View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Home / Publications / Research / Economic Brief / 2024

A Model-Based Perspective on In ation and the
Distribution of Relative Price Changes
By Alexander L. Wolman and Francisco Ruge-Murcia

Economic Brief
September 2024, No. 24-30

Key T akeaways
From 1995 to 2020 (the prepandemic period), there was a close relationship
between monthly in ation and the cross-sectional share of relative price
increases, which is a measure of asymmetry.
We estimate a 15-sector DSGE model and can match that relationship fairly closely.
According to the model, the aspect of the data that drives this relationship is
heterogeneity in the volatility of shocks to supply or demand across sectors.

A series of articles starting in 2022 has discussed the empirical relationship between
in ation and the distribution of relative price changes: In the stable regime from 1995 until
the pandemic era, the monthly in ation rate was closely related to a measure of asymmetry
or skewness in the distribution of relative price changes. In this article, we describe related
research that uses a dynamic macroeconomic model to study how in ation is jointly
determined by monetary policy and "relative price shocks," as well as other shocks.1 We use
that model to help us understand the factors that lead to the relationship emphasized in
previous articles.

The Empirical Relationship Between In ation and the
Distribution of Relative Price Changes
Macroeconomics has a long tradition of viewing large price changes for particular
categories of goods and services as especially important in determining the in ation rate
over short periods. T his thinking underlies the popularity of core in ation and (more
recently) trimmed mean in ation:

/

Core in ation removes the contribution of food and energy — which usually are
categories with high volatility of price changes — from the in ation rate.
T rimmed mean in ation removes a xed share of categories, those with the highest
and lowest price changes (with the categories varying across periods).
Instead of focusing on one of these alternate measures of in ation as a way of controlling
for large price changes, we look at a statistic from the distribution of relative price
changes. Before explaining the statistic, we need to provide some basic information about
the distribution of relative price changes.
For any category of good or service, the nominal price change is the change in the dollar
price of that item. T he relative price change is the change in the "goods" price of that item,
where the goods price means the price of the item relative to the price of the overall
consumption basket.2 Because the change in the nominal price of the overall consumption
basket is the in ation rate — and the in ation rate is also the average nominal price
change — it follows that the average relative price change is zero every month.
T he fact that the average relative price change is zero has a notable implication for the
distribution of relative price changes. In each month, we can split the distribution of
relative price changes into positive and negative changes:
T he positive changes are for the items whose nominal price changes are greater than
the in ation rate.
T he negative changes are for the items whose nominal price changes are less than the
in ation rate.
If the expenditure share of items with positive relative price changes is very large (close to
1), then it must be that the average of those positive relative price changes is small
compared to the average of the very small share of relative price decreases. If this were
not the case, the relative price changes as a whole could not average out to zero.
T o summarize: By the de nition of relative price changes, the average relative price change
must be zero. And because the average relative price change is zero, there is a tight
relationship between the share of expenditures with relative price increases and the
average size of relative price increases compared to the average size of the relative price
decreases. T hat relationship says that the higher the share of relative price increases, the
lower the ratio of the average relative price increase to the average relative price decrease.

How the Share of Relative Price Increases Can Help Us
Understand In ation
How is this useful? We began by talking about the potential role of large price changes in
accounting for monthly in ation rates. T he second property of the relative price change
distribution suggests that the share of relative price increases in a given month could be a
/

good summary statistic for the role of large price changes in a given month. And to the
extent these large price changes are important for in ation, then the summary statistic
may be useful for understanding in ation.
Figure 1 bears out this logic. T he scatterplot comprises dots representing monthly data
from January 1995 to February 2020. T he vertical axis is the monthly in ation rate
(measured in percent), and the horizontal axis is the monthly share of expenditures with
relative price increases (nominal price increases that are greater than that month's
in ation rate).3 When the share of relative price increases is high, the in ation rate tends
to be low. And we know from properties of the relative price change distribution that these
are months when the average relative price increase is much smaller than the (absolute
value of the) average relative price decrease.
It's important to emphasize that Figure 1 covers a period when in ation and the monetary
policy regime were generally stable. Over periods when those stability conditions do not
hold, it would be surprising if a tight relationship exists like in Figure 1.

Enlarge
In previous articles, we used the relationship in Figure 1 as a benchmark against which to
interpret data from the pandemic/postpandemic period beginning in March 2020.4 T o the
extent that high in ation readings were consistent with the prepandemic relationship to

/

the share of relative price increases, they were less worrisome than a generalized upward
shift in the relationship. In this article, we concentrate on the prepandemic relationship
and interpret it through the lens of an economic model.

Description of the Macroeconomic Model
It may seem obvious that large price changes would be important for determining the
in ation rate, but this is not a foregone conclusion in many standard macroeconomic
models. T o see this, we only need to note that it is feasible for monetary policy to perfectly
stabilize the in ation rate in those models.5 If monetary policy can perfectly stabilize
in ation regardless of the shocks to demand and supply for di erent categories of goods
and services, then it follows directly that in ation can be unrelated to the distribution of
relative price changes. Although the relationship in Figure 1 does not necessarily arise in a
macroeconomic model, because it is present in the data it ought to be present in the
models we use for policy analysis. In a particular model, one can then ask what features are
important for replicating the empirical relationship.
We proceed in this manner. In economics jargon, our model is a multisector New Keynesian
model. T here are 15 categories of consumption goods and services, and the producers of
those items face a cost to adjusting their prices. T his means that prices do not
instantaneously adjust fully to shocks, and it generates a role for monetary policy.
Monetary policy is characterized by a so-called "T aylor rule," which sets the short-term
interest rate as a function of in ation and output. T he model's policy rule also includes
interest rate smoothing, which means that the interest rate only partially adjusts toward
its T aylor-rule target each period. An important feature of the model for matching the data
is the rich set of sectoral shocks: In each sector, there are shocks that a ect both the
supply and demand for that sector's output.

What Features of the Model Are Important for Matching
the Relationship?
We estimate the model by maximum likelihood, using U.S. data on prices and quantities for
15 categories of consumption. Because these categories are broader than the 206
subcategories used to plot Figure 1, Figure 2 plots the analogous relationship based on 15
categories. It is of course not identical, but it shares the same broad properties as Figure 1.

/

Enlarge
In Figure 3, we plot simulated data from the model, using the estimated parameters. T hat
is, we simulate the model for the same number of months as plotted from the data, and for
each month we plot the in ation rate against the share of expenditures for which the
relative price increased in that month. T he simulated data has fewer observations close to
zero and 1, but the relationship between in ation and the monthly share of relative price
increases is similar in the model and in the data.

/

Enlarge
Next, we investigate what features of the model are important for generating this
relationship. In the 1995 paper "Relative Price Changes as Aggregate Supply Shocks,"
Laurence Ball and Gregory Mankiw brought attention to the relationship between in ation
and the distribution of relative price changes. Using data from a period when in ation was
not stable, they argued that in ation was related to interaction between shocks to the
distribution of relative price changes and frictions in price setting. If rms faced xed costs
of price adjustment, then price adjustment will occur disproportionately from rms with
large desired relative price adjustments. T hus, the distribution of desired relative price
adjustments will a ect the observed in ation rate.
Our model does not include xed costs of price adjustment, so its ability to match the
relationship in Figure 1 must lie elsewhere. In Figures 4-7, we display the same scatterplot
from simulated data for versions of the model in which speci c sets of parameters are
changed while the rest are held xed at their estimated values. In this manner, we hone in
on the crucial feature of the model that allows it to be consistent with the observed
relationship between in ation and the distribution of relative price changes.

/

T o begin, Figure 4 plots simulated data from a version of the model in which all prices are
exible. In the estimated model, there are separate price stickiness parameters in each
sector and the estimates feature large variation in this parameter across sectors. Some
research has suggested that di erential price stickiness across sectors can be important
for models' behavior, so we might expect that making price stickiness common across
sectors by shutting it down entirely would have a large e ect. In fact, Figure 4 shows that
the e ect is small: With exible prices, there is little change to the relationship between
in ation and the share of relative price increases.

Enlarge
Next, we restore the estimated price stickiness parameters and shut down variation in
demand shock volatility across sectors in Figure 5. T here are still sector-speci c demand
shocks, but they are all drawn from the same distribution.

/

Enlarge
Analogously, we shut down variation in productivity shock volatility in Figure 6. While there
are di erences between gures 5 and 6 — in ation is less volatile with common demand
shock volatility — the overall patterns in both cases are not much di erent from Figure 3.

/

Enlarge
We have now ruled out three aspects of the model as being necessary ingredients to
generate the negative relationship between in ation and the share of relative price
increases. Because that relationship involves heterogeneity across sectors, it must be that
it comes about from some combination of the inherent heterogeneity across sectors with
the way the model "processes" that heterogeneity and spits out an in ation rate.
For Figure 7, we take up one of the few remaining aspects of inherent heterogeneity: We
combine elements of Figures 5 and 6 by making demand-shock heterogeneity as well as
productivity-shock heterogeneity common across sectors. And this does the trick: T here is
virtually no relationship between in ation and the share of relative price increases, as the
correlation between the two variables is -0.06 in Figure 7, compared to -0.48, -0.47, -0.43
and -0.38 in Figures 3-6, respectively.

/

Enlarge
Heterogeneity in sectoral shock volatility, either for demand or productivity, is necessary
for generating the negative empirical relationship between in ation and the share of
relative price increases in our model. When shock volatility is identical across sectors, there
is little opportunity for the kind of large idiosyncratic price changes that could
meaningfully a ect in ation. Large price changes could occur, but they would likely be met
by o setting price changes in the opposite direction. In contrast, with heterogeneity
across sectors, the large price changes typically come from sectors with high volatility, and
thus will not be met by o setting price changes from other sectors. T his helps to explain
why it is necessary for volatility to di er across sectors to replicate the empirical
relationship.
However, from the perspective of the model, it is not su cient that there be
heterogeneous volatility. As mentioned above, the relationship depends not only on
sectoral heterogeneity but also on how the model processes that heterogeneity. And the
most important feature of the model for that processing is monetary policy. From the
standpoint of the model, monetary policy chooses to allow unusually large relative price
changes for particular sectors to pass through to in ation, instead of responding in a way
that generates large enough o setting nominal price changes for all other categories to
leave in ation una ected.

/

Francisco Ruge-Murcia is a professor of economics at McGill University. Alexander L.
Wolman is a vice president in the Research Department of the Federal Reserve Bank of
Richmond.

1 For additional information, see our 2022 working paper "Relative Price Shocks and In ation,"

which we revised in March.
2 Relative prices can be measured in terms of the price of any good or service, but we adopt the

standard convention of using the relative price in terms of the entire consumption basket.
3 The share of relative price increases in Figure 1 is based on a breakdown of the PCE

consumption basket into 206 categories. These data come from the National Income and
Product Accounts produced by the U.S. Commerce Department's Bureau of Economic Analysis.
4 These articles include the 2022 article "Relative Price Changes Are Unlikely to Account for

Recent High In ation," the 2023 article "Detecting In ation Instability" and the 2024 article
"In ation and Relative Price Changes Since the Onset of the Pandemic."
5 One might object that it is unrealistic to imagine (and to model) that monetary policy can

perfectly stabilize the in ation rate. That is an objection worthy of discussion, but we do not
address it here.

To cite this Economic Brief, please use the following format: Wolman, Alexander L.; and Ruge-

Murcia, Francisco. (September 2024) "A Model-Based Perspective on In ation and the
Distribution of Relative Price Changes." Federal Reserve Bank of Richmond Economic Brief, No.
24-30.
T his article may be photocopied or reprinted in its entirety. Please credit the authors,
source, and the Federal Reserve Bank of Richmond and include the italicized statement
below.
Views expressed in this article are those of the authors and not necessarily those of the Federal
Reserve Bank of Richmond or the Federal Reserve System.

Topics
Business Cycles

Economic Growth

In ation
/

Subscribe to Economic Brief
Receive a noti cation when Economic Brief is posted online.
By submitting this form you agree to the Bank's Terms & Conditions and Privacy Notice.

Email Address

Subscribe

Contact Us
RC Balaban
(804) 697-8144

© 1997-2024 Federal Reserve Bank of Richmond

/