Pass‐through in a concentrated industry: empirical evidence and regulatory implications

• DOI: http://doi.org/10.1111/1756-2171.12168
• Published date: 01 March 2017
• Date: 01 March 2017
• Author: Nathan H. Miller,Matthew Osborne,Gloria Sheu

RAND Journal of Economics

Vol.48, No. 1, Spring 2017

pp. 69–93

Pass-through in a concentrated industry:

empirical evidence and regulatory

implications

Nathan H. Miller∗

Matthew Osborne∗∗

and

Gloria Sheu∗∗∗

We estimate pass-through with 30 years of data from the portland cement industry. Robust

econometric evidence supports that fuel cost changes are morethan fully transmitted downstream

in the form of price changes. This validates an implicit pass-through assumption made in recent

academic research and regulatory analyses. We combine the econometric results with estimates

of competitive conduct obtained from the literature to evaluate the incidence of market-based

CO2regulation. Producersbear roughly 11% of the regulatory burden and could be compensated

with 16% of the revenues obtained.

1. Introduction

The distributional consequences of taxation in imperfectly competitive markets depend

on pass-through and markups (Weyl and Fabinger, 2013; Atkin and Donaldson, 2015). This

theoretical result poses a substantial challenge for applied research on incidence: reduced-form

estimates of pass-through are insufficient if firms exercise market power, whereas structural

models that allow for the recovery of markups often rely on tractable functional forms that

predetermine pass-through. This issue is particularly salient for the study of emissions-intensive

industries because production in these sectors often requires investments and fixed costs that

necessitate high markups in equilibrium.

∗Georgetown University; nhm27@georgetown.edu.

∗∗University of Toronto; matthew.osborne@utoronto.ca.

∗∗∗Depar tment of Justice; gloria.sheu@usdoj.gov.

We havebenefited from conversations with David Atkin, John Asker, Ron Borkovsky, Andrew Ching, Keith Head,Mar

Reguant, and Glen Weyl,as well as the seminar participants at Boston College, Department of Justice Antitrust Division,

London School of Economics, University of British Columbia, UCLA, and University of California, Berkeley. Wethank

Sam Larson, Erika Lim, Conor Ryan, and Jiayi Zhang for research assistance. The views expressed herein are entirely

those of the authors and should not be purported to reflect those of the US Department of Justice.

C2017, The RAND Corporation. 69

70 / THE RAND JOURNAL OF ECONOMICS

In this article, we estimate pass-through in the portland cement industry using reduced-form

regression techniques. We combine the results with estimates of markups derived from an earlier

study of the industry (Miller and Osborne, 2014) to evaluate the distributional consequences

of market-based regulation that places a price on CO2emissions. The portland cement industry

accounts for roughly 5% of global anthropogenic CO2emissions and also is a source of local

pollutants such as particulate matter and mercury (Van Oss and Padovani, 2003). Market power

arises due to spatial differentiation: cement is transported by truck to ready-mix concrete plants

and large construction sites, and the associated costs typically account for a sizable portion of

purchasers’ total expenditures.

Our regression equation derives from a general class of oligopoly models. We show how

regressors can be constructed in a manner that preserves the oligopoly interactions, even if only

aggregated prices are observed. In our application, we rely on variation in region-specific average

prices over the period 1980–2010. We calculate the fuel costs of each plant based on fossil fuel

prices and kiln technology and estimate pass-through by regressing observed prices on fuel costs

that are averaged to the region level. Unbiasedness is obtained under standard orthogonality

assumptions that are defensible given the institutional details of the empirical setting. To provide

one such detail, there are no viable substitutes for fossil fuel in the production process because

inputs (e.g., limestone and heat) are used in fixed proportions. As such, fuel costs plausibly are

uncorrelated with unobserved costs, especially once plant fixed effects absorb any plant-specific

factors related to the kiln technology.

The primary econometric result is that market-wide cost changes are more than fully passed

through to cement prices. The result is robust across a range of specifications and modelling

choices, and the confidence intervals are sufficiently tight to reject that pass-through is substan-

tially incomplete. Weyl and Fabinger (2013) demonstrate that, in a general symmetric model of

oligopoly, for pass-through to exceed unity it is sufficient that marginal costs are constant, firms

exercise market power, and demand is log-convex. The model provides the testable auxiliary

prediction that competition reduces pass-through under the same conditions. We find support

for this auxiliary prediction in the data, using measures of competition based on the number,

proximity, and capacity of nearby plants. Given the presence of plant fixed effects, identification

of this interactive effect rests on entry, exit, and fluctuations in the diesel price, the last of which

affects transportation costs and the magnitude of localized market power.

We evaluate the incidence of market-based CO2regulation using the Weyl and Fabinger

(2013) model of symmetric oligopoly.The effect of regulation on consumer surplus is determined

by pass-through. The effecton producersurplus depends on pass-through and a conduct parameter

that equals the multiplicative product of firm margins and the elasticity of market demand. Our

econometric estimates inform pass-through but not the conduct parameter. By contrast, recently

estimated structural models of the industry have bearing on the conduct parameter but are

insufficiently flexible to inform pass-through. Thus, wecalculate distributional effects for a range

of pass-through that reflects the confidence intervals obtained in this article, and a range of

conduct parameters that reflects confidence intervals obtained from a bootstrap of the Miller and

Osborne (2014) structural model. At the middle of these distributions, the pass-through that we

estimate implies that cement producers bear 11% of the regulatory burden and could be fully

compensated with 16% of the revenues obtained from regulation.

This counterfactual analysis has limitations, and we highlight three here. First, our approach

does not enable us to put a confidence bound on the harm to producers because there is no

straightforward wayto compute the statistical covariance between pass-through and conduct. Both

theory and our reduced-form results suggest this covariance is positive for pass-through in excess

of unity,and we discuss the implications of such a relationship qualitatively.Second, our approach

does not inform the extent of heterogeneity that arises among producers and consumers. We rely

on the symmetric model because it comports with our econometric estimates, which capture the

average effect of fuel costs on prices. Finally, the results may not convey to other sectors. A new

Working Paper finds that portland cement is one of the few US manufacturing sectors in which

C

The RAND Corporation 2017.