How to think about supply shocks

In a previous post, I argued that more than 100% of the inflation since late 2019 has been demand side. There were some adverse supply shocks around 2021-22 that led to significant inflation, but there have also been major positive supply shocks (notably immigration) that have tended to depress inflation. In net terms, the cumulative inflation is all demand side

I regard nominal GDP growth as a useful proxy for the contribution of demand. Because real GDP tends to rise at about 2%/year, on average, a 4% NGDP growth rate is a useful benchmark for appropriate monetary policy. Since late 2019, there’s been roughly 11% cumulative excess NGDP growth (i.e., above 4%), which can more than fully explain the roughly 9% cumulative excess PCE inflation (above 2%).

Most economists clearly don’t look at things this way. Most economists seem to regard the high inflation of 2020-24 as resulting from a mix of supply and demand shocks.  A recent San Francisco Fed working paper by Adam Hale Shapiro provides a decomposition of supply and demand side inflation this is broadly consistent with estimates I’ve seen from a number of economists:

Notice that both negative supply shocks and positive demand shocks play a major role, with negative supply shocks being especially important for headline inflation (which includes food and energy prices.)

Shapiro uses an interesting technique to tease out the contributions of supply and demand shocks:

Since inflation is constructed as the weighted sum of category-level inflation rates, it is straightforward to divide inflation by category, or groups of categories. I separate categories each month into those where prices moved due to a surprise change in demand from those where prices moved due to a surprise change in supply. The methodology is based on standard theory about the slopes of the supply and demand curves. Shifts in demand move both prices and quantities in the same direction along the upward-sloping supply curve, while shifts in supply move prices and quantities in opposite directions along the downward-sloping demand curve.

To say I have mixed feelings about this is an understatement.  I strongly support the technique of looking at co-movements of prices and output to identity supply and demand shocks, but I strongly oppose making inferences about aggregate price changes by aggregating sectoral price changes.

One of my first published papers (JPE, 1989, co-authored with Steve Silver) looked at real wage cyclicality.  We tried to estimate how real wage cyclicality depended on whether the economy was hit by supply shocks or demand shocks.  We identified these two types of shocks by looking at periods where prices and employment went in the same direction (demand shocks) and periods where prices and employment went in opposite directions (supply shocks). So I’m completely on board with that sort of identification strategy.  One can also compare changes in inflation with changes in real GDP growth rates.  Indeed my view that 2019-24 is all demand side inflation is due to the fact that growth was above trend—both prices and output were moving in the same direction.

Shapiro looks at price and output data for more than 100 categories of goods and services.  This is the part I don’t agree with (or perhaps don’t adequately understand.)  In any complex economy, some markets will show positive price/output correlations and some markets will show negative price/output correlations.  I fear that this technique will lead to overestimates of the role of supply, as even in an economy where 100% of inflation was demand generated you would find individual markets with negative price/output correlations (indicating supply shocks.)

Consider a thought experiment with an economy featuring stable but high rate of inflation, generated by fast money growth.  Also assume the public has become used to the rapid inflation, so wage and financial contracts factor in the inflation.  I.e., assume that money is roughly neutral.  You could imagine an economy where the money supply doubled every 12 months, and all wages and prices rose at a similar rate.  Output is (by assumption) at the natural rate.  By assumption, this would be an economy where almost 100% of inflation is demand side (from monetary policy).  And yet the price/output correlations would vary a great deal between sectors, as you would still have all sorts of changes in relative prices due to a variety of local supply and demand shocks.  In other words, the factors that affect relative prices in individual markets are radically different from the factors that affect the overall price level (monetary policy in this case, although velocity is another possibility.)

Shapiro directed me to a new study of Turkish inflation that leads me to believe that my thought experiment is more than just a hypothetical concern.  Before considering their study, think about how much inflation is likely to result from supply side factors.  If monetary policy generates 4% NGDP growth, then you will end up with 2% inflation if output grows at its 2% trend rate.  But if adverse supply shocks reduce output growth to negative 1%, and NGDP continues growing at 4%, then inflation will rise to 5%.  Thus I have no problem with the claim that supply shocks could briefly push inflation 3 percentage points above trend.  But what would it take for supply shocks to add 30% or 50% to a nation’s inflation rate?

The Turkish study by Okan Akarsu and Emrehan Aktu ̆g produced this graph:

Notice that Turkish inflation peaked at about 80% in 2022, and generally runs well ahead of the US.  Also note that the proportion attributed to supply and demand shocks is similar to the estimates shown in Shapiro’s graph for headline inflation.  You might think that fact is not surprising–the Turkish authors used a similar model—citing Shapiro’s work.  But I’d expect the contribution of supply shocks in an absolute sense to be relatively similar in the two countries—say low to mid-single digits.  Then if Turkey has a monetary policy that generates extremely high NGDP growth, I’d expect almost all of the inflation in Turkey to be demand side.

Here’s the abstract of the Turkish paper:

We document the demand and supply-driven components of inflation in Turkiye by following the decomposition method of Shapiro (2022). The results suggest that the recent hike in inflation, which started with the Covid-19 pandemic but deviates significantly from global inflation rates, was initially driven by supply factors, but over time it transitioned into an inflationary environment predominantly driven by demand forces. Consistent with theory, oil supply and exchange rate shocks increase the supply-driven contribution, while monetary policy tightening reduces the demand-driven contribution to inflation. This decomposition can potentially serve as a useful real-time tracker for policymakers.

Perhaps the phrase “exchange rate shocks” is one source of disagreement.  In my thought experiment where the money supply doubled each year, I assumed that wages and prices also doubled.  And one very important price is the price of foreign exchange—aka “the exchange rate”.  Thus one year it might take 100 Turkish lira to buy a US dollar, then a year later 200 lira, then 400 lira, then 800 lira.  I suppose that could be viewed as an “exchange rate shock”, but to me it is just one aspect of demand side inflation—which pushes all prices higher, including the price of foreign exchange.

We are so far apart that I wonder if the problem here is terminology.  The terms “supply” and “demand” were developed to explain relative price changes in specific markets for goods and services, not aggregate price changes.  There’s always been a split between those who prefer to think about inflation as depreciation in the purchasing power of money, caused by shifts in money supply and demand, and those who think about inflation more in terms of the sum of individual price rises, caused by supply and demand factors in a wide range of markets.  I’m on the monetarist side of that divide.

For the concept I’m interested in, we might be better off using entirely different terminology.  Thus I could use the term “nominal inflation” for any variation in inflation associated with variations in NGDP growth.  And I could use the term “real inflation” for any variations in inflation caused by real output changes, holding NGDP constant.  Of course, these terms would then merely represent accounting, and have no causal implications.  I do think that NGDP growth is ultimately determined by monetary policy (including monetary policy errors of omission), but that sort of causal claim does require evidence, it’s not just a tautology.

In any case, I might be missing something obvious here, and would be interested in how other view claims such as the estimate than half of Turkey’s 80% inflation in 2022 was supply side.  Does that seem plausible?  If so, what’s your definition of “supply driven”?

I’ve never seen a clear definition of supply side and demand side inflation.  In the absence of a consensus view, each empirical study of the question becomes a de facto definition.  Perhaps there’s no real debate at all, just differing definitions.

PS.  There is a method that makes supply inflation seem even lower than my estimates.  There’s an argument that any increase in real output tends to depress prices.  Thus if RGDP rises by 2% and NGDP rises by 4%, you could argue that the supply side has depressed the price level by 2%, ceteris paribus, and the demand side has raised prices by 4%, yielding 2% net inflation.