Energy prices and the expansion of world trade

Review of Economic Dynamics 11 (2008) 904–916

Contents lists available at ScienceDirect

Review of Economic Dynamics

www.elsevier.com/locate/red

Energy prices and the expansion of world trade ✩

Benjamin Bridgman ∗

Bureau of Economic Analysis, U.S. Department of Commerce, Washington, DC 20230, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 20 October 2006Revised 18 July 2008Available online 9 August 2008

JEL classification:F1L9

Keywords:Trade growthTransport costsOil shocks

The share of merchandise output that is internationally traded has significantly increasedwhile tariffs have fallen. However, standard trade models have surprising difficulty linkingthese two facts. Trade growth slowed in the 1970s as tariffs fell relatively sharply whileafter the late 1980s trade grew quickly as tariffs fell slowly. This pattern implies thatthe price-import elasticity has changed over time. Also, tariffs have not fallen enoughto generate such a large increase in trade given estimates of this elasticity. Changes intransport costs can resolve both puzzles. I present a vertical specialization trade modelwith an energy-using transportation sector. In the simulated model, trade growth slowsfrom 1974 to 1985. The oil shocks raised transport costs, offsetting falling tariffs, so theprice-import elasticity no longer needs to change. It also generates the observed volume oftrade growth since transport costs have fallen over the long run.

Published by Elsevier Inc.

1. Introduction

The share of merchandise output that is internationally traded has significantly increased since World War Two. At thesame time, successive rounds of General Agreement on Tariffs and Trade (GATT) negotiations have reduced tariffs. It isintuitive that the two phenomena are related.

Fig. 1 shows U.S. merchandise export share of GDP and industrial world merchandise tariffs.1 Superficially, the twoappear related. However, standard trade models have not been able to deliver the result that lower tariffs were responsiblefor postwar trade growth. They fail to deliver both the magnitude and pattern of trade expansion.

Tariffs have not fallen enough to generate such a large increase in trade. An aggregate elasticity of substitution betweendomestic and foreign goods or Armington elasticity (Armington, 1969) of around 11 is required to generate actual tradeexpansion in standard models. Estimates of this elasticity fall in a wide range of values depending on the data and method-ology, but most fall well below 11. (Erkel-Rousse and Mirza (2002) provide a summary of this literature.) High frequencyestimates, such as Blonigen and Wilson (1999) and Broda and Weinstein (2006), typically estimate low price-import elastic-ities of 1 or 2. Lower frequency estimates, such as those found in Head and Ries (2001) and Feenstra (1994), tend to findhigher elasticities although they are still only about half of the level required.

✩ I thank Sami Alpanda, Adam Copeland, Thomas Holmes, Adrian Peralta-Alva, James Schmitz Jr. and seminar participants at BEA, Federal Reserve Board,CNA Corporation, Louisiana State University and the Fall 2005 Midwest International Economics Group meetings at the University of Kansas for helpfulcomments. Comments from the editor Tim Kehoe and three referees substantially improved the paper. Kei-Mu Yi generously provided data. The viewsexpressed in this paper are solely those of the author and not necessarily those of the U.S. Bureau of Economic Analysis or the U.S. Department ofCommerce.

* Address for correspondence: Bureau of Economic Analysis, U.S. Department of Commerce, 1441 L Street NW, Washington, DC, USA. Fax: +1 (202) 6065366.

E-mail address: [email protected] Detailed information about data sources is available in Appendix A.

1094-2025/$ – see front matter Published by Elsevier Inc.doi:10.1016/j.red.2008.07.006

http://www.ScienceDirect.com/

http://www.elsevier.com/locate/red

mailto:[email protected]

http://dx.doi.org/10.1016/j.red.2008.07.006

B. Bridgman / Review of Economic Dynamics 11 (2008) 904–916 905

Fig. 1. Tariffs and U.S. export share.

In addition, tariffs have fallen steadily since the late 1960s while trade does not show steady growth. Trade growth showsno sustained growth from 1974 to 1986. The response of trade of falling tariffs is much stronger beginning in the late 1980s,which requires either a strong non-linear response to trade costs (a feature absent from standard models) or a large shift inimport-price elasticity in the mid-1980s, perhaps by as much as sevenfold (Yi, 2003).

A number of explanations for these puzzles have been proposed, but have failed to fully account for trade growth. Yi(2003) suggests trade in intermediate goods, or vertical specialization (VS), as a solution. However, the model simulations inYi (2003) capture only about half of trade growth from 1962 to 1999, leaving a great deal of trade to be explained. Bergoeingand Kehoe (2003) show that “New Trade Models” emphasizing increasing returns and monopolistic competition, such asKrugman (1979) and Helpman (1981), or non-homothetic preferences (Markusen, 1986) are also quantitatively unable todeliver the observed trade increase.

Changes in transport costs can resolve both of these puzzles. Transport costs have fallen since the 1960s, which explainsincreasing trade volumes. The slowdown coincides with a period of elevated oil prices. Higher energy prices raised the costsof transportation firms, which led to higher trade costs and slower trade growth. While the transport cost explanation maybe intuitive, the crucial test it must pass is whether the changes are quantitatively important to trade growth.

Oil shocks must be large enough to significantly increase transport costs. Oil shocks have often been suggested as asource of a number of economic disturbances, including the productivity slowdown and stagflation.2 However, it has beendifficult to develop models where the effects of energy price changes are quantitatively important, since energy expensesrepresent a small portion of production costs. Without a significant magnifying mechanism, energy is usually too narrow achannel to have large effects on macroeconomic variables. (See Barsky and Kilian (2004) for a recent discussion.) This is nottrue for transportation since energy expenses represent a much larger portion of inputs than for the overall economy.

Changes in trade costs also have to be sufficiently large to account for the observed increase in trade. As discussed above,standard models with tariffs fail to predict the amount of trade growth. Incorporating changes in transportation into a VSmodel passes both tests and can resolve both puzzles in trade growth.

This paper presents a tractable general equilibrium model with Ricardian trade in intermediate goods, similar to Dorn-busch et al. (1977) and Eaton and Kortum (2002). There is a continuum of intermediate goods that can be assembled intoconsumption goods. There are two countries that each produce a distinct consumption good. Trading both types of goodsrequires a shipping technology that uses energy as an input. I calibrate the model and run simulations using data on energyprices and tariffs.

In the simulated model, trade growth slows from 1974 to 1985. The increase in oil prices led to higher transport coststhat offset the decline in tariffs. Total trade costs (tariffs plus transport costs) were constant beginning in the mid-1970s.When energy prices began to fall in 1982, total trade costs also fell. The pattern of trade growth generated by the model ismuch closer to the observed pattern of trade expansion. Once transport costs are accounted for, the price-import elasticityno longer needs to change.

Accounting for falling transport costs significantly increases the predicted amount of trade growth. The model predictstrade expansion equal to observed trade growth. Transportation industries have developed a number of new technologiesthat have increased their productivity faster than goods producing industries and reduced the cost of shipping goods. Themodel also matches the composition of trade growth, generating an increasing role for VS trade.

2 The literature on the effect of the oil shocks on the economy is very large. Earlier work includes Hamilton (1983). More recent work includes Rotembergand Woodford (1996), Barsky and Kilian (2002), and Nordhaus (2004).

906 B. Bridgman / Review of Economic Dynamics 11 (2008) 904–916

A number of papers have modeled transportation as a production sector as opposed to an iceberg cost. Examples includeFalvey (1976) and Ruiz-Mier (1990) for trade and Ravn and Mazzenga (2006) for international business cycles. They differfrom this paper since they do not focus on the quantitative implications of transportation on trade.

While many authors have investigated postwar trade growth, including Rose (1991), Krugman (1995), Baier andBergstrand (2001), Yi (2003), and Bergoeing and Kehoe (2003), no work that I am aware of considers the effect of theoil shocks on the pattern of trade expansion. Backus and Crucini (2000) examine the effects of oil shocks on terms of trade,drawing some implications for trade balance, but do not focus on the historical expansion of trade.

The organization of the rest of the paper is as follows: Section 2 presents empirical evidence on transportation. Section 3presents the model. Section 4 defines equilibrium and solves the model. Section 5 calibrates the model and presents theresults. Section 6 concludes.

2. Transportation facts

This section establishes the empirical foundation for the model choice. The argument rests on the following observations:1. Transport costs are quantitatively important; 2. Energy is an important input to transportation; 3. Energy is difficult tosubstitute away from; and 4. Higher energy costs are associated with higher transport costs.

Transport costs have been a central theme in the study of international trade. For example, North (1958) and Estevade-ordal et al. (2003) argue that transport costs are a key factor in trade patterns in the late 19th Century and interwar periodrespectively. Despite this, transport costs have received less attention than tariffs. An important reason for this neglect is thelack of good data. Systematic measures are either lacking or of poor quality (Hummels and Lugovskyy, 2005). Notably forfocus of this paper, the U.S. time series begins after the onset of the oil crisis (Feenstra, 1996). The available data indicatethat the barriers due to transport costs are a similar magnitude to those from tariffs. For example, average import weightedU.S. tariffs for manufacturing in 1992 were 4.2 percent compared to transport costs of 4.1 percent (Bernard et al., 2006).

Energy price changes have a large impact on transport costs since they represent an important part of expenses inthe transportation business. Table 1 shows nominal expenditure on energy as a share of revenue for various modes oftransportation. Transportation is more energy intensive than the rest of the economy, especially for water and air, the mostimportant modes for international trade.

Energy prices pass through to the final cost of transportation since it is difficult to substitute away from energy. Fig. 2shows energy share in various transportation modes compared with the price of fuel oil in water transportation. The patternof energy’s share of revenue matches the price of energy very closely, indicating that shippers cannot substitute away fromenergy. This is a feature of most industries (Atkeson and Kehoe, 1999).

Table 1Energy share of revenue in transportation, 2000

Mode Share (percent)

Water 10.5Air 16.6Trucking 10.5Rail 9.5Business sector 7.6

Fig. 2. Energy price and revenue share: Various transportation industries.


The oil shocks in the 1970s led to large and sustained increases in energy prices. After a slow decline during the 1960s,the relative price doubled in response to the Oil Embargo in late 1973. (See Fig. 2.) Energy prices remained elevated untilthey collapsed in 1986. Recently, prices have begun to climb.

Fuel prices are an important determinant of transport costs. Lundgren (1996) regresses bulk goods freight rates on fuelprices and finds that the coefficient on fuel prices is significant and positive in explaining transit prices. For the coal andgrain trades between the U.S. and Europe, a 1% increase in fuel prices lead to about a 0.4% increase in freight rates. Sletmoand Williams (1981) examine the accounts of container liners engaged in the North Atlantic trade and find evidence thatthe energy price increases of the 1970s led to a substantial increase in cost for shippers.

While there were increases in the 1970s, overall transport costs have declined since 1960. Finger and Yeats (1976) reportthat U.S. trade-weighted transport costs in 1965 were 9 percent of import value while Hummels (2001b) finds that the 1994value to be 3.8 percent. At a more disaggregated level, Lundgren (1996) examines freight rates in bulk commodity trade andfinds that rates have declined by almost 70 percent since the early 1950s.

3. Model

3.1. Households

There are two countries each with a representative household. Households have preferences over a consumption goodrepresented by:

U =[ ∑

j=1,2

(C i

j

)1−σ] 1

1−σ

(1)

where C ij denotes consumption good j ∈ {1,2} for country i ∈ {1,2}. The associated prices are P i

j . Each country is endowed

with labor Ni . The wage is W i .

3.2. Intermediate goods sector

There is a continuum of intermediate goods xi(z) with a price P i(z) for z ∈ [0,1]. Each country is endowed with tech-nologies to produce the intermediate goods, where total intermediate output yi(z) is given by yi(z) = Ai(z)Ni(z). Theproductivity parameters are given by A1(z) = 1

(1+z)θand A2(z) = 1

(2−z)θ, a variant of the mirror image technology in Yi

(2003).

3.3. Consumption goods sector

Intermediate goods can be assembled into consumption goods. Each country can only produce the good with its name:j = i. The total output is given by the technology:

Y ij =

1∫0

ln(xi(z)

)dz (2)

for i = 1,2 and j = i.

3.4. Energy sector

Each country owns a technology to produce an energy good Ei using labor: Ei = AiE Ni

E . The price of energy is P iE .

3.5. Transportation sector

The countries may trade the goods they produce with each other using a transportation technology specific to that good.Transportation of the good requires both energy E and the good itself, both of which must be purchased at the countryof origin. The amount of energy and the good satisfies: xi(z) = Min{φi(z)Ei(z), Bxi

T (z)} where xi(z) is the amount of goodimported, φi(z) and B are productivity parameters, Ei(z) is the energy used and xi

T (z) is the amount of intermediate goodconsumed in shipping. In each country, φi(z) = φ Ai(z). This functional form is selected so that the ad valorem transportcost is the same for all goods.

There is a similar unit transportation requirement for consumption goods given by Min{φC EiC , BC i

T }.

3.6. Government

The countries each have a government that can impose an ad valorem (net of transport fees) tariff τ i on traded goods.The government gives the domestic representative household transfers T i and maintains budget balance.


4. Equilibrium

4.1. Definition

Households sell labor and purchase goods. They maximize U subject to the budget constraint∑j

P ij C

ij = W i Ni + T i (3)

Energy firms buy labor and sell energy. They face competitive markets and solve:

Max P iE Ai

E NiE − W i Ni

E (4)

Intermediate goods firms face competitive markets and solve:

Max P i(z)Ai(z)Ni(z) − W i Ni(z) (5)

For j = i, consumption goods firms solve:

Max P ij

1∫0

ln(xi(z)

)dz −

1∫0

P i(z)xi(z)dz (6)

Transportation firms buy domestic goods and energy and sell exports. Intermediate goods exporters face competitivemarkets and solve:

Max P−i(z)Min{φi(z)Ei(z), Bxi

T (z)} − P i(z)xi

T (z) − P iE Ei(z) − P i(z)Min

{φi(z)Ei(z), Bxi

T (z)}

(7)

where P−i(z) is the price of the good in the other country. Consumption goods exporters solve a similar problem.Feasibility for each consumption good requires that for j = 1,2:

C jT +

∑i=1,2

C ij = Y j

j (8)

There is a corresponding feasibility constraint for intermediate goods that are exported.Feasibility for energy production requires that:

AiE Ni

E = Ei = EiC +

1∫0

Ei(z)dz (9)

The definition of equilibrium is standard.

Definition 4.1. Given tariffs, an equilibrium is consumption, intermediate goods and energy allocations and prices in eachperiod such that:

1. Households solve their problem.2. Consumption and intermediate goods, transportation and energy firms solve their problem.3. The government balances its budget.4. The allocation is feasible.

4.2. Solution

The two countries are mirror images in intermediate goods production. There is a symmetric equilibrium with a closedform solution when the parameters are the same in the two countries. Specifically, if the parameters Ni, τ i , and Ai

E areconstant across the two countries, there exists an equilibrium where C1

1 = C22 , C2

1 = C12 , P 1

E = P 2E , W 1 = W 2, P 1

2 = P 21 and

P 11 = P 2

2 . Prices and quantities in the intermediate goods sectors across the countries mirror each other: P 1(z) = P 2(1 − z),etc. In the rest of the paper, I examine this symmetric equilibrium.

I denote the common parameters and quantities (for example, Ni and W i ) by omitting the i superscript (for example,τ 1 = τ 2 = τ ) and normalize price of country one’s most productively produced intermediate good to one (P 1(0) = 1). Thisimplies that the wage W 1 = 1. Define zi as the cutoff industry in country i such that intermediate goods z > z1 and z < z2will be imported. Given the functional forms,

z1 = 1 − z2 = 2(1 + τ + 1B + 1

AE φ)

1θ − 1

(1 + τ + 1 + 1 )1θ + 1

(10)
B AE φ


Intermediate exports are given by:

z2

(1 + τ + 1B + 1

AE φ)

(N + T )[1 + 1B + (1 + τ + 1

B + P EφC P )

1σ ]

[1 + τ + 1B + P E

φC P + (1 + τ + 1B + P E

φC P )1σ ]

(11)

where P 11 = P 2

2 = P and P 1E = P 2

E = P E .Consumption goods exports are given by:

C12 = C2

1 = N + T

P [1 + τ + 1B + P E

φC P + (1 + τ + 1B + P E

φC P )1σ ]

(12)

Tariffs in the United States are collected on the FOB value of goods (the value before transport costs are added). Therefore,

N + T = N[1 + τ + 1B + P E

φC P + (1 + τ + 1B + P E

φC P )1σ ]

[1 + 1

B + P EφC P + (1 + τ + 1

B + P EφC P )

1σ − z2(1+ 1

B +(1+τ+ 1B + 1

AE φ)

1σ )

(1+τ+ 1B + 1

AE φ)

] (13)

5. Simulations

5.1. Parameters

In this section, I parameterize the model and solve a sequence of the static model (setting P 1(0) = 1 in each period)using the observed time series of tariffs and energy prices over the period 1960 to 2005. Model simulations are reported in1960 dollars to remove (very minor) changes in the relative prices of final and intermediate goods.

As in Yi (2003), the two countries in the model are interpreted as the United States and the rest of the industrializedcountries (European Community and Japan). The calibration moments are selected to match U.S. data.

The measure of export share I use for empirical comparison is merchandise export share of GDP. U.S. share of worldGDP has declined, which would tend to increase export share in a symmetric model. On the other hand, merchandise(Agriculture, Mining and Manufacturing) output has declined as a share of GDP, which should reduce exports. As noted inBergoeing and Kehoe (2003), these effects are largely offsetting so this measure is less vulnerable to composition effects.3 Inaddition, real annual data is available for GDP over the entire sample period, which is not true of merchandise production.

The measure of tariffs is an adjusted version of the world tariff series from Yi (2003), which was constructed to estimatemanufacturing tariffs. Merchandise tariffs start at a lower rate (12.075 percent vs. 13.95 for manufacturing) and declineslightly less (10.2 percentage points vs. 10.9 for manufacturing), so I adjust Yi’s series to reflect merchandise tariffs, thedetails of which are in Appendix A. I set all tariffs to be the same across goods as a baseline assumption. The implicationsof this assumption are discussed below.

When selecting parameters for the transportation sector, I use U.S. data when available. When it is not available, I usedata from Canadian water transportation. I concentrate on water transportation since it is the most important mode ininternational trade. About two thirds of U.S. imports (by value) arrived by water over most of the period considered.4 Watertransportation is also in the middle of pack in energy expenditure, with energy expenses being a higher share of revenuefor air and a smaller share for rail and trucks. Therefore, it should be a reasonable proxy for the sector as a whole. I useCanadian data since Canada is the only country that I am aware of that provides detailed statistics on revenue and costs inwater transportation extending back to 1960.

I set AE such that the price of energy in each country is equal to the real price of fuel oil used in Canadian watertransportation in each period (AE = 1

P E). Nearly all modes use petroleum-based fuels to move goods, so changes in fuel oil

prices are a good proxy for the energy price changes faced by other modes. The data also reflect actual transaction pricesrather than spot rates.

I normalize the labor endowment for each country Ni = 1. The value of parameters φ and B in 1960 are selected tomatch the energy share in the transportation industry ES and the transport cost TC. The parameters are selected so thatthe ad valorem cost of transport is the same for all goods. This abstracts from good specific transport costs, just as a singletariff abstracts from good specific tariffs.

I calibrate the parameters in the transportation sector for the most productive intermediate good (z = 0 for country 1and z = 1 for country 2). The functional forms selected means that all intermediate goods face the same ad valorem tradecosts. I set the parameters in consumption goods transportation so that this sector also generates the same ad valorem tradecosts. That is, φC is set so that P E

PφC= P E

φ1(0)= P E

φ2(1)= P E

φ.

3 A rough estimate of the effect is (1 − U.S. share of world GDP) ∗ (Goods output share of GDP). Using United Nations national accounts data for worldGDP and BEA data for U.S. goods output share, this calculation produces (1 − 0.45) ∗ 0.31 = 0.17 in 1960 and (1 − 0.28) ∗ 0.15 = 0.11 in 2005.

4 By the 1990s, water’s share declined to a half as air transportation became more important. Concentrating on water will tend to underestimate theimpact of energy prices at the end of the period since the air mode is more energy intensive.


Table 2Parameters

φ(60) B(60) σ γφ γB θ

79.25 3.92 0.1613 1.0053 1.0053 0.449

The calibration conditions are given by:

φ1(0) = φ2(1) = P E

TC ∗ ES(14)

and

B = 1

TC(1 − ES)(15)

I use a value of 0.043 for ES, to match the 4.3 percent fuel oil share in Canadian Water Transportation in 1960. I use0.2665 as the value of TC in 1960. Waters (1970) examines data from the 1958 input–output tables and finds trade-weightedtransport costs to be 10 percent of imports. This number is adjusted twice.

First, I make an adjustment to account for the additional domestic leg of transportation in international trade. TheCIF/FOB ratio underestimates transport costs since it misses the cost of moving cargo from the domestic production pointto the port of exit and from the point of entry to its point of consumption. International shipments face additional costscompared to domestic ones since they must be carried on two inland legs: One each in the country of origin and destination.I divide the initial transport costs number by 1−0.366 based on the finding in Rousslang and To (1993) that domestic freightcosts account for 36.6 percent of international transport costs for the United States in 1977.

Second, it is well known that trade-weighted measures underestimate total costs since the goods that are the most costlyto trade are traded the least. A measure of the size of this bias for tariffs is the Mercantilist Trade Resistance Index (MTRI)proposed by Anderson and Neary (2003), which is the estimated uniform tariff equivalent that generates the observed levelof trade. I scale up transport costs by an additional 1.69, the ratio of MTRI that Kee et al. (2005) estimates to trade-weightedtariffs for the United States in 2002.5 These estimates only cover tariffs. I am not aware of any MTRI estimates for transportcosts. Anderson and van Wincoop (2004) note that transport costs are similar to tariffs in magnitude and variability, so atariff based estimate is likely to be a reasonable proxy for bias in transport cost measures.

There is productivity growth in the transportation technology. I use difference in average annual total factor productivity(TFP) growth rates in U.S. water transportation and that of the manufacturing sector over the period 1960 to 1999 fromO’Mahoney (1999) and updated in O’Mahoney and de Boer (2002). TFP is measured as the growth rate of real output lessthe growth of capital and labor inputs. The growth rate of parameter B , γB , is 0.53 percent a year. This is a conservativeestimate since TFP growth in other goods industries was slower than in manufacturing. This is comparable to TFP growth inair transportation, which has increased in importance, and a quarter of the growth in railroads.

Murtishaw and Schipper (2001) find that energy intensity (the amount of energy required to move a good) in trans-portation in the United States has remained almost constant. To keep energy intensity constant, I set the growth rate of theenergy parameter φ, γφ , equal to γB in each period.

To select a value for the Armington elasticity, I draw on Ruhl (2005). He argues that fixed costs of exporting can explainlow elasticities in response to short run shocks and higher elasticities for long run changes, such as falling tariffs. Changesin energy prices incorporate both long and short term movements. The major changes during the period examined canlegitimately thought of as long term. Energy prices were elevated for over a decade starting in 1974. Therefore, I use a valueof 0.1613 for σ , which corresponds to an elasticity of 6.2, the value in Ruhl (2005) for permanent changes in trade costs.

I select θ , the parameter governing relative productivity in intermediate goods production, to match intermediate goodsimports. I use the Bureau of Economic Analysis’s supplemental estimate of the imported share of commodities in the 1997Input–Output table, which assumes that each industry imports each commodity in the same proportion as the aggregateshare imported. I use 1997 because it is the only year available.6 The imported share of intermediate goods of merchandiseindustries is 12.5 percent. Given model estimates of trade costs in 1997, I did a grid search of θ so that 1 − z1 = z2 = 0.125.I use θ = 0.449.

Table 2 summarizes the parameters used in the calibration.

5.2. Results

Trade expansion predicted by the baseline calibration exhibits a slowdown in the 1970s and matches the overall amountof trade growth. The model also closely matches observations in the transportation sector.

5 Irwin (2007), using the closely related Trade Resistance Index, estimates the ratio in 1960 was 1.74 which suggests the bias has not changed too muchover the sample period.

6 Below I show that calibrating to earlier years does not affect the results.


Fig. 3. Model trade share.

Table 3Model error analysis

1960–2005 1960–1973 1974–1985 1986–2005

RMSE (tariffs only) 0.287 0.058 0.204 0.402RMSE (constant energy price) 0.239 0.097 0.276 0.281RMSE (full model) 0.171 0.096 0.180 0.204

5.2.1. Trade growth slowdownFig. 3 compares trade expansion data with that generated by the simulated model and two counterfactual scenarios. The

pattern of trade growth predicted by the model matches the data quite closely. After increasing rapidly in the late 1960s,model trade share shows no sustained growth from 1974 to 1985 but expands rapidly after 1985.7 Both model and datatrade growth slows in the 2000s.

Trade costs are flat from 1974 to 1982. Transport costs rise since the oil crises drive up the cost of fuel, which counteractsfalling tariffs.8 The increase in energy cost is large enough to cause transport costs to rise despite growing productivity.Energy costs decline from their 1982 peak and collapse in 1986, which causes trade costs to fall and trade to expandrapidly.

While model fits the pattern of trade well, it misses the timing somewhat. The major shifts trade growth in the data(the late 1960s and 1980s expansions, the 1974 slowdown) lag those predicted by the model by 1 to 3 years. The Armingtonelasticity used comes from Ruhl (2005), like Melitz (2003), a model with fixed costs of trading. A lag is consistent withsuch models. It may take time to build an export channel, by advertising new brands to consumers (Arkolakis, 2006) ordeveloping distribution chains. Exporters may wait to see whether an unanticipated change is long lasting before adjustingits trade patterns.

To show the contribution of transportation to total trade costs, Fig. 3 also shows trade expansion predicted by the modelunder two counterfactual scenarios. The “Constant Energy Price” series shows the predicted trade share if energy prices heldconstant at the 1960 level. The “Tariffs Only” series in shows trade expansion if transport cost was held constant at its 1960level.

Without the contribution of the oil shocks, trade costs in the model fall throughout the late 1970s and trade continues togrow. The impact of the oil shocks is the difference between the full model and “Constant Energy Price” simulations. In theyear of the largest difference between the two series (1982, the year fuel prices peak), “Constant Energy Price” trade shareis almost 25 percent higher than that of the full model. The second oil shock is strong enough to counteract almost tenyears of falling tariffs and transportation productivity growth. The improvement from including energy does not just occurduring the oil shock years. Higher oil prices during the 2000s keep the model’s export share flat by offsetting transportationproductivity growth.

One way to evaluate the improvement in predicting trade growth is to examine error statistics. Table 3 reports RootMean Square Error (RMSE) for both the three versions of the model. Comparing the full model to the “Constant EnergyPrice” version reduces RMSE by 28 percent. Table 3 also reports RMSE for three subperiods, corresponding to the yearsbefore, during and after the oil crises. The full model reduces error in each subperiod, with significant improvements during

7 There is a temporary spike in trade share the late 1970s and a decline in the mid-1980s. This variability may be due to idiosyncratic effects not capturedby the model, such as currency fluctuations. World export share shows a slowdown during the oil crisis period without the variability.

8 Below I show that the transport costs predicted by the model, as well as other aspects of the transportation sector, are consistent with the data.


Table 4Trade growth

Time period Data Tariffs only Full model

1960–1982 52.8% 48.4% 44.4%1982–2000 58.9 21.5 51.7

Table 5VS trade: Model and data

Observation Model Data

Sh. interm. input imported 1975 3.8 4.1Sh. interm. input imported 1985 6.6 6.2Sh. interm. input imported 1995 11.6 8.2Interm. gds. import share 1997 0.30 0.34VS share gross output 1972 0.9 0.2VS share gross output 1990 2.2 1.0VS share total trade 1990 7.3 7.4

both the oil shock years and the most recent period. Even missing the timing of major trade shifts and some year to yearvariation, the model is a significant improvement over a model without transportation.

The price-import elasticity no longer needs to change radically to explain the pattern of trade growth. Table 4 summa-rizes trade growth in periods before and after 1982. For the period 1960–1982 the tariffs only version of the model predictsthe observed volume of trade growth, though not the year-to-year pattern since there is no slowdown. However, after 1982it severely underestimates trade growth. With only falling tariffs, the elasticity has to increase from 6 in the early periodto about 8 in the later period. While intermediate goods trade introduces a non-linear response to falling trade costs evenwithout transportation, accounting for the energy shocks improves the model’s performance. The elasticity does not needto change to fully account for trade growth.

At the end of the period, oil prices increase to real levels similar to those of the oil shock period without causingthe dislocations in trade that occurred in that period. An important difference is that transport costs are now smaller.Productivity growth reduced the resources, both goods and energy, required to ship a unit of goods. Therefore, oil shocksnow have a smaller impact on the delivered price of traded goods.

In addition, the higher Armington elasticity used in the calibration reflects the response of trade to long term changesin trade costs. Some of the high profile increases have been due to temporary supply shocks, such as Hurricanes Katrinaand Rita. Large, sustained increases did not occur until the very end of the sample period. Historically, there has been a lagbetween big oil price and trade growth changes. If these oil prices are sustained, trade may suffer in the near future.

5.2.2. Trade expansionProductivity growth in transportation reduces the cost of moving a unit of goods. Trade costs fell by more than the

amount generated by falling tariffs alone so trade expands more for a given elasticity. The model predicts trade sharegrowth of 119 percent from 1960 to 2000, which is close to the 143 percent in the data. As shown by the “Tariffs Only”series in Fig. 3, omitting transport costs only generates trade growth of 80 percent.

In addition to predicting total trade growth, the model is also effective at matching a number of trade composition facts.In the model, intermediate goods are not traded until 1969 and expand to almost a third of exports in 2005. The increasingimportance of intermediate goods trade fits the data well (Feenstra, 1998; Hummels et al., 2001).

Table 5 compares trade composition model moments with the data. Campa and Goldberg (1997) estimate the share ofU.S. intermediate goods that are imported, the empirical counterpart of z.9 As can be seen from the first three rows ofTable 5, the model does a good job of capturing the increasing importance of imported inputs.

The model also captures the share of trade in intermediate goods. Ramanaryanan (2006) reports that intermediate goodsmade up 34 percent of U.S. merchandise trade in 1997, close to the model’s 30 percent.

The model can account for the pattern of VS trade: intermediate imports that are re-exported as part of final goodexports. Hummels et al. (1998) estimate that VS trade (in the model, (1 − z1) ∗ C1

2 for country one) accounted for 0.2 and1.0 percent of gross output in 1972 and 1990 respectively. The model generates slightly higher values of 0.9 and 2.2 percent.Hummels et al. (1998) report that VS trade as a share of total trade was 7.4 percent in 1990, nearly the same as the model’sprediction of 7.3 percent.

An implication of the rising importance of intermediate goods trade is that the comparative advantage parameter θ canhave as much impact on trade as the Armington elasticity σ . (Eaton and Kortum (2002) also point out this implication.)When there is little difference between the two countries technologies (θ is small), small trade costs can shut down VStrade. When θ = 0.2, with all other parameters at their baseline values, trade only grows 54 percent between 1960 and

9 Their calculation only covers manufacturing imports of manufacturing goods. Since the substantial majority of merchandise trade is manufacturinggoods, the comparison is still instructive.


2000 as opposed to 119 percent in the baseline model. Trade costs never fall enough to induce intermediate goods trade.With strong comparative advantage (high θ ), there is substantial intermediate trade even with high trade costs. Whenθ = 0.8, trade also grows less over this period (56 percent) but for the opposite reason. There is already a high level of VStrade in 1960 so there is less scope for falling trade costs to increase that trade. An intermediate value of θ is required tomatch the observed pattern of little VS trade early and a significant amount late in the period.

The results are robust to using earlier years to select the value of θ . Using the same grid search method as above tomatch the 0.2 percent VS trade share of gross output in 1972 moment from Hummels et al. (1998), θ = 0.43 is selected. Theresults are nearly identical, with trade growth of 112 percent from 1960 to 2000.

In the calibration, ad valorem trade costs, both tariff and transport costs, are held constant across goods. This assumptionabstracts from product level variation. Different goods and partners face different trade costs. Transport costs are affectedby distance, infrastructure quality (Micco and Serebrisky, 2004), the value-weight ratio and a number of other factors.Specific goods or locations may get preferential treatment (e.g. the U.S.–Canada Auto Agreement), which may be particularlyimportant for VS trade (Hummels et al., 1998).

The available data does not allow for direct measurement of trade costs of intermediate and final goods separately. Tariffstend to increase with levels of fabrication, with primary goods paying low duties and tariffs escalating with the degree ofprocessing (Balassa, 1968). If traded primary goods are more likely to be intermediate goods than processed goods, thentariffs on intermediate goods should be set lower than consumption goods in the model.

The uniform assumption may be a reasonable approximation. Yeats (1977) finds that transport costs tend fall with thedegree of processing, counteracting tariff escalation. Further, highly processed intermediate goods, such as auto parts, havebecome an important part of VS trade. The largest source of an industry’s purchases of inputs is typically itself. Therefore,intermediate goods and final goods may not be that different in their trade costs.

Using lower tariff on intermediate goods would reduce the impact of VS trade on total trade growth. The model predictsthat the majority of trade expansion is in final goods trade. Even without any expansion in intermediate goods, the modelwould capture a substantial portion of trade expansion.

5.2.3. Transportation industryThe model also matches data from the transportation industry along two margins: energy share and transport cost.Fig. 4 shows the energy share of transportation expenses ( P E E

P E E+P CT) from the model against data from Canadian wa-

ter transportation. (Data from 2001 are the most recent available.) The model’s predictions match the data very closely,indicating that the Leontieff production function for the transportation sector is a reasonable assumption.

The model matches the level of transport costs at the beginning of the period well. For example, the model predicts atransport cost of 9.6 percent in 1965 (removing both inland leg and trade-weighting bias adjustments to place them on atrade-weighted FOB/CIF basis) compared to 9 percent in the data (Finger and Yeats, 1976). This is unsurprising given that themodel is calibrated to match transport costs in 1960. In the 1980s and 1990s, observed transport costs begin to fall belowthose predicted by the model. In 1994, U.S. aggregate trade-weighted CIF/FOB ratio is 3.8 percent while the model predicts8.7 percent (Hummels, 2007). Transport costs fall too slowly. Hummels (2007) finds that import weighted transport costs(on a FOB/CIF basis) in the United States fell by 1.2 percentage points between 1974 and 1994, while the model predicts a0.8 percentage point decline. Trade weighted transport costs in the data fall about 5 percentage points from the 1960s tothe 1980s whereas they fall less than 2 points in the model.

Some of the discrepancy is likely due to the crude way in which the compositional effects are controlled for. Tradein goods with lower transport costs, particularly manufacturing, has grown more than those with high transport costs.Improvements in transportation quality may help explain the increasing importance of manufacturing trade relative toother goods. To maintain a sufficient supply of parts, VS production puts a premium on timeliness of deliveries. Improved

Fig. 4. Model energy share.


transportation has also allowed retailers to hold fewer inventories, increasing the demand for timeliness in some final goods(Evans and Harrigan, 2005). Transportation has become faster and more reliable, making it easier to break up the productionprocess across locations. Hummels (2001a) estimates that increased timeliness is equivalent to a 20 percentage point dropin trade costs for manufacturing.

Not all of the discrepancy may be to be due to shifts in trade composition. Transport costs may be falling too slowly inthe model since productivity growth in international shipping is likely to have been faster than that of the U.S. transportationindustry. An increasing amount of ocean shipping was done under flags of convenience (FOC). The share of the internationalfleet under a FOC has increased from 21.6 percent in 1970 to 46.6 percent in 2003 (UNCTAD, 2004). FOC shippers are freefrom the regulation that shippers under traditional flags face so they were freer to mechanize and reduce crewing levels.There is little data available on how much more productive FOC shippers are although some (rather crude) estimates havesuggested that they may operated at as little as half the cost of traditional flag shippers (Tolofari et al., 1986).

6. Conclusion

Accounting for transport costs can resolve the puzzle of why falling tariffs do a poor job of explaining trade growth.Transportation may be a fruitful avenue of inquiry to investigate the effect of the oil shocks on other aspects of the economy.For example, it could be used to examine the effects of the oil shocks on productivity growth. The model indicates thatenergy price shocks have significant effects on trade share: Without the energy shocks, trade share in the early 1980swould have been over 10 percent higher. In the model, such disruptions do not affect the industrial counties output. Inreality, they may have. Trade competing firms and industries are typically found to be more productive than protected ones.An interesting topic to address is the role of the trade slowdown in the worldwide productivity slowdown. In the contextof the 19th Century United States, Herrendorf et al. (2007) show that reductions in transport costs can increase productivityby allowing more specialization. Changes in transport prices may have had similar effects in the 20th Century.

Appendix A. Data sources

Figure 1

Export Share Index calculated as the growth rate of goods exports quantity index from Bureau of Economic Analysis NIPATable 1.1.3 line 15 minus the growth rate of real GDP quantity index from NIPA Table 1.1.3 line 1.

Tariffs “World Tariffs” from Yi (2003) adjusted to reflect merchandise tariffs.I recalculate the beginning and end points using Yi’s weights. (Data constraints prevent me from fully recal-

culating this series.) For the pre-Kennedy Round tariffs using Yi’s source (UNCTAD, 1968), I obtain 12.075, 1.825percentage points lower. For post-Uruguay Round using All Products trade-weighted tariffs data from World Devel-opment Indicators 2006 Table 6.7, I obtain 1.8775, 1.2 percentage points lower. I use these observations to adjustthe Yi series.

I subtract 1.825 percentage points for the years 1962–1967 and 1.2 percentage points for 1973–2000. For theKennedy Round implementation years (1968–1972), I subtract 1.5, the average of the two differences. I allocate thetransition this way since the Kennedy Round reductions were strongest for manufacturing tariffs. Post-KennedyRound tariffs are about a percentage point lower than manufacturing tariffs (1.1 percentage points for the UnitedStates). Values for 1960–1961 are set equal to that of 1962 and for 2001–2005 equal to 2000.

Figure 2

Water Transportation Energy Share Fuel Oil Expenditures (Table 2.7) divided by Transportation Revenues (Table 2.4), StatisticsCanada’s Shipping in Canada. 1960–1973: Class I and II For-hire Carriers. 1975–1996: All Carriers. 1997–1998, 2000–2001: For-hire Carriers. (2001 is the most recent edition available.)

Civil Aviation Total Fuel Expenditures (Table 4.1) divided by Total Revenue for all civil aviation (Levels I–IV, scheduled andcharter, Table 3.1), including passenger transportation. Statistics Canada’s Canadian Civil Aviation. The survey wasdiscontinued in 2000.

Railroads Total Diesel Fuel Expense divided by Operating Revenue of Class I Railroads, Association of American Railroads’Yearbook of Railroad Facts.

Transportation Fuel Price Fuel oil expenditures over quantity purchased (Table 2.7) for Canadian water transportation fromStatistics Canada’s Shipping in Canada, deflated by Canadian CPI from Bureau of Labor Statistics’s Foreign LaborStatistics. Carrier coverage is the same as water transportation energy share. Missing observations in 1974, 1999and from 2002–2005 are filled in using the growth rate of U.S. Residual Fuel price to Final Users (Table 5.22) fromthe Annual Energy Review deflated by U.S. GDP deflator.

Figure 3

Export Share Same as Fig. 1.


Figure 4

Energy Share Water Transportation from Fig. 2.

Table 1

Trucking Fuel Expenses divided by Total Operating Revenues, Canadian For-hire Freight Trucking: Table 2.1 in StatisticsCanada’s Trucking in Canada.

Water Transportation and Railroads Same as Fig. 2.Business Sector Expenditures in the U.S. commercial, industrial, and transportation sectors measured by U.S. Energy Con-

sumption by source (Table 2.1) times Consumer Price per BTU (Table 3.3) from Annual Energy Review divided bynominal GDP from the business sector (NIPA Table 1.3.5) from the Bureau of Economic Analysis.

References

Anderson, James E., Neary, Peter, 2003. The mercantalist index of trade policy. International Economic Review 44 (2), 627–649.Anderson, James E., van Wincoop, Eric, 2004. Trade costs. Journal of Economic Literature 42 (3), 691–751.Arkolakis, Costas, 2006. Market access costs and the new consumers margin in international trade. Mimeo. Yale University.Armington, Paul S., 1969. A theory of demand for products distinguished by place of production. International Monetary Fund Staff Papers 16 (1), 159–178.Atkeson, Andrew, Kehoe, Patrick J., 1999. Models of energy use: Putty-putty vs. putty-clay. American Economic Review 89 (4), 1028–1043.Backus, David K., Crucini, Mario J., 2000. Oil prices and the terms of trade. Journal of International Economics 50 (1), 185–213.Baier, Scott L., Bergstrand, Jeffrey H., 2001. The growth of world trade: Tariffs, transport costs, and income similarity. Journal of International Economics 53

(1), 1–27.Balassa, Bela, 1968. Tariff protection in industrial countries and its effects on the exports of processed goods from developing countries. Canadian Journal

of Economics 1 (3), 583–594.Barsky, Robert B., Kilian, Lutz, 2002. Do we really know that oil caused the great stagflation: A monetary alternative. In: Bernanke, B., Rogoff, K. (Eds.),

NBER Macroeconomics Annual 2001. MIT Press, Cambridge, MA, pp. 137–182.Barsky, Robert B., Kilian, Lutz, 2004. Oil and the macroeconomy since the 1970s. Journal of Economic Perspectives 18 (4), 115–134.Bergoeing, Raphael, Kehoe, Timothy J., 2003. Trade theory and trade facts. Staff report 284. Federal Reserve Bank of Minneapolis.Bernard, Andrew B., Jensen, J. Bradford, Schott, Peter K., 2006. Trade costs, firms and productivity. Journal of Monetary Economics 53 (5), 917–937.Blonigen, Bruce A., Wilson, Wesley W., 1999. Explaining Armington: What determines substitutability between home and foreign goods. Canadian Journal

of Economics 32 (1), 1–21.Broda, Christian, Weinstein, David E., 2006. Globalization and the gains from variety. Quarterly Journal of Economics 121 (2), 541–585.Campa, Jose, Goldberg, Linda S., 1997. The evolving external orientation of manufacturing: A profile of four countries. FRBNY Economic Policy Review 3 (2),

53–81.Dornbusch, Rudiger, Fischer, Stanley, Samuelson, Paul, 1977. Comparative advantage, trade, and payments in a Ricardian model with a continuum of goods.

American Economic Review 67 (5), 823–839.Eaton, Jonathan, Kortum, Samuel, 2002. Technology, geography, and trade. Econometrica 70 (5), 1741–1779.Erkel-Rousse, Helene, Mirza, Daniel, 2002. Import price elasticities: Reconsidering the evidence. Canadian Journal of Economics 35 (2), 282–306.Estevadeordal, Antoni, Frantz, Brian, Taylor, Alan M., 2003. The rise and fall of world trade, 1870–1939. Quarterly Journal of Economics 118 (2), 359–407.Evans, Carolyn L., Harrigan, James, 2005. Distance, time, and specialization: Lean retailing in general equilibrium. American Economic Review 95 (1), 292–

313.Falvey, Rodney E., 1976. Transport costs in the pure theory of international trade. Economic Journal 86 (343), 536–550.Feenstra, Robert, 1994. New product varieties and the measurement of international prices. American Economic Review 81 (1), 157–177.Feenstra, Robert, 1996. U.S. imports, 1972–1994: Data and concordances. Working paper 5515. NBER.Feenstra, Robert, 1998. Integration of trade and disintegration of production in the global economy. Journal of Economic Perspectives 12 (4), 31–50.Finger, J. Michael, Yeats, Alexander J., 1976. Effective protection by transportation costs and tariffs: A comparison of magnitudes. Quarterly Journal of

Economics 90 (1), 169–176.Hamilton, James D., 1983. Oil and the macroeconomy since World War II. Journal of Political Economy 91 (2), 228–248.Head, Keith, Ries, John, 2001. Increasing returns versus national product differentiation as an explanation for the pattern of U.S.–Canada trade. American

Economic Review 91 (4), 858–876.Helpman, Elhanan, 1981. International trade in the presence of product differentiation, economies of scale and monopolistic competition: A Chamberlin–

Heckscher–Ohlin approach. Journal of International Economics 11 (3), 305–340.Herrendorf, Berthold, Schmitz, James A., Teixeira, Arilton, 2007. Exploring the implications of large decreases in transportation costs. Mimeo. Federal Reserve

Bank of Minneapolis.Hummels, David, 2001a. Time as a trade barrier. Mimeo. Purdue University.Hummels, David, 2001b. Toward a geography of trade costs. Mimeo. Purdue University.Hummels, David, 2007. Transportation costs and international trade in the second era of globalization. Journal of Economic Perspectives 21 (3), 131–154.Hummels, David, Lugovskyy, Volodymyr, 2005. Are matched partner trade statistics usable measures of transportation costs? Review of International Eco-

nomics 14 (1), 69–86.Hummels, David, Rapoport, Dana, Yi, Kei-Mu, 1998. Vertical specialization and the changing nature of world trade. FRBNY Economic Policy Review 4 (2),

79–99.Hummels, David, Ishii, Jun, Yi, Kei-Mu, 2001. The nature and growth of vertical specialization in world trade. Journal of International Economics 54 (1),

75–96.Irwin, Douglas A., 2007. Trade restrictiveness and deadweight losses from U.S. tariffs, 1859–1961. Working paper 13450. NBER.Kee, Hiau Looi, Nicita, Alessandro, Olarreaga, Marcelo, 2005. Estimating trade restrictiveness indices. Mimeo. World Bank.Krugman, Paul, 1979. Increasing returns, monopolistic competition, and international trade. Journal of International Economics 9 (6), 469–479.Krugman, Paul, 1995. Growing world trade: Causes and consequences. Brookings Papers on Economic Activity 1995 (1), 327–377.Lundgren, Nils-Gustav, 1996. Bulk trade and maritime transport costs: The evolution of global markets. Resources Policy 22 (1/2), 5–32.Markusen, James R., 1986. Explaining the volume of trade: An eclectic approach. American Economic Review 76 (5), 1002–1011.Melitz, Marc J., 2003. The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica 71 (6), 1695–1725.


Micco, Alejandro, Serebrisky, Tomas, 2004. Infrastructure, competition regimes and air transport costs: Cross country evidence. Working paper 510. Inter-American Development Bank.

Murtishaw, Scott, Schipper, Lee, 2001. Disaggregated analysis of us energy consumption in the 1990s: Evidence of the effects of the internet and rapideconomic growth. Energy Policy 29 (15), 1335–1356.

Nordhaus, William, 2004. Retrospective on the 1970s productivity slowdown. Working paper 10950. NBER.North, Douglas, 1958. Ocean freight rates and economic development 1750–1913. Journal of Economic History 18 (4), 537–555.O’Mahoney, Mary, 1999. Britain’s Productivity Performance 1950–1996: An International Perspective. NIESR, London.O’Mahoney, Mary, de Boer, Willem, 2002. Britain’s relative productivity performance: Updates to 1999. Mimeo. NIESR.Ramanaryanan, Anath, 2006. International trade dynamics with intermediate inputs. Mimeo. University of Minnesota.Ravn, Morton O., Mazzenga, Elisabetta, 2006. International business cycles: The quantitative role of transportation costs. Journal of International Money and

Finance 23 (4), 645–671.Rose, Andrew, 1991. Why has trade grown faster than income? Canadian Journal of Economics 24 (2), 417–427.Rotemberg, Julio J., Woodford, Michael, 1996. Imperfect competition and the effects of energy price increases on economic activity. Journal of Money, Credit

and Banking 28 (4), 549–577.Rousslang, Donald J., To, Theodore, 1993. Domestic trade and transportation costs as barriers to international trade. Canadian Journal of Economics 26 (1),

208–221.Ruhl, Kim J., 2005. Solving the elasticity puzzle in international economics. Mimeo. University of Texas.Ruiz-Mier, L. Fernando, 1990. Transport costs in a Ricardian model with multistage production. Southern Economic Journal 56 (4), 943–960.Sletmo, Gunnar K., Williams, Ernest W., 1981. Liner Conferences in the Container Age: U.S. Policy at Sea. MacMillan, New York.Tolofari, S.R., Button, K.J., Pitfield, D.E., 1986. Shipping costs and the controversy over open registry. Journal of Industrial Economics 34 (4), 409–427.UNCTAD, 1968. The Kennedy Round Estimated Effects on Tariff Barriers, United Nations, Geneva.UNCTAD, 2004. Review of Maritime Transportation, United Nations, New York.Waters, W.G., 1970. Transport costs, tariffs and the pattern of industrial protection. American Economic Review 60 (5), 1013–1020.Yeats, Alexander J., 1977. Do international transport costs increase with fabrication? Some empirical evidence. Oxford Economic Papers 29 (3), 458–471.Yi, Kei-Mu, 2003. Can vertical specialization explain the growth of world trade? Journal of Political Economy 111 (1), 52–102.

Documents

Energy prices and the expansion of world trade