Trade Policy and Market Power: The Case of the US Steel Industry

Abstract

A primary function of trade policy is to restrict imports to benefit the targeted domestic sector. However, a well-established theoretical literature highlights that the form of trade policy (e.g., quotas versus tariffs) can have a significant impact on how much trade policy affects firms’ abilities to price above marginal cost (i.e., market power). The US steel industry provides an excellent example to study these issues, as it has received many different types of trade protection over the past decades. We model the US steel market and then use a panel of data on major steel products from 1980 through 2006 to examine the effects of various trade policies on the steel market. We find that the US steel market is very competitive throughout our sample with the exception of the period in which they received comprehensive voluntary restraint agreements (i.e., quotas) and were able to price substantially above marginal cost. All other forms of protection were in tariff form and had little effect on market power, consistent with prior theoretical literature on the nonequivalence of tariffs and quotas. We also find evidence that market power eroded over time in steel products where mini-mill producers gained sizeable market share, highlighting the role of technology in the market as well.

Full text

NBER WORKING PAPER SERIES
TRADE POLICY AND MARKET POWER:
THE CASE OF THE US STEEL INDUSTRY
Bruce A. Blonigen
Benjamin H. Liebman
Wesley W. Wilson
Working Paper 13671
http://www.nber.org/papers/w13671
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
December 2007
This research was supported by NSF grant 0416854. We thank Erika Bundy for excellent research
assistance. Any remaining errors are our own. The views expressed herein are those of the author(s)
and do not necessarily reflect the views of the National Bureau of Economic Research.
© 2007 by Bruce A. Blonigen, Benjamin H. Liebman, and Wesley W. Wilson. All rights reserved.
Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided
that full credit, including © notice, is given to the source.

Trade Policy and Market Power: The Case of the US Steel Industry
Bruce A. Blonigen, Benjamin H. Liebman, and Wesley W. Wilson
NBER Working Paper No. 13671
December 2007
JEL No. F13,F23,L11,L61
ABSTRACT
A primary function of trade policy is to restrict imports to benefit the targeted domestic sector. However,
a well-established theoretical literature highlights that the form of trade policy (e.g., quotas versus
tariffs) can have a significant impact on how much trade policy affects firms’ abilities to price above
marginal cost (i.e., market power). The US steel industry provides an excellent example to study these
issues, as it has received many different types of trade protection over the past decades. We model
the US steel market and then use a panel of data on major steel products from 1980 through 2006 to
examine the effects of various trade policies on the steel market. We find that the US steel market
is very competitive throughout our sample with the exception of the period in which they received
comprehensive voluntary restraint agreements (i.e., quotas) and were able to price substantially above
marginal cost. All other forms of protection were in tariff form and had little effect on market power,
consistent with prior theoretical literature on the nonequivalence of tariffs and quotas. We also find
evidence that market power eroded over time in steel products where mini-mill producers gained sizeable
market share, highlighting the role of technology in the market as well.
Bruce A. Blonigen
Department of Economics
1285 University of Oregon
Eugene, OR 97403-1285
and NBER
bruceb@uoregon.edu
Benjamin H. Liebman
Department of Economics
Saint Joseph's University
5600 City Avenue
Philadelphia, PA 19131-1395
bliebman@sju.edu
Wesley W. Wilson
Department of Economics
1285 University of Oregon
Eugene, OR 97403-1285
wwilson@uoregon.edu

1
I. Introduction
For over 50 years, GATT and WTO rounds have substantially reduced trade barriers
across the world. Yet, there are a number of key industries in which trade barriers are present,
topical, and used strategically to improve the position of domestic industries including
agriculture, lumber, textiles and apparel, automobiles, and steel. Of note, the form of trade
policy measures varies widely across sectors and countries, and the theoretical literature often
highlights important potential differences between various trade policies on market outcomes.
Yet most prior empirical literature focuses only on the effects of a specific trade liberalization
event or policy. In this paper, we evaluate and compare the effects of trade policy on the US
steel industry over the past three decades, where a range of different trade policies provides an
excellent opportunity to examine their differential market impacts.
The US steel industry has been the recipient of practically every form of trade protection
in the past four decades (see Table 1). Long dominant in world markets, US steel became a net
importer of steel in the early 1960s. By 1969, rising import shares and declining employment led
to voluntary restraint agreements (VRAs) with Japan and the European Community (EC) which
lasted through 1974. Subsequent years have included further rounds of negotiated VRAs, trigger
mechanisms, antidumping (AD) and countervailing duty (CVD) cases, and the safeguard action
by the Bush administration in the early 2000s. Over these decades, the steel industry has been
one of the largest and most frequent users of US trade protection programs. Steel industry
protection programs account for over one-third of the over 1400 US AD and CVD cases since
1980, and steel is one of only a few high-profile industries that have been allowed VRAs or
safeguards.
Previous literature on the steel industry and trade protection is mainly characterized by
papers that analyze the impact of only a particular trade policy during a relatively short period of

2
time. Crandall (1981) and Canto (1984) examine the effect of the US VRAs from 1969-1974,
finding that these VRAs had only a very modest effect in raising import prices for US steel firms,
with no discernible impact on rates of return or employment. Nieberding (1999) examines
whether the withdrawal of 1982 US AD cases against imported steel affected market power for
three large US steel producers and finds only mixed evidence for any positive effects. Lenway,
Rehbein and Starks (1990) and Lenway, Morck, and Yeung (1996) undertake event study
analysis of abnormal returns for US steel firms upon announcements of major trade policy
implementation that occurred in the late 1970s and early 1980s. Their evidence suggests that the
stock market reacted positively to a number of these events, especially the 1982 and 1984 VRA
announcements, but this experience varied significantly across firms, depending on such factors
as previous lobbying activity, size, financial leverage, and whether the firm was a mini-mill or
integrated producer of steel.
1
More recently, Chung (1998) finds that AD and CVD duties from
1982 through 1993 had only modest impacts on import penetration, while Bown (2004) and
Durling and Prusa (2006) find that AD and safeguards significantly decrease trade in targeted
products. Liebman (2006) finds little evidence that the 2002-2003 safeguard actions affected US
steel prices. Given the wide-varying methodologies, periods of study, and sometimes contrary
results, it is very difficult to get a complete and comparable evaluation of the effects of various
trade policies on imports and performance.
2
This paper provides a comprehensive look at the effect of trade policies on the steel
industry’s performance over the past three decades. We do this by estimating a general model of
the US steel market consisting of equations representing demand, import supply, and US pricing,
1
Integrated firms produce steel by combining iron ore and coking coal in blast furnaces, while mini-mills produce
steel from recycled scrap that is melted down in electric arc furnaces.
2
In addition to the literature cited that uses econometric techniques to evaluate trade policies ex post, there is also a
significant literature that examines these policies with computable general (or partial) equilibrium models, including
de Melo and Tarr (1990) and many US federal government agency reports. These studies also typically focus on
only single trade protection instances and vary in their model specifications.

3
using annual observations of a panel of 20 major steel products from 1980 through 2006.
3
From
this model, we are able to not only estimate the effects of different forms of protection on both
import penetration and market power in the US steel industry, but also control for many factors
the previous literature has claimed may contribute to the US steel industry’s (poor) performance
– including macroeconomic shocks, technological change, exchange rate movements, and
various supply and demand shocks.
4
Importantly, our examination of trade policy on market outcomes can provide evidence
for a well-established, but mainly theoretical literature, on the non-equivalence of tariffs and
quantitative restrictions. Beginning with the seminal paper by Bhagwati (1965), a theoretical
literature has shown that quantitative restrictions can affect market power of one or more firms
competing in a market, whereas tariffs have no such impact.
5
To our knowledge, there has been
surprisingly little to no prior work to explore this empirically.
Our paper’s econometric results provide a number of important results. First, we find
that the US steel market is generally a highly competitive one, as we are unable to reject the
hypothesis that baseline markups are zero over our sample; that is, pricing which is consistent
with perfectly competitive conditions. However, there is significant evidence that the voluntary
restraint agreements (VRAs) in the 1980s led to both substantial decreases in import penetration
and increases in the US industry’s ability to price above marginal cost, especially during the
1985-1989 period when virtually all import sources were covered under negotiated agreements
and data suggest that quotas were binding. On the other hand, we find that while other forms of
trade policies, such as tariffs and AD duties significantly lower import volumes, there is no
3
See Bresnahan (1989) for a general discussion of estimating price-cost margins.
4
For example, Crandall (1996) and Moore (1996), among others, have argued that mini-mill production may be
more important for explaining the decline of large integrated steel producers in the US than imports.
5
Other related papers in this literature include Mai and Hwang (1988), Krishna (1989), Rotemberg and Saloner
(1989). Feenstra (1988) provides theory and evidence for a different type of nonequivalence of quantitative
restrictions and tariffs: quantitative restrictions encourage firms to upgrade their quality, unlike tariffs.

4
evidence that these policies impact the departure of price above marginal cost. These trade
policy results provide some of the only evidence of which we are aware that confirms prior
theoretical literature hypothesizing that quantitative restrictions can lead to quite different pricing
responses by domestic firms (Bhagwati, 1965), and perhaps even facilitate collusion (e.g.,
Krishna, 1989). To the extent that market power is a significant potential inefficiency, it
suggests that the WTO’s stance against quantitative restrictions in favor of tariffs is an important
one on these grounds.
We also find significant differences in market power for mini-mill products versus
integrated products. Significant markups over cost are initially observed at the beginning of our
sample for steel products that mini-mills ultimately dominate by the end of our sample.
However, markups in these products gradually erode over the sample as mini-mill shares
increase, becoming statistically insignificant from zero by the 1990s. This suggests that mini-
mill technology may indeed be a significant reason for market power loss in the industry. For
other products that are largely the domain of integrated producers, we find little evidence of
market power throughout the sample with the exception of the VRA period. In fact, we find that
the comprehensive VRA period was only beneficial to these “integrated” steel products in terms
of increasing their ability to price marginal costs for their products, not for mini-mill products.
The remainder of the paper proceeds as follows. In the next section, we briefly describe
the US steel industry and its substantial history of trade protection. Section 3 takes a first look at
the data to understand the effect of trade policies on import penetration and industry profitability
before describing our formal model and empirical specification in section 4, and our empirical
results in section 5.

5
II. US Steel Industry and Its History of Trade Policies
A. US Steel Producers
The US steel industry is composed of two major types of producers:
integrated mills and mini-mills. Integrated mills use large blast furnaces to make pig iron from
iron and coke, which is then melted into raw steel in basic oxygen furnaces. Until recently,
integrated mills accounted for the majority of steel production in the United States. Their
production process is relatively capital- and energy-intensive and, thus, characterized by large
plant-level scale economies. Integrated mills often include on-site or nearby finishing and
rolling mills that further finish the semi-finished steel forms, such as ingots, slabs, and billets,
into finished products, such as bars and sheets. Over time, a process of “continuous casting”,
whereby molten steel is formed directly into finished products has spread throughout the
industry. Examples of integrated steel companies include US Steel and Bethlehem Steel.
The past three decades have also seen an ever-increasing share of steel production due to
mini-mill steel plants which melt recycled steel scrap with electric arc furnaces (EAFs) into raw
steel and steel products.
6
There are a number of cost efficiencies possible from mini-mill
production, particularly in the much smaller plant size and hence, capital costs, required for an
EAF. Historically, mini-mill producers have primarily produced lower-quality steel products,
such as wire rods and steel bar products, because of the greater impurities in steel made from
recycled scrap steel, rather than iron ore. However, over time, technologies have been developed
that have begun to allow mini-mill producers to break into higher-quality steel markets, such as
plate and sheet products. While Nucor is the well-known example of a mini-mill-based steel
company, there are scores of smaller mini-mill steel plants across the United States.

6
B. Brief History of US Steel Trade Policies
Prior to the 1960s, the US steel industry was far more concerned with fending off anti-
trust charges than securing trade relief from the federal government.
7
A string of factors,
however, led to the industry’s permanent shift from dominant world exporter to net importer.
8
In reaction to pressure from the large, integrated steel producers and the United Steel Workers
Union (USW), President Johnson negotiated the industry’s first VRA with Japan and the
European Community (EC) in 1969. While the VRA expired in 1974, a surge of imports in 1977
led to renewed calls for quantitative restrictions as well as AD and CVD petitions. In order to
avoid either outcome, President Carter implemented the Trigger Price Mechanism (TPM) in
1977. Under the TPM the domestic industry agreed to refrain from filing AD and CVD petitions
as long as import prices did not fall below Japanese production costs (the world’s lowest-cost
industry) plus an 8 percent profit margin.
The TPM was renewed in 1980, but the industry was convinced that the policy was
failing to provide sufficient protection from subsidized European imports and began filing
petitions for AD and CVD protection. This, in turn, led to a suspension of the TPM program for
much of the period after 1980, and which culminated in the filing of over one hundred AD and
CVD petitions primarily against European producers in January of 1982. In order to avoid trade
frictions that would result from significant AD and CVDs, President Reagan negotiated VRA
agreements across a wide range of steel products with the EC in October of 1982.
6
Data from various issues of the American Iron and Steel Institute’s Annual Statistical Yearbook show that percent
of US domestic steel produced by using EAFs has increased from about 15% in 1970 to around 50% today.
7
This confrontation even led to President Truman’s unsuccessful attempt to nationalize the industry in 1952.
8
These factors included: 1) a crippling strike in 1959 that required downstream users to seek non-domestic sources,
2) increasingly efficient, subsidized European and Japanese operations, 3) the discovery of large iron ore deposits
outside the US, and 4) a strong dollar. As such, between 1960 and 1968, US import penetration climbed from 4.7
percent to 16.7 percent of total US steel consumption. See Moore (1996) for a more detailed discussion of the
history of steel trade protection in the US through the early 1990s.

7
Although European steel imports were not permitted to exceed 5.5 percent of the US
market, overall import penetration remained high due to a strong dollar and import diversions to
non-EC sources. This likely contributed to the industry filing a large set of AD and CVD
petitions in early 1984 and ultimately filing a safeguard petition (historically known as a Section
201 Escape Clause action in the US) in 1984. These trade protection actions led to the
negotiation of a comprehensive VRA for all finished steel products and limiting total import
market share to 18.4 percent in the last couple months of 1984. The VRAs were put into place
for a roughly five-year period to end in October of 1989.
In late 1989, citing the industry’s strong performance, President George H. Bush decided
to renew the VRAs for only two-and-a-half additional years, rather than the full five years
requested by the industry. When the VRAs ended in early 1992, the steel industry immediately
filed a large number of AD and CVD petitions once again. While many industry observers
expected intervention by the administration, President Bush instead allowed the cases to reach
their completion. In July of 1993, affirmative AD and CVD determinations were ruled in favor
of the domestic industry in only about a half of the value of imports under review. In several
instances, competition from mini-mills, rather than imports, were seen as the real cause of injury
by the US International Trade Commission. The ruling was perceived as a major defeat for the
industry and was cited by Moore (1996) as an indication of the industry’s loss of political clout.
Through the rest of the 1990s, steel producers used AD and CVD actions targeted at a
limited number of specific products to secure trade relief. One possible reason for such limited
action was the strong economy and modernized US operations. For the first time in decades,
integrated producers were globally competitive, touted by some experts as an industry that had
survived its austere, rationalization period and which was now enjoying a much-deserved
“renaissance.” (Ahlbrandt, Fruehan, Giarratani, 1996) Unfortunately, a string of unexpected

8
shocks in 1998 brought this period quickly to an end. Most notable were currency crises in East
Asia and Russia which led to import surges and subsequent AD and CVD filings in the late
1990s. By the early 2000s, about one-third of the industry had fallen into bankruptcy, leading
President George W. Bush to implement another safeguard action on behalf of the US steel
industry in March of 2002, which placed tariffs ranging from 8-30% on many major steel
products in the first year. However, a number of major import sources were excluded including
Canada and Mexico, as well as less-developed countries. Secondly, downstream industries
successfully lobbied for exceptions over the ensuing safeguard period further watering down the
amount of affected imports. Finally, the safeguard tariffs were terminated prematurely in early
2004 due to a WTO dispute panel ruling against the US safeguard action.
Our purpose is to examine the effects of policy on pricing power in the industry. To do
so, we next turn to an analysis of the impact of these trade policies on the US steel industry’s
fortunes, first with a descriptive time series analysis, and then with formal estimation of their
impact on the US industry’s ability to price above marginal cost.
III. A First Look at the Data
One of the most striking characteristics of the US steel industry over the past decades is
the large drop in employment in the sector. As show in Figure 1, employment in the steel sector
has fallen from close to 600,000 employees in 1960 to less than 100,000 by the 2000s.
9
A
significant portion of this decline occurred in the first half of the 1980s, when industry lobbying
for protection led to comprehensive VRAs. The 1990s then saw a much more gradual decline in
the number of employees.
9
Data used in this section come from issues of the Annual Statistical Yearbook of the American Iron and Steel
Institute, with the exception of the Industrial Production Index which comes from statistical tables in the Economic
Report of the President.

9
It is quite informative, however, to match this trend with the associated trends in US
domestic production of steel, which has been relatively constant over the four-decade period.
There is a decline in output in the early 1980s as well, but it is much smaller in magnitude than
the decline in employment and had leveled off and begun to increase by 1983, whereas
employment continued to decline until 1987 before leveling off. In fact, the correlations between
output and employment both before and after 1980 are slightly negative and insignificant,
suggesting that other factors (such as technological advances) are extremely important for
understanding economic outcomes in this industry.
The effect of trade protection programs on domestic employment and production are not
clear from Figure 1 either. The period of the TPM was the period of greatest decline in both.
The ensuing VRA periods may have led to leveling off of the variables. However, the TPM
period was the beginning of a substantial recession in the United States, and the VRA period was
one of recovery.
Figure 2 provides a snapshot over the same period of some variables that get closer to the
heart of our statistical analysis, graphing out a measure of the steel industry’s annual profitability
(net income as a percent of sales), import penetration (the value of imports as a percent of
apparent supply to the US market), and an index of industrial production (1997=100). Over the
four decades import penetration has risen from around 5% in the early 1960s to between 25-30%
in the early 2000s. It is hard to argue that any of the trade protection programs affected this
import market share except for the comprehensive VRA period from the mid-1980s through the
early 1990s, when import shares decreased from about 25% to less than 20%. Interestingly,
profitability of steel firms during this period is not clearly helped by falling import shares and
actually is negative on average. Pair-wise correlations indicate a strong negative correlation
between import share and profitability of US firms before 1980 (-0.61), but an essentially zero

10
correlation after 1980. Taken at face value, this evidence suggests that in the post-1980 period,
even if trade protection programs effectively limit imports, profitability of US firms is
unaffected. However, these simple trends cannot uncover the independent effect of trade
policies from the effect of many other market forces on the steel firms’ performance, including
changes in exchange rates, technological progress, demand preferences and input prices. Thus,
we next turn to our econometric analysis that more formally examines the effect of trade policy
on US steel firms’ market power.
IV. Methodology and Data
A. Theory and Empirical Specification
Our approach is to estimate an econometric model of the industry that allows an
examination of the effects of various trade policy regimes, while controlling for demand, cost
and technology considerations.
10
We model US firms as a dominant firm that competes with a
competitive fringe. The fringe is composed of foreign-country exporters to the US. Evaluation
of the structure of imports suggests that these foreign firms can be modeled as a set of price-
taking suppliers. Specifically, while the market share of imports in the US steel market has
ranged from 15-25% since the mid-1970s (see Figure 2), the market shares of even individual
foreign countries (much less individual foreign firms) are very small and makes the assumption
of a fringe import supply quite reasonable. Table 2 provides market share of the top foreign-
country import sources for the US steel market in 1980, 1990, and 2000. The largest foreign-
country source is Canada with less than 5% of the market (spread across multiple steel firms),
10
Generally, studies using this methodology estimate price-cost margins use firm-level data. We choose industry-
level data for our analysis because there are very few US steel firms that provide publicly available data over the
length of time period we wish to consider. For example, Nieberding’s (1999) study over the 1978 to 1989 period
was able to find only three steel companies “with enough available historical data to allow for the implementation of
the market power test.” (p. 79) Our study covers a much longer period, including one that saw many bankruptcies in
the industry (early 2000s) which leads to unusable data for estimating market power at the firm level.

11
while very few countries have even 0.5% of the US market. Varying efficiencies across these
foreign import sources, as well as capacity constraints, means that we expect an upward-sloping
supply function in response to US steel prices, a relationship we estimate in our empirical
specification. Importantly, we note that the market structure features we observe across all
imports are generally true for each of our twenty products as well. Indeed, for seventeen of the
twenty products, more than half of the top twenty US import source countries ship the product in
an average year. Fifteen of the twenty products have a total import share (across all sources) of
25% or less in an average year. Finally, the results reported below are qualitatively identical
when we eliminate the products that have both high import market shares (25% or above) and
relatively few import source countries out of the top twenty (less than 12).
11
Given our market structure assumptions, our focus is on the ability of the US steel firms
to price above marginal cost and how this ability varies with trade policy changes and
technological changes. Thus, our model of the US steel market consists of a US demand
equation, an import fringe supply equation, and a domestic pricing equation
12
:
(, )
(, )
1
(,) (),
(, )
(, )
DD D
S
IMP IMP
DC D
US IMP IMP
S
DD
IMP
QQPX
QQPX
PMC Q Q X Q Q
QPX
QPX
PP
θ
=
=
⎛⎞
⎜⎟
=−− −
⎜⎟
∂
∂
⎜⎟
−
⎜⎟
∂∂
⎝⎠
(1)
where P is the US domestic price of steel; Q
D
(P,X
D
) is the domestic market demand function
with associated shifters, X
D
; Q
IMP
(P,X
IMP
) is the import supply with associated shifters, X
IMP
;
MC
US
(Q
US
-
Q
IMP
, X
C
) is the marginal cost equation for the US producing Q
D
-
Q
IMP
with cost
shifters X
C
; and
θ
represents an index of the degree to which prices depart from marginal costs.
We note that the bracketed term is the slope of the residual demand function used for pricing
11
These products are black plate, rails, tool steel, and wire-drawn.

12
decisions by US firms, and that as either the market demand and/or import fringe supply
functions become more responsive to price changes, it dampens the departure of price from
marginal costs.
The parameter, θ, indicates the degree to which US firms in the industry are able to set
prices above marginal cost is usually thought to range in values from 0 to 1. If θ is equal to 0,
the market yields no ability to price above marginal cost. As θ increase in values, the aggregate
ability of the industry to price above marginal cost increases up to a possible markup of 100%.
We note that, as is common in this literature (e.g., Bresnahan (1989), marginal costs are not
observed and are accounted for with other measurable variables. Further, our focus is on
whether or not θ is statistically important or not and, if important, whether it is affected by policy
and other variables.
We begin by assuming linear forms for the demand, import supply, and marginal cost
relationships and estimate the three equations as a system using 3-Stage Least Squares regression
techniques to control for potential endogeneity and correlation of error terms. However, as
discussed by Bresnahan (1982), identification of θ rests on interactions in the demand equation.
Alternatively, Gallet (1997) achieves identification of θ in a similar empirical specification by
assuming non-linearity in the marginal cost function and interacting the factor price terms. We
employ both sets of interactions for identification, as statistical tests always strongly support
inclusion of both in our specification. Below we describe our matrix of exogenous regressors for
each equation.
Before proceeding, there is one final issue we address. Corts (1999) levels a major
criticism at the new empirical industrial organization (NEIO) approach to estimating market
power – an approach we generally follow here. Specifically, he argues that the NEIO approach
12
The median US export volume for a steel product category in our data is less than 5% of US production in the

13
is estimating marginal changes in market power, not the level of market power, which need not
correspond to each other if the inherent strategic market game is dynamic. However, recent
work examining the inherent bias in estimating market power in the NEIO framework for
industries where one has precise estimates of marginal cost, and can therefore calculate market
power directly, finds that the bias from estimating market power in an NEIO framework is often
not large.
13
In addition, the focus of our paper will not be on the magnitude of market power per
se, but whether market power exists or not. Our main results are characterized by little evidence
of market power in the steel industry for virtually the entire sample period, with the exception of
a very large and discrete jump in estimated market power to collusive levels during the height of
the VRA period. It seems unlikely that such a large, discrete, and temporary jump in market
power would be due to some other unobserved change in the underlying dynamic strategic game
between firms.
B. Variables and Data
Our data set consists of annual observations of a panel of 20 major steel products
covering the years 1980 through 2006. Exploiting cross-product variation has rarely been
employed in previous studies of the steel industry, but the significant variation in how trade
policies, technological change, and other market forces impact these various products over time
allows us to more cleanly identify the impacts of focus variables. A full list of the steel products
we include in our sample appears in the data appendix. It includes all major categories, such as
plate, hot-rolled and cold-rolled sheet, wire rod, various types of bars, and drawn wire. While
category, and thus we do not model export behavior.
13
These examples include sugar (Genesove and Mullin, 1998), whiskey (Clay and Troesken, 2003), and electricity
(Puller, 2007).

14
our data appendix also describes construction of variables and data sources in detail, this section
briefly discusses the variables, beginning with our endogenous price and quantity variables.
The main data limitation that prevents us from going back farther with our sample than
1980 is due to data issues with our main endogenous variable, the US steel price. Crucial issues
in accurately measuring market steel prices are, first, the difference between listed and actual
trade prices and, second, the complication of agreed-upon contractual prices that may run out
many months or even longer than a year. While the US Bureau of Labor Statistics has historical
price series that go back well before 1980, these are often based on listed prices by steel firms
which may have little relationship with actual transaction prices (see Crandall 1981).
14
As a
result, Crandall (1981) uses unit prices derived from Current Industrial Reports for the steel
industry published by the US Census. These data are problematic for two reasons. First, the
prices reflect revenues derived on sales that may be bound by prior contractual arrangements, not
current market conditions. Second, this publication is quick to censor data for confidentiality
problems and, thus, missing values begin to accumulate quickly, especially toward the end of our
sample as the number of US steel firms declined. The price data we use come from Purchasing
Magazine, which surveys monthly US spot prices of major steel products. These data more
likely reflect current market conditions for a given period, but have the limitation that they only
first began in January of 1980.
15
We note that we construct a real US price by deflating by the
US GDP deflator. Our domestic and import steel quantity variables, the other endogenous
variables in our empirical specification, are measured in tons and these data come from annual
yearbooks of the American Iron and Steel Institute.
14
Our own correspondence with staff of the US Bureau of Labor Statistics suggests that this methodology has
ultimately been changed to better reflect market prices, but that this change began well after our sample begins in
1980.
15
In practical terms, we find that the BLS prices give us qualitatively similar answers to those reported in the paper
that use the Purchasing Magazine data. Price data constructed as unit values from the US Census’ Current

15
Exogenous variables in our US demand equation include the real price of a common
substitute for steel, aluminum, as well as an index of industrial production because steel is an
important intermediate input for so many industrial sectors in the economy. We expect the
coefficient on both of these terms to be positive. As mentioned earlier, we also include
interaction terms between these exogenous variables and the US steel price variable, which also
appears as a regressor on the right-hand side of the US demand equation.
Exogenous variables in the import fringe supply equation include the real exchange rate
(expressed in terms of US dollar per foreign currency), rest-of-world (i.e., non-US) real GDP,
real world input prices (including iron ore, oil, and coal), and trade policies. Our trade policies
include, first, a dummy variable indicating the years of the VRAs on steel products from 1983
through 1991. We also include trade-weighted average antidumping and countervailing duties,
which vary by product and year, as well as product-specific ad valorem tariffs. Finally, we
include a dummy variable for the recent safeguard period from 2002 and 2003. We expect an
increase in the real exchange rate measure (depreciation of the dollar) to increase US imports of
steel. Rest-of-world real GDP and US imports are expected to be negatively correlated if excess
capacity effects are present.
16
Increases in world prices of steel production inputs are expected to
decrease the supply of steel products to the US market from the rest of the world. Finally, US
trade policies are expected to restrict US imports with an obvious interest in how large these
trade policy effects are, ceteris paribus.
Exogenous variables in the final equation begin with real factor prices, including wages,
iron ore, scrap steel, electricity and coal. We also include interaction terms between these factor
prices to aid identification of our market conduct parameter, as discussed earlier. These factor
Industrial Reports generally perform quite poorly with coefficient signs on the price term often reversed from that
predicted by theory.
16
See Blonigen and Wilson (2005) for a more detailed analysis of this issue with respect to US imports of steel.

16
prices are expected to be positively correlated with marginal costs of the firms in the industry
and, hence, the real price of steel.
We also include an exogenous variable in the pricing equation that captures important
technological factors in the steel industry, namely, a measure of the percent of US steel that is
being “continuously-cast”. Continuous casting is a process that allows molten steel to be directly
shaped into semi-finished steel products. Prior to continuous casting, the metal was poured into
standing casts to form ingots, which then required many more steps of reheating and working the
metal to produce steel products. Over the time period of our sample, the amount of steel
produced through continuous casting increases from around 15% to over 95%. Wider prevalence
of this technology is expected to decrease real US steel prices, ceteris paribus. The other obvious
technological change in the industry has been the rise of mini-mill production. Below we discuss
some specific mini-mill technology events that we can use to examine the effect of mini-mill
technology on the industry’s ability to price above marginal cost.
The final variables on the right-hand side of the pricing equation are domestic production
of steel (Q
US
– Q
IMP
) and a term that interacts the domestic production of steel with the inverse of
the difference in the slopes between the demand equation and the import fringe supply
equation.
17
The coefficient on the first term indicates the extent to which marginal costs vary
with increasing production in the industry. The coefficient on the latter term, which we will refer
to as the “weighted domestic quantity” term identifies the parameter θ, which indicates the
ability of firms in the industry to price above marginal cost. Finally, we note that we include an
annual trend term and product fixed effects in all three equations.
The above fully describes our base specification. From this base specification our
analysis below proceeds by examining how various policy and market forces affect the US steel

17
industry’s ability to price above marginal cost. These analyses break down into two main
categories: trade policies and the impact of mini-mill production. With respect to trade policies,
we analyze the separate impact of various VRA periods, antidumping and countervailing duties,
standard import tariffs, and the 2002-2003 safeguard period. With respect to mini-mill versus
integrated steel producers, we investigate whether the ability to price above marginal costs varies
across products that are traditionally produced by integrated producers versus those produced by
mini-mills, as well as whether trade policies had differential impacts on these two sets of
products/producers.
V. Empirical Results
Column 1 of Table 3 displays three-stage least-squares estimates for our benchmark
three-equation system using our annual panel of 20 products and the years 1980-2006. The
model fits the data quite well, with R
2
statistics fairly high for each of the three equations (0.92,
0.73, and 0.85, respectively) and chi-squared statistics (not reported) easily rejecting the null
hypothesis of that the coefficients are jointly zero. We do not report the coefficient estimates on
the product fixed effects or factor price interaction terms, but note that the chi-squared tests
strongly support their inclusion throughout our analysis.
Most of the coefficients in our benchmark model are of expected sign and many are
statistically significant as well. The price terms in both our domestic demand and import fringe
supply equations are estimated with correct sign and are statistically significant; domestic
demand is inversely related to the market price, while fringe supply increases when the market
price increases. Industrial production and price of the substitute good (aluminum) have correct
sign in the demand equation, though not statistically significant. However, we note that these
17
The slopes in the construction of this variable are simultaneously estimated in the prior two equations, generating
cross-equation restrictions in our parameters.

18
variables are connected with interaction terms, which can affect their sign and significance.
Importantly, the coefficient on the interaction term between industrial production and steel price
is statistically significant, which is important for identification of the market power parameter, as
discussed above.
Control variables in the import fringe supply equation are generally of expected sign as
well, with many statistically significant. The exchange rate and rest-of-world GDP variables are
highly significant, suggesting that imports increase when the US dollar appreciates and when
foreign countries economies see reduced demand in their own markets. The various trade
policies, with the exception of countervailing duties, all are estimated to decrease import supply
of steel. The poor performance of the countervailing duty variable is not surprising in light of the
fact that such duties are almost always applied simultaneously with antidumping duties on the
same products and import sources for US steel cases, and that these countervailing duties are
invariably much smaller in magnitude than antidumping duties.
Control variables are often of correct sign in the pricing equation as well, though we note
that the interaction terms we include amongst all the factor price terms (electricity, ore, scrap
steel, coal, and wage prices) can easily affect sign and significance of the coefficient on that
factor price. The marginal effects of these factor prices, including the direct and interacted
effects, are generally of correct sign and statistically significant. The coefficient estimates on the
interaction terms, though not reported to save space, are jointly statistically significant. Thus,
they are likely quite important for identifying the departure of price from marginal cost, θ. Our
estimates also show that the industry is characterized by economies of scale, with a statistically
significant negative coefficient on the quantity variable. As expected, we find that greater use of
continuous-cast technology has a significant effect in reducing costs and, hence, domestic prices.

19
Finally, our base model estimates a positive coefficient for the weighted domestic
quantity variable in the domestic pricing equation; in other words, our estimate of θ. While this
coefficient is of expected sign, it is also statistically insignificant, indicating that we cannot reject
(perfectly) competitive pricing. We can also show that the magnitude of the implied markup
associated with this estimated coefficient is small. In particular, we can use the estimated
coefficients from our model (including θ) to generate an estimated price-cost margin. Recalling
that the domestic pricing relation is P = MC
Dom
- θ·P’(Q
Dom
)Q
Dom,
we can reorganize and divide
both sides by price, to obtain the markup solution:
'( )
,
D
om Dom Dom
PMC PQ Q
Markup
PP
θ
−
−
≡=
(2)
where P’(Q
Dom
) is defined as the term in large brackets of the last equation in (1).
18
Using our
estimated coefficients, as well as sample mean values of domestic steel output, steel prices,
industrial production, and the aluminum price, we calculate the average markup for our sample
as 0.5%. Thus, the magnitude and statistical significance of the markup in our base model
argues against any ability by the US steel industry to price above marginal cost.
A. Effects of Trade Policies on Market Power
We next explore how various trade policies affect market power. As we can see in the
benchmark estimates, all of the trade policies are estimated to decrease imports, with the
exception of countervailing duties. In column 2 of Table 3, we re-estimate our system of
equations and interact the trade policy variables with the weighted domestic quantity variable in
the domestic pricing equation, so that we can estimate how these trade policies may structurally
affect the domestic industry’s ability to price above marginal cost. Panel A of Table 4 provides

20
markup calculations for these various trade policies based on our estimates in column 2 of Table
3. The estimated markups are noteworthy in that there is a strong increase in the domestic
industry’s ability to price above marginal cost during the VRA period from 1982 through 1991,
but no evidence of effects on markups by any other trade policy. In fact, we cannot reject a
mark-up of 100% during the VRA period, which would indicate collusive pricing by the US steel
industry. While these results are quite dramatic in terms of differences in the industry’s ability to
price above marginal costs, they are consistent with prior theoretical literature showing that
quantitative restrictions can lead to significant market power for domestic firms (Bhagwati,
1965) and even facilitate collusion (e.g., Krishna, 1989), since the VRA period saw quotas as the
main form of protection, whereas the other trade policies we examine were characterized by ad
valorem duties. In fact, this is the most direct evidence for this well-established theoretical
literature of which we are aware.
It may be surprising that we do not find some evidence that AD duties can affect market
power, given that Konings and Vandenbussche (2005) find evidence that AD duties positively
affect market power in their sample of European firms enjoying AD protection. However, as
Konings and Vandenbussche point out, we may expect AD duties to be less likely to affect
market power in the US due to previous findings of significant trade diversion of imports from
targeted products to alternative import sources in US AD cases (Prusa, 1997). Such trade
diversion can substantially weaken the effects of the trade policy. Unlike the US situation,
Konings et al. (2001) document quite limited trade diversion effects in European AD cases.
In column 3 of Table 3 we take a closer look at the ability of the domestic industry to
price above marginal cost during the VRA period by splitting these years into three periods. The
18
In particular, from our benchmark model, we calculate P’(Q
Dom
) = 1/(-6.164 + 0.078*Industrial Production -
0.0005*Aluminum Price - 1.626), where we use the sample mean values for the Industrial Production and
Aluminum Price variables.

21
first couple years of this period (1983-1984) saw VRAs only with European Community
members, with allegations by the steel industry of significant trade diversion to other import
sources.
19
Comprehensive VRAs on virtually all import sources and steel products did not occur
until 1985 after further trade protection petitions by the industry. These “comprehensive” VRAs
lasted until late 1989, at which point they were only renewed through 1991. In this latter period
(1990-1991), there is evidence that many of the quotas were no longer binding as the economy
slid into the recession of the early 1990s. (Crandall, 1996) Thus, we hypothesize that the
industry’s ability to price above marginal cost was greatest from 1985-1989, when
comprehensive VRAs were in place.
Our estimates in Column 3 of Table 3 and the associated markup calculations in Panel B
of Table 4 show that the VRAs effects on market power were different across these three VRA
periods. We estimate statistically significant markups around 100% for both the EC VRA period
from 1983-1984 and for the comprehensive VRA period from 1985 to 1989. The latter VRA
period (1990-1991) is associated with a fairly small markup estimate of less than 30% and is
statistically insignificant. This last result is consistent with Crandall’s (1996) assessment that
market conditions were such that the quantitative restrictions were not binding towards the end
of the VRA period.
One potential issue with our VRA variables is that they are simply dummy variables for a
given time period. Are we picking up some spurious correlation or is the effect simply being
driven by other market factors during this particular time period? We are fairly confident that it
is not simply a spurious correlation. Table 5 shows our estimates of θ for every possible
continuous three-year of our sample, beginning with 1980-1982, and using the same
19
While these initial VRAs only involved EC countries, they covered a wide range of steel products. Thus, we
continue to proxy this period’s effects with a dummy variable taking the value of “1” during this period for all
products in our sample.

22
specification we employ in Column 2 of Table 3. The only three-year period that displays a
statistically significant estimate of θ is the 1987-1989 period, the height of the VRA period. The
market power estimate associated with the estimated θ for 1987-1989 is 92.7%, which is in line
with the large magnitude we estimate above for the VRA period. There is always the possibility
that there is some factor driving our VRA estimate for which we have not accounted, though we
have been careful to control for many market factors, including exchange rates, technology, and
demand shocks.
B. Differential Market Power for Mini-mill and Integrated Products
As discussed earlier, integrated and mini-mill steel producers are quite different in their
production processes, plant-level scale economies, and the types of products they typically
produce. In this section, we examine whether there are differing markup abilities across these
producers, by interacting relevant variables with an indicator variable for whether the product is
primarily produced by mini-mill producers. These products are reinforcing bar, hot-rolled bar,
bar and light structural shapes under three inches, and heavy structural shapes.
In Panel C of Table 4, we report estimated markups for both mini-mill and integrated
product during the entire sample and for the VRA period from 1983 through 1991.
20
Recall that
in the first section we found a significant effect of VRAs on markups. Panel C’s results show
that while we cannot reject a baseline estimate of no ability to price above marginal costs for
both sets of products, the positive impact of the VRAs is limited to the integrated products,
where we estimate a statistically significant markup over marginal cost of 76.5%. In contrast, we
estimate a markup of -21.8% markup for mini-mill products during the VRA period that is not
statistically different from zero. There are a number of possible explanations for this difference.

23
First, unionized workers are probably better able to organize lobbying efforts for trade protection
that covers the products they specifically produce and integrated mills were typically unionized,
whereas most mini-mills were not. Second, the mini-mill sector is characterized by many
smaller firms relative to the integrated sector. Thus, the integrated firms may have been better
able to lobby for stronger trade protection for their products and/or better able to achieve and
maintain tacit collusion once the VRAs were in place.
A final issue that we are able to explore, at least indirectly, is the extent to whether mini-
mill competition has eroded integrated firms profitability over time. A direct way to estimate
this would be if we had data on mini-mill market shares by product over time. These data are
unavailable. However, we have some indirect pieces of evidence. First, we know that overall
mini-mill market shares have steadily increased over time in the industry and the products where
this mini-mill market share has increased to take over the majority of market share by the end of
our sample – hot-rolled bar, cold-rolled bar, reinforcing bar, light bars and shapes, and structural
shapes.
21
In Panel D of Table 4 we examine the evolution of estimated market power in the two
types of products by interacting relevant terms by a trend variable.
22
The results show that
products mainly dominated by integrated producers throughout the sample have actually shown
little market power even from the beginning of the sample with a negative, but insignificant,
change in market power over time. In contrast, products that have seen a large rise in mini-mill
production initially display significant market power at the beginning of the sample (56.7%), but
see it significantly eroded over the sample with an annual decline of 1.6%. By the 1990s, our
estimates cannot reject the hypothesis that market power was zero for these products as well.
20
For the sake of brevity, we do not report the parameter estimates that we base our markup calculations on in this
section. These estimates are available upon request.
21
This information comes from US Census’ Current Industrial Reports: MA331B – Steel Mill Products, which
began to report such data toward the end of our sample.

24
This suggests that mini-mills broke into more lucrative steel products and quickly eroded profits
for the steel industry in these products over time.
VI. Conclusion
This paper estimates a model of the US steel market that allows analysis of the effect of
various trade policy programs on imports and market power in a comprehensive fashion for the
first time. Using a panel data set of 20 major steel products from 1980 through 2006, we find
substantial differences in the effects of various trade programs. While most protection programs
have had at least moderate impacts on reducing imports, only the comprehensive VRAs during
the latter half of the 1980s had any significant impact on the ability of the US steel industry to
price above marginal cost. Moreover, the impact of these VRAs was large and we cannot reject
the hypothesis that the industry was able to perfectly collude during this period. To our
knowledge, this is the most direct confirmatory evidence of the well-established theoretical
literature showing that a significant difference between quantitative restrictions and tariff (price)
policies are the ability of quantitative restrictions to allow firms to the ability to price above
marginal cost. Interestingly, we find that the market power effects of the VRAs occurred only
with steel products that were mainly produced by integrated steel producers. Steel products for
which mini-mill producers had a significant market share saw no discernible changes in market
power during the VRA period, though we also find that these “mini-mill” products see gradual
erosion of market power over time.
22
Specifically, we interact our “weighted domestic quantity” variable in the domestic pricing equation with 1) a
dummy variable indicating whether the product was mainly produced by mini-mills (bar products and structural
shapes), 2) a trend term, and 3) a trend term multiplied by the mini-mill product dummy.

25
References
Ahlbrandt, Roger S., Fruehan, Richard J., and Giarratani, Frank. (1996) The Renaissance of
American Steel. New York, NY: Oxford University Press.
Bhagwati, Jagdish N.. (1965) “On the Equivalence of Tariffs and Quotas,” in R.E. Baldwin et al.,
eds., Trade, Growth and the Balance of Payments: Essays in Honor of Gottfried Haberler,
Amsterdam: North-Holland.
Blonigen, Bruce A. and Wilson, Wesley W. (2005) “Foreign Subsidization and Excess
Capacity,” NBER Working Paper No. 11798.
Bown, Chad P. (2004) “How Different Are Safeguards from Antidumping? Evidence from US
Trade Policies Toward Steel,” Mimeo, Brandeis University.
Bresnahan, Timothy F. (1982) “The Oligopoly Solution Concept is Identified,” Economic
Letters, Vol.10: 87-92.
Bresnahan, Timothy F. (1989) “Empirical Studies of Industries with Market Power,” in Richard
Schmalensee and Robert Willig (eds.), Handbook of Industrial Organization, Volume 2.
Amsterdam, Oxford, and Tokyo: North-Holland, 1011-57.
Canto, Victor A. (1984) “The Effect of Voluntary Restraint Agreements: A Case Study of the
Steel Industry, Applied Economics, Vol. 16(2): 175-86.
Chung, Jae W. (1998) “Effects of US Trade Remedy Law Enforcement under Uncertainty: The
Case of Steel,” Southern Economic Journal, Vol. 65(1): 151-9.
Clay, Karen, and Werner Troesken. (2003) “Further Test of Static Oligopoly Models: Whiskey,
1882-1898,” Journal of Industrial Economics, Vol. 51(2): 151-166.
Corts, Kenneth S. (1999) “Conduct Parameters and the Measurement of Market Power,” Journal
of Econometrics, Vol. 88(2): 227-250.
Crandall, Robert W. (1981). The US Steel Industry in Recurrent Crisis. Washington, DC:
Brookings Institution.
Crandall, Robert W. (1996). “From Competitiveness to Competition: The Threat of Minimills to
National Steel Companies,” Resources Policy, Vol. 22(1-2), 107-18.
de Melo, Jaime, and David Tarr. (1990) “Welfare Costs of US Quotas in Textiles, Autos and
Steel,” Review of Economics and Statistics, Vol. 72(3): 489-97.
Durling, James P., and Thomas J. Prusa. (2006) “The Trade Effects Associated with an
Antidumping Epidemic: The Hot-Rolled Steel Market, 1996-2001,” European Journal of
Political Economy, Vol. 22(3): 675-95.

26
Feenstra, Robert C. (1988) “Quality Change under Trade Restraints in Japanese Autos,”
Quarterly Journal of Economics, Vol. 103(1): 131-46.
Genesove, David, and Wallace P. Mullin. (1998) “Testing Static Oligopoly Models: Conduct and
Cost in the Sugar Industry, 1890-1914,” RAND Journal of Economics, Vol. 29(2): 355-377.
Gallet, Craig A. (1997) “Cyclical Fluctuations and Coordination in the US Steel Industry,”
Applied Economics, Vol. 29 (1997) 29, 279-285.
Grossman, Gene M. (1986) “Imports as a Cause of Injury: The Case of the US Steel Industry,”
Journal of International Economics, Vol. 20 (1986), 201-223.
Konings, Jozef, and Hylke Vandenbussche. (2005) “Antidumping Protection and Markups of
Domestic Firms,” Journal of International Economics, Vol. 65(1): 151-165.
Konings, Jozef, Hylke Vandenbussche, and Linda Springael. (2001) “Import Diversion Under
European Antidumping Policy,” Journal of Industry, Competition and Trade, Vol. 1(3): 283-299.
Krishna, Kala. (1989). “Trade Restrictions as Facilitating Practices,” Journal of International
Economics. 26: 251-270.
Lenway, Stephanie, Rehbein, Kathleen, and Starks, Laura. (1990) “The Impact of Protectionism
on Firm Wealth: The Experience of the Steel Industry,” Southern Economic Journal, Vol. 56(4):
1079-93.
Lenway, Stephanie, Randall Morck, and Bernard Yeung. (1996) “Rent Seeking, Protectionism
and Innovation in the American Steel Industry,” Economic Journal, Vol. 106(435): 410-21.
Liebman, Benjamin H. (2006) “Safeguards, China, and the Price of Steel,” Weltwirtschaftliches
Archiv, Vol. 142(2): 354-73.
Mai, Chao-cheng, and Hong Hwang. (1988) “On the Equivalence of Tariffs and Quotas under
Duopoly: A Conjectural Variations Approach,” Journal of International Economics, Vol.
24(3/4): 373-80.
Moore, Michael O. (1996) “Steel Protection in the 1980s: The Waning Influence of Big Steel?”
in Anne O. Krueger, ed., The Political Economy of Trade Protection. National Bureau of
Economic Research Project Report Series. Chicago and London: University of Chicago Press,
15-34.
Nieberding, James F. (1999) “The Effect of US Antidumping Law on Firms’ Market Power: An
Empirical Test,” Review of Industrial Organization, Vol. 14 (1): 65-84.

27
Prusa, Thomas J. (1997) “The Trade Effects of US Antidumping Actions," in Robert C. Feenstra,
ed., Effects of US Trade Protection and Promotion Policies, NBER Project Report Series.
Chicago,IL: University of Chicago Press, 191-213.
Puller, Steven L. (2007) “Pricing and Firm Conduct in California’s Deregulated Electricity
Market,” Review of Economics and Statistics, Vol. 89(1): 75-87.
Rotemberg, Julio J. and Garth Saloner. (1989) “Tariffs vs Quotas with Implicit Collusion,”
Canadian Journal of Economics, Vol. 22(2): 237-44.

28
Figure 1
US Domestic Steel Output and Employment, 1960-2006
0
100
200
300
400
500
600
700
1960
1962
196
4
1
966
1968
1970
1972
197
4
1
976
1978
1980
1982
198
4
1
986
1988
1990
1992
199
4
1996
1998
2000
2002
2004
2006
Comprehen-
sive VRAs
VRA
with
EC
Trigger
Price
Mech.
VRAs with
EC & Japan
Domestic
Employment
(in thousands)
Domestic Production
(in millions of net tons)
S
a
f
e
g
u
a
r
d
s
Notes: Data come from Annual Statistical Yearbook, American Iron and Steel Institute,
various issues.

29
Figure 2
Steel Industry Profitability, Steel Import Penetration and US Industrial Production, 1960-2006
-20
-10
0
10
20
30
40
1
960
1962
1
964
1966
1
968
1
97
0
1
972
1
974
1976
1
978
1980
1
982
1984
1
986
1988
1
990
1992
1
994
1996
1
998
2
00
0
2
002
2
004
2006
Percent
0
20
40
60
80
100
120
Indus. Prod. Index
Comprehen-
sive VRAs
VRA
with
EC
Trigger
Price
Mech.
VRAs with
EC & Japan
Imports as % of Market
US Industrial
Production Index
(1997=100)
Profitability
(Income as
% of Sales)
S
a
f
e
g
u
a
r
d
s
Notes: Import share and profitability data come from Annual Statistical Yearbook,
American Iron and Steel Institute, various issues. The US industrial production index
comes from Table B-51 of the most recent issues of the Economic Report of the
President.

30
Table 1: US Steel Trade Protection Events
1969-1974
Voluntary Restraint Agreements
(VRAs) with Japan and the EC.
1977-1981 Trigger Price Mechanism applied to
all imports.
1982 Antidumping (AD) and countervailing
duty (CVD) cases filed against EC
countries. Subsequently terminated
for VRAs on EC imports.
1984 AD and CVD cases filed against non-
EC countries. Subsequently termina-
ted for comprehensive VRAs.
1984-1989 Comprehensive VRAs with all signi-
ficant import sources.
1989-1992 Extension of VRAs.
1992-1993 AD and CVD cases filed against sig-
nificant import sources after VRAs
expire. AD and CVD remedies
applied to only subset of products.
1998-2000 Multiple AD and CVD cases against
Japan and other Asian countries.
2002-2003 Safeguard remedies in form of tariffs
placed on steel imports, excluding
FTA partners and developing countries.

31
Table 2: Share of US Steel Market by Top Foreign Country Import
Source for Selected Years
US Market Share (in percent)
Countries 1985 1995 2005
Japan 6.2 2.2 1.1
Canada 3.0 4.0 4.9
Germany 2.5 1.4 1.2
Korea 2.0 1.2 1.5
Brazil 1.8 1.9 2.1
France 1.7 0.8 0.5
Belgium-Luxembourg 1.1 0.3 0.5
Sweden 0.7 0.2 0.3
United Kingdom 0.6 0.6 0.6
Italy 0.6 0.3 0.4
Netherlands 0.6 1.0 0.5
Spain 0.6 0.3 0.3
Mexico 0.3 2.0 3.4
China 0.1 0.4 2.0
Russia 0.0 1.3 1.3
Turkey 0.0 0.4 1.1
India 0.0 0.0 0.6
Notes: Market shares calculated as foreign country's imports of semi-finished
and finished steel mill products (in tons) as a percent of US apparent supply
(shipments by domestic firms minus exports plus imports) of steel mill products
(in tons). We list any countries that were one of the top ten import sources for
any of the three reported years and order by the largest import sources in 1985.
Source: Annual Statistical Report of the American Iron and Steel Institute
(various issues). New York, NY: The American Iron and Steel Institute

32
Table 3: Three-stage Least Squares Estimates of a Three-Equation Model of the US Steel Market
(Demand, Import Supply, and Domestic Pricing): The Effects of Various Trade Policies
Predicted
Sign
Base Model
Effect of
Various
Trade
Policies
Effect of
Various VRA
Periods
Equation 1: Domestic Demand
Steel Price -
-6.164
***
-6.767
***
-6.667**
(1.637) (1.535) (1.585)
Industrial Production
+ 17.639
25.058
*
26.081
*
(13.567) (13.481) (13.950)
Aluminum Price
+ 0.498
*
0.340 0.345
(0.272) (0.242) (0.274)
Steel Price * Industrial Production
? 0.078
***
0.080
***
0.082
***
(0.019) (0.016) (0.018)
Steel Price * Aluminum Price
? -0.0005 -0.0003 -0.0004
(0.0003) (0.0003) (0.0003)
Trend
? -24.427 -48.449 -56.172
(54.850) (53.183) (55.234)
Equation 2: Import Fringe Supply
Steel Price +
1.626
**
1.719
**
1.916
**
(0.809) (0.734) (0.811)
Real Exchange Rate
+ 22.047
***
22.417
***
23.350
***
(5.378) (5.120) (5.282)
Rest-of-World Real GDP
- -263.652
***
-270.212
***
-245.077
***
(78.187) (73.403) (76.91)
Foreign Ore Price
- 3.698 3.807 7.609
(3.100) (3.014) (3.678)
Foreign Oil Price
- -1.235 -1.005 -2.219
(1.473) (1.441) (1.573)
Foreign Coal Price
- -2.283 -2.584 -5.487
*
(2.448) (2.386) (2.924)
Tariff Rate
- -48.668
*
-47.367
*
-54.861
**
(26.067) (26.295) (26.442)
Antidumping Duty Rate
- -20.078
***
-20.342
***
-20.080
***
(3.618) (3.607) (3.640)
Countervailing Duty Rate
- 5.212 6.725 5.716
(4.646) (4.693) (4.688)
Safeguard Tariffs
- -19.179
***
-17.448
***
-16.991
***
(6.064) (6.144) (6.173)

33
VRA Period, 1983-1991
- -179.883 -203.343
**
(115.8) (112.9)
VRA Period, 1983-1984
- -167.293
(146.5)
VRA Period, 1985-1989
- -360.554
**
(145.9)
VRA Period, 1990-1991
- -238.465
*
(142.4)
Trend
? 186.769
***
189.451
***
158.558
***
(43.812) (41.815) (45.338)
Equation 3: Domestic Pricing
Electricity Price
+ 438.095 552.735 577.249
**
(229.8) (272.6) (255.6)
Ore Price
+ -9.800 -1.289 -1.052
(25.111) (27.089) (26.721)
Scrap Price
+ 3.924
*
3.586 2.648
(2.381) (2.717) (2.687)
Coal Price
+ -95.952 -128.297 -133.236
(75.497) (85.508) (81.983)
Steel Wage
+ 28.301
***
24.901
***
22.793
**
(10.239) (9.321) (10.350)
Continuous Cast
- -4.575
***
-3.992
**
-3.839
**
(1.701) (1.907) (1.932)
Trend
? 0.653 3.027 3.127
(6.536) (3.985) (6.426)
Domestic Quantity
? -0.029
***
-0.019
***
-0.020
**
(0.006) (0.004) (0.008)
Weighted Domestic Quantity
+ 0.0003 0.0015 0.0007
(0.0007) (0.0010) (0.0015)
Weighted Domestic Quantity × Tariff
+ -0.0006
(0.0007)
Weighted Domestic Quantity × AD Duty
+ 0.00004
(0.00004)
Weighted Domestic Quantity × CVD Duty
+ -0.0010
*
(0.0005)
Weighted Domestic Quantity × Safeguard
Tariffs + -0.00003
(0.00007)
Weighted Domestic Quantity × VRA Period
(1982-1991) + 0.0332
**
(0.0150)

34
Weighted Domestic Quantity × VRA Period
(1982-1984) + 0.0438
*
(0.0236)
Weighted Domestic Quantity × VRA Period
(1985-1989) + 0.0417
**
(0.0208)
Weighted Domestic Quantity × VRA Period
(1990-1991) + 0.0104
(0.0159)
Factor-price Interactions in Equation 3 Yes Yes Yes
Product Fixed Effects in Each Equation
Yes Yes Yes
Observations 540 540 540
R
2
- Domestic Demand 0.92 0.92 0.92
R
2
- Import Fringe Supply 0.73 0.73 0.73
R
2
- Domestic Pricing 0.85 0.85 0.86
Notes: Data sample is panel of annual observations across 20 major steel products from 1980
through 2006. Product fixed effects are included in each equation and as part of the instrument
set. The dependent variable for the first equation is total steel products demanded for a given
product-year combination in the US market in tons; the dependent variable for the second
equation is total imports of a product-year combination in tons; the dependent variable for the
third equation is the price for a product-year combination in US dollars per ton. The data
appendix provides details on data construction and sources. Standard errors are in parentheses
and ***, **, and * denote statistical significance of a coefficient at the 1%, 5% and 10% levels,
respectively. Chi-squared statistics are not reported in the table, but easily reject the null
hypothesis of jointly insignificant coefficients for all equations across all specifications. Fixed-
effects coefficient estimates for the twenty steel products are jointly statistically significant in
each of the three equations, but not displayed in order to save space. Likewise, the cost
interaction terms in the pricing equation are jointly significant in all specifications, but not
displayed to save space.

35
Table 4: Estimated Mark-ups from Various Regression Specifications
A. Mark-ups for Various Trade Policies Effects Estimates from Column 2 of Table 3
Baseline 3.4%
Tariff - For Each Percentage Point Increase -1.0%
AD Duty - For Each Percentage Point Increase 0.1%
CVD Duty - For Each Percentage Point Increase -1.7%
*
Safeguard Tariff - For Each Percentage Point Increase -0.1%
VRA Period from 1983-1991 82.8%
**
B Mark-ups for Various VRA Periods Using Estimates from Column 3 of Table 3
Baseline 1.1%
VRA Period from 1983-1984 111.6%
*
VRA Period from 1985-1989 98.0%
**
VRA Period from 1990-1991 27.1%
C. Mark-ups during the VRA Period: Integrated versus Minimill Products
Integrated Products
Baseline 0.6%
VRA, 1983-1991 76.5%
**
Minimill Products
Baseline -0.4%
VRA, 1983-1991 -24.2%
D. Evolution of Mark-ups Over Time: Integrated versus Minimill Products
Integrated Products, Controlling for VRA Period Effects
Baseline – 1980 16.8%
Annual Change in Market Power after 1980 -0.6%
Minimill Products, Controlling for VRA Period Effects
Baseline - 1980 56.7%
**
Annual Change in Market Power after 1980
-1.6%
**
Notes: ***, **, and * denote statistical significance of a coefficient at the 1%, 5% and 10%
levels, respectively.

36
Table 5: Estimates of θ Over Three-Year Windows
Three-Year Period θ P-Value of θ Estimate
1980 - 1983 -0.0347 0.321
1981 - 1984 -0.0235 0.215
1982 - 1984 0.0216 0.184
1983 - 1985 0.0123 0.375
1984 - 1986 0.0071 0.577
1985 - 1987 -0.0036 0.793
1986 - 1988 0.0098 0.383
1987 - 1989 0.0356 0.030
1988 - 1990 0.0189 0.276
1989 - 1991 0.0121 0.323
1990 - 1992 -0.0138 0.285
1991 - 1993 -0.0010 0.904
1992 - 1994 0.0086 0.282
1993 - 1995 0.0046 0.538
1994 - 1996 -0.0002 0.964
1995 - 1997 -0.0053 0.202
1996 - 1998 -0.0020 0.609
1997 - 1999 0.0006 0.672
1998 - 2000 0.0013 0.238
1999 - 2001 0.0009 0.189
2000 - 2002 0.0007 0.389
2001 - 2003 -0.0004 0.482
2002 - 2004 -0.0009 0.125
2003 - 2005 -0.0018 0.207
2004 - 2006 0.0135 0.218
Notes: These estimates are of our market power parameter (θ) over
all continuous three-year periods in our sample using the same
specification as that in Column 2 of Table 3.

37
Data Appendix
In this appendix, we first describe data construction and sources for the respective dependent
variables in our three-equation system, and then provide similar details for the control
variables in each equation in the order they are listed in Table 3 of the paper. At the end of
the section, we provide a table of basic descriptive statistics for each variable.
Dependent variables
1) Domestic Demand Equation: US market quantity. Annual Statistical Report, American
Iron and Steel Institute, various issues. To get total shipments by domestic US steel
producers, we record “Net Shipments” for each product-year observation which is “Gross
Shipments” by US steel plants minus “shipments from one reporting company to another
reporting company for conversion, further processing or resale.” We then add imports
(discussed next) and subtract export shipments using data from the same publication to get
total domestic quantity sold in the US measured in thousands of tons.
2) Import Fringe Supply Equation: Import quantity. Annual Statistical Report, American
Iron and Steel Institute, various issues. Galvanized and Other Metal Coated sheets were
combined figures from 1980-1982. We use the 1983 percent (88% is Galvanized) to estimate
separate quantities for these two products over the 1980-1982 period. This variable is
measured in thousands of tons.
3) Domestic Markup Equation: US steel price. For a variety of reasons described in the
text, we rely on steel prices provided by Purchasing Magazine, available for a fee at
http://www.purchasingdata.com/
. These data provide monthly spot prices for a number of
specific steel products (in US dollars per ton) back to January 1980. Table A.1 shows the
concordance between the steel products in our data sample and the price series used from
Purchasing Magazine. Where there was not an exact match, we used the price series that had
the largest correlation between the two products in another price series discussed in the text –
unit values from Current Industrial Reports. These correlations were always above 0.8. We
took the simple average of the monthly price series to create annual prices and deflated by
the US GDP deflator available from the Economic Report of the President, Table B-3, to
convert into real US dollars per ton.
Table A.1: Concordance for our product-level US price data
Product Categories in our Data
Purchasing Magazine Series Used
Wire Rods Wire Rod
Plates Hot Rolled Plate (coiled)
Hot-rolled Sheet and Strip Hot Rolled Sheet
Cold-rolled Sheet and Strip Cold Rolled Sheet
Sheet & strip - Galvanized Hot Dipped Galvanized Sheet
Tin Plate Cold Rolled Sheet
Bars - Hot-rolled Wide Flange Beams
Bars - Light structurals (under 3'') Reinforcing Bar
Bars - Reinforcing Reinforcing Bar
Bars - Cold finished Cold Finished Bar

38
Black plate Hot Rolled Plate (coiled)
Ingots, Blooms, Billets, Slabs, etc Wire Rod
Rails (Standard and Other) Hot Rolled Plate (coiled)
Sheet piling Wide Flange Beams
Sheets & Strip - Metallic Coated Cold Rolled Sheet
Sheets & Strip - Electrical Cold Rolled Sheet
Structural Shapes - Heavy Wide Flange Beams
Tin Free Steel Cold Rolled Sheet
Tool Steel Reinforcing Bar
Wire - Drawn Cold Rolled Sheet
Explanatory variables – Domestic Demand Equation
Our measure of US production is the production index for the manufacturing sector reported
in the Economic Report of the President, Table B-51. Data on US aluminum prices through
2004 come from the US Geological Survey, which are available online at
http://minerals.usgs.gov/ds/2005/140/#steelscrap
. All prices are expressed in terms of 1998
dollars per ton. We then used price data from recent volumes of the US Geological Survey’s
Minerals Yearbook to construct prices in 1998 dollars per ton for years 2005 and 2006. The
trend term in all three equations is constructed as the year minus 1979, the year prior to the
start of our sample.
Explanatory variables – Import Fringe Supply Equation
Our measure of the US real exchange rate is the “Broad Index of Real Exchange Rate of the
US Dollar” from US Federal Reserve Board official statistics. Monthly data were averaged
to generate yearly observations. These data are available at
http://www.federalreserve.gov/releases/h10/summary/
. Real world GDP measured in
constant 2000 US trillions of dollars and taken from the World Bank’s World Development
Indicators. Data on world iron ore, oil, and coal prices comes from the International
Monetary Fund’s International Financial Statistics data on world commodity prices. The
iron ore price is an index of North Brazilian port prices, coal is an index of Australian prices,
and oil is an average index of three spot prices – Dubai Fateh, U.K. Brent and West Texas
Intermediate. These prices are deflated using the US GDP deflator to express them in real
terms. Steel tariffs are measured as trade-weighted import tariffs and calculated from official
US Census trade statistics on US 7-digit Tariff System of the United States (prior to 1989)
and 10-digit Harmonized System product codes. Various issues of the US Census’s Current
Industrial Reports: Steel Mill Products were used to concord import product codes into
consistent classifications over the sample. US antidumping and countervailing duties are
constructed as trade-weighted averages across import sources for a given steel product. Data
for these came from Blonigen and Wilson (2005). The VRA and safeguard measures are
dummy variables that take the value of “1” for the years noted in the tables.
Explanatory variables – Import Fringe Supply Equation
Our measure of the US real exchange rate is the “Broad Index of Real Exchange Rate of the
US Dollar” from US Federal Reserve Board official statistics. Monthly data were averaged
to generate yearly observations. These data are available at
http://www.federalreserve.gov/releases/h10/summary/
. Real world GDP measured in

39
constant 2000 US trillions of dollars and taken from the World Bank’s World Development
Indicators. Data on world iron ore, oil, and coal prices comes from the International
Monetary Fund’s International Financial Statistics data on world commodity prices. The
iron ore price is an index of North Brazilian port prices, coal is an index of Australian prices,
and oil is an average index of three spot prices – Dubai Fateh, U.K. Brent and West Texas
Intermediate. These prices are deflated using the US GDP deflator to express them in real
terms. Steel tariffs are measured as trade-weighted import tariffs and calculated from official
US Census trade statistics on US 7-digit Tariff System of the United States (prior to 1989)
and 10-digit Harmonized System product codes. Various issues of the US Census’s Current
Industrial Reports: Steel Mill Products were used to concord import product codes into
consistent classifications over the sample. US antidumping and countervailing duties are
constructed as trade-weighted averages across import sources for a given steel product. Data
for these came from Blonigen and Wilson (2005). The VRA and safeguard measures are
dummy variables that take the value of “1” for the years noted in the tables.
Explanatory variables – Domestic Markup Equation
Measures of electricity and coal prices come from the US Energy Information
Administration. Real prices of electricity for industrial purposes are available at
http://www.eia.doe.gov/emeu/aer/elect.html
as the “Industrial – Real” column of Table 8.10.
Coal prices are available at http://www.eia.doe.gov/emeu/aer/coal.html
, and are expressed in
terms of 1998 dollars per ton. Data on US iron ore and scrap steel prices through 2004 come
from the US Geological Survey, which are available online at
http://minerals.usgs.gov/ds/2005/140/#steelscrap
. All prices are expressed in terms of 1998
dollars per ton. We then used price data from recent volumes of the US Geological Survey’s
Minerals Yearbook to construct prices in 1998 dollars per ton for years 2005 and 2006.
Wages for the steel industry come from various issues of the
Annual Statistical Report,
American Iron and Steel Institute, various issues. We use “Total Employment Cost per
Hour” as reported in Table 6 of most recent issues and deflated by the US GDP deflator to
convert into real terms. Data on the percent of all products produced with continuous casting
methods are available from the Annual Statistical Report, American Iron and Steel Institute,
various issues. The domestic quantity variable comes from Annual Statistical Report of the
American Iron and Steel Institute, various issues. To get total shipments by domestic US
steel producers, we record “Net Shipments” for each product-year observation which is
“Gross Shipments” by US steel plants minus “shipments from one reporting company to
another reporting company for conversion, further processing or resale.” The weighted
domestic quantity variable is the domestic quantity weighted by the inverse of the partial
derivative of the US market quantity with respect to the steel price (from the domestic
demand equation) minus the partial derivative of the import quantity with respect to the steel
price (from the import fringe supply equation) as noted in the text. We constrain the
coefficients in estimation of the domestic demand and import fringe supply to be the same as
those used in constructing the partial derivatives for this weighted domestic quantity variable.
In table A.2 we provide (in order) descriptive statistics for all the variables appearing in
column 1 of table 3, our base model. We don’t report the weighted domestic quantity
variable in the last equation, since it depends on estimated parameters.

40
Table A.2. Basic Descriptive Statistics of Variables Used for Estimation
Variable | Mean
Standard
Deviation Minimum Maximum
Domestic Demand Equation
US Steel Market Quantity 5155.57 5863.91 43.00 26283.00
Steel Price 486.06 147.88 221.76 948.93
Industrial Production 76.90 20.53 48.50 113.00
Aluminum Price 2470.53 1305.84 1215.43 6476.33
Steel Price * Industrial Production
36084.40 11425.53 16749.74 85012.94
Steel Price * Aluminum Price
1315896.78 1104371.94 323895.10 6041762.35
Trend 14.00 7.80 1.00 27.00
Import Fringe Supply Equation
Import Quantity 1096.54 1509.94 1.00 9320.00
Steel Price 486.06 147.88 221.76 948.93
Real Exchange Rate 99.14 9.50 86.67 122.59
Rest-of-World Real GDP
18.42 4.09 12.56 26.46
Foreign Ore Price
131.52 33.37 97.79 231.04
Foreign Oil Price
108.90 49.89 48.00 234.02
Foreign Coal Price
177.84 58.05 100.00 310.27
Tariff Rate 3.21 2.53 0.00 10.14
Antidumping Duty Rate 7.03 13.07 0.00 52.40
Countervailing Duty Rate 2.17 9.53 0.00 56.17
VRA Period, 1983-1991
0.33 0.47 0.00 1.00
Safeguard Tariffs
0.69 4.43 0.00 30.00
Trend 14.00 7.80 1.00 27.00
Domestic Markup Equation
Steel Price 486.06 147.88 221.76 948.93
Electricity Price 7.20 2.84 4.39 12.77
Ore Price 52.13 31.65 22.07 126.20
Scrap Price 155.51 55.50 67.29 334.55
Coal Price 34.08 19.83 15.77 84.40
Steel Wage 36.09 2.66 32.25 41.52
Continuous Cast 73.14 26.15 20.30 97.30
Domestic Quantity 4256.27 5097.45 18.00 22767.00
Trend 14.00 7.80 1.00 27.00