels. In the medium term activity levels will be determined from the supply side i.e. the output gap will tend to zero. So if we think we know about this supply side, and we know what the current account will be in the medium term, we can back out the medium term real exchange rate. We can then form a view about the extent to which current exchange rates are misaligned (or, more precisely, what expected interest rate differentials would have to be to justify current exchange rates). I’ve used this approach on a number of occasions in the past: perhaps most notably, to try and assess what Euro/Sterling exchange rate the UK should have entered the EuroZone at if it had decided to join in 2003.
The main problem with this approach is working out what the medium term current account should be. Actual current accounts are a poor guide, because they are influenced by both noise and short term factors, like the economic cycle and currency misalignment. In long term equilibrium it is reasonable to assume that the current account should be zero, because the current account is the change in national wealth. However we know that current accounts can show persistent surpluses or deficits over many years. Intertemporal consumption theory gives us some ideas, but on its own it is not that helpful. Many other factors may matter, such as countries having different demographic profiles. With no clear encompassing theory to use, empirical studies of the kind cited above may be our best guide.
Incidentally, the New Open Economy Macro (NOEM) approach, which is currently the most widely used microfounded open economy framework, essentially uses the same idea as the FEER: see for example this study by Obstfeld and Rogoff. It is more concerned with microfoundations, and less with data, but it shares with the FEER approach a focus on imperfectly competitive markets for internationally traded goods. As far as I know these authors have never acknowledged Williamson as a precursor, and I’m not sure why. As a result, many macroeconomists think NOEM invented this way of thinking about medium term exchange rates.
The details of which variables the authors of the ECB study find are important in determining medium term current accounts are probably not of wide enough interest to discuss in this post. What is more topical is that they use their robust models to estimate what underlying current accounts currently are for the US, UK, Japan and China. Perhaps unsurprisingly they find that, although the US would be in deficit and China in surplus, the numbers are much smaller than the deficits and surpluses observed in the recent past. More controversial, perhaps, is that they find Japan should also be running a deficit. In the past I and others have tended to assume surpluses for Japan, but this was always partly based on demographic features which were coming to an end, which is maybe what has now happened.
One slightly disappointing aspect of the study is that they did not look at Germany. There is some debate about the extent to which German surpluses represent a temporary misalignment of real exchange rates within the Eurozone, or whether they may be partly structural. The answer is rather important in assessing the extent to which deflation is required outside Germany, and it would have been very interesting to know what this study had to say on this issue.