The uptick in inflation has made headlines left and right. Why? For the longest time, inflation was below the European Central Bank's target rate of 2%, and now it is at twice this level at 4.1% in October 2021. The question now is: Why is inflation up? Is it the supply chain issues? Is it energy prices? Is it due to the labor market? In this post, I will present some of the disaggregated consumer inflation data for Europe, and see if we can explain the price movements.
Inflation in the News
Let's address where the discourse around inflation has taken us (for an excellent write-up see this speech by Shin of the Bank for International Settlements). The discourse around inflation explains the increase through two main channels:
First, we have the effects of bottlenecks in supply chains that happened early in 2020 and are still making headlines in the news. Here is what Shin had to say, which I think is quite appropriate:
Bottlenecks that started as supply disruptions due to the initial COVID shock in 2020 have this year morphed into something more persistent. Widespread reports of shortages have gone hand in hand with supply that appears to be running at full speed. Production of key manufactured inputs, such as semiconductors, now comfortably exceeds pre-pandemic levels. The same is true for many raw materials and freight volumes on key shipping routes. As well as supply, two additional factors are key in understanding where we are. The first is the shift in the composition of demand away from services to manufactured goods, which tend to be more dependent on the smooth functioning of supply chains than are services. The second is the endogenous behavioral changes that have given rise to so-called bullwhip effects, whereby supply chain participants react to perceived shortages by ordering more, ordering earlier, and hoarding inputs. This kind of reaction is prudent and rational when considered in isolation but can lead to aggregate outcomes that are ultimately self-defeating.
There are two aspects worth highlighting. First, there is imperfect coordination of the firms ordering goods, so each single firm is seeing the overall picture of rising prices and adding to this by buying more, thus causing more shortages and price changes. However, this is not uniform across goods, rather there are some goods that have had extreme price swings (think of lumber earlier this year) and some that have not been volatile. To this, we should also add the speculative trading that occurs in commodity markets when they get volatile. What does this mean? Inflation is differential, with some components showing strong inflation (or deflation), and others not moving much. We will show this in the data later on. A key facet of the bottleneck reason for inflation, is that this is temporary.
The second aspect is the effect of labor markets and wages. The way in which the labor market recovers from the pandemic will be partially reflected in wage growth. New wage levels will affect whether the purchasing power of households has increased, if wage growth outpaces inflation, or decreased, if wage growth is less than inflation. These are long-term structural issues, and not the heart of this post. Rather, I will now get into the consumer price index data for Europe to see about the aforementioned bottlenecks.
The Harmonized Index of Consumer Prices
The European Central Bank's (ECB) metric when considering inflation is called the Harmonized Index of Consumer Prices (HICP). To be precise, the ECB's mandate states:
Price stability is defined as a year-on-year increase in the Harmonised Index of Consumer Prices (HICP) for the euro area of below 2%.
The harmonization refers to the fact that all European Union countries follow the same method. In other regards, the index is similar to the Consumer Price Index of the United States. It is a weighted average of price indices for a greater variety of goods that us consumers purchase. The weights are proportional (roughly speaking) to the expenditure of an average household on these various subcategories. Unfortunately, these weights are updated infrequently. Thus, the overall inflation of the aggregate index may be misleading. After all, during the pandemic, the basket of goods we have bought changed quite a bit.
I downloaded the full HICP data from Eurostat, so let's have a look. Note that the HICP is also calculated for some non-Eurozone countries (e.g. Poland, Turkey, and the U.S.) to provide a comparison. It is important to note here that I downloaded the HCPI index itself (i.e. all price indices set to 100 in 2015), as opposed to the inflation data, because there is a wider coverage of the individual components. Accordingly, I had to seasonally adjust the inflation estimates myself. In doing so, I took a much simpler approach than the so-called X13 adjustment of the U.S. Census Bureau (not available in Python at the moment). Consequently, my estimates are slightly different to the official ECB values, but the overall trend and inference remain the same.
In Figure 1, I plot the time-series of inflation since 2005 for the European continent.
Focusing on the Euro Area countries, the inflation rate is only moderately concentrated around the mean inflation, with strong variation in the local inflation. For instance, in October 2021, Malta had a HICP inflation of ~1%, while Lithuania had a rate of ~7%, and the Euro Area as a whole had ~4%. The remaining countries' inflation rates are distributed uniformly between these two points. As Blair Fix noted recently, inflation is differential, in this case also across countries and not just goods. This means that the average inflation, such as the Euro Area HICP inflation, may not be reflective of what is happening in the components (or in your own consumption basket for that matter). In Figure 1, I split this by country, which already revealed a large range of behavior dependent on which countries are under consideration, let's look at the individual components next.
Not only are there significant differences in inflation between the European countries, the inflation in individual HICP components varies strongly within each of the countries. The HICP is divided into multiple different category levels, I took the third level of disaggregation. In Figure 2, I show the time-series of the three-digit HICP components, which include categories like: vegetables (CP01.1.5), motor cars (CP07.1.1) or games, toys and hobbies (CP09.3.1). Again we note that inflation is differential, and the aggregate index hides this variation in sub components. As such, looking only at the index can be strongly misleading.
For all of the countries shown there are a couple of components which have shown extreme variation. In the Eurozone, Germany and to a lesser extent France, there are roughly four components that have inflated sharply. In contrast, for the second row of countries including Spain, Poland, and Latvia, there are both more components with extreme price changes, as well as a wider spread of components overall (thicker band around the HICP inflation in black).
Energy prices have risen distinctly across countries, and make up most of the extremely inflationary components of the HICP. In Figure 3, I have highlighted the difference in the energy component inflation (red) compared to the overall CPI inflation when excluding energy.
To show the energy components in more detail, in Figure 4, I've colored only the five series relating to the energy components of the HICP for the same countries as Figure 2. As Brian Romanchuk just noted, "wholesale energy prices have undergone a correction, which will blow a hole in the rate-of-change of the energy component of the [consumer price index]". This is precisely what we see here. COVID-19 led to a steep decrease in energy usage (e.g. less driving to and from work, minimal air travel), which led to a low price. Right now, energy usage has increased sharply with the economic recovery and reopening, there is a cold winter, and there was less OPEC supply increase than anticipated. Since there isn't a quick-to-adapt spare capacity, prices are again skyrocketing (a longer thread by the International Energy Agency here).
What does this mean going forward? Historically, household energy demand has been inelastic. This means, even when prices rise people tend to consume the same amount of fuel. For instance, they continue to drive to and from work on a daily basis, and continue to heat their homes in winter. Therefore, if prices of energy including gasoline increase, then a larger proportion of the household's spare budget goes towards purchasing energy, which leaves less to spend on other goods and services. If high energy prices persist, we would then expect a drop in the demand for non-energy goods, such that less items are sold and prices for these goods may stabilize.
Another implication that is often overlooked is that energy is an essential input to production. Accepting the survey data, we know that firms typically price goods with a markup over the cost of their production. If it costs the publisher 20 dollars to produce Kalecki's Theory of Economic Dynamics, they might charge me 25 dollars, representing a 25% markup. If the markup rate is constant, then increases in the unit cost due to energy or commodity price surges will translate into higher goods prices. Thus, one effect of the inflation of energy prices and commodities may be overall inflation due to increased unit costs, in particular the material's heavy manufacturing sector would be affected by this.
The Remaining Factors
Now that we have isolated the energy components, where does the remaining spike in the overall inflation rate come from? The answer seems to be durable goods. The inflation rate of services, like your Netflix subscription, have remained at a steady rate of between 0% and 3% (see Figure 4), on the other hand, durable goods like furniture or cars have gone from a stable or deflationary trend to a strongly positive push. In Figure 4, you can see how the dynamics of non-durable goods (blue) and services (green) have remained relatively stable, while the durable inflation (red) has seen a strong increase in 2021. This has, naturally, increased the overall index. Intriguingly, this is also in contrast to the last several decades where goods prices have increased at a much lower rate than services.
Why we should be wary about indices
Now that we have an overall impression of what is going on, I am here to tell you why we shouldn't trust aggregate HICP data too much. Specifically, there are some measurement biases that we should be aware of. These are primarily due to the fact that the weights for calculating the aggregate CPI are updated so infrequently:
- Substitution bias: The consumption basket changes as prices change. If the prices of some goods become extremely high, we might switch to using different goods. Thus the weights that were relevant in 2015, may, due to inflation, not be as relevant today
- Quality change bias: The prices of some goods might reflect different iterations of these goods that have a superior (or inferior) quality that justifies a higher price for the good. It is unclear what portion of a price increase should be attributed to the fact that we have a better item.
- New product bias: There are tons of new products introduced to us frequently. Often these are expensive at first, with price reductions as sales and improvements are made. Eventually, the price reaches a low value, and then increases again as (potentially) the firm gains in market share (monopoly power). Thus the moment that the product is introduced into the index matters for whether we see inflation or deflation. This is particularly important with the wealth of new gadgets that are being rapidly introduced and made oblivious nowadays.
- Formula bias: Price data is collected for individual items and aggregated by the statistical office in a complicated procedure. This includes whether items are on sale for instance, such that some items might see inflation simply because they are no longer on sale. Also the outlets from which price data is collected is changed, so this itself might have price effects.
Finally, each of us has a personal inflation rate based on what we purchase and where. These indices are based on weights relative to the total personal consumption expenditures by categories for the European Union (I haven't checked the exact weighting function here). For instance if you only buy food items at the discount store and take public transport, you are likely to personally experience much lower inflation. Thus the overall HICP inflation, may not reflect your personal experience at all