I hate teaching Short Run Aggregate Supply shocks.

1. It's easy to teach them wrong.

2. I don't understand them very well.

3. I don't think anyone understands them very well.

Take the standard introductory or intermediate macro textbook. Put real income Y on the horizontal axis. Put the price level P on the vertical axis. (Or put inflation there instead, if you want). Draw a downward-sloping Aggregate Demand curve. Draw a vertical Long Run Aggregate Supply curve.

If you've only got those two curves, AD and LRAS, most of us think the model doesn't work very well in the short run. Because the model says that AD shocks will affect only P and not Y. Because the model says that fluctuations in Y can only be caused by LRAS shocks. And most of us think the world doesn't look like that.

So most of us add an upward-sloping (or maybe even horizontal) curve to the model. We call that third curve the "SRAS curve". We talk about nominal rigidities to motivate that third curve. We now think the model works much better in predicting the effects of AD shocks. Shifts in the AD curve affect Y in the short run, and only affect P in the long run.

But then we want to talk about SRAS shocks. That's when the trouble starts.

It is very easy to convince students that a short spell of bad weather causing a bad harvest could cause the SRAS curve to shift up and left while leaving the LRAS curve unchanged. After all, the weather is a short run supply shock that won't affect supply in the long run. 99% of students will happily accept that example.

But it's totally wrong. The LRAS curve tells us where the economy would be if there were no nominal rigidities. A short spell of bad weather would cause the LRAS curve to shift left for a short time, and would cause Y to drop even if there were no nominal rigidities. The long run right now is not just an average of short runs, even if the long run on average is the same as the short run on average (which it might or might not be, depending on whether the SRAS curve is linear).

Starting in full equilibrium, where all three curves intersect, what sort of shock would cause the SRAS curve to shift without causing the LRAS curve to shift?

1. Suppose a government replaces an income tax with a sales tax, or cuts the first and raises the second, in a revenue-neutral way. To a first approximation, if both taxes are roughly equally distorting, there will be no shift in the LRAS curve (yes, this is debatable, but set that aside). If there were no nominal rigidities, this change from one tax to another would have no effect on the price level. But if wholesale prices are sticky, and retail prices inclusive of tax are flexible, retailers may simply raise prices by the amount that sales taxes have increased. So the SRAS curve shifts up. The central bank should accommodate that upward shift in the SRAS curve by shifting the AD curve upwards by the same amount, to prevent a recession.

2. The classic example of the supply shock is the OPEC oil price increase of 1973. It is not obvious how this should affect the LRAS curve. It depends on whether the country is an oil importer or oil exporter. The terms of trade could either worsen or improve. It depends on whether the Y on the horizontal axis measures real output in terms of output prices or real income in terms of consumption prices. Either way, it is not at all obvious whether the SRAS curve will shift left or right, and whether it will shift left or right more or less than the LRAS curve shifts.

Take a country that is neither an oil exporter nor importer. There should be no first-order effect on that country's terms of trade and LRAS curve. All we know is that the exercise of monopoly power by a foreign producers' cartel restricting output of oil will increase the equilibrium relative price of oil. That says nothing about the effect on the average price level. It is perfectly compatible with all other prices falling, while the price of oil in domestic currency stays the same.

Here's the story I told my students. Suppose the price of oil in domestic currency is flexible, and the prices of non-oil products are sticky. Then the only way the relative price of oil can rise is if the nominal price of oil rises. So the price of gas at the pumps rises, but all other prices stay the same. So average prices rise. So the SRAS curve shifts vertically up. Again, the central bank should accommodate that upward shift in the SRAS curve by shifting the AD curve up by the same amount, to prevent a recession.

That's what I tell my students. But I am really not happy with my stories. Each one depends on a very ad hoc assumption about which prices are sticky and which prices are flexible. If I made the opposite ad hoc assumption I would get no shift in the SRAS curve.

As Brad DeLong notes, by adding a perceptive title to one of my previous blog posts, this is one reason why we need better microfoundations. If I understood nominal rigidities better, I would better understand what sort of supply shocks would shift the SRAS curve relative to the LRAS curve, and would better understand how monetary policy should respond to supply shocks.

How do you teach SRAS shocks?

[Update. Greg Mankiw responds to this post, and links to a 1995 paper (pdf) he wrote with Larry Ball that provides a theory of SRAS shocks. I had read that paper, ages ago, and should have referenced it in this post. I really like that paper. It shows a clear understanding of the SRAS puzzle, and puts forward a fully worked out solution that makes logical sense and is not obviously wrong. (In a lot of macroeconomics, "not obviously wrong" is serious praise.)

Here's the gist of their answer. Real shocks cause changes to equilibrium relative prices. By definition, since we are talking about relative prices, the mean of the distribution of equilibrium relative price changes is zero. But the distribution may be skewed right (a small number of firms should increase their relative prices a lot and a large number of firms should decrease their relative prices a little), or it may be skewed left (vice versa). Menu costs mean firms will only change their prices if the equilibrium relative price change is bigger than a threshhold. So if the distribution skews right, some firms increase their prices and the SRAS curve shifts up. If the distribution skews left, the SRAS curve shifts down.

I think the Ball-Mankiw theory is the best fully worked out theory we've got. Which makes me feel churlish to say — I still don't really like it. Especially since I can't articulate my reasons very well. But let me try. I don't see how it works in the case of changes in economy-wide indirect taxes, which seem to feed right through into shifting the SRAS curve. I can't help but think about the coordination problems in getting to the new Nash equilibrium in a monetary economy, especially when normal people (i.e. non-economists) can barely think though partial equilibrium experiments, barely understand the difference between relative and nominal prices, and think it goes without saying that of course an increase in oil prices or indirect taxes will cause inflation. And if everybody thinks it will, it will.

How the hell to teach that? "OK, students. You all think that an increase in oil prices will automatically cause the price level to rise. Because the price level is just an average, right? Well, you are all totally wrong. But, because everybody thinks like you, you are in fact right!" And how could I teach Ball and Mankiw in intro or intermediate macro? "Let me first explain the first, second, and third moments of a distribution of equilibrium relative price changes.Yes, this will be on the exam."]