My earlier article on price indices - The Price Level Does Not Exist - was tentative, and needed a more detailed explanation of my thinking. This article is an attempt to give a more detailed background, and I am tying it to inflation targeting. It should be noted that this is not just a jab at mainstream economics, it also applies to some post-Keynesian thinking. For example, one might summarise the policy goals of Modern Monetary Theory (MMT) as "full employment with price stability" - which is somewhat awkward if we cannot define a price level that is supposed to be stable. (This observation about MMT came to me during the recent MMT conference, and the article reflected my on-the-spot thinking.)
Long History Of Attacks On Price Indices
The problem with the snappy title of my previous article is that it blunders into a fairly ancient set of controversy in economic theory. I knew this, but I could not longer remember who wrote what about the topic. I noted that I am not attempting to write an academic article, and I did mention the most corrosive attacks on the existence of a price level by the Austrian economists.
Luckily, I have well-informed readers, and they did some of the requisite literature survey for me. (Thanks!) I would highlight the article by Nathan Tankus, "Are General Price Level Indices Theoretically Coherent" (spoiler: no) which fills out the historical background of my critique from a Keynesian viewpoint. His comments on modern policy comments reflect my thinking:
This is a well known set of issues; yet strangely heterodox economists have not really taken it up. You can still find plenty of heterodox economists uncritically speaking of "inflation". A fruitful avenue of future research would be reorganizing data to be coherent from a heterodox economics point of view. From here we can ask what different policy implications would heterodox economists derive from multiple price indices, including multiple indices of asset prices?
Naïve Theoretical View
The perspective I was writing from was that I usually take, starting from the theoretical hodge-podge that is financial market economics. (It may sound "mainstream," but I doubt that any mainstream academics want to take the blame for the market economic framework.) The common sense attitude one would take is this: "Sure, the Keynesians/Austrians came up with these theories decades ago saying that we cannot use price indices, but we use them all of the time. What's up with that?"
This is yet another area where the mainstream completely ignores heterodox critiques, and only someone with a knowledge of the history of economic thought would know what the mainstream response is. Since I have no idea what justification the mainstream uses to justify the use of unitary price indices, I am just going to look at the problem from a starting point that is close to where the mainstream is, and illustrate the issue for the reader using familiar terms.
Modelling the 1970s
Imagine that we are building a simple economic model that captures the experience of most developed economies in the 1970s. The model is aimed to be largely aimed at the domestic economy, so it will be an (almost) closed economy model. (I am ignoring the real issues that European countries had with their currency regimes.) I am not attempting to describe the entire model, only the model variables that refer to inflation.
- The first variable is the price of oil, which we will assume is largely set in world markets, that is, external to the model of the domestic economy. (The United States is an extremely large consumer of oil, and thus it mattered on the demand side. That said, the OPEC cartel did have a limited ability to stick the price of oil where it wanted.)
- We will assume that we can aggregate all wages in the domestic economy into a single variable (which matches the usual assumption). Since we will assume that this model is either monthly or quarterly, the reality is that this series should react to most economic variables relatively slowly (since wage agreements typically have a span of a year). Since there was a fair amount of indexation of wages in that period, there would be a component of wages that will react mechanically to the (historical) development of the model's aggregate consumer prices. As I discuss below, the assumption that we can do this aggregation is heroic.
- We will assume that domestic non-energy retail prices are largely determined by a markup over the wage bill, which matches the reality of how almost all retail prices are set. There may be a small mechanical leakage of energy prices into these domestic prices.
- If we do model imports, we will assume that foreign exporters are price takers, and thus conditions in foreign economies will have no effect on non-energy prices.
- The model's aggregate consumer price index is constructed in the usual fashion, aggregating the energy and non-energy retail prices.
To recap, there are three underlying variables associated with prices.
Energy prices (determined outside the model).
Domestic non-energy retail prices. (This corresponds to "core CPI" in this model.)
(There is a derived variable - the domestic retail price level - that is determined by non-energy retail and energy prices using an index formula.)
This setup is not greatly different than that of most mainstream macro models. For now, the only apparent difference of note is that energy prices have shown up, and the language around price markups may shock hardcore believers in general equilibrium.
If you are a believer in the importance of expectations in the determination of the price level, I would point out that they would show up in how markups are determined by businesses. Even in the 1970s, businesses generally avoided constant price changes (the exceptions would be things like retail gasoline, which is showing up in the energy price component of this hypothetical model). They would therefore set current prices in a fashion that incorporates the expected rate of inflation of their input costs.
The difference between my description of these variables and standard mainstream models is that they are vague about how prices are determined ("equilibrium"), and they largely assume away the business sector's free will in determining markups. In the mainstream story, markups are determined mechanically by marginal productivity, and thus we can essentially eliminate either wages or domestic prices from the model dynamics. They choose to eliminate wages, and so we end up with the familiar story that everything revolves around the aggregate consumer price index.
What happens in this hypothetical model when we try to simulate the 1970s? We would want to see how the model economy reacts to an upward spike in the price of energy. The interesting part of the model dynamics - which is different than the current environment - is that wage indexation will raise the domestic wage index. Unless businesses are willing to absorb this in their profit margins, domestic non-energy prices get marked up. Depending on what behavioural assumptions we stick into the model, this should provide a fertile ground for an inflationary outcome.
Even this simple framework - which is extremely close to what is used in conventional economic models - it is hard to see what inflation targeting means. What is the central bank supposed to be targeting? The overall consumer price index? Ex-energy ("core") consumer prices? Domestic wages?
Nobody serious believes that central banks (possibly excluding the one for Saudi Arabia...) can set the level of energy prices. We can try to label a rise in energy prices "a relative price shock" (I am fairly sure I have written that myself), but a factor of 10 increase in oil prices goes beyond that. Until we have some experience with such shocks, we have no idea how much this oil price rise will mechanically feed into the overall retail price level. (We might have such omniscience in a model, but that is not a realistic feature that we can base policy on.)
We can try looking at ex-energy prices, which appears to be the consensus view (outside of the European Central Bank, which was proud to hike rates in 2008 in response to the oil price spike). However, I believe that this strategy was evident in the 1970s, where all sorts of prices were being excluded from consideration due to "special factors." (Anchovy prices were a big deal.) If we construct price indices to exclude all of the things that are going up, we tend to miss inflationary events (which probably explains the ECB's attitude in 2008).
Furthermore, a lot of "core" prices are things that central bank policy can do little about. In recent years, the major drivers of core inflation have been tuition and medical prices - which are prices set in markets that are far removed from the mainstream ideal of competitive equilibrium. We then run into the question of needing to split up retail prices into cyclically sensitive components, and those that are insensitive (e.g., medical and education components of the CPI).
Finally, we end up with the solution that is closest to practice: the central bank de facto targets the general wage level. This is probably the only variable that monetary policy (or fiscal policy) can hope to influence.
Why do we have central banks targeting consumer price inflation, and not wages? Politics seems to be the only answer. It's a lot more acceptable for the central bank to say that it is trying to keep down the prices you (the voter) pay than it to tell you the truth - it does not want your wages going up to quickly. It also appears gives the central bank a free pass on the whole question of the division of income into shares between capital and labour.
This political expediency probably explains why say that central banks are targeting something they really cannot directly influence, and we end up with muddled models.
One interesting point to keep in mind that one of the major selling points of modern mainstream macro (e.g., DSGE models) was that they could explain the 1970s inflation. However, these models would have a very hard time trying to deal with the moving parts of my minimally realistic model setup, with the three price components.
It Gets Worse...
One might argue that I have really only added oil prices (since everyone knows that markups are entirely determined by marginal productivity). However, the minimal model set up described above was too simplistic.
It is not difficult to see that the jobs market is not homogeneous. At the minimum, there is a divide between skilled and unskilled workers (with the definition of "skilled" being "background that fits employer skill demands exactly"). As the well-known inequality studies have shown, wage growth is not uniform between these groups.
In "Policy and Poverty", Hyman Minsky discussed the effects of such a split using a model by Baumol. (Reprinted in Ending Poverty: Jobs, Not Welfare.) It should be noted that Minsky observed the inflationary effects of the old Keynesian policies in real time, and attributed them to this split in the labour market.
Even splitting the labour market in two causes modern mainstream macro to break down. Economic agents are somehow supposed to have expectations for all economic variables on an infinite time horizon, and these expectations are to be brought into equilibrium. The key assumption is that every household is the same. When we split the labour market into two, how exactly is that assumption going to work?
Furthermore, we see that wage growth differentials between groups are persistent; hot sectors or regions typically continue to run hot. How can the central bank manage expectations about average wage growth, if no sector is growing at the aggregate average?
Anyone attempting to micromanage inflation is in a fog. Different sectors have differing levels of productivity; at what point are wage gains unacceptably high?
Backwards-Looking Inflation Targeting
This problem is largely untenable. However, inflation targeting can work - so long as we ignore mainstream economic theory. All we need is a central bank that reacts to historical aggregate price data. If realised inflation numbers are "too high," it just hikes rates until it causes a recession. We just need to rely on the historical tendency for recessions to drive down prices across the board, and we do not really need to know what is happening at the sector level.
This makes inflation targeting look less politically palatable. Does it make sense to induce a recession if consumer price inflation creeps up to 3%? Central banks need the political cover of pretending to painlessly micromanage expectations, and not say what they are really doing.
If we cannot rely on aggregate price indices, how policy can interact with "inflation" is not particularly clear.