terms of trade shocks: what are they and what do they do?

Terms of Trade Shocks: What Are They andWhat Do They Do?

JARKKO P. J €A €ASKEL €A and PENELOPE SMITH

Reserve Bank of Australia, Sydney, NSW, Australia

This article describes and quantifies the macroeconomic effects ofdifferent types of terms of trade shocks and their propagation in theAustralian economy. Three types of shocks are identified based ontheir impact on commodity prices, global manufactured prices andglobal economic activity. The first two shocks, a world demandshock and a commodity-market-specific shock, are fairly standard.The third shock, a globalisation shock that may result, for instance,from the increasing importance of China, India and eastern Europein the global economy, is more novel. The globalisation shock isassociated with a decline in manufactured prices, a rise incommodity prices and an increase in global economic activ-ity. Determining the underlying source of variation in the termsof trade is shown to be important for understanding the impact onthe Australian economy as all three shocks propagate through theeconomy in different ways. The relative contribution of each shockto inflation, output, interest rates and the exchange rate has alsovaried over time. The main conclusion of the article is that a higherterms of trade tends to be expansionary but is not alwaysinflationary. A key result is that the floating exchange rate hasprovided an important buffer to the external shocks that move theterms of trade.

I IntroductionOne of the most significant influences on the

Australian economy over the past decade hasbeen the booming terms of trade, which has risenby almost 80 per cent. The terms of trade arecurrently around their highest level of the 140-year history of the series. The current boom iscomparable in magnitude to the wool booms ofthe 1950s and the 1920s but has been sustainedfor much longer. It more than reverses the trenddecline in the terms of trade observed over thesecond half of the 20th century.As discussed by Battellino (2010), previous

terms of trade booms have been associated with

significant economic volatility, creating complexchallenges for economic policy. The booms of1920s and 1950s, although initially expansionary,were followed by falling output and risingunemployment. With a fixed exchange rate inplace during the 1950s, the concurrent rise inexport receipts and capital inflows caused a surgein money and credit, resulting in year-ended CPIinflation of almost 20 per cent. In the periodsince the float of the Australian dollar theexchange rate response to a change in the termsof trade has become central to the macroeco-nomic outcome. Gruen and Shuetrim (1994)found evidence that in the period after the floatof the Australian dollar, rises in the terms oftrade have tended to reduce domestic inflation inthe short run.We argue that the consequences of a rising

terms of trade will ultimately depend on the

JEL classifications: E32, C32Correspondence: Jarkko P. J€a€askel€a, Reserve Bank

of Australia, GPO Box 3947, Sydney, NSW, Australia.Email: [email protected]

145

© 2013 Economic Society of Australiadoi: 10.1111/1475-4932.12039

ECONOMIC RECORD, VOL. 89, NO. 285, JUNE, 2013, 145–159

characteristics of the underlying shock, as well asthe policy response. In related work Kilian (2009)and Peersman and Van Robays (2009) demon-strate that the response of the USA and euro areaeconomies to oil price shocks also depends on thenature of the shock. The recent Australianempirical literature, however, largely ignores thisidea.We identify the impact of global shocks on the

terms of trade by estimating a sign-restrictedvector autoregression (VAR). Three stylisedterms of trade shocks are identified: a worlddemand shock, a commodity-market-specificshock, and a globalisation shock. A positiveworld demand shock is associated with a pick-upin global economic activity and an increase inexport and import prices. A positive commodity-market-specific shock increases export priceswithout a corresponding pick-up in global eco-nomic activity. Finally, a globalisation shockcaptures the increasing integration of emergingeconomies, notably China and India, into theworld economy.Increased globalisation has been associated

with a fall in the relative price of manufactures.As Australian imports are concentrated in man-ufactured goods this is also likely to increase theterms of trade. At the same time, as morecountries integrate into the world economy, thedemand for raw materials increases, exertingupward pressure on (commodity) export prices.Such a shock is also associated with an expansionin world output.We find that positive terms of trade shocks tend

to be expansionary but are not always inflation-ary. Whether these shocks are inflationarydepends on both the source of the shock and theresponse of monetary policy. We also find that thefloating exchange rate provides an importantbuffer against foreign shocks that move the termsof trade: the bulk of variation in the real exchangerate is attributed to world shocks, but the worldshocks have very little impact on other Australianvariables.The remainder of the article is structured as

follows. In Section II, we outline our benchmarkVAR and discuss identification of the shocks thatmove the terms of trade. In Section III, weestimate the effects of these shocks on Australianinflation, output, short-term nominal interest ratesand the real exchange rate. We also obtain thehistorical contribution of each identified shock tofluctuations in the domestic variables. Theconcluding remarks are in Section IV.

II Measuring the Effects of Terms of TradeShocks

A standard framework for estimating theeffects of international relative price shocks ona small open economy, like Australia, is toestimate a VAR model that is partitioned into anexogenous foreign block, designed to captureconditions in the world economy, and a domesticblock that includes output, inflation, interest ratesand the real exchange rate. Movements in inter-national relative prices are captured in severaldifferent ways, Dungey and Pagan (2000, 2009)and Otto (2003), for instance, include the terms oftrade.To capture fluctuations in international relative

prices that are not explained by movements inworld output or world interest rates, a shock istypically applied directly to the relative pricevariable. The implication of this approach is thatall relative price shocks are treated equally. Forexample, a rise in the terms of trade associatedwith a fall in the price of manufactured goods isassumed to have similar consequences for a smallopen economy as a rise in the terms of tradeassociated with higher world commodity pricesand higher world manufactures prices.We also adopt a VAR framework but depart

from the standard approach in allowing theresponse of domestic variables to vary dependingon the nature of the terms of trade shock. Liu(2010) examines the effect of different types ofinternational shocks on the Australian economy.Like Dungey and Pagan, Liu assumes that termsof trade shocks do not influence world variables.In contrast, we maintain the small open economyassumption that the prices of tradable goods aredetermined in world markets, implying that allterms of trade shocks originate in (and affect) theworld economy.Before outlining our identification scheme, it is

helpful to briefly review the key influences oncommodity prices over recent years. Theseinclude the global business cycle, the rapiddevelopment of Asian economies, adverse com-modity supply shocks and the sticky response ofsupply in commodities markets more generally.As discussed by Kilian (2009), there is little

direct evidence of how the global business cycleaffects industrial commodity markets, althoughthere is typically a positive correlation betweenglobal output growth and commodity prices. Forexample, the 2000s boom in global commodityprices was associated with strong economicgrowth worldwide, particularly in Asia. The

© 2013 Economic Society of Australia

146 ECONOMIC RECORD JUNE

long-term horizons of capital-intensive invest-ment in the mining sector meant supply could notquickly expand to meet unexpected changes todemand.Strong growth in commodity prices in recent

years has also been linked to the unusually high,and rising, intensity of use of metals in industri-alising Asia. In 2007, GDP metal intensities inChina were 7.5 times higher than in developedcountries and four times higher than in otherdeveloping countries (World Bank, 2009). Thisreflects the resource intensity of the infrastructureinvestment required to support rural–urbanmigration, investment in physical capital such asplants and infrastructure, as well as lower rates ofefficiency in the use of these resources.The composition of Chinese exports also

appears to have been important. Roberts andRush (2010) provide evidence that China’s(mainly manufacturing) exports have been atleast as important as construction as a driver ofChina’s demand for resource commodities. In thissense, the integration of East Asia into the globaleconomy could be interpreted as a commodity-market-specific shock.For much of the past decade, globalisation also

appears to have dampened world import prices.The sharp expansion in the supply of manufac-tured goods accompanying the integration of EastAsia into the world economy and the transfer ofmanufacturing activities to these economies hasplaced downward pressure on the world price ofmanufactured goods. As imports of developedcountries are usually concentrated in manufac-tured goods, this globalisation shock may be anadditional factor supporting the terms of trade,particularly for commodity exporting countries.

(i) The Benchmark VARTo distill the various global shocks underlying

movements in the terms of trade we estimate thefollowing sign-restricted VAR:

wt

dt

� �¼ axt þ

Xpi¼1

Aiwt�i

dt�i

� �þ B

�wt�dt

� �; ð1Þ

where wt and dt are vectors of endogenous worldand domestic variables, xt is a vector of exoge-nous variables and B is the contemporaneousimpact matrix of the vectors of mutually uncor-related world �wt and domestic �dt disturbances.There are three variables in the world block

wt ¼ ðpxt ; pmt ;Dywt Þ which broadly capture condi-

tions in the world economy that are exogenous fora small open economy commodity exporter (likeAustralia) but relevant to their terms of trade: pxtis export price inflation; pmt is import priceinflation; and ywt is quarterly growth in the outputof Australia’s major trading partners.1 To abstractfrom fluctuations in export and import pricescaused by movements in the exchange rate, theexport and import price series are converted toworld prices using the trade weighted index.2

The second group of variables dt ¼ ðDydt ; pdt ;rdt ;DqtÞ are specific to the Australian economy:Dydt denotes domestic output growth; pdt is con-sumer price inflation; rdt is the nominal short-terminterest rate; and Dqt is the log difference of thereal exchange rate. Appendix I contains a fulldescription of the data and their sources.The sample used for estimation runs from

1984:Q1 to 2010:Q2. It was selected to includethe earlier period of strong commodity pricegrowth in the late 1980s. The start date isconstrained by the float of the Australian dollarin December 1983. To capture the move toinflation-targeting, a constant and a dummyvariable are included in the model in xt . Thedummy variable is equal to 1 during the inflation-targeting period from March 1993 onward, and 0otherwise.Augmented Dickey-Fuller and Phillips-Perron

tests indicate that all variables in the benchmarkmodel as specified in wt and dt are I(0). Further-more, trace tests failed to find evidence of a co-integrating relationship amongst the levels of thevariables in the model, so the model is specifiedin differences rather than in levels. The resultspresented are for a lag length of P = 3.

(ii) Identification Using Sign and ParametricRestrictionsIdentification of structural shocks in VAR

models is typically achieved by placing restric-tions on the model parameters. However, anincreasingly popular alternative is to placerestrictions on the direction that key variableswill move (over a given horizon) in response to

1 We also estimated the model with a measure ofworld industrial production but found that it did notsubstantially alter the results.

2 We are assuming instantaneous and perfect passthrough of the nominal exchange rate into export andimport prices. Indices of world manufacturing orcommodity prices are typically constructed under thisassumption.


2013 TERMS OF TRADE SHOCKS 147

different types of shocks. VAR models identifiedusing this technique are known as sign-restrictedVARs. Sign restrictions have been used as amethod of identifying structural shocks in VARmodels by Faust (1998), Canova and De Nicol�o(2002) and Uhlig (2005). Peersman and VanRobays (2009) demonstrate the use of signrestrictions in identifying different types of oilprice shocks.3

We use a combination of sign and parametricrestrictions. Sign restrictions are imposed on theestimated responses of the level of export prices(pxt ), import prices (pxt ) and world output (ywt ) toidentify different types of global shocks thatmove the terms of trade. This amounts to placingrestrictions on the signs of the accumulatedimpulse responses. The specific shocks that weconsider are a ‘world demand’ shock, a world‘commodity-market-specific’ shock and a ‘glob-alisation’ shock. The sign restrictions adopted inthis article are presented in Table 1. The restric-tions are imposed for four periods following ashock, as is standard in the literature. Therestrictions also guarantee that all shocks aredistinctly identified and that the identificationscheme does not suffer from the ‘multiple shocks’described by Fry and Pagan (2011).The interpretation of the world demand shock

in Table 1 is straightforward. It captures move-ments in export and import prices associated withthe global business cycle. A positive worlddemand shock increases pxt , pmt , and ywt . Theglobalisation shock captures the integration ofemerging market economies into the world tradesystem, resulting in stronger world growth andhigher world commodity prices, while at the same

time placing downward pressure on the price ofmanufactured goods (import prices). Finally,innovations to export prices that are not explainedby world demand or globalisation shocks areattributed to the commodity-market-specificshock. This shock accounts for periods of risingcommodity prices that are not associated with apick-up in world output growth. As this shockalso captures financial investment in commoditiesand precautionary demand, its impact on worldoutput may be positive or negative.Note that the restrictions used to identify the

globalisation shock are the only ones that alsorestrict the response of the terms of trade.However, because pxt is more volatile than pmtwe expect that positive world demand and com-modity-market-specific shocks will also increasethe terms of trade.In keeping with the small open economy

assumption, we restrict the contemporaneousimpact matrix B and lag matrices Ai to be blocklower triangular. This implies that fluctuations inthe world price of Australian exports and imports,and world output are only driven by shocks to theworld block (�wt ), and domestic shocks (�dt ) have noimpact onwt . Even though there are seven shocks inthe model, we only need to identify the three worldshocks to avoid the multiple shocks problem.Because we are mainly interested in the responseof the domestic variables to different types offoreign shocks, we do not place restrictions on thedomestic variables or identify domestic shocks. Asimilar approach is taken by Peersman and VanRobays (2009), although they do not place para-metric restrictions on the Ai matrices.

III Results

(i) Impulse ResponsesWe follow Peersman (2005) in using a Bayes-

ian approach for estimation and inference andhence capture both sampling and model uncer-tainty.4 A total of 100,000 successful draws fromthe posterior are used to show the median, 84thand 16th percentiles of the impulse responses.5

TABLE 1VAR Restrictions on Shocks that Improve the Terms of

Trade

px pm yw d

World demand shock + + + n.a.Globalisation shock + � + n.a.Commodity-market-specific shock + n.a. � n.a.Domestic shocks 0 0 0 n.a.

Note: + (�) means a positive (negative) response of thevariable in the column to the shock in the row. 0 means noresponse as implied by the small open economy assumption.n.a. means no restriction is imposed on the response.

3 Fry and Pagan (2011) provide a comprehensivereview of literature on sign restricted VARs.

4 Detailed descriptions of algorithms used forBayesian inference in these models can be found inUhlig (2005), Peersman (2005) and Rubio-Ramirezet al. (2010).

5 Because the point-wise distribution of each impulseresponse is approximately normal, the 16th and 84thpercentiles roughly correspond to a one-standard-devi-ation error band.



Figure 1 shows the accumulated impulseresponses of the world variables to the threeidentified shocks. The direction of theseresponses (for the first four periods) correspondto the restrictions outlined in Table 1. Themagnitudes of these responses, however, areunrestricted. Figure 1 also shows the impulseresponses for the terms of trade, which areconstructed from the responses of export andimport prices.Fry and Pagan (2011) criticise the practice of

using the median response as a measure of centraltendency because it mixes the responses ofdifferent candidate models. They suggest select-ing a single model and hence a unique set ofimpulse responses from the Monte Carlo drawsthat is closest to the set of median responses.

Accordingly, we also report their ‘median target’measure. To locate this measure we use theimpulse responses from all seven variables in themodel.6

FIGURE 1Responses of World Variables to Terms of Trade Shocks

2

4

6

Percentile bands Median Median target

0 5 10 15 20–2

0

2

4

2

4

6

5 10 15 20–2

0

2

4

0

1

0

1

–1

0

1

2

Worlddemand shock

Quarters after shock

5 10 15 20

Commodityprice shock

Globalisationshock

Exp

ort p

rice

sIm

port

pri

ces

Wor

ld o

utpu

tTe

rms

of tr

ade

2

1

0

–1

Notes: Figures show the median and median target accumulated impulse responses to the different types of terms of trade shocks,together with the 16th and 84th percentile bands. The impulse responses for the terms of trade are constructed from the impulseresponses for export prices and import prices.

6 The median target (MT) solution selects a (stan-dardised) model draw d that minimises MT ¼ hðdÞhðdÞ,where hðdÞ ¼ ðIRFi;d;s

t � IRFi;median;st Þ=stdðIRFi;median;sÞ.

In our model, there are seven variables (denoted by i),three identified shocks (denoted by s) and the number ofperiods (denoted by t) for impulse responses (IRF) is20, so the size of hðdÞ is 420 9 1. A difference betweenthe median responses and the median target responsesindicates that the shocks associated with the medianimpulses are correlated, suggesting that they do notcome from a single model.



The responses of variables in the world block toworld shocks appear to be permanent. This isconsistent with pretesting which indicated thateach series was I(1). The effect of each shock onthe terms of trade also appears to be permanent,providing additional evidence that export pricesand import prices are not co-integrated. Althoughthe magnitudes of the responses of export andimport prices to each shock are different, theresponse of the terms of trade is similar. Eachidentified shock permanently increases the termsof trade by around 2–4 per cent according to themedian target measure.The response of import prices to the commod-

ity market-specific-shock is the only unrestrictedresponse in the world block. The response isclearly positive for the first two quarters, beforereturning to zero. The response of world output tothe commodity-market-specific shock is clearlynegative. Overall, the responses to a commodity-

market-specific shock are consistent with a text-book ‘commodity supply’ shock, where a disrup-tion to the supply of commodities is followed byan increase in commodity prices, a rise in theprice of goods that use those commodities as aninput and a fall in world output. However, withouta reliable long-run measure of the world supply ofAustralia’s major export commodities such ascoal and iron ore, it is impossible to identify acommodity supply shock.Figure 2 shows the unrestricted responses of

domestic variables to the shocks identified inTable 1. The first column shows the response ofthe domestic variables to a positive world demandshock. In response to this shock, there is apermanent appreciation in the real exchange rateof 2–3 per cent. The median impulse responsesuggests that level of output is 0.25 per centhigher for three quarters, although the mediantarget measure suggests a somewhat weaker

FIGURE 2Responses of Domestic Variables to Terms of Trade Shocks

0

2

4


Worlddemand shock


5 10 15 20


Globalisationshock

Rea

lex

chan

ge r

ate

Dom

estic

out

put

Infl

atio

n ra

teIn

tere

st r

ate

4

2

0

0.5

0.0

0.5

0.0

0.0

–0.1

0.0

–0.1

0 5 10 15 20

0.5

0.0

–0.5

–1.05 10 15 20

0.5

–0.5

0.0

–1.0

Notes: Figures show the median and median target impulse responses to the different types of terms of trade shocks, together withthe 16th and 84th percentile bands. The accumulated impulse responses are shown for output and the real exchange rate.



response. Nominal interest rates increase byabout 50 basis points and remain elevated foraround two years. Inflation is slightly higher onimpact (due to higher import prices), but rapidlyreturns to zero as the exchange rate appreciatesand the interest rate increases.

The central column of Figure 2 contains theresponses to a commodity-market-specific shock.In response to this shock, the real exchange rateappreciates by around 3 per cent on impact andthen depreciates by the same amount over the

TABLE 2Asymptotic Variance Decomposition

px pm Dyw Dq Dyd pd rd

World demand shockMedian target 43 36 68 39 20 3 18One std. dev. interval (21,51) (19,55) (35,75) (15,41) (9,23) (2,16) (4,26)

Commodity-market-specific shockMedian target 51 45 16 36 6 18 18One std. dev. interval (39,68) (29,66) (10,39) (18,45) (4,14) (3,17) (2,16)

Globalisation shockMedian target 6 19 15 1 3 6 4One std. dev. interval (4,17) (7,24) (5,37) (3,11) (3,12) (1,10) (1,10)

Shocks to domestic blockMedian target – – – 24 70 73 59One std. dev. interval – – – (24,39) (59,76) (63,87) (57,84)

FIGURE 3Historical Decompositions of Export and Import Price

Inflation

–10

0

10

2010

–6

0

6

Dat

a

–10

0

10

–6

0

6

–10

0

10

–6

0

6

–20

–10

0

10

–12

–6

0

6

Wor

ldde

man

d sh

ock

Com

mod

itypr

ice

shoc

kG

loba

lisat

ion

shoc

k

20001990201020001990

tx

tm

FIGURE 4Historical Decompositions of World Output and the

Terms of Trade

–1.5

0.0

1.5

2010

–6

0

6

Dat

a

–1.5

0.0

1.5

–6

0

6

–1.5

0.0

1.5

–6

0

6

–3.0

–1.5

0.0

1.5

–12

–6

0

6

Wor

ldde

man

d sh

ock

Com

mod

itypr

ice

shoc

kG

loba

lisat

ion

shoc

k

20001990201020001990

Change interms of tradeyt

w



next three quarters. The higher real exchange rateappears to be passed through to lower consumerprices for three quarters after the shock, althoughthe magnitude of the inflation response is small(less than 0.1 per cent per quarter). This suggeststhat any domestic inflationary pressure that mightarise from the higher import prices is offset by thehigher real exchange rate and the interest ratefalls accordingly. According to the median andmedian target, the level of domestic output ispermanently higher following a commodity-mar-ket-specific shock; however, the strength of thisresponse is imprecisely measured, with the 16thpercentile band touching zero.The last column of Figure 2 shows the

response of the domestic variables to a positiveglobalisation shock. In this case, the realexchange rate appears to depreciate, althoughthe estimated magnitude of this depreciation is

small. Domestic inflation falls on impact beforereturning to trend in 4–8 quarters, perhaps due tothe depreciation of the real exchange rate. Thereis no significant response by the short-termnominal interest rate or output.To summarise, the evidence presented here is

that terms of trade shocks are not necessarilyexpansionary or inflationary, with the net effectdependent on the nature of underlying shock. Thereal exchange rate appears to have provided animportant buffer against foreign shocks over thesample period. Overall, monetary policy – ascaptured by the short-term market interest rate –has tended to respond to the foreign shocks insuch a way that it reduces the inflationary effectof these shocks.

FIGURE 5Historical Decompositions of the Real Exchange Rate

and Domestic Output

–10

0

10

–1.5

0.0

1.5

Dat

a

–10

0

10

–1.5

0.0

1.5

–10

0

10

–1.5

0.0

1.5

Wor

ldde

man

d sh

ock

Com

mod

itypr

ice

shoc

kG

loba

lisat

ion

shoc

k

–10

0

10

–1.5

0.0

1.5

2010–20

–10

0

10

–3.0

–1.5

0.0

1.5

20001990201020001990

Dom

estic

shoc

ks

qt ytd

FIGURE 6Historical Decompositions of Inflation and the Nominal

Interest Rate

–1.5

0.0

1.5

0

8

16

Dat

a

–1.5

0.0

1.5

0

8

16

–1.5

0.0

1.5

0

8

16

Wor

ldde

man

d sh

ock

Com

mod

itypr

ice

shoc

kG

loba

lisat

ion

shoc

k

–1.5

0.0

1.5

0

8

16

2010–3.0

–1.5

0.0

1.5

–8

0

8

16

20001990201020001990

Dom

estic

shoc

ks

td rt

d



(ii) Variance DecompositionTable 2 reports the contribution of the three

identified shocks to the unconditional variance ofthe data. We report the results for the mediantarget draw rather than the median because theconstruction of variance, and historical, decom-positions requires that shocks come from aunique model. We also examined variancedecompositions at shorter horizons, but theresults were broadly similar and are not reportedhere.The variance decomposition indicates that the

commodity-market-specific shock and the worlddemand shock account for the largest share of thevariance in the world price of exports (51 and 43per cent respectively) and imports (45 and 36 percent respectively). The globalisation shockappears to be less important but notably has alarger impact on Australian export prices: themedian target measure suggests that the globali-sation shock accounts for 6 per cent of thevariance in export prices and 19 per cent of thevariance in export prices. Most of the samplevariation in world output growth is attributed tothe world demand shock (68 per cent), withroughly similar contributions from the commod-ity-market-specific shock (15 per cent) and theglobalisation shock (16 per cent).The variance decomposition can also be used to

quantify the extent to which the floating exchangerate has provided a shock absorbing mechanismagainst global shocks. If the exchange rateeffectively stabilises foreign shocks, then these

FIGURE 7Responses of Tradable and Non-tradable Inflation to

Terms of Trade Shocks

Quarters after shock Percentile bands Median Median target

Wor

ldde

man

d sh

ock

Com

mod

itypr

ice

shoc

kG

loba

lisat

ion

shoc

k

–0.2

0.0

0.2

Tradable inflation Non-tradable inflation

–0.2

0.0

0.2

–0.2

0.0

0.2

–0.2

0.0

0.2

0 5 10 15 20

–0.2

–0.4

0.0

0.2

5 10 15 20

–0.2

–0.4

0.0

0.2

Notes: Figures show the median and median target responsesto the different types of terms of trade shocks, together withthe 16th and 84th percentile bands.

TABLE 3Asymptotic Variance Decomposition: Tradable and Non-tradable Inflation

Dq Dyd rd pnt pt

World demand shockMedian target 22 30 13 6 9One std. dev. interval (18,43) (8,20) (3,20) (4,13) (5,14)

Commodity-market-specific shockMedian target 44 8 11 13 18One std. dev. interval (15,41) (4,13) (2,13) (4,14) (6,17)

Globalisation shockMedian target 3 3 4 8 13One std. dev. interval (3,11) (4,11) (2,11) (3,12) (4,12)

Shocks to domestic blockMedian target 31 60 72 73 60One std. dev. interval (27,41) (62,78) (63,85) (67,83) (62,78)

Note: Results for the world block are unchanged from Table 2 and are not repeated here; pt is tradable inflation) and pnt is non-tradable inflation.



TABLE 4Asymptotic Variance Decomposition; Inflation Targeting Sample

px pm Dyw Dq Dyd pd rd

World demand shockMedian target 24 21 55 19 15 10 39One std. dev. interval (21,45) (12,38) (30,69) (14,37) (12,30) (6,25) (15,52)

Commodity-market-specific shockMedian target 69 61 22 49 16 5 5One std. dev. interval (42,68) (40,69) (15,42) (18,47) (5,18) (5,21) (4,28)

Globalisation shockMedian target 7 18 24 8 11 16 4One std. dev. interval (6,20) (12,29) (7,37) (4,15) (4,17) (5,20) (2,17)

Shocks to domestic blockMedian target – – – 24 58 69 53One std. dev. interval – – – (21,42) (45,68) (45,71) (25,59)

FIGURE 8Responses of World Variables to Terms of Trade Shocks Inflation-Targeting Sample

2

4

6


2

4

6

5 10 15 20–2

0

2

4

0

1

0

1

–1

0

1

2

Worlddemand shock

Quarters after shock5 10 15 20


Globalisationshock

Exp

ort p

rice

sIm

port

pri

ces

Wor

ld o

utpu

tTe

rms

of tr

ade

2

1

0

–1

0 5 10 15 20

4

2

0

–2

Notes: Figures show the median and median target accumulated impulse responses to the different types of terms of trade shocks,together with the 16th and 84th percentile bands. The impulse responses for the terms of trade are constructed from the impulseresponses for export prices and import prices.



shocks should have a substantial effect on theexchange rate, but much less of an impact on theother domestic variables. Table 2 shows that thevariance of the real exchange rate is explainedpredominantly by the world shocks, with themedian target draw indicating that only 24 percent of variation in the real exchange rate isexplained by ‘domestic’ factors.7 Other, uniden-tified shocks explain substantially more of thevariation in the remaining domestic variables.According to the median target measure, 70per cent of the variance in domestic output and73 per cent of the variance in trimmed-mean

inflation is explained by domestic factors. Theseresults indicate that the exchange rate does indeedprovide an important buffer against the force ofglobal shocks.

(iii) Historical DecompositionHistorical decompositions can be used to esti-

mate the contribution of the identified shocks toeach variable in the model over time. Figure 3shows the contributions of each shock to quar-terly growth in export prices and import pricesvariables over the sample period using the mediantarget measure.Export prices appear to have been subject to

several sizeable commodity-market-specificshocks throughout the sample period, most nota-bly in mid-2008. The commodity-market-specificshock appears to be capturing a shift in thecomposition of world growth in the first half of

FIGURE 9Responses of World Variables to Terms of Trade Shocks Inflation-Targeting Sample


Worlddemand shock


5 10 15 20


Globalisationshock

Rea

lex

chan

ge r

ate

Dom

estic

out

put

Infl

atio

n ra

teIn

tere

st r

ate

2

0

–2

2

0

–2

0.5

0.0

1.0

0.5

0.0

1.0

0.0

–0.1

0.0

–0.1

0 5 10 15 20

0.5

0.0

–0.5

–1.05 10 15 20

0.5

–0.5

0.0

–1.0

Notes: Figures show the median and median target impulse responses to the different types of terms of trade shocks, together withthe 16th and 84th percentile bands. The accumulated impulse responses are shown for output and the real exchange rate.

7 We are cautious in the use of the word ‘domestic’ inthis situation because own shocks to the real exchangerate may not be entirely domestic in origin. Forexample, global risk-premia shocks are often cited asimportant sources of variation in the exchange rate.



2008. Although major trading partner growth fellwell below the sample average in the June quarterof 2008, Chinese GDP growth remained a littlehigher than the sample average. A sluggishsupply response may also have contributed tothe run-up in commodity prices. However, with-out adequate data on the supply of mineralcommodities, we cannot formally evaluate thispossibility.8

The decline in export prices that followed inlate 2008 is, unsurprisingly, explained by anegative world demand shock and subsequentrecovery is attributed mainly to a commodity-market-specific shock. The recovery broadlycoincides with the Chinese fiscal stimulus, whichhad a strong emphasis on infrastructure projects.Although the globalisation shock explains arelatively small proportion of the variation ofexport prices, it is estimated to have contributed

to the decline – and the subsequent increase – inexport prices during the Asian financial crisis.Figure 4 shows the contributions of each shock

to quarterly growth in world output and the termsof trade. The decline in world output growthduring the Asian financial crisis is explained bythe combined effects of a negative globalisationshock and a negative commodity-market-specificshock. In contrast, the sharp decline in worldoutput in late 2008 and early 2009 is explained bya negative world demand shock with a smallercontribution from a commodity-market-specificshock.Figure 5 plots the contribution of terms of trade

and ‘domestic’ shocks to quarterly growth in thereal exchange rate and domestic output andFigure 6 plots the contribution of these shocksto quarterly inflation and the interest rate. Fluc-tuations in the real exchange rate are explained bya combination of global and domestic shocks,with a very limited role for the globalisationshock.The world shocks appear to have made only

modest contributions to historical variations in

FIGURE 10Historical Decompositions of Export and Import Price

Inflation Inflation-Targeting Sample

–10

0

10

2010

–6

0

6

Dat

a

–10

0

10

–6

0

6

–10

0

10

–6

0

6

–20

–10

0

10

–12

–6

0

6

Wor

ldde

man

d sh

ock

Com

mod

itypr

ice

shoc

kG

loba

lisat

ion

shoc

k

20041998201020041998

tx

tm

FIGURE 11Historical Decompositions of World Output and the

Terms of Trade; Inflation-Targeting Sample

–1.5

0.0

1.5

2010

–6

0

6

Dat

a

–1.5

0.0

1.5

–6

0

6

–1.5

0.0

1.5

–6

0

6

–3.0

–1.5

0.0

1.5

–12

–6

0

6

Wor

ldde

man

d sh

ock

Com

mod

itypr

ice

shoc

kG

loba

lisat

ion

shoc

k

20041998201020041998

Change interms of tradeyt

w

8 Annual data from ABARES and the US GeologicalSurvey suggest that growth in the world supply of somecommodities, such as iron ore, slowed sharply in 2007and 2008.



output, inflation and short-term interest rates.However, a significant proportion of the late 2008contraction in real GDP growth, and much of thesubsequent recovery, is attributed to the com-bined effects of a negative world and demandshock and a commodity-market-specific shock.This contrasts with the recession of the early1990s, which the sign-restricted VAR attributesmostly to shocks to the domestic block.

(iv) Additional Results and Robustness Checks

Estimation over the inflation-targeting periodAs a robustness check, we estimated the

benchmark VAR over the inflation-targetingperiod. The estimated impulse responses arebroadly similar, although the one-standard-devi-ation error bands are wider in some cases (seeFigures 8 and 9). The variance decomposition isqualitatively similar to the longer sample (seeTable 4).

The historical decompositions are shown inFigures 10–13. When the model is estimated overthe inflation-targeting period, the commodity-market-specific shock appears to be more impor-tant for explaining the variation in import prices,world output and the real exchange rate, and theworld demand shock appears to be less importantthan in the longer sample. Moreover, globalshocks are estimated to explain a greater propor-tion of the variation in interest rates. Of the globalshocks, the globalisation shock is estimated to bemore important in accounting for variation ininflation, while the world demand shock and thecommodity-market-specific shock explain moreof the variation in interest rates.

Direct versus indirect effects on inflationThe consumer price index captures the prices of

both tradable and non-tradable goods. Terms oftrade shocks will have a direct impact on CPI

FIGURE 13Historical Decompositions of Inflation and the Nominal

Interest Rate; Inflation-Targeting Sample

0

1

0

5

Dat

a

0

1

0

5

0

1

0

5

Wor

ldde

man

d sh

ock

Com

mod

itypr

ice

shoc

kG

loba

lisat

ion

shoc

k

0

1

0

5

2010–1

0

1

–5

0

5

20041998201020041998

Dom

estic

shoc

ks

td rt

d

FIGURE 12Historical Decompositions of the Real Exchange Rateand Domestic Output; Inflation-Targeting Sample

–10

0

10

–1.5

0.0

1.5

Dat

a

–10

0

10

–1.5

0.0

1.5

–10

0

10

–1.5

0.0

1.5

Wor

ldde

man

d sh

ock

Com

mod

itypr

ice

shoc

kG

loba

lisat

ion

shoc

k

–10

0

10

–1.5

0.0

1.5

2010–20

–10

0

10

–3.0

–1.5

0.0

1.5

20041998201020041998

Dom

estic

shoc

ks

qt ytd



inflation via the price of tradable goods, but therecan be indirect (second round) effects as well. Toevaluate the relative importance of direct andindirect effects on inflation, we extend thedomestic block of the benchmark VAR to includeboth non-tradable and tradable inflation in placeof trimmed-mean inflation.The impulse responses of tradable and non-

tradable inflation to the three global shocks areshown in Figure 7. The impulse responses of theremaining variables in the extended VAR aresimilar to those presented for the benchmarkVAR in Figure 3 and are not shown.The world demand shock has very little impact

upon tradable or non-tradable inflation, consistentwith the results for trimmed-mean inflation shownin Figure 2. For the commodity-market-specificshock, the negative response of consumer priceinflation appears to be mainly driven by theresponse of the tradable component, which is

slightly offset by the positive response from thenon-tradables component. Inflation is initiallylower after a globalisation shock. Figure 7 indi-cates that this is likely due to the decline in tradableinflation. However, tradable inflation rises insubsequent periods, perhaps due to the deprecia-tion of the real exchange rate. The effect of thisincrease on trimmed-mean inflation is, however,offset by amodest decline in non-tradable inflation.The variance decompositions for tradable and

non-tradable inflation are shown in the last twocolumns of Table 3. We do not show the variancedecomposition for the world block as this isidentical to results presented in Table 2. Notsurprisingly, domestic shocks explain a largershare of non-tradable inflation than tradableinflation. This provides further evidence that thefloating exchange rate has provided an effectivebuffer to external shocks.

IV ConclusionThis article has provided empirical evidence on

the effects of terms of trade shocks on inflation,output, interest rates and the real exchange rate inthe Australian economy. Three shocks to theterms of trade were identified using sign restric-tions in a VAR: a world demand shock thatincreases export prices, import prices and worldoutput; a commodity-market-specific shock thatpushes up export prices, without a correspondingpick-up in world output; and a globalisationshock that increases export prices and worldoutput, but reduces import prices.The main conclusion of this article is that

increases in the terms of trade tend to beexpansionary but are not always inflationary,with the exchange rate providing an effectivebuffer to external shocks that move the terms oftrade. Inflation and output are found to risefollowing a world demand shock, but this effectis relatively short-lived due to higher interestrates and the appreciation of the real exchangerate. A commodity-market-specific shock, mean-while, is found to increase import prices, whichresults in higher trimmed-mean inflation onimpact. The appreciation of the real exchangerate, however, offsets the impact on inflation,which is lower in subsequent periods. Finally, theglobalisation shock results in lower domesticinflation on impact but also a depreciation of thereal exchange rate, mitigating the deflationaryeffect in the quarter following the shock.The terms of trade shocks were found to

explain around two thirds of the variation in the

FIGURE 14Historical Decompositions of Tradable and Non-

tradable Inflation

–2

0

2

–1.5

0.0

1.5

Dat

a

–2

0

2

–1.5

0.0

1.5

–2

0

2

–1.5

0.0

1.5

Wor

ldde

man

d sh

ock

Com

mod

itypr

ice

shoc

kG

loba

lisat

ion

shoc

k

–2

0

2

–1.5

0.0

1.5

2011–4

–2

0

2

–3.0

–1.5

0.0

1.5

20011991201120011991

Dom

estic

shoc

ks

tnt

tt



real exchange rate over the sample, but less thanone fifth of variation in the other domesticvariables. The globalisation shock was found tobe of more limited empirical importance inexplaining movements in the modelled variablesthan the world demand shock or the commodity-market-specific shock.

REFERENCES

Battellino, R. (2010), Mining Booms and the AustralianEconomy. Address to The Sydney Institute, Sydney.

Canova, F. and De Nicol�o, G. (2002), ‘MonetaryDisturbances Matter for Business Fluctuations inthe G-7’, Journal of Monetary Economics, 49,1131–59.

Dungey, M. and Pagan, A. (2000), ‘A Structural VARModel of the Australian Economy’, EconomicRecord, 76, 321–42.

Faust, J. (1998), ‘The Robustness of Identified VARConclusions about Money’, Carnegie-RochesterConference Series on Public Policy, 49, 207–44.

Fry, R. and Pagan, A. (2011), ‘Sign Restrictions inStructural Vector Autoregressions: A CriticalReview’, Journal of Economic Literature, 49,938–60.

Gruen, D. and Shuetrim, G. (1994), ‘Internationalisa-tion and the Macroeconomy’, in Lowe, P. and Dwyer,J. (eds), International Integration of the AustralianEconomy. Reserve Bank of Australia, Sydney;309–63.

Kilian, L. (2009), ‘Not All Oil Price Shocks are Alike:Disentangling Demand and Supply Shocks in theCrude Oil Market’, American Economic Review, 99,1053–69.

Liu, P. (2010), ‘The Effects of International Shocks onAustralia’s Business Cycle’, Economic Record, 86,486–503.

Otto, G. (2003), ‘Terms of Trade Shocks and theBalance of Trade: There is a Harberger-Laursen-Metzler Effect’, Journal of International Money andFinance, 22, 155–84.

Peersman, G. (2005), ‘What Caused the Early Millen-nium Slowdown? Evidence Based on Vector Auto-regressions’, Journal of Applied Econometrics, 20,185–207.

Peersman, G. and Van Robays, I. (2009), ‘Oil and theEuro Area Economy’, Economic Policy, 24, 603–51.

Roberts and I. and Rush, A. (2010), Sources of ChineseDemand for Resource Commodities, Research Dis-cussion Paper 2010–08, RBA.

Rubio-Ramirez, J.F., Waggoner, D.F. and Zha, T.(2010), ‘Structural Vector Autoregressions: Theoryof Identification and Algorithms for Inference’,Review of Economic Studies, 77, 665–96.

Uhlig, H. (2005), ‘What are the Effects of MonetaryPolicy on Output? Results From an Agnostic Iden-tification Procedure’, Journal of Monetary Econom-ics, 52, 381–419.

World Bank. (2009), Global Economic Prospects.Commodities at the Crossroads. The InternationalBank for Reconstruction and Development / TheWorld Bank, Washington D.C.

Appendix

I Data Description and SourcesExport prices (px): seasonally adjusted impli-

cit price deflator for expenditure on exports ofgoods and services (ABS) multiplied by thequarterly average of the nominal trade weightedindex (RBA)Import prices (pm): seasonally adjusted impli-

cit price deflator for expenditure on imports ofABS multiplied by the quarterly average of thenominal trade weighted index (RBA)World output (yw): export weighted, season-

ally adjusted real GDP of Australia’s majortrading partners (RBA)Domestic output (yd): seasonally adjusted,

chain volume measure of non-farm gross domes-tic product (ABS)Trimmed-mean CPI (pd): seasonally adjusted,

trimmed-mean consumer price index, 1989/90 = 100, excluding interest charges and adjustedfor the tax changes of 1999–2000 (RBA)Short-term nominal interest rate (rd): quar-

terly average of the 90-day bank bill rate (RBA)Real exchange rate (q): real trade-weighted

index (RBA)Non-tradables CPI (pnt): seasonally adjusted,

non-tradables component of the consumer priceindex, excluding interest charges and adjusted forthe tax changes of 1999–2000 (RBA)Tradables CPI (pt): seasonally adjusted, trad-

ables component of the consumer price index,excluding interest charges and adjsted for the taxchanges of 1999–2000 (RBA)



terms of trade shocks: what are they and what do they do?

Documents