I am going to Russia next week. It will be good to be back in wonderful Saint Petersburg. In connection with my trip I have been working on some econometric models for Russia. It is not exactly work that I enjoy and I am deeply skeptical about how much we can learn from econometric studies. That said, econometrics can be useful when doing practical economics – such as trying to forecast Russian growth and inflation.
So I have been working on this model for the Russian economy. The main purpose of the model is to learn about what I would would call the petro-monetary transmission mechanism in the Russian economy. It is my thesis that the primary channel for how oil prices are impacting the Russian economy is through the monetary transmission mechanism rather than through net exports.
Here is my theory in short: The Russian central bank (CBR) dislikes – or at least used to dislike – a freely floating exchange rate. Therefore the CBR will intervene to keep the ruble stable. These days the CBR manages the ruble within a band against a basket of the US dollar and the euro. Today the ruble is much more freely floating than it used to be, but nonetheless the ruble is still tightly managed and the ruble is certainly not a freely floating currency.
So why is that important for my econometric models for Russia? Well, it is important because it means quite a bit to the causality I assume in the model. Lets look at two examples. One where the ruble is completely pegged against another currency or a basket of currencies and another example where the ruble is freely floating and the central bank for example targets inflation or nominal GDP.
Pegged exchange rate: Causality runs from oil to money supply and NGDP
If we are in a pegged exchange rate regime and the price of oil increases by lets say 10% then the ruble will tend to strengthen as currency inflows increase. However, with a fully pegged exchange rate the CBR will intervene to keep the ruble pegged. In other words the central bank will sell ruble and buy foreign currency and thereby increase the currency reserve and the money supply (to be totally correct the money base). Remembering that MV=PY so an increase in the money supply (M) will increase nominal GDP (PY) and this likely will also increase real GDP at least in the short run as prices and wages are sticky.
So in a pegged exchange rate set-up causality runs from higher oil prices to higher money supply growth and then on to nominal GDP and real GDP and then likely also higher inflation. Furthermore, if the economic agents are forward-looking they will realize this and as they know higher oil prices will mean higher inflation they will reduce money demand pushing up money velocity (V) which in itself will push up NGDP and RGDP (and prices).
Now lets look at the case where we assume a freely floating ruble.
Floating ruble: Oil prices and monetary policy will be disconnected
If we assume that the CBR introduce an inflation target and let the ruble float completely freely and convinces the markets that it don’t care about the level of the ruble then the causality in or model of the Russian economy changes completely.
Now imagine that oil prices rise by 10%. The ruble will tend to strengthen and as the CBR is not intervening in the FX market the ruble will in fact be allow to strengthen. What will that mean for nominal GDP? Nothing – the CBR is targeting inflation so if high oil prices is pushing up aggregate demand in the economy the central bank will counteract that by reducing the money supply so to keep aggregate demand “on track” and thereby ensuring that the central bank hits its inflation target. This is really a version of the Sumner Critique. While the Sumner Critique says that increased government spending will not increase aggregate demand under inflation targeting we are here dealing with a situation, where increased Russian net exports will not increase aggregate demand as the central bank will counteract it by tightening monetary policy. The export multiplier is zero under a floating exchange rate regime with inflation targeting.
Of course if the market participants realize this then the ruble should strengthen even more. Therefore, with a truly freely floating ruble the correlation between the exchange rate and the oil price will be very high. However, the correlation between the oil price and nominal GDP will be very low and nominal GDP will be fully determined by the central bank’s target. This is pretty much similar to Australian monetary policy. In Australia – another commodity exporter – the central bank allows the Aussie dollar to strengthen when commodity prices increases. In fact in Australia there is basically a one-to-one relationship between commodity prices and the Aussie dollar. A 1% increase in commodity prices more or less leads to a 1% strengthening of Aussie dollar – as if the currency was in fact pegged to the commodity price (what Jeff Frankel calls PEP).
Therefore with a truly floating exchange rate there would be little correlation between oil prices and nominal GDP and inflation, but a very strong correlation between oil prices and the currency. This of course is completely the opposite of the pegged exchange rate case, where there is a strong correlation between oil prices and therefore the money supply and nominal GDP.
Do I have to forget about econometrics? Not necessarily
So what do that mean for my little econometric exercise on the Russian economy? Well, basically it means that I have to be extremely careful when I interpret the econometric output. The models I have been playing around with I have estimated from 2000 and until today. I have done what is called Structural VAR analysis (with a lot of help from a clever colleague who knows econometrics much better than me). Some of the results we get are surely interesting, however, we got one major problem and that is that during the 12 years we are looking Russian monetary policy has changed significantly.
In the early part of the estimation period the Russian central bank basically maintained a quasi-pegged exchange for the ruble against the dollar. Later, however, the CBR started to manage the ruble against a basket of dollars and euros and at the same time the CBR would “adjust” the ruble rate to hit changing nominal targets – for example an inflation target. The CBR have had multiple and sometimes inconsistent targets during the past decade. Furthermore, the CBR has moved gradually in the direction of a more freely floating ruble by allowing for a wider “fluctuation band” around the euro-dollar basket.
So basically we would expect that causality in the Russian economy in 2000 would be pretty much as described in the pegged exchange rate case, while it today should be closer to the floating exchange rate case. That of course means that we should not expect the causality in our model to be stable causal structure. Econometricians hate that – to me it is just a fact of life or as Ludwig von Mises used to say “there are no constants in economics” (I am paraphrasing von Mises from my memory). This of course is also know as the Lucas Critique. Some would of course argue that we could take this into account when we do our econometric work, but regime changes do not necessarily happen from day to day. Often regime change is gradual, which makes it impossible to really to take into account in econometric studies.
And this is one of my problems with econometrics – or rather with how econometric studies often are conducted. They do not take into account regime change and therefore do not take into account expectations. As a result well-known correlations tend to breakdown. The best example is of course the disappearance of the Phillips curve relationship in the 1970s and 1980s. Another example is the breakdown of the causal relationship between money supply growth and inflation in 1990s.
So what do I do? Should I give up on my little econometric venture? No, I don’t think so. Econometrics can clearly be useful in determining the magnitude and importance of different shocks in the economy and surely some of our econometric results on the Russian economy seems to be pretty robust. For example over the estimation period it seems like a 10% increase in the oil prices have increased the M2 and nominal GDP by around 2%. That is nice to know and is useful information when you want to do forecasting on the Russian economy. But it would be completely naive to expect this relationship to be constant over time. Rather the Russian central bank is clearly moving in the direction of a more and more freely floating ruble so we should expect the correlation between oil prices one the one hand and M2 and NGDP on the other hand to decrease going forward.
Concluding, econometrics can be useful in doing “practical” economics like macroeconomic forecasting, but one should never forget to do the homework on the institutional structures of the economy and one should never ever forget about the importance of expectations. Economic reasoning is much more important than any statistical results.
Next stop Moscow
International monetary disorder – how policy mistakes turned the crisis into a global crisis
Fear-of-floating, misallocation and the law of comparative advantages
PEP, NGDPLT and (how to avoid) Russian monetary policy failure
Should small open economies peg the currency to export prices?