Does High Market Concentration Contribute to Inflation?

Ford, George

Inflation is a top policy concern for Americans, and for good reason.¹ When President Biden took office in January 2021, inflation was holding steady at 1.4%. But by June 2022, the inflation rate skyrocketed to 9%—a level not seen since 1981. As of this writing, inflation remains near 7%.² Covid-related supply chain problems were (and continue to be) a contributor to high prices across the globe. However, another significant contributor to excess inflation in the United States is the “fiscally-induced inflation” caused by the trillions in government spending injected into the US economy over the past two years, which has produced sizable increases in both disposable income and the money supply.³

Prices rise when too much money chases too few goods. Empirical analysis by economists at the Federal Reserve estimate that the high inflation in the US relative to peer nations is explained by the United States’ higher government spending on social programs during the Covid pandemic. Specifically, the American Rescue Plan Act of 2021 (ARP) “resulted in an unprecedented injection of direct assistance with a relatively short duration.”⁴

Several scapegoats have been offered for rising prices, with one being price hikes by firms operating in concentrated markets (e.g., the oil industry, drug makers, meatpackers, and internet service providers).⁵ Economist Hal Singer argues that market power (industry concentration) is a key contributor to the nation’s inflation woes, claiming that high inflation provides cover for concentrated industries to raise prices through tacit collusion.⁶

Attributing price increases on high industry concentration has become a pretext for the Biden administration to call for both aggressive antitrust enforcement and expansive new economic regulation by the Federal Trade Commission (FTC).⁷ Lina Khan, FTC Chair, is taking up this charge.⁸ There are some obvious problems with this economic argument, however. Coordinating pricing, explicitly or tacitly, would seem more difficult during times of economic chaos.⁹ Likewise, supply constraints make even competitive industries behave in ways that look monopolistic, because output expansions and entry are precluded by the lack of necessary inputs, including labor.¹⁰

In this paper, I consider the question of market power and inflation by analyzing trends in the Producer Price Index (PPI) for several hundred industries defined by a six-digit North American Industry Classification System (NAICS) code.¹¹ The data span from 2015 through February 2022 (the latter marking the Russian invasion of Ukraine and the decoupling of the Russian economy). The PPI data are linked to the four-firm concentration ratio (CR4) for each industry from the 2017 Economic Census, the latest data available. The analysis compares the change in prices during the Covid shutdowns and after President Biden’s inauguration when inflation rose rapidly.¹² I find no evidence that prices rose more in highly concentrated industries during either the heat of the pandemic or Biden’s tenure as president.¹³ As such, more concrete evidence of actual collusion is required to justify FTC intervention in markets.¹⁴

Data

PPI data are collected from the US Bureau of Labor Statistics at the six-digit NAICS code over the 2014–2022 period.¹⁵ Data for the petroleum industries is excluded from the sample as prices for oil and gas were extremely volatile during this period.¹⁶ Inflation determined from these data is computed using a one-year percentage change in the PPI. Thus, the final sample spans 2015 through February 2022, the latter date marking the Russian invasion of Ukraine and the subsequent decoupling of the Russian economy.

Note that the PPI is different from the Consumer Price Index (CPI), the latter typically used in the reporting of inflation numbers. The CPI measures the prices paid by consumers, while the PPI measures the price received by producers, usually in wholesale transactions. To link the concentration of producers in industries to prices, the PPI is the better representation of price changes (and concentration data cannot be linked directly to CPI data). Inflation measured by the two indexes are highly correlated. Over this period, the correlation coefficient is 0.95.¹⁷ PPI inflation is, however, larger than CPI inflation in recent years, and some research suggests that the PPI and CPI are only weakly related over short time periods.¹⁸

Concentration statistics for six-digit NAICS codes in 2017 are obtained from the latest-available Economic Census (a rolling five-year survey).¹⁹ The concentration statistic used here is the four-firm concentration ratio (CR4), which equals the market share of the four largest firms in the industry. Merging the two datasets provides a balanced sample of 465 six-digit industries with nearly eight years of monthly data (86 months).

It is feasible to use the five categories for analysis, though only 23 industries are in the “very high” group. As an alternative, industries are divided into three groups based on the CR4 using Bain’s criteria summarized in table 1. Most of the analysis is based on these three groups:

Low-concentration group (< 35%)
Moderate-concentration group (35%–65%)
High-concentration group (≥ 65%).²⁰

Results are also provided for the less aggregated concentration groups.

Examining Inflation and Concentration

My analysis divides the sample into three periods demarcated by the beginning of the Covid-19 epidemic (March 2020) and the first full month of the Biden administration (February 2021). The first, or pre-period, begins in January 2015 and ends in February 2020. I refer to the three periods as follows:

Pre-period (period 0 from January 2015 to February 2020)
Covid-19 interval (period 1 from March 2020 through February 2021)
Biden period (period 2 from February 2021 through the end of the sample).

Much of the analysis focuses on the changes in the mean PPI-based inflation rates for a large group of industries (defined at the six-digit NAICS) between the pre-period (when inflation was unremarkable) and the latter two periods of material price volatility. Let inflation be $p_{t c}$ , where $t$ is one of three periods and $c$ the concentration group indicator with values low, moderate, and high $(L, M, H)$ . My primary difference measure is $\left(p_{2 H}-p_{0 H}\right)-\left(p_{2 L}-p_{0 L}\right)$ , which quantifies the change in inflation, in this particular case between the high-concentration group $(c=H)$ and the low-concentration group $c=L)$ between the Biden period $(t=2)$ and the pre-period $(t=0)$ .²¹ Linear regression is used for hypothesis testing with standard errors that are robust to both arbitrary heteroskedasticity and arbitrary autocorrelation.

PPI Inflation Trends

To begin, trends in inflation for the low and highly concentrated industries from the three-category grouping are illustrated in figure 1. The shaded area in the figure is the spread between the two series (higher minus lower concentration). Inflation rates for the high-concentration group are lower than the low-concentration group during the Covid interval and, to a lesser extent, during the Biden period (as shown by the shaded area). The figure does not suggest a meaningful difference in price increases between the two groups, with inflation in the high-concentration group remaining mostly below that of the low-concentration group. Statistical testing is conducted later in the paper.

Figure 1. Inflation Trends; Low & High Concentration

In figure 2, the trends between the low and moderately concentrated group are illustrated. These trends are more closely related than are the low and high-concentration groups (as illustrated by the shaded areas). Inflation in the moderately concentrated group is slightly higher during the Biden period, but the spread is comparable to the pre-Covid interval (as detailed later with hypothesis testing).

Figure 2: Inflation Trends; Low & Moderate Concentration

From a graphical perspective, there appears to be no evidence of higher inflation in highly concentrated industries. In the next section, a statistical analysis is conducted to test for differences in mean inflation among the concentration groups.

Means Difference Analysis

In table 2, the mean inflation rates for the three periods and three concentration groups are summarized $\left(p_{0 L}, p_{0 M}, p_{0 H}\right)$ . Prior to March 2020, inflation rates are low and near 1.6% for the low and moderate-concentration groups, but near zero for the high-concentration group. Inflation is statistically equal between the low and moderate-concentration groups $\left(p_{0 M}-p_{0 L}\right)$ . Inflation for the high-concentration group is much lower than the low-concentration group (-1.80 percentage points, $p_{0 H}-p_{0 L}$ ). The difference is statistically significant.

During the Covid-19 interval, mean inflation rates remain relatively table for the low-concentration group (falling 0.29 percentage points, $p_{1 L}-p_{0 L}$ ), but larger declines are observed in moderate and high-concentration groups (-1.07 and -1.26 percentage points, $p_{1 M}-p_{0 M}$ and $p_{1 L}-p_{0 L}$ ). For the latter two groups, the inflation rates (0.60 and -1.49) are statistically different from zero. Their changes from the pre-period are also statistically different from zero. The Covid-19 pandemic had more meaningful effects on industries in the moderate and high-concentration groups than it did for the low-concentration group.

During the Biden administration as the nation emerged from Covid-19 lockdowns, inflation rose sharply across the board. In the low-concentration group, inflation during the Biden period rises to 7.33%, an increase of 5.75 percentage points ( $p_{2 L}-p_{0 L}$ ). Similarly, mean inflation rates for the moderate and high-concentration groups are 8.06% and 4.71%, respectively. The increase in mean inflation for the moderate concentration group ( $p_{2 M}-p_{0 M}$ ) is comparable to the low-concentration group (6.39 versus 5.75 percentage points). The difference in these changes for the two groups from the pre-period is relatively small (0.73 percentage points, \left(p_{2 M}-p_{0 M}\right)-\left(p_{2 L}-p_{0 L}\right)).

The difference in the mean inflation rate during the Biden period between the high and low-concentration groups $\left(p_{2 H}-p_{0 H}\right)$ is more substantial (-2.62 percentage points) and statistically different from zero. The difference in the differences in inflation rates between the Biden period and the pre-period $\left[\left(p_{2 H}-p_{0 H}\right)-\left(p_{2 L}-p_{0 L}\right)\right]$ are 0.64 [= 0.73 – 0.09] for the moderate-concentration group and -0.82 [= (-2.62) – (‑1.80)] for the high-concentration group, though neither is statistically different from zero.

These results suggest inflation and industry concentration are largely unrelated. With the inflation rate in the highest concentration group being well below that of the low-concentration group, it appears that high concentration is not a contributor to the higher inflation in recent months.

Alternative Samples

When looking at the CPI, it is common to exclude food and energy prices in what are called “core inflation.” I mimic that practice here by excluding the remaining energy (petroleum industries were already excluded from the sample) and food industries (i.e., core inflation) from the PPI data. Results are summarized in table 3. For the low and moderate-concentration groups, the results are mostly comparable to those in table 2.

The most significant change from table 1 is that inflation spread between the high and low-concentration groups is larger (increasing to -3.74 percentage points, a statistically significant difference). Consequently, the difference in the differences in inflation between the high and low-concentration groups between the pre-period and the Biden period has increased (-1.65 versus -0.82).

The comparison of mean inflation may also be influenced by outliers.²² Two approaches are used to address such extreme values. First, the data is winsorized at the upper 5% tail.²³ These results are summarized in table 4. As expected, the mean inflation rates during the Biden period are smaller when the data is winsorized. Aside from that difference, the results are mostly comparable to the full sample: The low and moderate-concentration groups have comparable inflation over time, but the high-concentration group has lower inflation during the Biden period. Again, inflation increased by less in the high-concentration group than the low-concentration group.

The second approach to address extreme value is median regression, where the regression is centered on the median rather than the mean. Results are presented in table 5. Relative to table 4, the largest differences are for the high-concentration group, where the mean inflation rates are larger than in table 4 for all three periods. Still, during the Biden period the mean inflation in the high-concentration group is lower than that of the low-concentration group.

Overall, these data do not support the claim of relatively higher inflation in more concentrated industries. Increases in the prices are generally comparable across the concentration groups. There is strong evidence to reject the claim that high prices broadly provide some type of cover for firms in more highly concentrated industries to raise prices. Concentration, it appears, is not driving inflation, at least as measured by the PPI.

Five Concentration Groups

In table 6, I conduct the same analysis using the five-category grouping of concentration. The low-concentration group is unchanged from table 1, while the moderate-low and moderate-high groups are portions of the moderate group from table 1, and the high and very high groups are portions of the high-concentration group from table 1.

Breaking apart the high-concentration group does offer some additional insights. For instance, the moderately high-concentration group has much higher inflation during the Biden period (10.54%) than the other groups, though the mean inflation rate is not statistically different from the low-concentration group (suggesting high variability in one or both groups). The spread between the pre-period and Biden period also has grown substantially (4.41 percentage points). This difference is statistically different from zero at the 10% level.

Inflation rates in the high and very high groups are much lower than the other groups during the Biden period and the two inflation rates are comparable (about 4.7%). This high-concentration group had much lower inflation during the Covid-19 Interval, while inflation was a bit higher for the very high group. The spread between the pre-period and the Biden period have widened for these two groups (-1.47 and -0.09), though neither difference is statistically different from zero. In all, the analysis of the PPI data indicate that, if anything, inflation is lower in the high-concentration groups, but this difference persists over time. It appears, at least from this data, that concentration is not to blame for high prices.

Synthetic Counterfactual

As shown in figures 1 and 2, the mean inflation rates across the concentration groups often do not follow similar trends. This disparity makes it difficult to visually interpret the differences in trends across time periods. As an alternative to using all the data from the groups, the synthetic difference-in-differences procedure (SDID) permits the construction of a counterfactual for the high-concentration group using data from the low-concentration group.²⁴

Normally, the SDID approach is intended to have a causal interpretation. Here it does not as the treatment period (the Biden period) applies to all industries. The SDID procedure is used to construct a weighted-average mean from the low-concentration group series where those weights are selected to provide the “best” fit to the high-concentration series during the pre-period and the Covid-19 interval. This weighted series is then extrapolated into the Biden period as a counterfactual. A hypothesis test of whether the actual high-concentration mean deviates materially from this counterfactual is determined by placebo effects.

Figure 3. Inflation Trends, Synthetic Counterfactual; Low & High Concentration

As seen in figure 3, the synthetic counterfactual provides an excellent fit for the high-concentration inflation series prior to the Biden period (i.e., compare the shaded areas in this figure to those of figure 1). This good fit extends into the Biden period (though the spread is more negative during the Biden period). There is no indication from this figure that inflation for the high-concentration group relative to the low-concentration group has changed. The null hypothesis of equality between the two series during the Biden period is not rejected (prob = 0.70) and the estimated means difference is -0.336.

The same procedure may be applied to the other concentration groups. The estimated means difference and their significance levels are provided in table 7. Effects are restricted to the Biden period in all cases. For all groupings, the effect sizes are small, and none are statistically different from zero. Consistent with the prior analysis, there is no evidence to suggest inflation is higher in high-concentration industries during the recent spell of very high inflation.

Conclusion

Inflation is one of many problems plaguing Americans these days. There are several first-order contributors to current inflation, mainly the massive deficit-spending by the US government occurring at precisely the wrong time (i.e., supply-side constraints, low unemployment, and the policies of the Federal Reserve Bank). Several second-order influences are being tossed around—some more valid than others. The analyses in this paper provide evidence that market concentration is not a determinant of current inflation. Using several modeling approaches to compare inflationary trends in hundreds of industries with varying degrees of market concentration, I find no evidence to support the claim that price increases are larger in highly-concentrated markets and thus providing an outsized contribution to inflation. If anything, inflation in the higher concentration industries during the Biden administration was slightly below that in the unconcentrated industries. Attempts to address inflation through antitrust and economic regulation are likely to impose more costs than benefits.

Does High Market Concentration Contribute to Inflation?

Executive Summary

Resources

Focuses

Federal Agencies

Data