1 s2.0 S0022199697000081 Main

Journal of International Economics 43 (1997) 437–461
Exchange rate pass-through for strategic pricing and

advertising: An empirical analysis of the U.S.
photographic film industry
Vrinda Kadiyali*
Johnson Graduate School of Management, Cornell University, 560 Malott Hall, Ithaca, NY 14853 -
4201, USA
Received 20 September 1996; revised 12 December 1996; accepted 6 February 1997
Abstract
This paper builds a model in the ‘‘new empirical industrial organization’’ framework
[Bresnahan, T., 1989. Industries and Market Power. In: Schmalensee, R., Willig, R. (Eds.),
Handbook of Industrial Organization. North Holland, Amsterdam.] to estimate exchange
rate pass-through in price and advertising. Price-cost margins, market structure and firm
conduct are estimated endogenously, allowing for precise estimation of the degree of, and
reasons for, incomplete pass-through. I demonstrate that price and advertising pass-through
are functions of demand, cost and market conduct in the short run. In the long run, the
market conduct and cost technologies are, in turn, determined by the exchange rate
movements. 1997 Elsevier Science B.V.
Keywords: Exchange rate; Pass-through; Pricing; Advertising; Industry structure
JEL classification: F12; F14; L13; L60
1. Introduction
The U.S. economy saw big swings in the real exchange rate in the 1980s. The
dollar started appreciating in 1980, and continued to do so till November 1985.
This was followed by a period of steady devaluation for most of the second half of
*Tel.: (607)-255-1985; fax: 607-254-4590; e-mail: [email protected]
0022-1996 / 97 / $17.00  1997 Elsevier Science B.V. All rights reserved

PII S0022-1996( 97 )00008-1
438 V. Kadiyali / Journal of International Economics 43 (1997) 437 – 461
the decade. These movements in exchange rate, however, were not fully reflected
in U.S. import prices. The phenomenon of import price movements not reflecting
exchange rate movements is termed ‘‘pricing to market’’ (Krugman, 1987), or as
‘‘incomplete pass-through’’.1 This phenomenon has both macro- and micro-
economic aspects. On the macro level, it has implications for the U.S. balance of
trade. This paper examines some microeconomic implications of the incomplete
pass-through phenomenon. The microeconomic aspects of pass-through include
how incomplete pass-through can be a profit maximizing strategy for firms and
tells us what types of market structures or exchange rate movements are likely to
generate incomplete pass-through.
In this paper, I study incomplete pass-through in pricing by Fuji Photo Film of
Japan in the U.S. photographic print film industry. Eastman Kodak and Fuji Photo
Film of Japan are the two major players in this industry. See Table 1 for market
shares, and Table 2 for descriptive statistics (see Appendix A for details on how
these data were collected). Fuji Photo produces print film in Japan, and the U.S.
subsidiary imports film into the U.S. Fuji Photo made a successful entry into the
U.S. in the early 1980s, a time when the dollar was overvalued. Fuji pursued
aggressive pricing and advertising policies to build a presence in the U.S. By the
time the dollar started to turn around, Fuji had already entrenched itself. In the
latter half of the decade, as importing film from Japan became more expensive,
Fuji did not raise prices to reflect the movements of the dollar–yen exchange rate.
Table 1
Annual market shares
Year Kodak Fuji
1979 79.92 4.12
1980 79.57 5.60
1981 77.95 5.75
1982 74.42 6.27
1983 69.25 8.38
1984 66.17 9.87
1985 63.96 11.48
1986 62.33 13.80
1987 63.57 16.36
1988 66.46 17.21
1989 65.14 17.33
1990 62.93 17.37
Industry analysts believe that a 5% market share defines a significant market presence. Store-brand
film, and other firms like Konika, 3M, etc. had an insignificant market share.
1
Pricing-to-market studies examine how prices for the same good differ in different geographic
markets, and whether these differences can be explained by exchange rate differences alone. Pass-
through relates the sensitivity of import prices to exchange rate movements, and is therefore, a broader
concept. This paper is a study of pass-through.
V. Kadiyali / Journal of International Economics 43 (1997) 437 – 461 439
Table 2
Descriptive statistics
Variable Mean Std. Dev. Minimum Maximum
Kodak quantity 5.9E107 1.2E107 3.7E107 8.2E107
Fuji quantity 1.1E107 6 330 601 2 527 917 2.1E107
Kodak price ($) 2.06 0.23 1.71 2.48
Fuji price ($) 1.66 0.25 1.26 2.09
Kodak advertising ($) 6 785 824 3 805 630 941 327 1.6E107
Fuji advertising ($) 869 311 629 452 2264.4 2 026 871
Income ($) 2397.05 277.75 2014.44 2806.88
Price of capital (U.S.) (%) 6.75 2.45 4.50 12.80
Price of labor (U.S.) ($) 6.54 0.41 5.63 7.21
Price of silver (U.S.) ($) 2371.34 1775.68 960.65 11 813
Price of capital (Japan) (%) 4.57 1.30 2.61 7.08
Price of labor (Japan) ($) 6.83 2.03 4.57 10.08
In addition, there was incomplete pass-through in Fuji’s advertising expenditures.

That is, as cost of importing film increased, Fuji prices did not rise in step. This
meant that variable profits were being squeezed. Despite this, Fuji’s advertising
budget was not scaled back.
This paper uses the ‘‘new empirical industrial organization’’ (NEIO) framework
(Bresnahan, 1989) to study pass-through. The econometric model for the film
market is from Kadiyali (1996). It includes firm-level specification of demand and
costs. I test a variety of alternate market structure regimes in Kadiyali (1996),
using the best-fitting one for the analysis of pass-through. Therefore, the analysis
of pass-through in this paper is directly connected to firm behavior and market
structure. In addition, this paper attempts to link exchange rate pass-through with
the advertising decisions of firms. This represents a step towards a more complete
examination of any firm’s multiple decisions in relation to exchange rate
movements.2 In addition, I conduct a simulation that reveals that in the short term,
market structure is a determinant of pass-through, but in the long run, market
structure is also a function of the exchange rate movements. Since market structure
and pass-through are determined simultaneously, it is important to model and
estimate them simultaneously.
The rest of the paper is organized as follows: Section 2 provides a brief
background. Section 3 contains a brief description of Kadiyali’s (Kadiyali, 1996)
structural estimation model, connecting that methodology to those currently used
in pass-through analysis.3 Section 4 discusses the empirical results on price and
advertising pass-through, and Section 5 concludes.
2
The decision to enter a foreign market is itself a function of the exchange rate movements. I do not
model this decision of Fuji endogenously. See Dixit (1989) for a formal model of this issue, and
Feinberg (1992) for an empirical test of the model with aggregate data.
3
For a literature review in this area, see Menon (1995).
2. Background
Krugman (1987) classifies pass-through models as either static or dynamic. In

static models, firms can price to market because they can price discriminate across
their domestic and export markets. This assumes that there is incomplete arbitrage
between these markets. Dynamic models, by definition, have multiperiod effects
on the demand or supply side. For example, on the demand side, reputation effects
could constrain firms from passing through every fluctuation in the exchange rate.
There may be other forms of intertemporal effects in demand, e.g., brand loyalty
on the part of consumers, network externalities, switching costs, multiperiod
pricing and advertising effects, any or all of which can cause incomplete pass-
through (see Froot, Klemperer, 1989).
Supply relationships can have an element of adjustment costs (Kasa, 1992), or
capacity constraints that cause price stickiness, especially if the exchange rate
fluctuation is either unanticipated or expected to reverse (Baldwin, 1988). It might
not be profitable for firms to expand their production and overseas sales and
distribution networks in response to every movement of the exchange rate. More
generally, there may be substantial costs of entry, expansion, and exit that force
producers to absorb exchange rate fluctuations in their mark-ups instead of passing
them on to the consumer as price changes (Dixit, 1989). Dynamic optimization
can also arise because of firms’ different profit maximizing horizons. Ohno’s
(Ohno, 1990) simulation analysis shows that the degree of pass-through is
inversely related to the length of firms’ planning horizons.
Another consideration relates to how the nature of exchange rate movements
affect the degree of pass-through. Studies on this aspect use dynamic models of
pass-through. For example, Froot, Klemperer (1989) and Kasa (1992) show that
firms are more likely to respond to large, permanent exchange rate changes than to
changes that are small and temporary. In the first case, a firm might respond by
changing its export price, expanding or contracting production and distribution. In
the second case, its response might be to absorb temporary changes in its product
mark-up rather than to adjust price.
A related question asks how much of price discrimination in the presence of
pricing to market is deliberate on part of the producers, and how much is because
of unexpected exchange rate movements. Giovannini (1988) and Marston (1990)
separate the two effects and conclude that even after allowing for exchange rate
surprises, there is still strong evidence to support deliberate pricing to market.
Researchers have documented two types of evidence of incomplete pass-through
for various industries across various countries. First, they have compared the price
of the domestic good in domestic and export markets, or across two export
markets. Second, they have compared the prices of imports and the domestic good
for the same product class. For example, Dornbusch (1987) offers empirical
evidence at the 2- and 4-digit SIC industry level on the extent of pass-through. He
does so by examining the price differential between export and import prices of
similar goods in the U.S. as a function of the exchange rate movements. Other
empirical studies, e.g., Mann (1986), have also compiled evidence at aggregated
industry levels. Feenstra (1989) estimates pass-through equations by regressing the
prices of the importing firm on regressors that include a time trend, a demand
shifter, a measure of the cost of production in the country of production, and as a
measure of competitive effects, the price of the same product produced domestical-
ly. His equations give a measure of how demand, supply, and competition affect
the import price, and hence the degree of pass-through. While this evidence amply
demonstrates the existence of incomplete pass-through, it does not account for
intra-industry differences in demand, cost, or the market structure of the industry
that are at the heart of any complete explanation of the microeconomics of
incomplete pass-through.
Knetter (1989); Schembri (1989); Goldberg (1995) are three exceptions to the
reduced-form estimation in the empirical pass-through literature. Knetter investi-
gates whether U.S. and German firms price-discriminate across various export
markets. He estimates marginal cost, and hence mark-ups, for these firms. He finds
that pricing to market is possible because these firms can price discriminate across
these markets. However, the data used in his study are at the 7-digit SIC level, and
the analysis of competitive effects in any industry is minimal. Schembri’s analysis
is more structural. However, he too does not model the impact of competitive
effects on pass-through. Goldberg estimates firm-level mark-ups, and hence
firm-level pass-through for the U.S. automobile industry. She imposes a Bertrand
specification on interactions of firms in the industry.
The NEIO literature provides a fully-structural microeconomic framework for
analyzing pass-through issues at an industry level. Specifically, these studies
estimate price–cost margins from market outcome data which is critical to
correctly identifying the degree of pass-through. In addition, these studies estimate
the underlying market structure and conduct of firms in the industry, incorporating
industry- and firm-specific features into the micro- and econometric modeling of
firm and consumer behavior in the industry. These features enable us to identify
the source of pass-through, i.e., to separate demand, cost, and market structure
explanations of pass-through.
I use different data from most aggregate studies, which use import unit values
instead of unit prices, or price indexes instead of actual prices. These data suffer
from the obvious problem of not capturing real market outcomes. Lawrence
(1990) has even argued that a large part of lack of pass-through arises because of
excessively aggregated data. In this study, because data are at firm level, there is
no aggregation problem. I estimate cost coefficients endogenously, not from
accounting data of firms. This adds to the credibility of results.
Further, if the industry being studied is subject to non-tariff, the observed price
and quantity data will not be representative of true pass-through (Bhagwati, 1988).
The photographic film industry is not subject to these barriers.
Another point of departure from other empirical studies of pass-through
(excepting Goldberg, 1995) is that I do not have data on the price of Fuji film in
Japan. Therefore, I cannot compare Fuji’s pricing in Japan, its domestic market, to
its pricing in U.S., its export market, as earlier studies have done. However, as
Marston (1990) points out, if the two markets are separate and the marginal cost of
production is constant, as I assume for my model, the analysis of the two markets
can be separated. Therefore, this is not a study of pricing to market, but of
incomplete pass-through.
3. The model
In Kadiyali (1996), I empirically estimate a structural model of industry

behavior for the duopolistic market of the U.S. photographic print film industry. In
this paper, I estimate the impact of that structure on pass-through of exchange rate
movements on price and advertising. Much of the following description of the
estimated model and results is taken from Kadiyali (1996).
The demand functions facing Kodak and Fuji are assumed to have the following
form:
O a Q 1 a Y,
3
q1 5 a11 P1 1 a12 P2 1 g11 A 11 / 2 1 g12 A 21 / 2 1 a0 1 i i 4 (1)

i 51
1 b 1 O b Q 1 b Y,
3
q2 5 a21 P1 1 a22 P2 1 g21 A 11 / 2 1 g22 A 21 / 2 0 i i 4 (2)

i 51
where q represents quantity, P price, A advertising, Q the quarterly dummies for

the first three quarters, Y for real income, subscripts 1 Kodak and subscript 2 for
Fuji.4
Advertising has been modeled to capture ‘‘advertisement fatigue’’, a widely
recognized phenomenon (see Lillien et al., 1992). Multiperiod advertising effects,
defined as advertising in one period influencing demands in that as well as future
periods, have been used extensively in the empirical I.O. literature (e.g., the
depreciating stock of ‘‘goodwill’’ formulation of Roberts, Samuelson, 1988).
However, a static specification is adequate for this study because tests on the
estimates of demand reject the null of first-order autocorrelation. This evidence is
consistent with many marketing studies that have found no carry-over effects of
advertising beyond one quarter for consumer nondurables (see Lillien et al., 1992).
Quarterly dummies are included in the demand specification to account for
seasonal patterns in quantities, prices and advertising. The Wolfman Report, an
4
Other specifications like the double logarithmic greatly complicate or violate conditions for
identification and estimation for various market structures (e.g., multiple solutions or noninvertibility
resulting from solving first-order conditions when demand is double logarithmic). Hence, in the absence
of industry evidence favoring one form over another, I adopt the linear specification for tractability.
annual trade journal, reports that the first quarter is the quarter with the lowest
sales, and highest in the fourth.
Real per capita income is the exogenous demand shifter. Amateur camera sales
boomed in the 1980s, contributing to the derived demand for film. Although
camera sales (or the stock of cameras in use) could be another potential demand
shifter, they have not been included as an explanatory variable in the demand
function. Given the high correlation between camera sales and prices, film prices,
advertising, and real income, it is not possible to include the price or quantity of
cameras to estimate the impact of camera sales on film demand. This exclusion
probably does not cause misspecification. Kodak’s market power in the amateur
camera market suffered serious loss in the 1980s, which meant that Kodak was not
in a position to use its market power in the camera market to gain market power in
the film market. Fuji also had limited market power in the amateur camera market
segment. Therefore, there is no danger of misstating market power of either Kodak
or Fuji by not including the price of cameras. If the price of cameras is only a
demand shifter, its role has been captured by including real per capita income,
with which it is highly (positively) correlated.
The cost structure of both the domestic firm, Kodak, and the importing firm,
Fuji are also explicitly modeled. Firms are posited to have constant marginal cost
of production, i.e., if C5total production costs, MC5marginal costs, q5quantity
produced, F 5fixed costs, then
Ci (qi ) 5 MCi qi 1 Fi . (3)
Marginal cost is specified to be a linear function of the material and factor

prices for both producers.
* pcap1 1 n *12 plab 1 1 n 13
MC1 5 c 1 5 n 11 * ag1 1 n 14
* t, (4)
* pcap2 1 n *22 plab 2 1 n 23

MC2 5 c 2 5 n 21 * ag2 1 n 24
* t, (5)
where pcap represents the price of capital, plab that of labor, ag of silver, and t a
time trend.
The duopolists face different factor and material prices, because Kodak
manufactures in the U.S. and Fuji in Japan. However, the price of silver in the two
countries is highly correlated. This makes the estimation of separate parameters in
Eqs. (9)–(12) below in which both Kodak and Fuji marginal cost enter, difficult.
Therefore, only the U.S. silver price is used for estimating the silver cost
parameter for both Kodak (n 13 ) and Fuji (n 23 ). Other material inputs include
gelatin sheets and cyanine dyes. Price series for these two inputs are not available
on a consistent basis for the U.S. and Japan. When they are available, have very
little variance. Moreover, these two form a nearly insignificant proportion of
material costs. Therefore, only the price of silver is included in the marginal cost
specification.
The time trend variable captures technological and other trend variables that
affect the marginal costs of production. Some of these trends are perhaps captured
more appropriately by fixed costs, e.g., R&D costs. Specifically, industry observers
believe the decrease in marginal cost over the period is a result of R&D
investments. Hence, marginal cost could be misspecified by not accounting for
R&D. The time trend is a simple way of capturing these missing elements.
To estimate the price and advertising pass-through, Fuji’s film import costs in
the U.S. must be compared to price and advertising levels for Fuji film in the U.S.
To do this, I convert the yen costs of production into dollar terms (see Appendix A
for details). In addition, the ad valorem tariff rate is added to Fuji’s dollar-
denominated cost of producing and importing film. This gives us in dollar terms
the final cost of importing film into the U.S. This cost number is used to estimate
Fuji’s (and Kodak’s) determination of optimal price and advertising levels in the
U.S. film market. Since fixed costs are not explicitly modeled, mark-up numbers
are variable profit numbers.
The model of industry behavior on both the demand and supply side in this
paper is static as are the models in Knetter (1989); Schembri (1989); Goldberg
(1995). Empirically, on the demand side, advertising and price persistence effects
seem to be absent. Network externalities and switching costs are not an issue in
this market. On the supply side, a search of the trade press reveals that capacity
constraints are not an issue. However, the empirical model does not account for
adjustment costs, or for investments in distribution channels or any other sunk
entry cost. Some of these costs could be captured in fixed costs of production or
serving the market. As noted earlier, even if a measure for these could be found,
fixed costs drop out of the optimization rules of the estimated model.
The use of a static framework for modeling demand and supply in this industry
assumes that any exchange rate change is expected to be permanent. This
assumption is clearly hard to justify (though the use of quarterly, instead of daily,
data may iron out some temporary movements). The current study can, therefore,
be seen as first-cut approximation to the actual process of pass-through.
Given the demand and cost structure described above, firms maximize profits:
S
Pi 5 (Pi 2 ci ) gi 0 1 aii Pi 1 aij Pj 1 gii A 1i / 2 1 gij A j1 / 2 1 O a Q 1 a YD
3
i 51
i i 4
2 A i 2 Fi . (6)
Japanese firms are thought to have a longer horizon of optimization than American
firms. Thus, the assumption of single-period profit maximization may not be
appropriate for Fuji. However, the lack of dynamic elements in either demand or
supply precludes the possibility of multiperiod optimization for Fuji in my model.
I estimate several game-theoretic interaction regimes between the duopolists.
Some of these games are leader–follower in both price and advertising, Nash in
both pricing and advertising, tacit collusion in one or both instruments, and
combinations of these behaviors across price and advertising setting. This method
is from Gasmi et al. (1992). This model cannot incorporate non-differentiable
strategies such as trigger strategies, nor are there sufficient data points in this study
to permit the estimation of endogenous switching models like Porter’s (Porter,
1984). The estimated equations are:
O aQ 2a Y
3
1/2 1/2
q1 2 a11 P1 2 a12 P2 2 g11 A 1 2 g12 A 2 2 a0 2 i i 4
i 51
5 u1 Demand for Kodak film, (7)
O bQ 2b Y
3
q2 2 a21 P1 2 a22 P2 2 g21 A 11 / 2 2 g22 A 21 / 2 2 b0 2 i i 4

i 51
5 u2 Demand for Fuji film, (8)
q1 1 w11 P1 1 w12 P2 2 f1 Z1 2 f5 Z2 5 u 3 Price FOC for Kodak, (9)
q2 1 w21 P1 1 w22 P2 2 f6 Z1 2 f2 Z2 5 u 4 Price FOC for Fuji, (10)
j 11 P1 1 j 12 P2 1 A 11 / 2 2 f3 Z1 2 f7 Z2 5 u 5 Advertising FOC for Kodak,

(11)
j 21 P1 1 j 22 P2 1 A 11 / 2 2 f8 Z1 2 f4 Z2 5 u 6 Advertising FOC for Fuji,

(12)
where Z1 is a vector5h pcap1 , plab 1 , ag1 , tj, Z2 is a vector5h pcap2 , plab 2 , ag2 ,
tj. hu 1 , u 2 , u 3 , u 4 , u 5 , u 6 ) are assumed to be joint-normally distributed errors.
Different games corresponds to different restrictions on the w, f and j parameters
in the system of Eqs. (7)–(12) above. Appendix B has conditions for the
identification of the model above, and the restrictions on the w, f, and j
parameters in the model above that lead to the two of the twelve estimated games.
Each estimated game can be distinguished econometrically from the others.
Marginal cost of production for both firms can be recovered from the estimates of
coefficients of the system of equations.
The various games described above are estimated by using quarterly firm-level
data on price, quantity, advertising expenditure, and elements of production costs.
Prices, advertising expenditures and cost elements are all expressed in dollar terms
(except the cost of capital, which is proxied by interest rate). As discussed earlier,
this means that the marginal cost of Fuji’s production is converted to marginal cost
of Fuji’s importing film into the U.S.
Table 3
Correlation coefficients
q1 q2 P1 P2 A1 A2 Y Pcap1 Plab 1 Ag1 Pcap2 Plab 2 T
q1 1.0 0.8 20.4 20.4 0.8 0.8 0.6 20.4 20.6 20.4 20.2 0.5 0.5
q2 1.0 20.7 20.8 0.6 0.5 0.9 20.8 20.9 20.6 20.5 0.9 0.9
P1 1.0 0.8 20.5 20.3 20.7 0.7 0.7 0.4 0.2 20.5 20.8
P2 1.0 20.4 20.2 20.8 0.8 0.7 0.6 0.4 20.6 20.9
A1 1.0 0.7 0.5 20.5 20.4 20.3 20.3 0.4 0.5
A2 1.0 0.3 20.3 20.3 20.3 20.2 0.2 0.3
Y 1.0 20.8 20.9 20.6 20.5 0.9 0.9
Pcap1 1.0 0.6 0.8 0.7 20.7 20.9
Plab 1 1.0 0.5 0.2 20.8 20.9
Ag1 1.0 0.5 20.4 20.6
Pcap2 1.0 20.6 20.5
Plab 2 1.0 0.8
T 1.0
4. Results and discussion
4.1. Results of the structural estimation analysis
As Table 3 shows, high multicollinearity exists between variables. This prevents

FIML estimation, where parameters and the variance–covariance matrix are
estimated simultaneously. The multicollinear variable series in my data do not
have enough variation in them to undertake this simultaneous estimation of
parameters and the matrix. Hence, the estimates for this study are iterated 3SLS
estimates, where the parameters and the variance–covariance matrix are estimated
sequentially over several iterations. For the post-entry period, I identify the market
regime as the game that generates estimates with lowest minimized sums of
squares (573.89, specifically). This is a tacitly collusive market outcome in which
firms collude on both price and advertising setting. None of other estimated games
give meaningful results.5 For example, the next-best-fitting game, which is
collusion in pricing with Nash advertising, has minimized sums of squared errors
of 89.51. However, this game results in a positive own-price, and negative
own-advertising coefficient in Kodak’s demand function. When I attempt to
impose economically intuitive parameter restrictions, i.e., negative price and
positive advertising, the estimates fail to converge (i.e., minimized sums of
squared errors51015). I encounter similar problems with other games.
Table 4 shows the estimates for the best-fitting post-entry game. The estimates
suggests that in the post-entry period, the duopolists tacitly collude on both prices
5
Estimates are available from the author.
Table 4
Estimates of the best fitting game: Tacit collusion in pricing and advertising
Minimized sums of squared errors573.90
Number of observations544
Parameter Estimate Std. Error T-Statistic
Kodak demand:
Constant 28.24E107 2.59E107 23.18
1st quarter 29977286 2755785 23.62
2nd quarter 4.88E106 1.61E106 3.03
3rd quarter 4.89E106 1.56E106 3.13
Income 34954.60 8226 4.25
Kodak price 25.61E106 1.93E106 22.90
(Kodak Ad)1 / 2 4901.12 1089.14 4.50
Fuji price 3.23E107 8191679 3.95
(Fuji Ad)1 / 2 4451.85 1451.37 3.07
Kodak costs:
Price of capital 2.26E202 4.26E202 0.53
Price of labor 0.35 5.91E202 5.94
Price of silver 5.04E204 2.20E204 2.29
Time trend 20.17 7.68E202 22.20
Weight of 0.24 6.61E202 3.67
Kodak in the (tacit) joint profit maximization exercise Fuji demand:
Constant 25.82E107 5552088 210.48
1st quarter 1250066 1043656 1.20
2nd quarter 410014.6 439184.3 0.93
3rd quarter 1180663 440303.8 2.68
Income 20695.2 1177.98 17.57
Kodak price 5497617 1541277 3.57
Fuji price 2215030.86 488706.5 20.44
(Kodak Ad)1 / 2 2357.39 537.50 4.39
(Fuji Ad)1 / 2 2483.94 594.68 4.18
Fuji costs:
Price of capital 0.22 0.10 2.15
Price of labor 0.18 0.12 1.51
Time trend 1.00E202 7.39E202 0.14
and advertising. Fuji enjoyed substantial cost and demand advantages over Kodak.
These advantages lead to a large weight of 76% on Fuji’s profits in the tacit joint
profit maximization market outcome.
On the cost side, the major advantage that Fuji has is that its cost of producing
and importing a roll of film into the U.S. is about $0.31, compared to Kodak’s
$0.78 for producing in the U.S. (calculated at mean values for factor prices over
the period). Demand advantages stem from favorable own- and cross-price
elasticities, and favorable own- and cross-advertising elasticities (see Table 5).
Note that Fuji’s cross-advertising elasticity is greater than Kodak’s, i.e., Fuji
Table 5
Elasticity estimates
1980–90 Kodak Own price 20.2021
Own ads 0.1066
Fuji price 0.9397
Fuji ads. 0.0327
Y 1.4665
1980–90 Fuji Own price 20.0323
Own ads. 0.0943
Kodak price 1.0727
Kodak ads. 0.2655
Y 4.4946
benefits more from Kodak advertising that Kodak does from Fuji advertising. The
difference in elasticities is significantly different from zero. Fuji’s own-price
elasticity is lower than Kodak’s.
Low prices (despite the high growth of demand, despite Fuji’s increasing costs
of imports in the second half of the decade, and despite the low price elasticities),
high advertising level (despite low advertising elasticities), low price and advertis-
ing elasticities, and the lack of entry despite significant profit margins in the
industry show that Kodak and Fuji undertook limit pricing and limit advertising
through the 1980s. The duopolists need to collude to keep prices low enough to
deter entry, but higher than if there were more competitive pricing outcomes.
Given collusion in advertising, profits are higher when firms collude on prices than
if they do not collude on prices. Why the duopolists need to collude on advertising
to erect an entry barrier remains to be explained. Given the lack of business
stealing effects of advertising (i.e., positive cross-advertising effects), colluding on
advertising levels is profitable because it internalizes the externality aspect of
advertising. Note, however, that the model does not explicitly account for entry
deterrence in the profit maximization exercise of firms. In order to do that, I would
need data to estimate the counterfactual demand and cost structure for a potential
entrant, which the incumbent duopolists use in their optimization routine. In the
absence of such data, the model specification is incomplete. Therefore, the
evidence of limit pricing and limit advertising cannot be substantiated further.
4.2. The relationship between pass-through and market structure
Exchange rate movements affect optimum prices and advertising in the

following way: Consider a duopolist firm 2, competing with firm 1, and
maximizing profits under a general specification of demand function, with constant
marginal cost. MC2 is the cost of production. To obtain cost of importing, MC2 is
multiplied by the exchange rate e. The assumption here is that all production costs
are incurred in the home country. If any fraction of product costs are incurred in
the importing country, then the pressure to pass-through exchange rate changes is
less.
p2 5 (P2 2 MC2 e)q2 2 A 2 , (13)
p1 5 (P1 2 MC1 )q1 2 A 1 , (14)
q2 5 q2 (P1 , A 1 , P2 , A 2 ), (15)
q1 5 q1 (P1 , A 1 , P2 , A 2 ), (16)
≠p ≠q ≠P
]2 5 0 ⇒ (P2 2 MC2 e) ]2 ]1 1 ]]
≠P2 S
≠q2 ≠A 1 ≠q2
]] 1 ] 1 q2 5 0,
≠P1 ≠P2 ≠A 1 ≠P2 ≠P2 D (17)
≠p ≠q ≠P
]1 5 0 ⇒ (P1 2 MC1 ) ]1 ]2 1 ]]
≠P1 S
≠q1 ≠A 2 ≠q1
]] 1 ] 1 q1 5 0,
≠P2 ≠P1 ≠A 2 ≠P1 ≠P1 D (18)
≠p ≠q ≠P1
]]2 5 0 ⇒ (P2 2 MC2 e) ]2 ]]
≠A 2 S≠q2 ≠A 1 ≠q2
1 ]] ]] 1 ]] 2 1 5 0,
≠P1 ≠A 2 ≠A 1 ≠A 2 ≠A 2 D (19)
≠p ≠q ≠P2
]]1 5 0 ⇒ (P1 2 MC1 ) ]1 ]]
≠A 1 S≠q1 ≠A 2 ≠q1
1 ]] ]] 1 ]] 2 1 5 0.
≠P2 ≠A 1 ≠A 2 ≠A 1 ≠A 1 D (20)
To see how price and advertising change when e, the exchange rate changes, we
would have to solve Eqs. (15)–(20) for prices, advertising and quantities for both
firms in terms of the exogenous variables i.e., in terms of income, marginal cost
components, and other quarterly dummies. The system above cannot be solved
without imposing some constraints on the functional form parameters, as well as
the market conduct.6 Imposing some functional form structure on the system above
(e.g., Nash pricing and advertising with demand and cost like Eqs. (7)–(12) above)
does help solve the system, but even then, the sign for expressions for dp2 / de and
dA 2 / de cannot be determined unambiguously.
The general lessons learnt from expressions (15) through (20) is that equilib-
rium prices and advertising levels (and hence, pass-through) depend on demand
elasticities, cost structures (the MC parameters), as well as firm conduct. Clearly,
imposing an incorrect market conduct regime when estimating the pass-through
will give an incorrect estimate of the pass-through. This underscores the
importance of estimating market structure endogenously in the NEIO framework.
Turning now to the pass-through in the film market, Fig. 1 plots Fuji prices,
advertising dollars, dollar profits and the yen–dollar exchange rate. I rescale
variables by dividing each series by its sample mean to fit into one graph. The
graph shows that the fluctuation in Fuji’s advertising budget does not seem highly
6
Even if the functional forms of Eqs. (1)–(5) are imposed on (15) through (20) above, closed-form
solutions for prices, advertising, and quantities cannot be found.
Fig. 1. Plot of Fuji’s prices.
correlated with the exchange rate movements (correlation of the dollar / yen and
advertising50.24, significant at 89%). Price and exchange rate seem to move
together, though not all the time (correlation of dollar / yen and price50.58,
significant at 99%). Hence, there is incomplete pass-through in both pricing and
advertising. This is the expected result, given that the market structure is far from
competitive.
Notice that Fuji profits fluctuate greatly (correlation of dollar / yen and profits5
0.66, significant at 98%). These fluctuations are in part driven by the incomplete
pass-through. For example, in the period 1988–90, Fuji profits and prices both
declined, even though the appreciating yen increased the cost of importing film.
Profit fluctuations are also driven by fluctuations in demand, e.g., in the period
1988–1990, demand dropped mostly in response to the recession in the U.S.
economy for much of this period.
There are four features of this pass-through that I will discuss: how large the
price pass-through is, whether is it symmetric in appreciation and depreciation,
how it varies with market share, and how the advertising pass-through behaves.
Table 6 documents Fuji price and advertising elasticities with respect to the
dollar–yen exchange rates. These have been calculated by solving the estimated
system of Eqs. (7)–(12) for equilibrium prices and advertising. Elasticities are
estimated by using the estimated derivatives of price and advertising with respect
to exchange rate, and sample means for variables.
For price pass-through, the best comparison point for these elasticity estimates
is Goldberg’s (Goldberg, 1995) estimates since her model for measuring pass-
Table 6
Exchange rate elasticity of Fuji price and advertising
Price elasticity Advertising elasticity
1980–1984 0.076 2.967
(Average market share of Fuji57.17%)
1985–1990 0.178 4.597
(Average market share of Fuji515.59%)
through is closest to the one employed here. Goldberg endogenously estimates

demand and cost structures, but imposes Bertrand–Nash pricing behavior in a
static optimization model. She finds that in the automobile industry, the pass-
through coefficient for Japanese cars is low (between 15–30%), the reason for
which is high demand elasticities. In the film market, the estimated own-price
elasticity for Fuji is lower than Kodak’s. Despite this, pass-through elasticities are
very small, especially in the pre-1984 period. This is despite collusion which
should make price increases easier to sustain. The threat of entry could be keeping
Fuji from increasing prices.
Note also that the price pass-through is not symmetric. This contrasts with
Lawrence’s (Lawrence, 1990) finding of symmetric pass-through during dollar
appreciation and depreciation. Goldberg’s (Goldberg, 1995) simulations reveal
asymmetric pass-through. Prices increase more in dollar depreciation than they
decrease in dollar appreciation because demand elasticities change asymmetrically
with price increases and decreases. In the film market too, prices do not decrease
as much in the dollar appreciation (elasticity50.076) as they increase in dollar
depreciation (0.178). An important reason for low pass-through is firm conduct.
Collusion makes it easier to increase prices, whereas a decrease in prices could
destabilize the collusion.
Fuji’s market share was increasing through the entire period 1980–90. Fuji’s
pass-through (elasticity) increases as its market share increases, as predicted by
Dornbusch (1987) and Feenstra et al. (1996). Both explanations rely on increasing
market power (and in the latter paper, similarity of incentives of importers to
pass-through as importers’ market share increases), which is applicable to this
market as well. Collusion provides additional reason to be able to increase
pass-through. Notice that the exchange rate movements in the second half of the
1980s are more likely to induce Fuji to increase its price, and this is also the time
when the collusive equilibrium was well established, making it easier to sustain
matched price increases.
However, it is not clear how long such a collusive equilibrium would last since
Fuji was under continuous pressure in the second half of the 1980s to pass on
increases in the cost of import. This is especially true because there are demand
and cost advantages underlying the 76% weight of Fuji profit in the joint profit
maximization exercise. Some of the cost advantages began to erode in the second
half of the decade, which reduced Kodak’s incentive to continue colluding. Hence,
it was in Fuji’s interest not to increase price. Unfortunately, there are too few data
points to test a model for switching regressions, e.g., one in which Fuji was the
leader in the first half of the 1980s when it enjoyed cost advantages of importing
film from Japan, to one in which the game was Nash in the second half when these
advantages were getting eroded.
Turning now to the advertising pass-through, Table 6 shows a large advertising
elasticity for the whole period, and an increase in this elasticity during dollar
depreciation. Once again, market structure affects the advertising pass-through
numbers. On the demand side, the estimated demand functions show that there are
no business-stealing effects of advertising. This feature benefits Fuji more than it
does Kodak. So, when exchange rates fluctuate, what strategy should Fuji use in
constructing its advertising budget?
In the first half of the decade, when costs of importing are low, Fuji enjoys a
greater mark-up. Hence, it can afford a large advertising budget. In the second
half, when mark-ups are getting squeezed, there is increasing pressure on the
advertising budget. Despite this, there is an increase in the advertising pass-
through elasticity. Therefore, counter to the argument made by Dornbusch (1987),
even though Fuji’s market share increases in the second half of the decade, its
ability to pass through cost increases in advertising is constrained. First, its
increasing market share was due to (low price and) high advertising, and a reversal
of these policies would decrease market share. Second, (increasing prices or)
decreasing advertising would make competitors’ entry viable, especially because
demand mostly increased through the period.
My analysis shows that pricing and advertising are determined by costs,
exchange rates, and demand and market structures. In the long-run, however, there
is a simultaneous determination of pass-through, cost, and market structures.
Therefore, the degree of pass-through both determines and is determined by the
market structure, demand, and cost parameters. For example, the cost structure
determines the pass-through, and in the long run, is itself a function of the
exchange rate movements. Similarly, market structure and pass-through are
determined by costs and demand, and yet in the long-run, the market structure can
itself be a function of the exchange rate movements.
One way to demonstrate the above argument is to simulate at what exchange
rate level the current collusive market structure changes to some other market
structure. For example, if the cost of importing film after 1984 increases at 10% or
50% more than the actual increase, does the collusive market structure break
down? 7 What market interaction would replace collusion if it were to break down?
There are several models of collusion breakdowns developed in industrial
organization economics (see Porter, 1984; Rotemberg, Saloner, 1986; Bagwell,
Staiger, 1995). In applying these models to international economics, Davidson
7
I thank a referee for suggesting that I include this discussion.
(1984) develops a model showing how tariffs (or any increase in cost of
importing) affect the ability of domestic and foreign, importing, firms to collude.
Unlike the papers mentioned above, the model in this paper does not have a
specific supergame model of firm conduct. Such a supergame model would include
punishment strategies for sustaining collusion, and entry deterrence behavior. In
the absence of a specific supergame model of firm conduct, it is not possible to
benchmark collusive profits with profits if collusion were to break down.
However, we can vary exchange rates and estimate profits for alternative games.
In particular, assuming that Fuji’s cost of importing film is higher than the actual
levels, I test if collusion is likely to break down because some other game profits
look more attractive to either Kodak or Fuji. The collusive game gives larger
profits for Kodak and Fuji than any other game, even if Fuji’s cost of importing
film for Fuji increases by 50%. The problem is that for these profits to be realized,
Fuji must increase its price, Kodak must drop its price, and both Kodak and Fuji
must cut back advertising. These actions lead to an increase in Kodak sales, and a
decrease in Fuji’s. How likely are these actions? This returns us to the issue raised
earlier on the punishment mechanism used to sustain collusion. Also, the analysis
ignores the effect of higher Fuji prices, lower Kodak prices, Kodak and Fuji
advertising on competitor entry, and the effect of such an entry on the sustainabili-
ty of collusion (see Harrington, 1989, for an analysis of the last issue).
Additionally, if the cost of importing had increased by 50%, Fuji would probably
have considered producing film in the U.S. (as it has done recently). This, in turn,
would have very different implications for the sustainability of collusion.
Given the difficulties in interpreting the results of this simulation, I conduct
another simulation exercise that demonstrates how market structure is a function of
exchange rates. I create a new series for exchange rate movements in which the
exchange rate movements are reversed for the time period 1980–90. For example,
the rate in 1980, quarter 1, is assigned to 1990, quarter 4; the rate in 1980, quarter
2, is assigned to 1990, quarter 3, etc. Thus, in new series, the yen in strong in the
first half of the 1980s, and the dollar in the second half. Under this altered
exchange rate movements, Fuji enters at a time when the exchange rate
movements are unfavorable to it, and gets a break only when it is already
reasonably well entrenched in the market. I then estimate various market structures
described in the previous subsection, and find the game that best explains the same
market outcomes of price, advertising, and quantities that were observed in the
time period.
In Table 7, the results of the simulation show that the market outcome would
have been very different had the exchange rate movements been reverse-mapped.
The best-fitting game under the new exchange rate regime turns out to be the Nash
game between Kodak and Fuji, in both pricing and advertising. This is intuitive,
given that Fuji’s cost advantage because of the overvalued dollar when it entered,
is not available to it in the simulated scenario. Specifically, the marginal cost of
(producing and) importing film increased from $0.31 in the real game to $1.08 in
Table 7
Estimates of best game for the simulated exchange rate movements: Nash in pricing and advertising
Minimized sums of squared errors5112.39
Number of observations544
Parameter Estimate Std. Error T-Statistic
Kodak Demand:
Constant 29.80E107 2.00E107 24.91
1st Quarter 2044181 6364619 0.32
2nd quarter 4691555.38 1549280.12 3.03
3rd quarter 4706977.94 1505321.56 3.13
Income 33621.0906 7909.85 4.25
Kodak price 25391317.14 1855294.28 22.91
(Kodak Advertising)1 / 2 4714.13 1048.27 4.50
Fuji price 3.35E107 5971331 5.62
(Fuji Advertising)1 / 2 21030.91 7838.34 2.68
Kodak costs:
Price of labor 9.54E202 1.61E202 5.94
Time trend 4.17E204 8.99E203 0.46
Fuji demand:
Constant 25.89E107 7115948 28.27
1st quarter 1814314 1037479 1.75
2nd quarter 336510.2 451387.8 0.75
3rd quarter 498249.5 505266.1 0.99
Income 20259.73 1324.15 15.3
Kodak price 5469182 1976340 2.77
Fuji price 26090.24 338.91 214.97
(Kodak Advertising)1 / 2 2594.04 611.63 4.24
(Fuji Advertising)1 / 2 3519.06 783.25 4.49
Fuji costs:
Price of labor 0.1153966 2.14E202 5.4
Time trend 1.69E202 6.10E203 2.77
the simulated game. The differences in demand and cost structure are responsible
for the change in the behavior regime from tacitly collusive to Nash. In particular,
notice that Fuji’s own-price response is much lower than before, implying that it
needs less tacit cooperation from Kodak to keep its profits high. Also note that
although Fuji still benefits more from Kodak’s advertising than Kodak does from
its own, Kodak is now in a stronger position, as it also benefits more from Fuji’s
advertising than it does from its own. Thus, had the exchange rate movements
been different as in this simulation exercise, tacit collusion would not be the profit
maximizing strategy for both firms. The results indicate that the profit maximizing
strategy is Nash.
Notice also that Fuji’s cost structure is very different—capital no longer figures
in the marginal cost, but labor does, and silver enters with a positive sign. The
Japanese price of labor in Japan rose from 1986 through 1988, then decreased till
1989, and then rose again in 1990. The price of capital, on the other hand,
decreased from 1986 through 1988 and increased from 1989 onwards. Given that
the reversed exchange rate movements implied a strong yen up to 1985, and a
strong dollar from 1986 onwards, it makes sense that Fuji’s optimal production
technology in the simulation should involve a higher expenditure on labor, and
lower expenditure on capital (Table 7) than the one actually used (Table 4)
(inference based on significance of coefficients).
The analysis above reveals that for small and short-term exchange rate
movements, the analysis of pass-through can take market structure and marginal
cost structure as exogenous. However, these are themselves functions of the
exchange rate in the long run. Thus, a complete analysis of pass-through should
account for its simultaneous determination with market, cost, and demand
structures.
5. Conclusion
This is the first industry study to systematically analyze the phenomenon of

pricing and advertising to market in the structural econometric framework used in
NEIO studies. I demonstrate the additional insights that are gained by estimating
such a model over estimating a simple, reduced-form, pass-through equation. I
also perform a simulation exercise to show how a long-run analysis of pass-
through must account for the optimal choice of firm conduct and cost technology,
given demand conditions. The exercise shows that firm conduct, market structure,
costs, and firm advertising and pricing—and hence, by implication, its pass-
through strategy—are all codetermined.
The structural model used in this study is static. Future work of this type might
address the dynamic aspects of demand and supply alluded to in Section 2.
Especially important aspects are those that provide a buffer to firms as they adjust
to macroeconomic changes like exchange rate movements. Some of these are
inventory, and hence capacity, decisions of firms, and other sunk costs. In
addition, an empirical model of industry structure and conduct should include a
model of equilibrium expectation formation to determine how firms (both
domestic and foreign) distinguish between temporary and permanent exchange rate
movements.
7. Notation
431 (Exchange rates); 442 (International business); 611 (Market structure and
industrial organization); 631 (Industry studies)
Acknowledgments
This paper is based on essay 3 of my dissertation. I am indebted to my advisor,

Rob Porter, for good comments and advice, and for financial support. Special
thanks to Kyle Bagwell and Warren Bailey for their comments, to Vivekanand
Chickermane for help with programming, and to the referees for suggestions that
have greatly improved the exposition of the paper. I am responsible for remaining
errors.
Appendix A
Description of data
I have used several sources to assemble quarterly data on all relevant variables
in the model for the period 1980–90.
I obtained price data from Popular Photography magazine, a monthly publi-
cation with advertisements from various mail order firms for film and other
photographic equipment. I chose 100 ASA film because this speed is uniformly
available for the decade under study. From Popular Photography I determined the
lowest price quote for a roll of 36-exposure color film.
The Wolfman Report provided quarterly sales numbers for the film market. I
obtained market share (in terms of quantities) estimates from the Study of Media
and Markets by Simmons Market Research Bureau and Product Data Series by
Mediamark Resource Inc. to determine quarterly sales for each firm.
Leading National Advertisers’ estimates provided quarterly advertising expendi-
tures by product line. Data on quarterly consumer incomes were from Survey of
Current Business.
Annual Wholesale Price Indexes /Producer Price Indexes issued by the U.S.
Bureau of Labor Statistics are the source for Kodak’s input prices. Silver haldide is
the most expensive material input for film manufacturing. Therefore, I checked
only for the monthly price of silver. Given the buying power of Kodak in the silver
industry, I performed appropriate tests to ensure the exogeneity of silver price.
Also, the correlation between Japanese and U.S. silver prices is about 0.98.
Therefore, I used only the U.S. silver price in the joint estimation of Kodak and
Fuji cost functions, thus avoiding the problem of multicollinearity. The Monthly
Labor Review and International Comparisons of Manufacturing Productivity and
Labor Costs Trends of the U.S. Bureau of Labor Statistics provided cost of labor
series for Kodak. International Financial Statistics published by the I.M.F.
provided a proxy for the price of capital. Bank of Japan’s Price Index Annual
provides monthly price indexes for the same materials in Japan.
To convert price indices into absolute price measures, I need the price of labor
and silver for any one time period. I use Wholesale Price Indexes /Producer Price
Indexes, Employment and Earnings Report of the U.S. Bureau of Labor Statistics
for Kodak. For Fuji, the sources are Yearbook of Minerals and Non-ferrous Metal
Statistics, MITI, and Yearbook of Labor Statistics, Ministry of Labor, Japan.
Monthly exchange rate data were drawn from various issues of Annual Statistical
Digest (1980–88), and Survey of Current Business (1989–90). These exchange
rate series were used to convert Fuji yen cost numbers to dollar terms. To this, I
added tariffs reported in the Harmonized Tariff Schedules to get the cost of
importing Fuji film into the U.S.
The nominal series by price indices were appropriately deflated to determine the
real series. I used seasonally adjusted C.P.I. series for U.S. to deflate income and
price data. For input costs of labor and silver, I used W.P.I. series for U.S. and
Japan. For deflating the price of capital, I used the rate of inflation as measured by
the W.P.I.. The source for the W.P.I. series is the International Financial Statistics
by I.M.F.
Appendix B
Conditions for identification of the model
Following Gasmi, Vuong, and Laffont, the general model is as follows:
O aQ 2a Y
3
q1 2 a11 P1 2 a12 P2 2 g11 A 11 / 2 2 g12 A 21 / 2 2 a0 2 i i 4

i 51
5 u1 Demand for Kodak film, (A.1)
O bQ 2b Y
3
q2 2 a21 P1 2 a22 P2 2 g21 A 11 / 2 2 g22 A 21 / 2 2 b0 2 i i 4

i 51
5 u2 Demand for Fuji film, (A.2)
q1 1 w11 P1 1 w12 P2 2 f1 Z1 2 f5 Z2 5 u 3 Price FOC for Kodak, (A.3)
q2 1 w21 P1 1 w22 P2 2 f6 Z1 2 f2 Z2 5 u 4 Price FOC for Fuji, (A.4)
j 11 P1 1 j 12 P2 1 A 11 / 2 2 f3 Z1 2 f7 Z2 5 u 5 Advertising FOC for Kodak,

(A.5)
j 21 P1 1 j 22 P2 1 A 11 / 2 2 f8 Z1 2 f4 Z2 5 u 6 Advertising FOC for Fuji,

(A.6)
where Z1 is a vector5h pcap1 , plab 1 , ag1 , tj, Z2 is a vector5h pcap2 , plab 2 , ag1 , tj
and hu 1 , u 2 , u 3 , u 4 , u 5 , u 6 j are joint-normally distributed errors.
For the model above, the list of variables are:
1. Endogenous: q1 , q2 , P1 , P2 , A 11 / 2 , A 12 / 2
2. Exogenous: pcap1 , plab 1 , ag1 , pcap2 , plab 2 , t, Y, Q i with i51, 2, 3.
Each demand equation has five included endogenous variables (one quantity,
two price and two advertising variables), and six excluded exogenous variables (all
the cost variables, including the time trend). Each price and advertising first-order
condition has three included endogenous variables (one quantity and two price,
and two price and one advertising variable respectively), and four excluded
exogenous variables (three quarter dummies, and the real income variable). In
addition, market power is identified as the advertising impact is modeled as
non-linear. Hence, the general model is identified. Therefore, games nested in this
general model, are also identified.
Gasmi et al. (1992) have calculated optimization rules and the restrictions on
the general model. Here, I examine two games: the Nash, and the collusion on
both prices and advertising (the best-fitting game). Similar analysis for other
games are available from the author.
In this appendix, n 1 is a vector5hn 11 , n 12 , n 13 , n 14 j, n 2 is a vector5hn 21 , n 22 ,
n 23 , n 24 j
First-order conditions are obtained by solving:
≠Pi / ≠Pi 5 0, ≠Pi / ≠A i 5 0, i 5 1, 2. (A.7)
Nash: In this game, the duopolists pick price and advertising decisions
simultaneously in one stage. The Nash equilibrium price and advertising optimi-
zation rules are:
1/2
a11 (P1 2 c 1 ) 1 g10 1 a11 P1 1 a12 P2 1 g11 A 12 1 g12 A 21 / 2 5 0, (A.8)
a22 (P2 2 c 2 ) 1 g20 1 a21 P1 1 a22 P2 1 g21 A 11 / 2 1 g22 A 21 / 2 5 0, (A.9)
21 / 2
1 / 2(g11 (P1 2 c 1 )A 1 ) 2 1 5 0, (A.10)
21 / 2
1 / 2(g22 (P2 2 c 2 )A 2 ) 2 1 5 0. (A.11)
Comparing these equations (Eqs. (A.8), (A.9), (A.10), (A.11)) to those listed in
the general model above (Eqs. (A.1)–(A.6)), the following restrictions on the
parameters of the general model give the equations above:
w11 5 a11 , w12 5 0, w21 5 0, w22 5 a22 , f1 5 a11 n 1 , f2 5 a22 n 2 , j 11

5 2 1 / 2(g11 ), j 22 5 2 1 / 2(g22 ), f3 5 2 1 / 2(g11 n 1 ), f4
5 2 1 / 2(g22 n 2 ), j 12 5 j 21 5 0, f5 5 f6 5 f7 5 f7 5 f8 5 0. (A.12)
Collusion on prices and advertising: The maximand in this case is a weighted

sum of joint profits, where l is the weight of firm 1 in the joint profit
maximization exercise, and (12 l) the weight of firm 2. Let the maximand be
denoted by F, where F is:
max F ; lp1 (P1 , P2 , A 1 , A 2 ) 1 (1 2 l)p2 (P1 , P2 , A 1 , A 2 ). (A.13)
(P 1 , A 1 ,P 2 , A 2 )
The first-order conditions are:

≠F / ≠P1 5 l[(P1 2 c 1 )a11 1 g10 1 a11 P1 1 a12 P2 1 g11 A 11 / 2 1 g12 A 11 / 2 ]
1 (1 2 l)(P2 2 c 2 )a21 5 0, (A.14)
≠F / ≠P2 5 l(P1 2 c 1 )a12

1/2 1/2
1 (1 2 l)[(P2 2 c 2 )a22 P2 1 a21 P1 1 g22 A 2 1 g21 A 1 ]
5 0, (A.15)
≠F / ≠A 1 5 l[(P1 2 c 1 )0.5*g11 A 121 / 2 2 1]

21 / 2
1 (1 2 l)[(P2 2 c 2 )0.5*g21 A 1 ] 5 0, (A.16)
≠F / ≠A 2 5 l[(P1 2 c 1 )0.5*g12 A 21
2
/2
]
21 / 2
1 (1 2 l)[(P2 2 c 2 )0.5*g22 A 2 2 1] 5 0. (A.17)
These can be rewritten as follows:
q1 1 a11 P1 1 ((1 2 l) /l)a21 P2 2 a11 c 1 2 ((1 2 l) /l)a21 c 2 5 0, (A.18)
q2 1 ( l /(1 2 l))a12 P1 1 a22 P2 ( l /(1 2 l))a12 c 1 2 a22 c 2 5 0, (A.19)
(g11 / 2)P1 1 ((1 2 l) / 2l)g21 P2 2 A 11 / 2 2 (g11 / 2)c 1 2 ((1 2 l) / 2l)g21 c 2 5 0,

(A.20)
( l / 2(1 2 l))g12 P1 1 (g22 / 2)P2 2 A 12 / 2 2 ( l / 2(1 2 l))g12 c 1 2 (g22 / 2)c 2 5 0.

(A.21)
Comparing Eqs. (A.18), (A.19), (A.20), (A.21) with those listed in the general
model above (A.1 through A.6), we see that the restrictions (A.22) below on the
parameters of the general model give the equations above:
w11 5 a11 , w11 5 (1 2 l)a21 /l, w22 5 a22 , w21 5 la12 /(1 2 l),
j 11 5 2 g11 / 2, j 12 5 2 (1 2 l)g21 / 2l, j 21 5 2 lg12 / 2(1 2 l), j 22

5 2 g22 / 2l,
f1 5 w11 n 1 , f2 5 w22 n 2 , f3 5 j 11 n 1 , f4 5 j 22 n 2 ,
f5 5 0, f6 5 0, f7 5 j 12 n 1 , f8 5 j 21 n 2 . (A.22)
Comparing the restrictions that are needed to generate parameters of the Nash
games (A.12) and collusive game (A.22) from the parameters of the general game
shows that the restrictions are not equal. Therefore, the two games are not
perfectly nested in each other, making it possible to distinguish econometrically
between these two games in the estimation. The same is also true of all the other
games. Further, as the general model is identified, the two games here (as well as
the other games) are identified since they are nested in the general model.
References
Bagwell, K., Staiger, R., 1995. Collusion over the business cycle. Mimeo, Northwestern University.
Baldwin, R., 1988. Hysteresis in import prices: the beachhead effect. American Economic Review
78,(4) 773–785.
Bhagwati, J., 1988. The pass-through puzzle: the missing prince from Hamlet. Mimeo, Columbia
University.
Bresnahan, T., 1989. Industries and Market Power. In: Schmalensee, R., Willig, R. (Eds.), Handbook of
Industrial Organization. North Holland, Amsterdam.
Davidson, C., 1984. Cartel stability and tariff policy. Journal of International Economics 17, 219–237.
Dixit, A., 1989. Hysteresis, import penetration, and exchange rate pass-through. Quarterly Journal of
Economics 104,(2) 205–228.
Dornbusch, R., 1987. Exchange rates and prices. American Economic Review 77,(1) 93–106.
Feenstra, R., 1989. Symmetric pass-through of tariffs and exchange rates under imperfect competition:
an empirical test. Journal of International Economics 27, 25–45.
Feenstra, R., Gagnon, J., Knetter, M., 1996. Market share and exchange rate pass-through in world
automobile trade. Journal of International Economics 40, 187–207.
Feinberg, R., 1992. Hysteresis and export targeting. International Journal of Industrial Organization 10,
679–684.
Froot, K., Klemperer, P., 1989. Exchange rate pass-through when market share matters. American
Economic Review 79,(4) 637–654.
Gasmi, F., Laffont, J., Vuong, Q., 1992. Econometric Analysis of Collusive Behavior in a Soft-Drink
Industry. Journal of Economics and Management Strategy 1,(2) 277–311.
Giovannini, A., 1988. Exchange rates and traded goods. Journal of International Economics. 24,(1)
45–68.
Goldberg, P., 1995. Product differentiation and oligopoly in international markets: the case of the U.S.
automobile industry. Econometrica, July.
Harrington, J., 1989. Collusion and predation under (almost) free entry. International Journal of
Industrial Organization, 381–401.
Kadiyali, V., 1996. Entry, Its Deterrence, and Its Accommodation: The Case of the U.S. Photographic
Film Industry. Mimeo, forthcoming in the Rand Journal of Economics 27,(3) 452–478.
Kasa, K., 1992. Adjustment costs and pricing-to-market: Theory and evidence. Journal of International
Economics 32, 1–30.
Knetter, M., 1989. Price discrimination by U.S. and German exporters. American Economic Review
79, 198–209.
Krugman, P., 1987. Pricing to market when the exchange rate changes. In: Arndt, S., Richardson, J.D.
(Eds.), Real and financial linkages among open economies. M.I.T. Press
Lawrence, R., 1990. U.S. current account adjustment: an appraisal. Brookings Papers on Economic
Activity 2, 343–383.
Lillien, G., Kotler, P., Moorthy, S., 1992. Marketing Models. Prentice Hall, New Jersey.
Mann, C., 1986. Prices, profits margins and exchange rates. Federal Reserve Bulletin, 366–379.
Marston, R., 1990. Pricing to market in Japanese manufacturing. Journal of International Economics
29, 217–236.
Menon, J., 1995. Exchange rate pass-through. Journal of Economic Surveys 9, 197–231.
Ohno, K., 1990. Exchange rate fluctuations, pass-through, and market share. IMF Staff Papers 37,(2)
294–309.
Porter, R., 1984. A Study of Cartel Stability: The JEC 1880–1886. Bell Journal of Economics 14,(2)
301–314.
Roberts, M., Samuelson, L., 1988. An Empirical Analysis of Dynamic, Nonprice Competition in an
Oligopolistic Industry. Rand Journal of Economics 19,(2) 200–220.
Rotemberg, J., Saloner, G., 1986. A supergame-theoretic model of price wars during booms. American
Economic Review 76,(3) 390–407.
Schembri, L., 1989. Export prices and exchange rates: An industry approach. In: Feenstra (Ed.), Trade
policies for international competitiveness. University of Chicago Press.

1 s2.0 S0022199697000081 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0022199697000081 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0022199697000081 Main

Uploaded by

Copyright:

Available Formats

Journal of International Economics 43 (1997) 437–461

Exchange rate pass-through for strategic pricing and

Keywords: Exchange rate; Pass-through; Pricing; Advertising; Industry structure

JEL classification: F12; F14; L13; L60

*Tel.: (607)-255-1985; fax: 607-254-4590; e-mail: [email protected]

0022-1996 / 97 / $17.00  1997 Elsevier Science B.V. All rights reserved

In addition, there was incomplete pass-through in Fuji’s advertising expenditures.

Krugman (1987) classifies pass-through models as either static or dynamic. In

In Kadiyali (1996), I empirically estimate a structural model of industry

q1 5 a11 P1 1 a12 P2 1 g11 A 11 / 2 1 g12 A 21 / 2 1 a0 1 i i 4 (1)

q2 5 a21 P1 1 a22 P2 1 g21 A 11 / 2 1 g22 A 21 / 2 0 i i 4 (2)

where q represents quantity, P price, A advertising, Q the quarterly dummies for

Marginal cost is specified to be a linear function of the material and factor

* pcap2 1 n *22 plab 2 1 n 23

5 u1 Demand for Kodak film, (7)

q2 2 a21 P1 2 a22 P2 2 g21 A 11 / 2 2 g22 A 21 / 2 2 b0 2 i i 4

5 u2 Demand for Fuji film, (8)

q1 1 w11 P1 1 w12 P2 2 f1 Z1 2 f5 Z2 5 u 3 Price FOC for Kodak, (9)

q2 1 w21 P1 1 w22 P2 2 f6 Z1 2 f2 Z2 5 u 4 Price FOC for Fuji, (10)

j 11 P1 1 j 12 P2 1 A 11 / 2 2 f3 Z1 2 f7 Z2 5 u 5 Advertising FOC for Kodak,

j 21 P1 1 j 22 P2 1 A 11 / 2 2 f8 Z1 2 f4 Z2 5 u 6 Advertising FOC for Fuji,

4. Results and discussion

4.1. Results of the structural estimation analysis

As Table 3 shows, high multicollinearity exists between variables. This prevents

4.2. The relationship between pass-through and market structure

Exchange rate movements affect optimum prices and advertising in the

p1 5 (P1 2 MC1 )q1 2 A 1 , (14)

Fig. 1. Plot of Fuji’s prices.

through is closest to the one employed here. Goldberg endogenously estimates

This is the first industry study to systematically analyze the phenomenon of

This paper is based on essay 3 of my dissertation. I am indebted to my advisor,

Conditions for identification of the model

Following Gasmi, Vuong, and Laffont, the general model is as follows:

q1 2 a11 P1 2 a12 P2 2 g11 A 11 / 2 2 g12 A 21 / 2 2 a0 2 i i 4

5 u1 Demand for Kodak film, (A.1)

q2 2 a21 P1 2 a22 P2 2 g21 A 11 / 2 2 g22 A 21 / 2 2 b0 2 i i 4

5 u2 Demand for Fuji film, (A.2)

q1 1 w11 P1 1 w12 P2 2 f1 Z1 2 f5 Z2 5 u 3 Price FOC for Kodak, (A.3)

q2 1 w21 P1 1 w22 P2 2 f6 Z1 2 f2 Z2 5 u 4 Price FOC for Fuji, (A.4)

j 11 P1 1 j 12 P2 1 A 11 / 2 2 f3 Z1 2 f7 Z2 5 u 5 Advertising FOC for Kodak,

j 21 P1 1 j 22 P2 1 A 11 / 2 2 f8 Z1 2 f4 Z2 5 u 6 Advertising FOC for Fuji,

≠Pi / ≠Pi 5 0, ≠Pi / ≠A i 5 0, i 5 1, 2. (A.7)

a22 (P2 2 c 2 ) 1 g20 1 a21 P1 1 a22 P2 1 g21 A 11 / 2 1 g22 A 21 / 2 5 0, (A.9)

w11 5 a11 , w12 5 0, w21 5 0, w22 5 a22 , f1 5 a11 n 1 , f2 5 a22 n 2 , j 11

Collusion on prices and advertising: The maximand in this case is a weighted

The first-order conditions are:

≠F / ≠P2 5 l(P1 2 c 1 )a12

≠F / ≠A 1 5 l[(P1 2 c 1 )0.5*g11 A 121 / 2 2 1]

q2 1 ( l /(1 2 l))a12 P1 1 a22 P2 ( l /(1 2 l))a12 c 1 2 a22 c 2 5 0, (A.19)

(g11 / 2)P1 1 ((1 2 l) / 2l)g21 P2 2 A 11 / 2 2 (g11 / 2)c 1 2 ((1 2 l) / 2l)g21 c 2 5 0,

( l / 2(1 2 l))g12 P1 1 (g22 / 2)P2 2 A 12 / 2 2 ( l / 2(1 2 l))g12 c 1 2 (g22 / 2)c 2 5 0.

j 11 5 2 g11 / 2, j 12 5 2 (1 2 l)g21 / 2l, j 21 5 2 lg12 / 2(1 2 l), j 22

You might also like