In my previous blog post I outlined a four-equation model to be used for real-world macroeconomic forecasting. In today’s post I will look a bit closer at one of these equations – or rather I will discuss how we can think about forecasting the supply side of the economy.

So we basically want to understand how and why the (short-run) aggregate supply (AS) curve shifts in our AS/AD framework. I will discuss the different kinds of supply shocks we can think of and then I will discuss how we can use financial market data to forecast the impact of such shocks on aggregate supply.

But first let us think about how most real-world macroeconomic forecasters think about supply shocks. The theoretical economist might be surprised to learn that there basically is no supply side in the “mental models” most forecasters are using.

Rather most real-world forecasters think of supply shocks as a kind of demand shock. Lets for example think of an increase in the price of oil. In an AS/AD framework an oil price hike would shift the AS curve (for an oil importing country) to the left causing real GDP growth to drop and inflation to increase. This is basically a story of higher production costs leading companies to reduce output.

However, in the thinking of real-world forecasters an oil price hike is something which increase inflation and as a direct consequence causes real disposable incomes to drop. As real disposable income drops so does private consumption growth and given that most of these real-world forecasters are doing what I have termed national accounting economics a drop in private consumption automatically leads real GDP to drop (remember Y=C+I+G+X-M).

In reality most real-world forecasters therefore assume two things. 1) Prices are not only sticky, but fixed and 2) the monetary policy regime is a fixed exchange rate regime meaning there will be no monetary offset to what effectively is an velocity-shock. Therefore “translating” what real-world forecasters are doing into a AS/AD framework when thinking about oil prices is to think of an oil price increase as a drop in aggregate demand and at the same time assume that the AS is flat. Hence, in the real-world forecaster’s model an oil price increase shifts the AD curve to the left as illustrated below.

We see that this causes real GDP growth to drop, while leaving inflation unchanged (due to the flat AS curve). The real-world forecaster obviously doesn’t think an oil price hike will not cause inflation to increase. However, she will model that in a separate (mostly) unrelated model.

In fact as a general rule the real-world forecaster will think of the determination of real GDP as 100% demand side driven and at the same time think of inflation as 100% supply side driven. In that sense I don’t think I am wrong when claiming that to most real-world forecasters inflation is never a monetary phenomena, but rather completely driven by supply side/cost factors.

Hence, she will probably use inflation models that mostly rely on developments in wages, oil prices, food prices and exchange rates, while probably also assuming some “mean-reversion” in the sense that over 2 or 3 years inflation will converge towards a given inflation target.

This kind of faulty thinking is also what leads some economists to believe that an earthquake or war is great for the economy. What they see is the effect on aggregate demand, while completely ignoring the supply side part of the story.

What I here will argue is that we should stop thinking about supply shocks as quasi-demand shocks. A supply shock is not the AD curve moving along the AS curve, but rather the AS curve shifting left or right and moving along the AD curve.

Different types of AS shocks

To be able to practically do forecasting of the impact of supply shocks we need first to think of what kind of supply shocks we have. To do this I think this rudimentary production function is useful:

(1) Y= A*f(K, L, O)

Here Y is the level of real GDP, while K, L and O are capital, labour and raw materials (such as oil). A is what we call the Total Factor Productivity (TFP) – or said in another way how clever we are at using the three productions factors in production.

Hence, we can basically think of shifts in the AS curve as shocks to one or more of these four variables (A, K, L or O). These shocks can be permanent of temporary.

Getting practical with supply shocks

Doing real-world macroeconomic forecasting often is about dealing with practical matters rather than with theoretical issues (that is why most real-world forecasters over time turn into pretty bad theoretical economists). What I am trying achieve here is how we can maintain a sober theoretical approach to forecasting while at the same time keeping things practical.

Furthermore, it should be noted that a lot of “forecasting” is not really forecasting in the sense of forecasting the future. Rather it is about “forecasting” how for example quarterly GDP data for the present quarter will come in when they are published at a later stage. So what we really are looking for are real-time indicators. Here I believe the financial markets are extremely useful.

Furthermore, financial markets are also giving us insight into the future impact on real GDP growth and inflation of shocks.

But lets get practical and that also means that I here only will focus on two of the four different types shocks outlined above – capital shocks and raw material shocks. I focus on these two primarily because I think those are the most important and common supply shocks that we empirically are facing.

Cost of capital shocks

Often real-world forecasters think of the cost of capital (the “interest rate”) as something which is determined by monetary policy and something, which is impacting private consumption and investments and hence “aggregate demand” (in the most vulgar Keynesian form).

However, this is not what I think of here. Rather to me the cost of capital is something that determines investments and hence the capital stock in the economy. That is a supply factor.

Let me illustrate what is going on in Russia at the moment. We are seeing massive capital outflows from Russia. The real-world forecaster probably sees this as a negative shock to demand. However, that in my view is not what it is. (The capital outflows, however, can have an indirect impact on aggregate demand if it causes the central bank to tighten monetary policy).

Rather what we are seeing is a drop in the supply of capital available in the Russian economy. Capital is simply becoming less easily available and hence more expensive.

The case of Russia today also provides us with a clue to what kind of financial market indicators we can use to identify shocks to the cost of capital.

I would particularly focus on three different measures of the cost of capital.

First, what is the stock market telling us? If the Russian stock market drops relative to earnings or relative to stock markets in the rest of the world then that is telling us the cost of capital is increasing in Russia.

Second, we can also look at the price of ensuring against a Russian government default – a so-called Credit Default Swap. This is also an indirect measure of the cost of capital in Russia.

Third and finally we can also look at the bond yield spread for Russian bonds – either government or corporate bonds – versus similar bond yields in the US or the eurozone.

Again in practical terms I think it would be useful to create an index of these different measures of the cost of capital to get a single “number” for the shocks to the cost of capital in a given country. I will try to do that in a later post when I try to estimate (or simulate) an AS curve for one or more countries.

Raw material costs – it is mostly about oil prices

Shocks to oil prices seem to be the most common shock to aggregate supply across countries in the world. Furthermore, the correlation between oil prices and other commodity prices historically has been fairly high so in my view for the practically oriented macroeconomic forecaster the most easy way to capture shocks to raw material costs is simply to look at the price of oil in local currency.

Hence, in reality there will be a quite high correlation between the oil price and most other commodity prices. Therefore, for practical purposes using the oil price measured in local currency will be entirely enough to capture most of the raw material shocks to most economies.

Putting it all together

In my previous post I outlined a simple AS curve, which I think would make both economic and econometric sense for most economies in the world.

(1) y = a0 + a1*n + SS

Where y is real GDP growth (%q/q), n is nominal GDP growth and SS is a supply shock. a0 and a1 are coefficients.

Given the discussion above we can now identify SS by thinking of SS as either cost of capital shocks or shocks to the oil price. There are certainly other supply shocks out there, but for practical purposes I think that these two supply side factors “capture” the large majority of of supply side shocks in most real-world economies.

This gives us an expanded AS curve:

(1)’ y = a0 + a1*n + a2*CoC + a3*OP + a4*Ygap(t-1)

Where CoC is an index for the cost of capital and OP is the oil price in local currency (de-trended). Note that I here also have included the output gap (Ygap) lagged one quarter to take into account the equilibrating process towards “full employment” in the economy. a2, a3 and a4 are coefficients.

Next step – lets try to estimate the model

I will try to do that in my next blog post, but my readers are obviously very welcome to do the econometrics on their own on whatever country they want.

PS I use the term “real-world macroeconomic forecasters” above as a generalisation. Some might argue that this is far too much of a generalisation and that the world of forecasting has changed. That might be it and I know a lot of central bankers are terribly proud of their DSGE models, but if they look themselves in the mirror they will soon acknowledge the fact most arguments made by central bank officials are not based on DSGE thinking, but rather on national accounting economics.