Marine Resource Economics, Volume 23, pp. 439–457 Printed in the U.S.A. All rights reserved

0738-1360/00 $3.00 + .00 Copyright © 2008 MRE Foundation, Inc.

The Effects of Exchange Rates on Export Prices of Farmed Salmon JINGHUA XIE University of Tromsø Shanghai Ocean University HENRY W. KINNUCAN Auburn University ØYSTEIN MYRLAND University of Tromsø Abstract The CBS inverse demand system is extended to include exchange rates. Applying the extended model to trade data for farmed salmon, results suggest export prices are at least as sensitive to changes in exchange rates as to changes in trade volume. Exchange rate pass-through (absorption into export prices) is complete for the Chilean peso and the British pound, but incomplete for the Norwegian kroner and the US dollar. This suggests producers in Chile and the United Kingdom (UK) are more affected by short-term movements in relative currency values than are producers in Norway and Rest of World (ROW). Model simulations suggest currency realignments, especially the depreciation of the Chilean peso, contributed to the 2003-04 collapse in world salmon prices. Key words Exchange rates, flexibilities, inverse demand system. JEL Classification Codes Q13, M30, F10.

“An overvalued currency is in effect an implicit export tax that, depending on the elasticity of foreign import demand and the elasticity of domestic factor supplies, has its incidence on the exporting sector” G. Edward Schuh (1976, p. 804).

Introduction Most farmed salmon enters international trade. Moreover, the currencies of the major exporting countries of Norway, Chile, and Great Britain differ from the currencies of the major importing countries/regions of the European Union, United

Jinghua Xie is a Ph.D. candidate and Oystein Myrland is a professor in the Department of Economics and Management, Norwegian College of Fisheries Science, University of Tromsø, Breivika N-9037, Tromsø, Norway, email: [email protected] and [email protected], respectively. Henry W. Kinnucan (corresponding author) is a professor in the Department of Agricultural Economics and Rural Sociology, Auburn University, Auburn, AL 36849-4201 email: [email protected] An earlier version of this paper was presented at the workshop “Farming the Sea” held in Oslo, Norway, 21-22 August 2007. Appreciation is expressed to the Norwegian Seafood Export Council for providing data and to Allen Klaiber and an anonymous journal reviewer for helpful comments. Responsibility for final content, however, rests strictly with the authors.

439

440

Xie, Kinnucan, and Myrland

States of America, and Japan. Currency realignments, therefore, have potentially important effects on export prices. Moreover, as shown in figure 1, exchange rates between the major producing and consuming nations are not static. Of particular note is Chile’s peso, which depreciated 35% on a trade-weighted basis between 1998 and 2003. As Schuh (1976) suggests, an undervalued currency acts as an implicit subsidy to the exporting sector. A major hypothesis to be examined in this study is the extent to which the subsidy contributed to the 2003-04 collapse in world salmon prices (see figure 2). Since the introduction of salmon aquaculture in the early 1980s, the world has been treated to large increases in supplies of fresh salmon (Anderson 2002; Knapp, Roheim, and Anderson 2007). Although the productivity growth underlying the supply increase is a boon to consumers, it has been a mixed blessing for salmon

Figure 1. Trade-weighted Exchange Rates for Major Salmon Exporters

Figure 2. Mean Centered Salmon Prices in Euros

Effects of Exchange Rates on Salmon Prices

441

producers, sparking inter alia trade disputes that have led variously to tariffs, feed and import quotas, minimum import prices, marketing fees, and safeguard investigations (Asche 1997a, 2001; Asche and Steen 2003; Kinnucan and Myrland 2000, 2002, 2005, 2006). The chief cause of the antagonism is periodic low prices, and these serve as the focal point for this analysis. Salmon prices are determined in an integrated world market (Asche, Bremnes, and Wessells 1999). The ability of world markets to absorb increased supplies without reductions in product value hinges on the market demand elasticity, and substantial effort has been devoted to estimating this parameter (Herrmann and Lin 1988; DeVoretz and Salvanes 1993; Herrmann, Mittlehammer, and Lin 1993; Asche 1997b; Asche, Salvanes, and Steen 1997; Asche, Bjørndal, and Salvanes 1998; Asche and Steen 2003). Exchange rates, however, are also potentially important and much less attention has been given to this factor. The only known research on exchange rates is the study by Kinnucan and Myrland (2002) who found salmon prices are much more sensitive to movements in exchange rates than to changes in the other factors studied; namely, a feed quota, marketing fees, generic advertising, and international shipping costs. The purpose of this research is to determine the relative importance of supply growth and currency realignments in the price formation of farmed salmon. It differs from Kinnucan and Myrland’s (2002) study in that exchange-rate effects are estimated econometrically instead of simulated using a partial equilibrium model. Specifically, we extend the differential inverse demand system of Laitinen and Theil (1979), also known as the inverse Central Bureau of Statistics or CBS model of Keller and van Driel (1985), to include exchange rates.1 The extended model is then estimated using 1998-2005 monthly data on fresh salmon exports and prices from Norway, the UK, Chile, and ROW. Flexibilities from the estimated model are computed and hypotheses are tested. Prior to model specification we present a simple comparative-static analysis of the problem. The paper concludes with a brief summary of the major findings.

Comparative-Static Analysis Analytical insight into the inverse demand system estimates to follow can be obtained by considering a simple situation in which a country exports commodity X to three markets. In equilibrium:

X S = X1D ( P1 ) + X 2D ( P2 ) + X 3D ( P3 )

(1)

Pi = Bi ⋅ P ,

(2)

where XS is the fixed total quantity supplied to the export market, XDi is the quantity demanded in the ith export market, Pi is the import price expressed in the importer’s currency unit (MCUi), P is the export price expressed in the exporter’s currency unit (XCUi), and Bi = MCU i/XCUi is the bilateral exchange rate. Dropping the superscripts and taking the logarithmic total differential of equations (1) and (2) yields:

1 Exchange-rate effects are commonly modeled using cointegration, VAR, or other time-series techniques (e.g., see Gervais and Khraief 2007, and the references cited therein). An advantage of the demand systems approach is that estimates of structural parameters can be obtained as a by-product of the analysis.

442

Xie, Kinnucan, and Myrland d ln X = k1 η1 d ln P1 + k 2 η2 d ln P2 + k 3 η3 d ln P3

(3)

d ln Pi = d ln Bi + d ln P ,

(4)

where ki = Xi/X is the ith market’s quantity share and ηi = ∂lnXi/∂lnPi < 0 is the export demand elasticity corresponding to the ith market. The inverse demand curve in terms of exchange rates is obtained by substituting equation (4) into equation (3) and solving for export price to yield:

d ln P =

1 η

d ln X −

k1 η1 η

d ln B1 −

k 2 η2 η

d ln B2 −

k 3 η3 η

d ln B3 ,

(5)

where η = (k1η1 + k2η2 + k 3η3) < 0 is the overall export demand elasticity. As expected, the inverse demand curve is downward sloping, and an increase in the value of the exporter’s currency shifts the curve to the left. Also, the market-specific exchange rate flexibilities have a lower limit of minus one; i.e., |∂lnP/∂lnBi| ≤ 1. Combining equations (4) and (5), an increase in the value of the exporter’s currency is split between a rise in the import price and a fall in the export price:

∂ ln Pi ∂ ln Bi ∂ ln P ∂ ln Bi

=1 −

=−

k i ηi η

k i ηi η

.

(6a)

(6b)

Equation (6) is the incidence relation alluded to by Schuh (1976). Several hypotheses can be deduced. With the maintained hypothesis that export supply is fixed, exporter’s incidence depends on the magnitude of the demand elasticity the exporting country faces in the market where its currency has strengthened. In particular, if the exporting country is a small player in that market such that ηi = –∞, then ∂lnP/∂lnB i = –1 and none of the increase in the exchange rate is passed on to foreign buyers. That is, ∂lnPi/∂lnBi = 0 and producers in the exporting country bear the full burden of currency appreciation. This is consistent with the “complete pass-through” result obtained by Miljkovic, Brester, and Marsh (2003, p. 644) for a price discriminating monopolist facing a perfectly elastic demand curve in the export market. If demand elasticities are uniform across export markets such that ηi = ηo, equation (5) reduces to:

d ln P =

1 ηo

d ln X − d ln Z ,

(7)

k k k where d ln Z = (k1d ln B1 + k2d ln B2 + k3d ln B3) and Z = B1 1 B2 2 B3 3 is an exchange rate index akin to the Stone price index. In this instance, exchange rate pass-through is always complete (∂ ln P/∂ ln Z = –1), provided export supply is fixed and markets are efficient. The relative effect of changes in export volume and exchange rates on export price depends on the magnitude of the demand elasticity. In particular, if ex-

Effects of Exchange Rates on Salmon Prices

443

port demand is price elastic such that |ηo| > 1, changes in the exchange rate index have a larger effect on export price than do changes in export volume; the opposite is true if export demand is price inelastic.

Model Differential inverse demand systems have a long history in the study of fish price formation. Examples include Barten and Bettendorf (1989); Eales, Durham, and Wessells (1997); Fousekis and Karagiannis (2001); Holt and Bishop (2002); Park (2004); and Park, Thurman, and Easley (2004). Besides being consistent with economic theory, 2 inverse demand systems permit quantities are to be treated as predetermined. This is an important advantage for perishable products, such as fish, where inventory behavior is negligible and production lags are long rendering supplies price inelastic over the observation interval. Among the more popular functional forms for these systems are the Rotterdam and the differential Almost Ideal Inverse Demand System (AIIDS)3 developed by Barten and Bettendorf (1989), the inverse Central Bureau of Statistics or CBS system originally proposed by Laitinen and Theil (1979), and the inverse analog of the National Bureau of Research or NBR system of Neves (1987). To discriminate among these forms, in preliminary work we estimated the generalized demand model developed by Eales, Durham, and Wessells (1997). Results indicated the CBS model provided the best fit. As noted by Keller and van Driel (1985, pp. 382–3), the CBS “allows for consistent aggregation over consumers, flexible Engel curves, and the imposition of all [theoretical] restrictions including concavity on the estimates.” The CBS is a cross between the Rotterdam and the AIDS in that the quantity effects are fixed (as in the Rotterdam) while the scale effects vary with budget share (as in the AIDS).

Basic Specification The basic inverse CBS has the form:

wi d ln π i = hi d ln Q +

∑h

ij

d ln q j

i = 1, ..., n,

(8)

j

where πi = pi/y is the normalized price of good i; pi and qi are the nominal price and quantity of good i; y = Σni=1p iqi is total expenditure; wi = piq i/y is the expenditure share for good i; d ln Q = Σ wid ln qi is the Divisia volume index; and hi = bi – wi is the scale effect that is assumed to decrease with budget share.4 Equation (8) expresses the normalized price of commodity i as a function of the scale of consumption (as measured by the Divisia volume index) and the quantities

2

For a good general discussion of the theoretical properties of these systems and their interpretation, see Anderson (1980); Park and Thurman (1999); Brown, Lee, and Seale (1995); Holt (2002); and Matsuda (2005a,b). 3 Moschini and Vissa (1993) and Eales and Unnevehr (1994) developed and illustrated the application of an Almost Ideal Inverse Demand System. 4 If the scale effect is assumed constant; i.e., invariant to budget share, equation (8) is the inverse Rotterdam system (e.g., Matsuda 2005b, p. 787).

444

Xie, Kinnucan, and Myrland

consumed of the individual commodities. Theory implies the following parametric restrictions:

∑h i

i

= −1

∑b

i

i

∑

j

∑h

= 0

i

hij = 0

hij = h ji

ij

= 0

(adding-up)

(homogeneity)

(9a)

(9b)

(Antonelli symmetry).

(9c)

Scale and quantity flexibilities corresponding to equation (8) are computed from the following formulas:

hi

fi =

f ij* =

wi

hij

=

bi wi

−1

(scale)

(10a)

(compensated quantity)

wi

f ij = f ij* + w j f i

(10b)

(uncompensated quantity).

(10c)

Denoting ei = ∂ ln qi/∂ ln y as the expenditure elasticity, Park and Thurman (1999) ≤ show that f i ≤ > −1 implies ei > 1. That is, an elastic (inelastic) scale response implies an inelastic (elastic) expenditure response.

Incorporating Exchange Rates To incorporate exchange rates into the basic CBS, first totally differentiate the budget constraint to yield: d ln y = d ln P + d ln Q,

(11)

where d ln P = Σ wid ln pi is the Divisia price index. Then define:

d ln pi = d ln pix + d ln Z i ,

(12)

where pi is the import price of good i in the importer’s currency, pxi is the export price of good i in the exporter’s currency, and Zi is the exchange rate that converts the export price into the currency of the import price. Substituting equations (11) and (12) into (8) and re-arranging terms yields: wi d ln π ix = hi d ln Q +

∑h

ij

j

d ln q j +

∑ c d ln Z ij

j

i = 1, ..., n,

(13)

j

where πxi = pxi/yx = pxi/PxQ is normalized price in the exporter’s currency and d ln Px = Σni=1wid ln pxi is the Divisia price index in exporters’ currencies.

445

Effects of Exchange Rates on Salmon Prices The exchange-rate coefficients in equation (13) are defined as: cij = wi w j − wi δ ij ,

(14)

where δij is the Kronecker delta. They obey the classical restrictions on consumer demand, namely:

∑c i

∑ cij = c ji

j

ij

= 0

cij = 0

(adding-up)

(15a)

(homogeneity)

(15b)

∀ i ≠ j

(Antonelli symmetry).

(15c)

Exchange rate flexibilities are computed using the formula:

cij

zij =

wi

.

(16)

The flexibilities given in equation (16) indicate the response of normalized price to changes in the exchange rates. The corresponding expressions for the response of absolute price to changes in exchange rates (derived in the appendix) are given by: cij

zijA =

wi (1 − wi )

.

(17)

Substituting equation (14) into equation (17), the absolute price flexibilities in terms of budget shares are:

ziiA =

zijA =

− wi (1 − wi )

= −1

wi (1 − wi ) wi w j

wi (1 − wi )

=

wj 1 − wi

(own effect)

(18a)

(cross effect).

(18b)

Equation (18a) is consistent with equation (7) of the comparative static results. Thus, to test whether markets are efficient in the sense that exchange rate passthrough is complete, it is sufficient to test whether estimated own-exchange rate flexibilities as defined by equation (17) equal minus one. Intercepts are generally included in differential demand systems to account for gradual changes in tastes (Deaton and Muellbauer 1980, pp. 69–70). Adding an intercept to equation (13) and imposing the restriction hi = bi – wi yields the inverse CBS to be estimated: ⎛ pix ⎞ wi d ln ⎜ x ⎟ = ai + bi d ln Q + ⎝P ⎠

∑h

ij

j

d ln q j +

∑ c d ln Z ij

j

j

i = 1, ..., n.

(19)

446

Xie, Kinnucan, and Myrland

Equation (19) differs from equation (13) in that relative price replaces normalized price as the dependent variable, and taste change is permitted via the inclusion of an intercept. These changes are innocuous with respect to interpretation of parameters and the computation of flexibilities. Adding-up is enforced by imposing the added restriction Σ iai = 0.

Empirical Specification A four-equation system is estimated to represent major exporters of fresh salmon, namely Norway (i = 1), UK (i = 2), Chile (i = 3), and ROW (i = 4). Thus, we implicitly assume fresh salmon is weakly separable from all other goods, including frozen salmon.5 Because monthly data are used, eleven monthly binary variables D kt(k = 2,…,12) are specified to account for seasonality in fish demands (Wessells and Wilen 1994). The coefficients of each of these dummy variables must sum to zero over equations to satisfy adding up. Exchange rates, which enter as indexes, are defined as follows: J

d ln Z i =

∑

j =1

k ij d ln Bij

i = 1, 2, 3, 4 ,

(20)

where Zi is the trade-weighted exchange rate corresponding to exporting country i, kij is the share of country i’s exports sold in market j (assumed to be a fixed constant), and Bij is the corresponding bilateral exchange rate. The bilateral exchange rates are expressed as the importer’s currency unit divided by the exporter’s currency unit (MCU i/XCUi); hence, an increase in Zi represents currency strengthening from the exporter’s perspective. Differentials are approximated using finite time differences. Thus, d ln Zi is approximated as Δ ln Z i = ln Zi,t – ln Zi,t–1, where subscript t indexes time.6 The final estimating equation takes the form:

⎛ pix, t ⎞ wi, t Δ ln ⎜ x ⎟ = α i + ⎝ Pt ⎠

12

∑

4

ϕ im Dm , t + βi Δ ln Qt +

m =2

∑γ

ij

Δ ln q j , t

(21)

j =1

4

+ φ ij

∑ Δ ln Z

j,t

+ εi , t

i = 1, 2, 3, 4 ,

j =1 4

4

where wi,t = ( wi,t + wi,t−1 )/ 2, Δ lnQt = ∑ j=1w j ,t Δ ln q j ,t , Δ ln Pt x = ∑ j=1w j ,t Δ ln p xj ,t , and ε i,t is a random disturbance term. Equation (21) was estimated using monthly data for the period January 1998 through December 2005. Data for export values and quantities were obtained from the Norwegian Seafood Export Council. These data were divided to obtain imputed prices, which are FOB and measured at the wholesale level. The imputed prices were converted to the exporting country’s currency using the appropriate exchange

5

Preliminary analysis based on a system that included frozen salmon as an aggregate fifth good indicated little substitution between fresh and frozen, supporting our separability assumption. 6 As noted by Matsuda (2005a), the process of first differencing is likely to make variables stationary. For this reason, and for the reasons given in Wang and Tomek (2007), we did not test for unit roots.

Effects of Exchange Rates on Salmon Prices

447

rate. (The US dollar is used as the representative currency for ROW.) The export prices and quantities, along with the exchange rate indices discussed next, were centered on data means prior to converting to logarithms. Exchange rate data were obtained from the website developed by Professor Werner Antweiler (Antweiler 2007). The exchange-rate indices were constructed using (nominal) bilateral exchange rates corresponding to the top four importers for each country. Specifically, the Z variables in equation (21) were constructed as follows (time subscripts suppressed): Z 1 = (USD/NOK) 0.10 (CAD/NOK) 0.05 (EUR/NOK) 0.65 (JPY/NOK) 0.20 Z 2 = (USD/GBP) 0.05 (CAD/GBP) 0.05(EUR/GBP) 0.85 (JYP/GBP) 0.05 Z 3 = (USD/CLP) 0.40 (CAD/CLP) 0.10 (EUR/CLP) 0.30 (JPY/CLP) 0.20 Z 4 = (CAD/USD) 0.10 (EUR/USD) 0.40 (JPY/USD) 0.35 (GBP/USD) 0.15, where USD = US dollars, CAD = Canadian dollars, EUR = European euros, JPY = Japanese yen, NOK = Norwegian kroner, GBP = British pound, and CLP = Chilean peso. The exponents in these formulas refer to approximate average quantity shares over the sample period. For example, since approximately two-thirds of Norway’s exports go to the European Union, the EUR/NOK exchange rate receives a weight of 0.65 in Norway’s exchange-rate index. Differencing results in the lost of one observation. Hence, the model is estimated with 95 observations (February 1998 through December 2005). The fourth equation is deleted during estimation to avoid singularity in the residual covariance matrix of the full model. The coefficients of the deleted equation in each instance were recovered using the adding-up conditions. The model was estimated using the SUR estimator in LIMDEP. Price homogeneity and symmetry are treated as maintained hypotheses, while exchange-rate homogeneity and symmetry were tested prior to imposition. The relative importance of exchange-rate movements and export growth in price formation is tested via the hypothesis: H N : φ ii = γ ii

i = 1, 2, 3, 4

H A : H N not true.

(22a) (22b)

Null hypothesis (22a) asserts that own exchange-rate effects are identical to (compensated) own quantity effects; i.e., z ii = f *ii . To test whether exchange-rate pass-through is complete we tested: H N : φ ii = − wi (1 − wi )

i = 1, 2, 3, 4

H A : H N not true.

(23a) (23b)

Null hypothesis (23a) asserts that absolute price own exchange-rate effects equal minus one; i.e., z Aii = –1. Hypotheses (22) and (23) were tested using the Wald statistic. In estimation it is standard to assume the coefficients of differential demand

448

Xie, Kinnucan, and Myrland

systems are fixed constants even though they are endogenous, dependent on budget shares. Keller and van Driel (1985, p. 379) note that this assumption is not innocuous in that, for example, it implies the Engel curves from q-dependent systems are linear. This caveat, and others suggested by the empirical work of Byron (1984), need to be borne in mind when interpreting the empirical results.

Results Exchange-rate homogeneity and symmetry are rejected at usual significance levels (table 1). Hence, the model was estimated with and without the restrictions to determine whether parameters are unduly affected. Results suggest not (table 2). In particular, imposition of the restrictions has little effect on parameters with the important exception that the estimated own exchange-rate effect in Norway’s price equation becomes significant. Moreover, the R2s and DW statistics are little affected, with the latter indicating either inconclusive serial correlation, or its absence. The remaining discussion, therefore, will be based on results with the restrictions imposed. Statistical significance is determined using the 5% probability level based on a two-tail t-test (critical value = 1.96) unless indicated otherwise. Results overall are satisfactory in that the estimated demand curves are downward sloping, own exchange-rate effects are negative, and cross exchange-rate effects are positive or insignificant. Norway’s equation has the best explanatory power (R2 = 0.84), followed by ROW (0.69), UK (0.43), and Chile (0.21). The intercepts are insignificant except for ROW, where it is positive. This suggests taste and preferences for fresh salmon from ROW may be strengthening over time. The estimated coefficients for the Divisia volume index are insignificant for the UK and Chile, indicating preferences for fresh salmon from these sources are homothetic (expenditure elasticity equal to one). The estimated coefficient of the Divisia volume index is positive and significant for Norway and negative and significant for ROW, indicating fresh salmon from Norway is a superior good (expenditure elasticity greater than one), while fresh salmon from ROW is a normal good (expenditure elasticity between zero and one). The own-quantity effects are estimated with a high degree of precision (t-ratio ≥ 6 in absolute value), except for Chile where the t-ratio is –0.99 (removing the theoretical restrictions raises the t-ratio to –1.25). Counterbalancing the imprecise own-quantify effect for Chile is a precise own-exchange rate effect (t-ratio = –5.63), which hints at the potential importance of exchange rates in price formation for Chilean salmon. The t-ratios for the remaining own-exchange rate effects indicate less precision at –2.10, –1.23, and –1.07 for Norway, the UK, and ROW, respectively.

Table 1 Tests of Theoretical Restrictions Hypothesis Exchange rate homogeneity Exchange rate symmetry Exchange rate homogeneity and symmetry

Number of Restrictions

Wald Test p value

Result

3 3

0.0289 0.0007

Reject at 5% level Reject

6

0.0041

Reject

4

3

2

1

–0.0063 (–1.26) 0.0002 (0.05) –0.0021 (–0.63) 0.0082 (2.25) 0.0291 (2.67) 0.0072 (1.04) –0.0049 (–0.68) –0.0314 (–3.89)

d ln Q

Intercept

4

3

2

0.0325 (2.96) 0.0069 (0.98) –0.0085 (–1.18) –0.0308 (–3.79)

–0.0083 (–1.63) 0.0001 (0.03) –0.0019 (–0.56) 0.0101 (2.72)

1

d ln Q

Intercept

Equation

–

–

–0.0489 (–5.97) –

d ln q1

–

–

–0.0549 (–6.39) –

d ln q1 0.0059 (1.41) –0.0028 (–1.23) –0.0042 (–1.25) –

d ln q 3 0.0341 (6.66) 0.0066 (2.47) 0.0011 (0.41) –0.0420 (–8.95)

d ln q 4 0.0207 (0.32) 0.0236 (0.56) –0.0631 (–1.45) 0.0188 (0.38)

d ln Z 1 0.1593 (2.28) –0.0536 (–1.22) –0.0664 (–1.46) –0.0390 (–0.76)

d ln Z 2 0.1094 (2.71) 0.0634 (2.45) –0.1622 (–6.07) –0.0109 (–0.36)

d ln Z 3

–

0.0148 (3.74) –0.0186 (–5.99) –

d ln q 2 0.0027 (0.68) –0.0021 (–0.96) –0.0033 (–0.99) –

d ln q 3 0.0313 (6.38) 0.0060 (2.30) 0.0027 0.97 –0.040 (–8.69)

d ln q 4

–

–

–0.0963 (–2.10) –

d ln Z 1

–

0.0315 (1.08) –0.0454 (–1.23) –

d ln Z 2

0.0328 (1.32) 0.0535 (2.62) –0.1352 (–5.63) –

d ln Z 3

Exchange Rate Homogeneity and Symmetry Restrictions Imposed

–

0.0150 (3.67) –0.0188 (–5.99) –

d ln q 2

Exchange Rate Homogeneity and Symmetry Not Imposed

0.0320 (1.06) –0.0395 (–1.47) 0.0489 (2.25) –0.0398 (–1.07)

d ln Z 4

–0.0493 (–0.92) –0.0413 (–1.23) 0.1278 (3.68) –0.0372 (–0.93)

d ln Z 4

Table 2 SUR Estimates of Inverse CBS Demand System for Fresh Farmed Salmon, 1998–2005 Monthly Data (1 = Norway, 2 = UK, 3 = Chile, 4 = ROW; Asymptotic t-ratio in parentheses)

0.69

0.21

0.43

0.84

R2

0.70

0.28

0.43

0.84

R2

2.24

1.54

2.84

1.54

DW

2.28

1.44

2.83

1.63

DW

Effects of Exchange Rates on Salmon Prices 449

450

Xie, Kinnucan, and Myrland

Scale and Quantity Flexibilities The scale and quantity flexibilities evaluated at data means are reported in table 3. Their t-ratios were computed using the Wald test in LIMDEP with budget shares treated as normally distributed random variables. Focusing first on the scale flexibilities, estimates range from –1.17 for ROW to –0.93 for UK. These estimates suggest a 1% increase in the scale of fresh salmon exports would cause normalized prices in exporters’ currencies to decline by between 0.93% and 1.17% depending on source origin and holding exchange rates constant. The t-ratios associated with these flexibilities range from –13.4 for the UK to –143 for Norway. Hence, the hypothesis that volume growth in toto does not affect normalized prices is firmly rejected. Overall, it appears fresh salmon from Norway and the UK is a superior good in international trade (since |fi | < 1 implies ei > 1), while fresh salmon from Chile and ROW is a normal good. Compensated flexibilities reflect the pattern commonly found in the literature. In particular, most of the off-diagonal elements are positive, implying net complements. The reason for this counterintuitive result, as explained by Fousekis and Karagiannis (2001), is that the homogeneity restriction on the compensated responses Σjhij = 0 combined with the negative definiteness of the Antonelli matrix prejudices results toward complementarity. To circumvent this problem, interaction effects are commonly discussed in terms of the uncompensated flexibilities.7 Indeed, as shown in table 3, the uncompensated flexibilities are uniformly negative, indicating all four products are gross substitutes. Given their intuitive signs, statistical significance (all t ratios exceed 2 in absolute value), and policy relevance, further discussion/analysis is confined to the uncompensated flexibilities. The estimated uncompensated own flexibilities range from –0.60 for Norway to –0.19 for Chile. Since the inverse of the absolute value of the own-quantity flexibility sets the lower bound on the own-price elasticity (Houck 1966), these estimates suggest export demands for source-specific fresh salmon are price elastic. The uncompensated cross flexibilities in Norway’s equation are smaller in absolute value than the own-flexibility. This suggests increases in exports from Norway have a larger depressing effect on Norway’s export price than similar increases from its international competitors. This might be expected since Norway dominates the export market. Indeed, Norway’s dominance is apparent, since in the remaining equations the cross-flexibilities corresponding to quantity from Norway exceed in absolute value the own flexibilities in those equations. For example, in the UK equation the own-flexibility is –0.28 compared to –0.36, –0.18, and –0.11 for cross-flexibilities for quantities from Norway, Chile, and ROW, respectively. Thus, an increase in exports from Norway has a larger depressing effect on the UK price than a similar increase in exports from the UK and its other international competitors.

Exchange Rate Flexibilities Exchange rate flexibilities for both normalized and absolute price are reported in table 4 along with their t-ratios. Most of the estimated flexibilities have the correct sign, and ten are significant at 10% level or better (using the critical t-value of 1.28). The normalized and absolute price flexibilities differ in expected ways. In particular, the absolute price exchange-rate flexibilities are larger in absolute value than their normalized price counterparts, but otherwise are identical in sign and sta7

An alternative approach is to compute “Allais coefficients” (e.g., Barten and Bettendorf 1989).

b

a

4

3

2

1

–0.946 (–143.5) –0.928 (–13.4) –1.029 (–24.5) –1.169 (–26.9)

fi –0.090 (–5.97) 0.147 (3.74) 0.016 (0.68) 0.168 (6.27)

fi1* 0.027 (3.74) –0.185 (–5.99) –0.013 (–0.96) 0.033 (2.33)

fi2* 0.005 (0.68) –0.021 (–0.96) –0.019 (–0.99) 0.015 (0.97)

fi3* 0.058 (6.38) 0.059 (2.30) 0.016 (0.97) –0.215 (–8.69)

fi4* –0.604 (–37.3) –0.357 (–7.22) –0.542 (–17.3) –0.467 (–14.2)

fi1

Evaluated at mean data points with symmetry and homogeneity imposed. Number in last column is the mean budget share used to compute the flexibilities. See text for formulas.

Equation

Compensated

–0.068 (–9.26) –0.278 (–8.38) –0.116 (–8.01) –0.085 (–5.48)

fi2

–0.156 (–21.9) –0.179 (–7.23) –0.194 (–9.48) –0.184 (–11.1)

fi3

Uncompensatedb

Table 3 Estimated Scale and Quantity Flexibilitiesa (1 = Norway, 2 = UK, 3 = Chile, 4 = ROW; Asymptotic t–ratio in parentheses)

–0.119 (–13.1) –0.114 (–3.93) –0.176 (–9.64) –0.433 (–16.3)

fi4

0.54 – 0.10 – 0.17 – 0.19 –

wi

Effects of Exchange Rates on Salmon Prices 451

452

Xie, Kinnucan, and Myrland

Table 4 Estimated Exchange Rate Flexibilitiesa (1 = Norway, 2 = UK, 3 = Chile, 4 = ROW; Asymptotic t-ratio in parentheses) Normalized Price Flexibilities Equation 1 2 3 4 a

z i1

zi2

zi3

zi4

–0.178 0.058 0.060 0.059 (–2.10) (1.08) (1.32) (1.06) 0.313 –0.451 0.530 –0.392 (1.08) (–1.23) (2.62) (–1.47) 0.193 0.314 –0.793 0.287 (1.32) (2.62) (–5.63) (2.25) 0.163 –0.215 0.266 –0.214 (0.98) (–1.48) (2.22) (–1.07)

Absolute Price Flexibilities zAi1

zAi2

z Ai3

zi4A

–0.388 0.127 0.132 0.129 (–2.10) (1.08) (1.32) (1.06) 0.348 –0.501 0.590 –0.436 (1.08) (–1.23) (2.62) (–1.47) 0.232 0.378 –0.956 0.346 (1.32) (2.62) (–5.63) (2.25) 0.200 –0.264 0.327 –0.263 (0.98) (–1.48) (2.22) (–1.07)

Evaluated at mean data points with symmetry and homogeneity imposed.

tistical precision as measured by t-ratios. Thus, for brevity we will restrict attention to the absolute price flexibilities, as they are somewhat easier to interpret. In particular, z Aii = –1 means exchange rate pass-through is complete. Of the four exchange-rate indices studied, the Chilean index has the largest influence on prices. The basis for this statement is that all four of the estimated coefficients for Z3 are significant at the 10% level or better, compared to just two coefficients each for Z 1, Z2, and Z4. Moreover, the estimated flexibilities for Z3 tend to be larger than for the other indices. As an example, according to these estimates an isolated 1% strengthening in the trade-weighted Chilean peso reduces the Chilean price 0.96% and increases the prices of its international competitors by between 0.13% (Norway) and 0.59% (UK). By way of comparison, an isolated 1% strengthening in the trade-weighted Norwegian kroner reduces the Norwegian price 0.39%, raises the Chilean price 0.23%, and has no effect on UK and ROW prices.

Hypothesis Tests Test results for hypotheses (22) and (23) are presented in table 5. Tests were performed with and without the homogeneity and symmetry restrictions (tested in table 1) imposed. Since inferences were unaffected, we report results for the restricted model only. Focusing first on hypothesis (22), results suggest currency realignments are at least as important as supply growth in the price formation of farmed salmon (table 5). Specifically, the hypothesis f*ii = z ii fails to be rejected in all cases except Chile, in which case the exchange rate effect is larger than the quantity effect. Turning to hypothesis (23), results are mixed in that the British pound and Chilean peso show complete pass through while the Norwegian kroner and the US dollar show less than complete pass through. Specifically, the hypothesis zAii = –1 is rejected for Norway and ROW but not for the UK and Chile. Thus, it would appear that markets are efficient with respect to fresh salmon trade from the UK and Chile, but less so with respect to trade from Norway and ROW. In the case of Norway, anti-dumping measures may help to explain the less-than-complete pass through (Asche 2001; Asche and Steen 2003; Kinnucan and Myrland 2005, 2006). Yet Chile has also been subjected to tariffs (albeit less onerous), and these appear not to have affected market efficiency.

453

Effects of Exchange Rates on Salmon Prices Table 5 Hypotheses Tests Equivalency of Exchange Rate and Quantity Effectsa Equation 1 2 3 4 a b

Computed Value t–ratio –0.087 –0.266 –0.774 0.000

–1.02 –0.74 –5.40 0.006

Complete Exchange Rate Pass–throughb

Result

Computed Value

t–ratio

Result

Fail to reject Fail to reject Reject Fail to reject

0.6118 0.4990 0.0437 0.7371

3.32 1.23 0.26 2.99

Reject Fail to reject Fail to reject Reject

Null hypothesis: |fii*| – |zii| = 0. Null hypothesis: z iiA + 1 = 0. See text for details.

Simulation Between 2000 and 2003 the trade weighted exchange-rate index for Norway’s currency increased 10%, while the corresponding indices for the currencies of UK, Chile, and ROW decreased by 9%, 26%, and 7%, respectively. Over the same period, exports of farmed salmon from Norway, the UK, Chile, and ROW grew 20%, 46%, 56%, and 50%, respectively. To assess the extent to which currency realignments and supply growth may have contributed to the 2003–04 collapse in world salmon prices, we simulated the model using the uncompensated flexibilities for quantity in table 3 and the normalized flexibilities for exchange rates in table 4. Results, disaggregated to show distributional impacts, are presented in table 6. Although most of the price collapse can be explained by supply growth, currency realignments are a contributing factor. This is most true for the Chilean peso, where the 26% devaluation increased Chile’s price 21% and decreased UK’s price 14%. Thanks to peso depreciation, Chile’s producers avoided the worst consequences of the market glut. Specifically, according to model simulations, supply growth and currency adjustments combined to decrease Chile’s price by 18% compared to decreases of between 34% and 47% for the prices of Chile’s international competitors. In light of the dumping complaints filed by UK producers against Norway and Chile following the price collapse, it is of some interest to note peso weakening (14%) and supply growth from Chile (10%) combined to account for most of the simulated 39% decrease in UK’s price. Thanks to kroner strengthening that offset, in part, the negative effects of its supply growth, Norway contributed a modest 4% to the decline in the UK price. These results underscore the potential relevance of monetary phenomena to dumping and safeguard investigations.

Concluding Comments This research supports Schuh’s (1974, 1976) hypothesis that exchange rates are an important determinant of farm prices. In the case of farmed salmon, the prices of major exporting countries were found to be at least as sensitive to changes in relative domestic currency values as to changes in export volume. For Chile, the estimated own-exchange rate flexibility (–0.79) was much larger in absolute value than the estimated own-quantity flexibility (–0.19). This suggests small percentage

454

Xie, Kinnucan, and Myrland

Table 6 Effects of Export Growth and Currency Realignments on World Salmon Prices (Effects in Percent) Export Price

Norway UK Chile ROW

Effects of Export Growth a

Norway

UK

Chile

ROW

All

–12 –7 –11 –9

–3 –13 –5 –4

–9 –10 –11 –10

–6 –6 –9 –22

–30 –36 –36 –45

Effects of Exchange Rate Changes

Norway UK Chile ROW

Norway

UK

Chile

ROW

–2 3 2 2

–1 4 –3 2

–2 –14 21 –7

0 3 –2 1

All

a

–4 –4 18 –2

Combined Effects

Norway UK Chile ROW a

a

Norway

UK

Chile

ROW

All

–14 –4 –9 –8

–4 –9 –8 –2

–10 –24 10 –17

–6 –3 –11 –20

–34 –39 –18 –47

Totals may not sum due to rounding error.

changes in the relative value of Chile’s peso have a much larger impact on Chile’s export price than equivalent small percentage changes in export quantity. Model simulations suggest the 26% depreciation in the trade-weighted peso between 2000 and 2003 acted as an important implicit subsidy to Chile’s salmon sector and contributed to the 2003–04 collapse in world salmon prices. Exchange rate pass-through was found to be complete for Chile and the UK and incomplete for Norway and ROW suppliers. Complete pass-through means the incidence of currency realignments are borne largely by the exporting sector, especially in the short run when export supply is inelastic. It also suggests markets are efficient at converting changes in relative currency values into price changes. Incomplete pass-through, on the other hand, suggests export prices are “sticky” and may be influenced by market power, non-tariff trade barriers, or both. The latter are clearly plausible in the case of the kroner, as Norway dominates the world salmon market and has instituted a variety of controls to adapt supply to market conditions. A caveat in interpreting our findings is that the exchange rate indices used in the econometric model are based on fixed quantity weights. In reality, the weights are endogenous and vary with export shares, which may introduce bias into the estimates. Clearly, more work is needed to address this issue and to refine our estimates. In the meantime, our results showing exchange rates to be statistically significant and empirically important suggest monetary phenomena are not to be overlooked in explaining salmon prices.

Effects of Exchange Rates on Salmon Prices

455

References Anderson, J.L. 2002. Aquaculture and the Future: Why Fisheries Economists Should Care. Marine Resource Economics 17:133–51. Anderson, R.W. 1980. Some Theory of Inverse Demand for Applied Demand Analysis. European Economic Review 14:281–90. Antweiler, W. 2007. Pacific Exchange Rate Service. The University of British Columbia, Sauder School of Business. http://fx.sauder.ubc.ca/data.html accessed 2 July 2007. Asche, F. 1997a. Dynamic Adjustment in Demand Equations. Marine Resource Economics 12:221–37. _ . 1997b. Trade Disputes and Productivity Gains: The Curse of Farmed Salmon Production. Marine Resource Economics 12:67–73. _ . 2001. Testing the Effect of an Anti-dumping Duty: The US Salmon Market. Empirical Economics 26:343–55. Asche, F., T. Bjørndal, and K.G. Salvanes. 1998. The Demand for Salmon in the European Union: The Importance of Product Form and Origin. Canadian Journal of Agricultural Economics 46:69–81. Asche, F., H. Bremnes, and C. Wessells. 1999. Product Aggregation, Market Integration, and Relationships between Prices: An Application to World Salmon Markets. American Journal of Agricultural Economics 81:568–81. Asche, F., K.G. Salvanes, and F. Steen. 1997. Market Delineation and Demand Structure. American Journal of Agricultural Economics 79:139–50. Asche, R., and F. Steen. 2003. When Anti-dumping Measures Lead to Increased Market Power: A Case Study of the European Salmon Market. Discussion Paper 27/03. Stavanger University College and Centre for Fisheries Economics, Norwegian School of Economics and Business. Stavanger, Norway. Barten, A.P., and L.J. Bettendorf. 1989. Price Formation of Fish: An Application of an Inverse Demand System. European Economic Review 33:1509–25. Brown, M.G., J.-Y. Lee, and J.L. Seale, Jr. 1995. A Family of Inverse Demand Systems and Choice of Functional Form. Empirical Economics 20:519–30. Byron, R.P. 1984. On the Flexibility of the Rotterdam Model. European Economic Review 24:273–83. Deaton, A., and J. Muellbauer. 1980. Economics and Consumer Behavior. Cambridge, MA: Cambridge University Press. DeVoretz, D.J., and K.G. Salvanes. 1993. Market Structure for Farmed Salmon. American Journal of Agricultural Economics 75:227–33. Eales, J., C. Durham, and C.R. Wessells. 1997. Generalized Models of Japanese Demand for Fish. American Journal of Agricultural Economics 79:1153–63. Eales, J.S., and L.J. Unnevehr. 1994. The Inverse Almost Ideal Demand System. European Economic Review 38:101–15. Fousekis, P., and G. Karagiannis. 2001. Wholesale Level Demand for Fish Grades in Greece. Applied Economics Letters 8:479–82. Gervais, J.-P., and N. Khraief. 2007. Is Exchange Rate Pass-Through in Pork Meat Export Prices Constrained by the Supply of Live Hogs? American Journal of Agricultural Economics 89:1058–72. Herrmann, M.L., and B.H. Lin. 1988. The Demand and Supply of Norwegian Atlantic Salmon in the United States and European Community. Canadian Journal of Agricultural Economics 38:459–71. Herrmann, M.L., R.C. Mittelhammer, and B.H. Lin. 1993. Import Demand for Norwegian Farmed Atlantic Salmon and Wild Pacific Salmon in North America, Japan, and the EC. Canadian Journal of Agricultural Economics 41:11–25.

456

Xie, Kinnucan, and Myrland

Holt, M.T. 2002. Inverse Demand Systems and Choice of Functional Form. European Economic Review 46:117–42. Holt, M.T., and R.C. Bishop. 2002. A Semiflexible Normalized Quadratic Inverse Demand System: An Application to the Price Formation of Fish. Empirical Economics 27:23–47. Houck, J.P. 1966. A Look at Flexibilities and Elasticities. Journal of Farm Economics 48:225–32. Keller, W.J., and J. Van Driel. 1985. Differential Consumer Demand System. European Economic Review 27:375–90. Kinnucan, H.W., and Ø. Myrland. 2000. Optimal Advertising Levies with Application to the Norway-EU Salmon Agreement. European Review of Agricultural Economics 27:39–57. _ . 2002. The Relative Impact of the Norway-EU Salmon Agreement: A Midterm Assessment. Journal of Agricultural Economics 53:195–219. _ . 2005. Effects of Income Growth and Tariffs on the World Salmon Market. Applied Economics 37:1967–78. _ . 2006. The Effectiveness of Antidumping Measures: Some Evidence for Farmed Atlantic Salmon. Journal of Agricultural Economics 57:459–77. Knapp, G., C.A. Roheim, and J.L. Anderson. 2007. The Great Salmon Run: Competition between Wild and Farmed Salmon. Report prepared for TRAFFIC North America, Washington, D.C. www.worldwildlife.org/trade/salmonreport.cfm Laitinen, K., and H. Theil. 1979. The Antonelli Matrix and the Reciprocal Slutsky Matrix. Economic Letters 3:153–57. Matsuda, T. 2005a. Differential Demand Systems: A Further Look at Barten’s Synthesis. Southern Economic Journal 71(3):607–19. _ . 2005b. Forms of Scale Curves and Differential Inverse Demand Systems. American Journal of Agricultural Economics 87(3):786–95. Miljkovic, D., G.W. Brester, and J.M. Marsh. 2003. Exchange Rate Pass-Through, Price Discrimination, and U.S. Meat Export Prices. Applied Economics 35:641– 50. Moschini, G., and A. Vissa. 1993. Flexible Specification of Mixed Demand Systems. American Journal of Agricultural Economics 75:1–9. Neves, P. 1987. Analysis of Consumer Demand in Portugal, 1958–1981. Memorie de Maitrise en Sciences Economiques, University Catholiqque de Louvarin, Louvain-la-Neuve, Belgium. Park, H. 2004. On Inferring Individual Behavior from Market Behavior in a Predetermined Quantities Model. Applied Economics 36:715–21. Park, H., and W.N. Thurman. 1999. On Interpreting Inverse Demand Systems: A Primal Comparison of Scale Flexibilities and Income Elasticities. American Journal of Agricultural Economics 81:950–58. Park, H., W.N. Thurman, and J.E. Easley, Jr. 2004. Modeling Inverse Demands for Fish: Empirical Welfare Measurement in Gulf and South Atlantic Fisheries. Marine Resource Economics 19:333–51. Schuh, G.E. 1974. The Exchange Rate and U.S. Agriculture. American Journal of Agricultural Economics 56(1974):1–13. _ . 1976. The New Macroeconomics of Agriculture. American Journal of Agricultural Economics 58(1976):802–11. Wang, D., and W.G. Tomek. 2007. Commodity Prices and Unit Root Tests. American Journal of Agricultural Economics 89:873–89. Wessells, C.R., and J.E. Wilen. 1994. Seasonal Patterns and Regional Preferences in Japanese Household Demand for Seafood. Canadian Journal of Agricultural Economics 42:87–103.

457

Effects of Exchange Rates on Salmon Prices

Appendix A Derivation of the Expression for Absolute Price Flexibilities The expression to indicate how absolute price (as opposed to normalized price) changes in response to changes in exchange rates is obtained by first rewriting text equation (13) as follows:

d ln pix − d ln P x − d ln Q =

hi wi

d ln Q +

hij

∑w j

d ln q j +

i

cij

∑w j

d ln Z j .

(A1)

i

Taking the logarithmic partial differential of equation (A1) with respect to Zj yields: ∂ ln pix ∂ ln Z j

−

∂ ln P x ∂ ln pix ∂ ln p

x i

∂ ln Z j

=

cij wi

.

(A2)

Noting that ∂ ln Px/∂ ln pix = wi equation (A2) reduces to: ∂ ln pix ∂ ln Z j

(1 − wi ) =

cij wi

,

or, more simply,

zijA =

cij wi (1 − wi )

Equation (A3) is identical to text equation (17).

.

(A3)