THE MIT JAPAN PROGRAM Science, Technology,

advertisement
THE
MIT JAPAN
PROGRAM
{4
t
1 a 7..
Science, Technology,
Management
0
-0A
j
%%.P
y;;?
b
DO MULTINATIONALS SHIFT PRODUCTION IN
RESPONSE TO EXCHANGE RATE CHANGES?
DO THEIR RESPONSES VARY BY NATIONALITY?
Subramanian Rangan
MITJP 96-03
Center for International Studies
Massachusetts Institute of Technology
Distributed courtesy of the
MIT JAPAN PROGRAM
Science · Technology * Management
E38-754
CenterforInternationalStudies
77 MassachusettsAvenue, Cambridge, M4 02139
Tel: 617-253-2839
Fax: 617-258-7432
Email: <robart(mit.edu>
©MIT Japan Program
DO MULTINATIONALS SHIFT PRODUCTION IN
RESPONSE TO EXCHANGE RATE CHANGES?
DO THEIR RESPONSES VARY BY NATIONALITY?
Subramanian Rangan
MITJP 96-03
-------I·
____--
About the MIT Japan Program
and its Working Paper Series
The MIT Japan Program was founded in 1981 to create a new generation
of technologically sophisticated "Japan-aware" scientists, engineers, and
managers in the United States. The Program's corporate sponsors, as well
as support from the government and from private foundations, have made
it the largest, most comprehensive, and most widely emulated center of
applied Japanese studies in the world.
The intellectual focus of the Program is to integrate the research
methodologies of the social sciences, the humanities, and technology to
approach issues confronting the United States and Japan in their relations
involving science and technology. The Program is uniquely positioned to
make use of MIT's extensive network of Japan-related resources, which
include faculty, researchers, and library collections, as well as a Tokyobased office. Through its three core activities, namely, education,
research, and public awareness, the Program disseminates both to its
sponsors and to the interested public its expertise on Japanese science
and technology and on how that science and technology is managed.
The MIT Japan Program Working Paper Series provides an important
means to achieving these ends.
_I
I
Do Multinationals Shift Production in Response to Exchange Rate
Changes? Do Their Responses Vary By Nationality?
Evidence From 1977-1993
No matter what the risk profile, the firm that is able to exploit...volatility possesses a
competitive advantage gained by its ownership of a global network...In the case
of...multinational(s this advantage]...might potentially consist of production shifting.
Bruce Kogut (1985: 37)
Exchange rate changes don't figure high on the list of reasons that multinational
enterprises locate operations abroad. But as the quote at the top suggests, once they
locate operations in two or more currency areas, multinational enterprises (MNEs)
may in many respects be well-positioned to exploit changes in exchange rates. In
addition to hedging in currency markets, flexing profit margins, and improving
productivity, MNEs may respond to exchange rate changes by also shifting
production within their networks to areas made more competitive by the exchange
rate change.
Although this simple concept has long held appeal to economists and
management scholars, skeptics have wondered whether multinational enterprises
really shift production--say between home and abroad--in response to currency
swings. After all, these skeptics note, even at the margin, economic, institutional, and
organizational factors (such as plant scale economies and insufficient coordination)
may make such switching suboptimal or unfeasible. Besides, considering the many
well established differences in the average operating practices of MNEs
headquartered in different countries, many observers including policy makers wonder
whether U.S., European, and especially Japanese MNEs respond equally flexibly to
exchange rate changes.
I have been exploring these questions with a data set constructed from the U.S.
Bureau of Economic Analysis' (BEA) annual surveys on the operations between 1977
1
__
I_
I_~~~~~~~~~~~~~~~~~---
and 1993 of United States multinationals abroad and foreign multinationals in the
United States. To anticipate, let me summarize the key findings that emerge from this
analysis: (i) Across the board, MINEs from the United States, Europe, and Japan
exhibit systematic and statistically significant responsiveness in the anticipated
manner to exchange rate changes. (ii) In terms of magnitude, Japanese MNEs exhibit
among the most vigorous responses, although, their estimated exchange rate elasticity
of 1.3 appears unlikely to be statistically different from those of MNEs headquartered
in Europe and the United States. l (iii) Finally, as one might expect, responsiveness
varies sharply across the sub industries within manufacturing. Accordingly, the paper
concludes that when it comes to exchange rate-induced production shifting within
MNE networks, industry matters but nationality doesn't.
I will present and discuss these findings after elaborating on the research
questions and hypotheses, the data and methodology, and the model that I estimate
empirically.
Production Shifting Within MNEs
Multinational enterprises, like most firms, seek to maximize their profits,
market share, and longevity. Exchange rate changes can influence all three of these
goals and it therefore stands to reason that MNEs should respond to them. Indeed, in
perfectly competitive markets, the question may be moot. But as Stephen Hymer
(1976 [1960]) argued in his pioneering work, MNEs operate in imperfectly
competitive markets where by virtue of certain firm-specific advantage these
enterprises enjoy rents. Add to the presence of rents uncertainty over the future
course of exchange rate changes, and institutional inertia becomes a feasible option.
Therefore, it is legitimate from a research standpoint to ask: Do MNEs shift
production in response to exchange rate changes?
I F-test for equality of coefficients is to be conducted.
2
_
·
CI_
Nearly three decades ago, Raymond Vernon (1966: 198) had noted that
multinational enterprises with multiplant locations might source from low cost
facilities when it became apparent that such facilities were cheaper net of transport
costs and tariffs. More recently, Jane Little (1987: 46) has written that:
Firms with production and marketing facilities on both sides of an exchange rate
possess an extra degree of flexibility in adjusting to a new competitive situation.
These multinationals can turn to existing plants in countries where the currency is
depreciating and, with comparative ease, expand output where relative production
costs are falling.
Although this proposition has long held appeal to economists and management
scholars (see Adler and Stevens, 1974; Ghoshal, 1989; Knetter, 1993; Kogut, 1985;
Lessard, 1986; Lipsey and Kravis, 1986; Vernon, 1966), there has been considerable
skepticism surrounding its feasibility. For instance, Bruce Kogut (1985: 32) has
asked rhetorically, "Do managers perceive and identify potential options generated by
being multinational?...Are there organizational mechanisms that permit the
coordination of the international activities essential to the exploitation of flexibility?"
David Goldsbrough (1981: 573-580) has argued that because "integrated plants"
within a multinational firns' network might produce specialized outputs which "have
fewer close substitutes,"
trade flows generated by the location decisions of a firm with large fixed investments
in several countries may not respond as rapidly to shifts in relative prices as those of
an independent producer...
What is the reality? Do multinational enterprises respond flexibly to exchanges rate
changes? Do they shift production within their internal networks--say between home
and abroad--in response to currency swings? Controlling for other economic factors,
do MNEs headquartered in different countries respond differently to exchange rate
changes? Does their at-the-margin behavior reflect the well established differences in
average tendencies?
It is fitting to explore these issues in the context of a conference entitled "Does
Ownership Matter?," because that question gets raised here in two important ways. In
3
_
I _II~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~---
____
the first instance, multinational enterprises (MNEs) are networks of affiliated
companies linked by ties of common "ownership." Thus trade between affiliates in an
MNE network has been dubbed "hierarchical" trade, and many scholars have been
interested in contrasting this trade to that conducted in arm's length "markets"
between unaffiliated firms. Accordingly, this set of scholars have pursued the
question from an institutional angle by asking: Are hierarchies as responsive as arm's
length markets in adjusting to exchange rate changes?.
The presumption has been that trade in arm's length markets is more
responsive because hierarchical or intrafirm trade "usually...[takes] place in
consequence of central commands rather than in response to price signals..."
(Helleiner, 1981: 3). If this indeed is the case, then "ownership" matters in the sense
that governance through internalization impedes flexibility and adjustment to relative
price changes. The implications for both firms and nations are obvious and important.
But in the context of examining MNE responsiveness to exchange rate
changes, the question "Does Ownership Matter?" may also be asked with an emphasis
on firms' nationality. This formulation of the question reads: Do national factors
influence the extent of production shifting that MNEs undertake in response to
exchange rate changes? In particular, do MNEs headquartered in Europe and
especially Japan respond as vigorously to exchange rate changes as MNEs
headquartered in the U.S. (even when what is called for is a substitution away from
home content into foreign content)?
The focus on Japan is important and interesting for at least two reasons. For
one, as Paul Krugman (1991: 1) points out, "There is...a widespread sense that as
Japan has moved from the periphery to the center of the world economy, it has
continued to play the game by somewhat different rules than other advanced nations."
This impression is based partly on casual empiricism, partly on some careful studies
that document significant differences in the operating practices of Japanese
enterprises (see Kreinin, 1988; and Lawrence, 1991), and partly on some well known
facts.
4
--
---
------ ·------ rr(···-r_^_..
-·
.
--
-sr
For instance, it is well known that for a variety of reasons including the
recency of their expansion abroad, the foreign affiliates of Japanese multinationals
rely, on average, more heavily on home operations than do their European and U.S.
counterparts. But considering the rise in global competition and the heightening of
trade tensions, managers and policy makers alike want to know whether such reliance
is, even at the margin, relatively more sticky and inflexible. Peter Petri (1991: 52) has
pointed out that some observers believe that the answer will turn out to be affirmative.
He writes:
[There is a thesis that] is challenging the view that Japan has become more open with
endaka. It emphasizes the relatively slow adjustment of the Japanese...bilateral trade
surplus with the United States...notwithstanding sharp improvements in U.S. price
competitiveness. The proponents of this view have argued that exchange rate
adjustments. no matter how large, cannot satisfactorily open Japan.
A second reason why it is interesting to explore whether Japanese, European,
and U.S. MNEs respond differently to exchange rate changes is because it provides
another way by which we can infer something about the relative inf ,-
of market
versus institutional forces.
The logic goes as follows: Exchange rate changes are a market-driven
exogenous force that buffets all MNEs regardless of their nationality. What is
different between Japanese and Western, especially U.S. enterprises, is the set of
institutional arrangements under which they operate. Therefore, if after taking
product mix differences into account it is found that Japanese, European, and U.S.
firms respond differently to exchange rate changes, then the presumption is
strengthened that institutional forces matter even at the margin. If significant
differences are not found, we may conclude that, at least at rt
forces (exchange rate changes in this case) supersede instittu..
, certain market
;s;.
Such a
finding will weaken wh1 ! importance the notion of path dependence has in
-
---.- l----L
international business and also imply that sufficiently large shifts at the margin may
bring about convergence even at the mean. 2
In the remainder of this paper, I want to focus on the question of whether and
to what extent MNEs shift production in response to exchange rate changes and
whether there are systematic and significant national differences in the vigor with
which they respond. I have addressed more directly in another paper the question of
whether "markets" respond more flexibly than "hierarchies" to common exchange rate
changes. 3
In thinking conceptually about the issue of production shifting within MNEs,
it will be useful to work with the following simple but not atypical scenario. Suppose
there is a U.S.-headquartered MNE engaged in the manufacture and sale of products
in the United States and Europe. For a variety of well-known reasons including
transport costs and immovability of certain value added activities (such as distribution
and service), the MNE co-locates the bulk of its operations near its markets. In other
words, what the MNE sells in Europe it produces, for the most part, in Europe. But
there remain some intermediate inputs that the European affiliates of this MNE source
from the United States--almost exclusively from the parent unit. The net result is that
the products that this MNE (or strictly speaking, its foreign affiliate) sells in Europe
contain a mix of local (read European) and home (read U.S.) content, and the question
at hand is: How do changes in real exchange rates influence the composition of this
mix?
Factors in the Decision
2 As the text indicates, in this research I treat exchange rate changes as exogenous. I assume, quite
reasonably I think, that the sourcing responses and strategies of MNEs don't cause exchange rate
changes, but, rather, that the chain of causality runs the other way. Also, consistent with a "random
walk" characterization of exchange rate movements, all changes in rates are considered permanent.
3 Rangan (1994).
6
__
----
_
~
From a managerial-microeconormic perspective, there are several obvious
factors over and above changes in real exchange rates that are likely to determine
whether and how much the mix between local and home content will shift. At the top
of the list may be the availability and lumpiness of suitable capacity in the location
favored by the exchange rate change. A second factor may be the importance of scale
economies. In the presence of sizable plant scale economies in the area where the
currency is appreciating, exchange rate changes will have to be sufficiently large
before switching becomes optimal.
Further, switching costs, such as severance payments and redesign charges,
are undoubtedly part of the equation. The magnitude of switching costs is likely to
differ from industry to industry (say being larger in autos than in electronics), and
even from host country to host country (depending on local regulations regarding
layoffs, local content, and export performance requirements). Like plant scale
economies, the presence of switching costs will tend to make production shifting
optimal only in response to relatively large shifts in exchange rates.
A factor that may enable MNEs headquartered in Europe and Japan to be more
responsive to exchange rat,: changes than MNEs headquartered in the United States is
the relative ease of "exit" in the United States compared to in Europe and Japan.
Consider what might happen when the dollar appreciates. When the dollar
appreciates, European and Japanese MNEs might shrink their U.S. operations and
shift production to existing facilities at home. Because they dominate home markets,
existing facilities back home are likely to be better placed to accommodate the
increased demand, but in some cases, marginal expansion may be called for. But the
response of U.S.-based MNEs to a dollar appreciation may appear more sticky
because expansion in Europe or Japan to serve the home (i.e. U.S. market) is fraught
with the risk of strained and costly exit (should the need arise). That is, getting in
may be easy but getting out may not be. Mindful of a dampened ability to respond to
future depreciations in the U.S. dollar, U.S. MNEs may not exhibit vigorous
7
·I·IL
ol
III
production shifting responses to appreciations in the dollar. Alternatively, they may
exhibit a relatively sluggish or lagged response.
Another important factor that may matter in a few cases is whether the firms'
competitive advantages are bound to a unique location (such as Silicon Valley, for
instance). For firms and industries where this is the case, even large appreciations in
the region's currency may not elicit a vigorous switching response. In the strategy
vocabulary popularized by Michael Porter, such firms compete on "differentiation"
not "cost," and introducing newer, more sophisticated products might be their best
response to low cost competitors based in regions where currencies have depreciated.
Of course, such a strategy typically entails staying put in the location where the
innovation occurs. Consequently, little or no production shifting may be observed.
Of course, even in such circumstances if the real exchange rate continues to
appreciate over a prolonged period, then, eventually, the changed competitive position
ought to get reflected in the sourcing patterns of these "location-bound" MNEs.
Presumably, such lagged adjustment also occurs in cases where an MNE has plants in
two or more currency areas but is unable to shift production readily among them
because the plants are specialized.
The issue of lags in international trade adjustment was developed by Helen
Junz and Rudolf Rhomberg (J&R) in a seminal paper written over two decades ago.
J&R (1973: 413) suggest a temporal taxonomy of lags consisting of: recognition lags
(the time taken to "become aware of the changed competitive situation"), decision
lags ("the time taken for new business connections to be formed and new orders to be
placed"), delivery lags (self explanatory), replacement lags (the time taken to wear out
or deplete existing stocks before new orders can be placed), and production lags (the
time taken by producers to decide to switch old or add new capacity to service foreign
markets).
Going down this taxonomy, it would appear that in terms of the speed with
which they can respond to common exchange rate changes, MNEs ought to be betterplaced than arm's length traders who operate in international markets. In particular,
8
----
--------- ·
_._1___--_11-_1
--_----
·111-·--
1-
--I_ ---
Junz and Rhomberg's "decision" lags (the time taken for "new business connections to
be formed and new orders to be placed") and "production" lags (the time it might take
"producers...to become convinced that a profit opportunity which they perceive in
certain markets is sufficiently large and permanent to warrant the expense and effort
of shifting from supplying one market to another or adding capacity in order to supply
the other market...") ought to be shorter for multinational enterprises than their solely
national counterparts.
Considering these information and scanning advantages that MNEs enjoy over
arm's length traders, we might anticipate that to the extent MNEs shift production in
response to exchange rate changes, they do so with shorter lags than those reported in
traditional empirical studies of trade adjustment. Indeed, in related previous work
(Rangan, 1994), I found that this hypothesis does receive support in the data. 4
Conceptual Models
Based on the preceding discussion, we can formulate three somewhat distinct
models of multinational sourcing adjustment to exchange rate changes. I have
portrayed these models in the three panels in Figure 1 which is drawn from the
viewpoint of a foreign affiliate of an MNE parent. For illustrative purposes, let us
suppose these pictures are drawn from the perspective of a German affiliate of an
MNE headquartered in the United States.
In each panel, the X-axis depicts changes in the bilateral dollar-mark real
exchange rate. Real depreciations in the dollar are indicated by shifts to the right of
zero, and real appreciations by shifts to the left of zero. That is, to the right of zero,
U.S. capacity is becoming relatively more competitive and to the left it is becoming
less competitive. The Y-axis depicts the U.S. (or home) content level in products sold
4 Please refer to that paper for a fuller discussion of the issue of lags.
9
-L
in Germany by the German affiliate of the U.S. MNE. With this context in mind, let
us consider each model in turn.
THE STRONG SUBSTITUTION MODEL. The "strong substitution" model places
heavy emphasis on factor costs and predicts that multinationals will adjust in a
smooth and symmetric manner to all changes in real exchange rates--regardless of
whether they are small or large, appreciations or depreciations. Thus, when the dollar
depreciates in real terms, the U.S.-content level in products sold abroad by U.S.-based
MNEs will rise monotonically. Indeed, if exchange rate changes are sufficiently
large, the MNEs may cease foreign production altogether choosing instead to service
foreign markets from their facilities at home (i.e. U.S.-content level reaches 100%).
The converse will hold when the dollar appreciates. Such a pattern of adjustment will
produce a smooth and unbounded adjustment curve as shown in the top panel in
Figure 1.
Of course, in reality the strong substitution model of smooth and unbounded
adjustment is unlikely to be a good representation of the manner in which MNEs shift
production because the model ignores many of the factors just discussed in the
preceding section.
THE THRESHOLD MODEL. The "threshold" model goes some distance toward
closing the gap between model and reality. It posits that because firms have to also
factor in plant scale economies and switching costs it will not be optimal for them to
shift production in response to small exchange rate changes. But, if the magnitude of
the exchange rate change is so large that the rate crosses a certain threshold (which is
likely to vary by firm, industry, and country), then, even after factoring in switching
costs and lost plant scale economies, it will be optimal for firms to shift production
and they will do so.
Of course, the size of the threshold may not be the same on both sides of the
initial rate. Even relatively small appreciations in the foreign currency may trigger a
switching response in the direction of home because in comparison with home,
individual foreign markets are likely to be small and hence unlikely to suffer major
10
__I_
V___·
losses in production scale economies. For opposite reasons, it would take a relatively
large appreciation in the MNE parent's home currency before switching becomes
optimal. Therefore, the switching threshold may be shorter (i.e. closer to zero) on the
home currency depreciation (or right hand) side.
Nevertheless, this model posits that the vigor of the response, once triggered,
would be the same on either side. Accordingly, it predicts a discontinuous but
unbounded and symmetric (i.e. same slope) adjustment curve as shown in the middle
panel in Figure 1.
But like the strong substitution model, the threshold model also ignores the
fact that when it comes to important and potentially costly decisions such as the one
to shift production, MNEs plot their strategies over long time horizons. In the words
of Raymond Vernon (1971: 119):
[T]he management is concerned with maintaining the loyalty, incentive, and
initiative over the long run and, if necessary, is usually prepared to modify the classic
return-on-investment calculations to keep the principal members...[of the
organization] in play.
Thus, it is unlikely that solely in response to changes in exchange rates, an MNE
would completely abandon its operations in countries that have become relatively less
cost competitive. This notion of bounded responses motivates the final model.
THE FLOOR AND CEILING MODEL. The "floor and ceiling" model maintains the
notion of thresholds but it also postulates that there is a certain level-call it the
"floor"--below which home-country content cannot be reduced in the medium run (i.e.
over the average exchange rate cycle). For instance, a critical or highly scaleintensive input may have to be fabricated in a single facility which, for historical and
market size-related reasons, is located at home. Or it may be that because of relative
newness, the input has to be fabricated near the site of innovation and ongoing
research--home. Under such circumstances, changes in exchange rates may not
trigger a shift in the locus of production of these inputs.
11
Likewise, there is also a "ceiling" above which the home-country content level
cannot rise in the medium run, say because certain value-added activities (such as
packaging, distribution, sales, and service) have to remain local. Among other
factors, value-to-weight ratio of inputs, tariffs, the degree to which value is added in
the provision of services, and local content regulations may all go into determining
the height of the ceiling.
This model suggest a discontinuous adjustment curve with kinks on either side
where the slope of the adjustment curve goes from being positive to zero. The kinks
imply two things. First, that the responses are bounded. That is, in the time horizon
contemplated here, the home-local mix cannot go to 1:0 or 0: 1 proportions. Second,
that given certain initial home content levels, responses may be asymmetric between
appreciations and depreciations. For instance, if the initial home content level is near
or at the floor, then even a relatively large appreciation in the home currency is
unlikely to elicit switching responses. But threshold breaching depreciations will
trigger a rise in home content. Likewise, if the initial home content level is near or at
the ceiling, then even relatively large depreciations in the home currency won't elicit
much of a response, but threshold-breaching appreciations will.
Among the three models sketched above, it would appear that the "floor and
ceiling" is most plausible, especially if one thinks of the floors and ceilings as being
endogenous over longer time horizons. To be sure, it will be difficult from an
empirical standpoint to distinguish between the "threshold" model and the "floor and
ceiling" model because: (a) the shortest time window over which changes can be
examined in the BEA data is one year--a period whose length may be sufficient for
firms to shift floors and ceilings; and (b) because in the sample interval (1977-1993)
firms may be operating within the adjustment band--the positive sloped area that is
away from either floor or ceiling--and may not brush up against either the upper or
lower bounds.
THE ROLE OF COMPETITION, LIBERALIZATION, AND TECHNOLOGY. Beyond
the factors considered in these simple models, three other factors are likely to bear
12
___111_1_
_____11_
1_
_____l)____m__ll_^__1
upon the phenomenon. First, the intensity of international competition has risen
considerably over the period covered in this study-- 1977 to 1993--and a priorione
would think that this is likely to have caused elasticities of substitution to rise over
this time. Second, over the last 15 years many countries have liberalized their trade
,egulations and taken other steps that make the transshipment of intermediate and
final goods relatively more attractive and feasible. This is also likely to have caused
elasticities to rise over this time period. Finally, considering the steep fa1l over the
last 15 years in transportation and particularly telecommunications costs, it ought to
have become easier and less expensive for multinationals to coordinate their
production networks. This too is likely to have caused elasticities to rise over this
time period.
The Empirical SpecificationAccordingly, the specifications I estimate are variations of the following:
ACij, = a + jAejj, + 7iiT, + eqj,,
where ACJ, denotes changes in home content level in products sold by MNE affiliates
operating in industry i, country j, in period t; a is the intercept; AeUj, denotes changes
in the industry-specific bilateral real exchange rate (explained below); T, is a time
trend variable whose coefficient is meant to capture the role of rising competition,
liberalization, and falling telecommunication costs; and 6Ei is an error term. Of
course, wiy is the exchange rate elasticity of the home content level and it can be
estimated with lags.
Considering the breadth of the earlier discussion, this specification is clearly
"parsimonious." I want to say, therefore, a word about omitted variable bias and
choice of functional form. First, firms don't report and the BEA doesn't gather data
on switching costs or plant scale, not to mention "floors" and "ceilings." Further, I
13
.-·1PI-"-I---"L--4U"W--i··1L·LD
am not aware of another source that provides these data specifically for multinational
enterprises. Ergo, the reduced form specification.
Having acknowledged this, let me point out that the left out variables-including unavailability of capacity, switching costs, plant scale economies, locationboundedness, and being near a floor or ceiling--only impede or dampen
responsiveness. This implies that the although the estimate of pfij may be biased, it is
likely to be biased only downward. Consequently, if Pij turns out to be a positive and
statistically significant, then we have a strong indication that the MNEs in the sample
do actually shift production in response to exchange rate changes. In this sense, the
parsimonious specification provides a strong test 6of the hypothesis that MNEs shift
production in response to exchange rate changes.
On the issue of functional form I will note three things. First, in the absence
of formal modeling on which to base a choice of alternatives, it is sensible to stick
with a linear specification. Second, in existing trade literature the linear model is
conventional. 5 Finally, the key dependent and independent variables are measured in
percentage changes, and this too supports the choice of a linear specification.
Variables, Methodology, and Data
As I noted earlier, I examine the production shifting responses for two sets of
MNE affiliates--the majority-owned foreign affiliates of MNEs headquartered in the
United States, and the U.S. (minority and majority) affiliates of MNEs headquartered
in Canada, Europe and Japan. Consequently, there are two sets of dependent
variables in my analysis.
In my analysis of the production shifting responses of U.S. MNEs, the
dependent variable is the percentage change in U.S. content levels in sales made
abroad by their majority-owned foreign affiliates. The dependent variable in my
analysis of European and Japanese MNEs is the percentage change in the foreign5 See Stern and colleagues, 197?; Hooper and Mann, 199?; Lawrence, 199? (get full cites>.
14
-1
1'' '
_1__1_1
__j^lXI__·___ll__l_-11--------1-.
(primarily home-) content levels in sales made in the United States by their U.S.
affiliates.
Thus, in both cases I examine changes from the viewpoint of affiliates as
opposed to parents (mainly because the estimation process is less prone to
measurement error). 6 The changes in content levels are volume not value measures.
To clarify, let me explain briefly how I estimate "volume" changes in content. Take
the U.S.-content level in sales made abroad by U.S. majority-owned foreign affiliates
(MOFAs). For the base year of the study I estimate U.S. content by dividing the U.S.
exports made to MOFAs in a particular country in a particular industry in the base
year by the sales made by MOFAs in that country in that industry in the same year.
For subsequent years, I estimate U.S. content levels in the same manner except
that I first deflate export values by the industry-specific U.S. export price, and
likewise deflate foreign sales values by the industry-specific producer price in the
host country. Then, in order to remove currency translation effects, I convert back
into national currencies all MOFA sales figures (which are reported in current U.S.
dollars), and rescale these national currency figures back into U.S. dollars at the
nominal exchange rate that was in effect during the base year. This procedure assures
that pure currency, pure price, and equivalent but opposing currency and price
changes will not influence the U.S. content measure.7
The key independent variable in my analysis is, of course, the percentage
change in industry-specific real exchange rates, which I estimate based on changes in
bilateral nominal exchange rates and changes in industry-specific producer prices in
the United States and the partner countries in the study.
6 In order to estimate production shifting responses from an MNE parents' point of view, we would
need quite detailed information on each of the many countries from which MNE parents source inputs.
Such information is not available annually on an industry by country basis and extrapolating based on
available information is not only a challenging task but also one that is likely to aggravate
measurement error.
7 Upon request I will be happy to furnish details and some examples of how I estimate the content
levels.
15
-
-·l(-·lllsar*ra·lqgII`ICI"L*"L----·-
--
The industries covered in this study include manufacturing as such, and within
manufacturing, food, chemicals (including pharmaceuticals), metals, non-electrical
machinery, electrical machinery, and motor vehicles and parts. Country coverage is
guided by relative importance of countries as homes or hosts of MNEs, and by the
availability of data. Thus, I examine the production shifting responses of U.S.
MOFAs in 9 countries including Canada, France, Germany, Italy, Netherlands,
Switzerland, United Kingdom, Japan, and Australia. And, on the flip side, I examine
the production shifting responses of the U.S. affiliates of Canadian, French, German,
Dutch, Swiss, British, and Japanese multinational parents.
The unit of analysis varies based on the question being addressed, but periodindustry-country triplets form the basic units of observation. For example, the
percentage change between 1985 and 1986 in real exchange rates and U.S. content
levels in sales made by U.S. majority-owned foreign affiliates in the chemical
industry in France is one such observation.
Exchange rate data and national price deflators come from the IMF's
InternationalFinancialStatistics;the OECD's Indicators of IndustrialActivity; and
the U.S. Bureau of Labor Statistics' U.S. Export and Producer Price series. Data for
estimating content levels come primarily from the U.S. Bureau of Economic Analysis'
annual surveys of United States multinationals abroad and foreign multinationals in
the United States. To date, annual surveys of foreign multinationals' operations in the
United States are available for 17 years from 1977 through 1993. Annual surveys of
U.S. multinationals' operations abroad are available for 12 years from 1982 through
1993. I supplement this latter series with the BEA's 1977 benchmark survey.
Although this is not a very long series, the coverage here in terms of time,
countries, and industries is wider than in any previous study that has considered these
issues, not to mention the fact that this is what the existing data allow. Besides, as
Figure 2 shows, the 1977-93 time interval encompasses at least one prolonged episode
each of dollar appreciation and depreciation along with other less pronounced shifts in
the exchange rate. So the results ought to be robust and generalizable.
16
____I_
_
1_1__
1-11_1-_1111-11(-----·
Empirical Results
U.S. MNEs Abroad
Let us begin by looking at how exchange rates and home content levels have
moved over the course of the study period. Consider Figure 3A which plots exchange
rate movements and sourcing patterns within the Canadian affiliates of U.S.headquartered MNEs.
To get oriented, look to the center of Figure 3A at the panel entitled
"Chemicals and allied products." The solid line in this panel plots the course of the
real exchange rate that is specific to the chemical and drug industry, and the dotted
line tracks U.S. content levels in sales made by the Canadian chemical and drug
affiliates of U.S. MNEs. Both the exchange rate, which is stated in terms of U.S.
dollars per Canadian dollar, and the home content levels are plotted as indexes. The
scale on the left pertains to exchange rates (1980 = 100) and the scale on the right
pertains to home content levels (1982 = 100).
Eyeballing the solid line in this panel, we can identify roughly three exchange
rate episodes over the 1977-93 period between the U.S. and the Canadian dollar. First
between 1977 and 1981 the U.S. dollar appreciated by about 10 percent, then between
1981 and 1991 it depreciated by about 30 percent, and finally between 1991 and 1993
it appreciated by about 10 percent. So in the index, we see a down, up, down pattern.
Likewise, keeping in mind that estimates of U.S. content levels between 1978
and 1981 are missing because the BEA did not conduct annual surveys on the foreign
operations of U.S. MNEs during those years, we can follow the dots in this panel to
see the shifts in the U.S. content levels in sales made by Canadian chemical and drug
affiliates of U.S. MNEs. By connecting the dots we see that (when the dollar was
appreciating) between 1977 and 1982 U.S. content levels fell by about 30 percent;
then (when the dollar was depreciating sharply) between 1982 and 1990 U.S. content
17
i
Il)ir
)111111II
..1..
I^.
·--·l-·-·r-
·I·.-·r*--L-.rwrrrrrlrur---r---ur
'L-
levels rose by about 125 percent; finally (when the dollar was again appreciating)
between 1990 and 1993 U.S. content levels fell by about 10 percent.
The pattern is clear enough to suggest that at least in the chemicals and allied
products industry, the Canadian affiliates of U.S. MNEs were shifting production
quite vigorously in lock step with exchange rate changes. Moreover, they were doing
so systematically and contemporaneously.
To get a sense for the absolute magnitudes of U.S. content, refer to the table
on the right hand top corner of Figure 3A. The table contains two columns of
information--one indicating the industry structure of U.S. multinational operations in
Canada, and the other indicating U.S. content levels in 1982--the year for which the
U.S. content index is set to 100. As we can see from this table, chemicals and allied
products have on average accounted for 13 percent of the total sales made by the
Canadian manufacturing affiliates of U.S. multinational parents. And, in 1982, U.S.
content levels in sales made by Canadian chemicals and allied products affiliates
stood at 11.9 percent. Now following the dotted line in the center panel, we can tell
that in the intervening years, especially between 1982 and 1990 when the dollar was
depreciating sharply, U.S. content levels in this industry rose to about 27 percent. So
during a period when the dollar fell by 30 percent, the U.S. content levels in this
industry rose by more than 100 percent.
Contrast the patterns in the "Chemicals and allied products" panel with those
in the "Motor vehicles and equipment" panel. In the latter case, there is virtually no
movement in either the real exchange rate or the U.S. content level. The U.S. content
level in motor vehicles and parts was 62 percent in 1982 and it remained virtually
unchanged in 1993. Not only is this pattern unlike the one we saw in chemicals and
allied products, but it also shapes the industry share-weighted aggregate pattern
shown in the panel on the center top of the exhibit. This is because (at 43 percent of
the total) motor vehicles and parts bulks largest in overall sales made by the Canadian
manufacturing affiliates of U.S. parents.
18
_ ___11
_·_
IE
I
11__
____I
Scanning the other panels, we can see coterninous movements in exchange
rates and U.S. content levels in "Food and kindred products," and perhaps in "Primary
and fabricated metals." But. in "Machinery, except electrical," and "Electric and
electronic equipment," no particular relationship is apparent. And as noted earlier, the
panel entitled "MOFA industry shares-weighted manufacturing" (in the top center of
the page) provides a summary picture in which the weights reflect the six subindustries' shares in total manufacturing sales. 8
Figures 3B through 31, which are included in the appendix, show how
exchange rates and U.S. content levels moved in the other countries studied (i.e.
France, Germany, Italy, the Netherlands, the United Kingdom, Switzerland, Japan,
and Australia--all countries where U.S. multinational enterprises have sizable foreign
operations). In flipping through these exhibits, look particularly at the center top
panel which shows the industry shar-' weighted aggregate patterns. Although there
are wide variations by industry, anm
..
nme panels where content levels are not shown
due to missing data, by and large, real exchange rates and U.S. content levels move
coterminously.
Now consider figures 4A and 4B which show scatterplots of movements in
exchange rates and U.S. content levels. If the foreign affiliates of U.S. MNEs were
flexing their U.S. content levels in response to shifts in real exchange rates then we
would expect the points in the plot to be arrayed in an upward sloping pattern.
Consider first figure 4A. Overall there appears to be a positive correlation.
During the dollar appreciation period of 1980-85 U.S. content fell (see points in the
lower left quadrant), and during the dollar depreciation period of 1985-89 U.S.
content rose (see points in the upper right quadrant). But it appears that the responses
during the dollar depreciation period are more systematically related to exchange rate
changes than those during the dollar appreciation period. Indeed a regression line
8 The category called "Other" includes manufacturing sales that occur outside the six product
categories shown here. As the table on the top right hand corner of the figure indicates, this "Other"
category accounts for only 16% of the sales made by the Canadian manufacturing affiliates of U.S.
MNEs.
19
-- ---r-·-rllg--·--·1·P111
1"41101L1-·L-
---
through the points on the top right quadrant is positively sloped and statistically
meaningful.
But this is not the case for the points on the bottom left quadrant of Figure 4A.
These points are estimates from the first half of the 1980s when the U.S. dollar was
appreciating. Reading from the X-axis, we can see that between 1980 and 1985 the
dollar had appreciated by between 20 and 40 percent against the currencies of the
countries considered. How did U.S. content levels move? Clearly since all the points
except the one for Germany fall to the south of the zero mark on the Y-axis, we know
that U.S. content levels fell over this period. But the extent to which content fell in
each country shows no relationship to the extent to which exchange rates changed.
For instance, whereas the Swiss manufacturing affiliates of U.S. MNEs dropped their
U.S. content by about 50 percent, the German manufacturing affiliates which faced an
even steeper dollar appear to have raised their U.S. content over this period.
Clearly, unless unobserved country effects dominate (and they are unlikely
to), there is no systematic relationship between changes in exchange rates and U.S.
content levels during this dollar appreciation period. I want to suggest two
possibilities that might explain the pattern or lack thereof. First, recall that since
annual data on U.S. content levels are available only after 1982, we don't know how
content levels moved between 1979 and 1982 when the dollar appreciated most
steeply. If firms' responses were contemporaneous and thus "front loaded," we
would have missed it. In my opinion, this is likely to be a major contributing factor.
Second, we can see from the little tables in Figures 3A-I that the absolute
levels of the U.S. content was already quite low in Germany (3.9 percent), the United
Kingdom (5.8 percent), and Italy (4.1 percent)--countries which showed the least
response. It is plausible that U.S. MNEs given their long history of being
multinationals had localized all but the most critical inputs. Thus they were operating
at or near the "floor."
In any event, turn to Figure 4B which covers at the 1989-93 period. Here, with
the exception of Japan, appreciations and depreciations are unmistakably correlated
20
___1
_1101
positively with drops and rises, respectively, in U.S. content. We will look at the
results of regressions after reviewing the production shifting patterns of the U.S.
affiliates of MNEs headquartered in Europe and Japan.
Foreign MNEs in the United States
Before reviewing the sourcing patterns of the U.S. affiliates of foreign
multinational enterprises, I must note two differences which are driven by data
availability. The helpful difference is that unlike in the case of U.S. MNEs abroad,
here the BEA data allow us to estimate content levels for the entire sample period
including 1978-81. 9 So we get a more complete picture here.
The second difference is that the content level tracked here is the foreign or
non-U.S. content level. Strictly speaking, since the independent variable is a bilateral
real exchange rate, we would like to track just the home content level. But the BEA
data do not allow us to decompose by country of origin imports made by the U.S.
affiliates of foreign MNEs. Fortunately, the damage done ought to be limited because
between 60 and 80 percent of the imports made by these affiliates are sourced from
their parents. 10
With this in mind, turn to Figure 5A which shows the movements in foreign
content levels in sales made by the U.S. affiliates of British multinationals. Note that
this figure is set up exactly as Figure 3A. And as was the case for the foreign
affiliates of U.S. multinationals, the food and chemical industry show the most
noticeable patterns, but as the center top panel entitled "MAffiliates' Industry Shares-
9The BEA extrapolates its data for 1977, 1978, and 1979 from the 1980 benchmark survey. Second,
even here there are years when the data are missing for one or another industry.
10 Two points on this. First, between 1977 and 1993, the dollar has moved in roughly the same
direction against other major currencies. This correlation between bilaterals (dollar-pound, dollar-DM,
etc.) ought to ameliorate to a great extent the problem just described. Second, the estimates of foreign
content have been made after taking into account all available information. For instance, in deflating
the import values, the share of imports that are sourced from parents are deflated by the industryspecific export price in the home country of the U.S. affiliate, and the balance is deflated by the
industry-specific, U.S. import price.
21
---allC--arrIll(lllllller___·.l-·IOII--L·--_-Y-
-
Weighted Manufacturing" suggests, even overall, the patterns are remarkable and the
story is clear--foreign content and exchange rates move in lock step.
Likewise, consider Figure 5B which shows the movements in foreign content
levels in sales made by the U.S. affiliates of Japanese multinationals. Although there
are several patches of missing data, the pattern is again clearly noticeable. Indeed,
eyeballing Figures 5C through 5H (enclosed in the exhibit appendix), one gets the
sense that the patterns are rather robust and consistent. Lastly, Figures 6A and 6B
show scatterplots for the U.S. affiliates of foreign MNEs and we can see quite clearly
the anticipated relationship.
Regression Results
So do Japanese MNEs shift production in response to exchange rate changes?
Has their response been less elastic than those of their European and U.S.
counterparts? Based on the pictures we saw so far we expect the answer to the first
question to be yes, and, indeed, this is what the regression results in Table 1 show.
The coefficient on exchange rates for Japan is positive and passes easily the
conventional test for statistical significance. And since the exchange rate coefficients
in the table are elasticities, the results for Japan imply that for every one percent
appreciation (depreciation) in the yen-dollar real exchange rate, the U.S. affiliates of
Japanese MNEs drop (raise) by 1.3 percent their foreign content level.
In terms of comparison, we can see from the other columns in the table that
Japanese MNEs respond at least as elastically as the British, French, or German firms.
In fact, as the first column in Table shows, the average exchange rate elasticity is
around 1.06 which is not very different from the number for Japan alone.
Before moving to look at the results for U.S. MNEs, let me say a word about
lags and the time trend variable. The inclusion of lagged exchange rates produced no
change in the results and the lagged variables themselves took neither sizable nor
even moderately significant coefficients. These results imply that the U.S. affiliates
22
-I_----(IIPP·LEII-ii_13
1__
_
___1________11____1_________I
of foreign multinational enterprises not only shift production contemporaneously with
exchange rate changes, but their response is also complete in the same year. This
absence of lags in sourcing adjustment is noteworthy because her unlike in
conventional trade adjustment, there appears to be no j-curve effect (at least none
spanning more than 12 months). But recall from the earlier discussion this finding is
not inconsistent with what Junz and Rhomberg's work on lags might suggest.
On the coefficient on the time trend variable, the story is the same. It is
neither large nor even remotely statistically significant. This is surprising because
one would expect that between 1977 and 1993, elasticities would hre risen. One
explanation could be that because I assigned the time trend variaole positive integer
values increasing from 1 to 16 (for the 16 years), the coefficient turned up as
unimportant because during the 16 periods some of the changes in the dependent
variable were positive others were negative. But the results shown in Table 2 suggest
that even if I were to somehow correct for this problem, the elasticities are likely to
largely unchanged over these last 15 years.
In Table 2, we see the results for multiyear windows. In column I, we see that
between 1979-85--a period when the home currencies were depreciating vis-a-vis the
dollar--the production shifting elasticity of foreign multinationals in United States
was 2.8. Then during the 1985-87 period, when their home currencies appreciated
sharply against the U.S. dollar, foreign multinationals exhibited a production shifting
elasticity of 1.17 (which if anything appears lower than the elasticity of the previous
years). And finally, bertween 1989 and 1993, the production shifting elasticity for
these entities was 1.39. So it appears from these results that there is no trend in the
elasticity coefficients.
Table 2 also shows (in columns 4 and 5) the regression results for U.S.
multinationals. At 1.49 and 1.88, there appears to be little to distinguish the U.S.
results from the ones we've seen already.
Conclusion
23
1
I11C·l
__^__ICCII____L1_31_
During the previous decade and a half real exchanges rates have moved quite
dramatically. Between the late 1970s and mid 1980s the U.S. dollar rose sharply and
then tumbled against major currencies (especially the Japanese yen) in the years
subsequent. Using these exchange rate episodes as a natural test bed, this paper has
examined whether multinational enterprises respond by shifting their sourcing, and if
they do so, whether their responsiveness differs by nationality of the parent firm.
The results just reviewed provide compelling evidence that multinational firms
shift production systematically in response to exchange rate changes and that the
vigor with which they do so is unaffected by the country in which they are
headquartered. In particular, Japanese multinationals respond at least as elastically to
exchange rate changes as MNEs headquartered in Europe and the United States.
Indeed, the appreciation of the yen over the last decade has decreased sharply the
reliance that the U.S. affiliates of Japanese MNEs place on their home operations.
This finding is consistent with the conclusion reached by other empirical
studies (see Lawrence, 1991; and Petri, 1991) "that access to the Japanese market is
not completely insensitive to incentives--that the implicit barriers to imports are more
like tariffs than quotas" (Krugman, 1991: 4).
An equally noteworthy finding of this study is the absence of lags in the
sourcing adjustment of European and Japanese multinationals. Why U.S. MNEs
exhibit lags in adjustment while foreign MNEs don't is a puzzle that remains to be
explored. One plausible reason is that U.S. MNEs face an asymmetric disadvantage
in entering and especially exiting European and Japanese labor markets. There may
also be factors related the degree to which firms outsource in responding to exchange
rate changes. But more work remains to be done before these hypotheses can be
sorted out, and in my future research on this topic I would like to distinguish between
responsiveness that is fully internal to the MNEs' own network and that which relies
on outsourcing.
24
-
r-
______
__ Illls···Y---^------
1___1
Figure 1. Conceptual models of multinational sourcing responses to exchange rate changes
l
,, -l
i
iii
:ii
100%
THE STRONG SUBSTITUTION MODEL
U.S.
Content
Level
0%
I
I
0
Exchange Rate Changes
_ ,
I
I
I
I
I
I
II
t.I. iAlI.
I
tI -........
'
I
THE THRESHOLD MODEL
I
THE FLOOR AND CEILING MODEL
I
I
I
I
I
I
I
I
I
I
tSutll.: ConcDual models o1 sourcrng
-- - -·
-·-------·------------
---
"L?'
I · ~~~~~~~
i..r
-I
2 ,4
I~·'·i
."~;~'.~L;
;·-_·
i·
f .r
I.:i:z·r-
-··C~~
\''''
'~t,~~A~~
....·;E'~l''jli^:
C?
0\
r-
i
-
c.
4
Itj
4
Z:
00
.
Lx,
00
^I
%J
z
:-z
Cg
C-r
I>-,
'.
.
-d
'-
t-I
stw#|t}PeleLX* <<**--ttt~·
.
- UL"s
·-.--
~~L"-.
1Z
";J,
_...
r.
)i~~~~~~~~~.u~
·
:i`~~~~~~~~~4t~~~S~~~s;-.
pl-,.-,-;-&----,.:61
;
R
I.C
~r3
C
C
C
",
,
"l
V)
/22L
__
___
CII_
C_I
o-O
o
-C
tj
4)
Col
*
4-
a~t~~~~
e~t
N-9~
(4(. n
IC
-.
a
')
V
C
2
I)
os
0\~
s
0.
~oO
;a
cr
ca
IC
,u *a
a8
1
r
.x
a
00
§
vW
a
00
z.
'i'
.I
o'I
V.
o
U.-
a
C
o
00
r00
I
4.)
.C
r
V~
C
0
a
C
~00
Q
CM
0
C
-
O
!
xs6o
u
O
4)
aLa
U Y r
na ~~~cr~
Y
aO
0
2
IC
,~:O
8W
C
U
a
4)
MI
*
~a
4)
0)
U
Eo
C
i
0
2>Co
.
_
00 aa
00c
'
e:
I
-
-I
.
*1-
4)
10
8
0
I
C4
u
0
00
in
-
mO
to
o
_
n
¢
0
f4
-
-
I
m
0'
I11
C
"C
0
C-
3Z
Ea
00
00
crl
%m
i
co
Q
-,
C)
C7
9
vI
00
'3
v;
4
0%
-0
C)
o,
0
0%
co
4)
tI
_
00
4,
.
V
'a.
4)V
10
oo
%n
a
w
c)
cc
(n
a,
Q
I
co
;L
t
ac8
4.-
0
E:
0
.C
U)
C)
U
'
o
3
0%
0
f.
o 3 Cu,"a
S
0
G
-a
.,
0
C)'c
4
3
;%
00
C
0
0%
0%
_cor
o
cr,
0e
V%
3a
c
....
OI
ta-·
0
,.
0
Z
0
=
o
>
0o
X
-
3e
4.CZ-
u
.4
C
CC
C)
w
..
_
-
a
_~
.
Ca
a
e
C, 00
Ca
.
J
-
-
U
-s
0%
Co
c
C
C
u
C%
0o
4-
c
CC
m
C
-
00
0o
-I
a
*5 a- C
FCO a
~
C
3
co
U
CO
Co
1o
z
f-v
co
03
,
00
0
ac
o
_
>
C
ro
4)
a~c
C
4i
C
c
--
1---11-···1
^IY-----
-
--
O84r.. #
; *C -.
o0..
Ci,
-
" ".mw"
-
.. - .
1-·-"..---!
----- -
*
a
4-A.
Figure :d. Change in U.S. content levels in sales made by all U.S. majority-owned
foreign affiliates in Manufacturing, 9 countries, 1982-1985 and 1985-1989*
Percentage change in U.S. content levels
in sales made by U.S. MOFAs
-At
,v
l
l
l
7
I
I
SwiMfg
I
I
Ipn Mfg
MYfg
-
40
_ . . . .
-. . . .
-.
. ..
... .
. .:..
:Ger_Mfg
I.
Ig
UK,.Mfg
...- I.. .. ..
20
·..
..
...
.....
...........
CanMfg
Ne't_Mfg
*
,
.
i Gcr_Mfg
0
**
rdLMfg-
ICan
.
Mfg
. ...
I
UK_Mfg
:AusMfg
a
Jpn..Mfg
JpnMfg !
I
FCMfg
I
I
!
-20
.
..-.
..
.. .. .....
....
,!
....
.
........................................-
.....
Fr.Mfg
I
I
Net.Mff
I
.
-40
I
I
Swi_Mfg
-60
*
II
__
-40
-60
-20
0
I
I
I
I
40
20
60
Percentage change in U.S. real exchange rates
Changes < 0 are dollar appreciations between 1980 and 1985;
Changes > 0 are dollar depreciations between 1985 and 1989
Sources: Author's estimates based on data obtained from the U.S. Bureau of Economic
Analysis, U.S. Direct Investment Abroad; IMF, InternationalFinancialStatistics;OECD,
Indicators of IndustrialActivityr U.S. Bureau of Labor Statistics, U.S. Export Prices.
*: For reasons stated in the text (lack of data), changes in U.S. content levels between 1982
and 1985 are plotted against changes in real exchange rates between 1980 and 1985.
uEdia:plou: mtg
--- --------
·-------
__
----------
I
T-
____^I_
ImCIP1_IJ
I_II
4.
Figure IL Change in U.S. content levels in sales made by all U.S. majority-owned
foreign affiliates in Manufacturing, 9 countries, 1989-1993
Percentage change in U.S. content levels
in sales made by U.S. MOFAs
_
60
_
· · L_
I
I
_C _
I
:
i
: ·
*.
* Ger_Mfg
40
,:
ai
Swi_Mfg
,
20
...
................
... ,. e. . e
.
_.e.._
_.
__
i.
.... _
: * UKMfg
Ndt_Mfg
;
I
AusMf a
a
·
0
* Can_Mfg
I:
-20
:
i
I
-40
-20
_
_C
_F
___
-10
0
10
Percentage change in U.S. real exchange rates
Changes < 0 are dollar appreciations between 1989 and 1993;
Changes > 0 are dollar depreciations between 1989 and 1993
20
Sources: Author's estimates based on data obtained from the U.S. Bureau of Economic Analysis,
U.S. Direct Investment Abroad; IMF, InternationalFinancialStatistics; OECD, Indicatorsof
IndustrialActivity; U.S. Bureau of Labor Statistics, U.S. Export Prices.
2
q
*
1o
Wo -I
ll:r
.'
rl
-t
-T
-
"
0
r-
'0
a,
0
M
el
C
el
r:
m
cc
*
Ce14(14
In
In
cn
r0
Q
V
(l
00
I
0
eS00
.E
ca
.
'S
(A
vO
00
0
0
U
I,
00
c,
o
C
U
00
-'
u
00
CC
M
.0
(U0
0%
_-
g
u
*L
,
cr
.2
I
.C
IC
-. 0
ca
0i
*E:
a
V
c
co
f-
0
a
c
0
o
00
-.
I
00
lu
f:-
a,
oC
,U
p
(A
v-u
0%
O%
00
U,
cn
CZ
0
-
C-,
0%
C%
I.co
-
0
U:
)
t.. <
0
8
V:E
0
_
I',
3
IU
0
8
0n
W'
_
0
.2
c
0
0
I
I
n
.
0
. 85
0
U
-
IS
M
b
.
_
_
.E
r -;
..xu ,
U
'.d
C
C
o
·
t
r-
-
C; .U
0
.r
o
-
0
-
0
'r
3
C
U
_~
8 ;88 S 8
0
_%
.~
D
-__~ 0
o
_\
M
0
.0
I-
z)
0%
co
z
CO
I.
5
C
,&
0
00
U,
4)
.Y
c
:E
s
U
4
V~~
.
Q *:
o
C w
00'E.
0%
ft-t-
u,
aa
co
'A
4)
c
2I
o
_14
o
>~~~~~a
o
.a
%
o
E
U
v7
0
rco
.2
0.
0
e
0.
)
W
cot
V
cu
'I,
00
r,
a
a0
00
oi-
-
_
aCe C
9
U
0%
5-
.W
Q)
4)
rr-
.O
-
(U
-uC
,
00
.
u
0
3 2
U
4)
o6
<5
.E
Z;=I -a
.0
0
f
=C*CM
,
_
U0
U
>.
00
c3
Ug
-C
ou
oo
C4WC
U
V,
a2U
Cca 0
_c
PC
U
o,
-
-.
II
-
<,
'3O
·,
0
.
._
-QEO
0
O
_
-r
%x
0
:'
U1
04)
co
l
30
-------
---
o
:
o
a
.. *
O
-
*
00
vl
-
o
e-
-
,-o -_
-
0o
en
C
0%
8
0%
I
0
E
cr
C
c>\
0%
aL
K
r
00
0%
z b
bm
b> tr >
U
-
.
oc
c. ;
C
v:
1:1
1:
C·
C
-~
(U
c
b
~0(
.Q-~:
o
0%
C
0
o
U
2o
Cl
E
s:
o
V_
I
ri -
-
-
-
-
DU
~cC_
-
(U
(U
a-
II
O,
4)-.to
-o
-
'
:
-
C;
Ir
1.
a,
zrO
0
C
=
M
(
u
Iroo
Q)
.r_
U
(U
.2vI
2
-
.e00
co
C-
-
_
C
U
o
00
r
.
~
.U
c
X C 00
_ _
UO
C)
C
c#:
C
_
6
.n
(C.'
-
-
I-
a
0%
-
=
I%
aU
o
0%
a
00
C
'='c
C3
a:
O*-a
ms
II
l)
2E
"C
co
%n
co
U
-o
C
fn
00
00
a- Ct.
0*
U
0
.U
4)
.t,c
4
_
Q
_
e
S
I
C._
.2
uj?
C e
cn
U
Zc
'c
3
0%
C-
.2
U
Q
4)
..
sg
,
Ln
0
In
C4
8
r"
c
_
V
0
-
rv
-
0o
'I,
I%
4)-
1
o
~
C
G
4ovU
-0.; O
_
0%
crk
C4
C_
I-
C
(0
CZ
4,
0
*Z -1
CM'
00
00
Woc
C-
co
00
W
0
la
.)
b
U
4)
o0%
Ic
00
ra
t,m
)
2
r-
04
z
0)
)
04
4)
t-
U
o
>e
.C
"v
Id.
CJ
5
0,
31
--
----- --··---------~-~1~-~~"~1-1'
U
*Z
0
CU
C
0
s,
0
oa
t
mr
ro ,,
t-
.
.:z
o
-·-----
!
s~
00
w
U
I00
v00
b)O
C
0%
C'
C.
00
'
Xv
O
-n
C-'
a
0
0
- *
(o
Figured. Changes in foreign content levels in sales made by the U.S. affiliates of
foreign multinationalsin manufacturing, 7 countries, 1979-1985 and 1985-1987*
Percentage change in foreign content levels in sales
made by the U.S. affiliates of foreign multinationals
200
I
a
Germany
I
Netherlands
France m
Japan
II
100
United Kingdom,
I
Switzerland.
I
0 ..... ''......
".~Jaiii '.
I
Switzerland
· United Kingdbm
Germany' France
France
· %
Canada
·
Netherlands
Japan
II
I
Sw.zrln
-100
_
_· _
·_
0
-50
100
50
Percentage change in home-country real exchange rates
Changes > 0 are depreciations in home country currency between 1979 and 1985
Changes < 0 are appreciations in home country currency between 1985 and 1989
Sources: Author's estimates based on data obtained from the U.S. Bureau of Economic
Analysis, U.S. Direct Investment Abroad; IMF, InternationalFinancialStatistics; OECD,
IndicatorsoflIndustrialActivity; U.S. Bureau of Labor Statistics, U.S. Export Prices.
*: For Japan, the yen depreciation period begins in 1978 and goes through 1985.
(di:plos: mIti only -1979-87
32
1__11__
11_ _
Figure S. Changes inforeign content levels in sales made by the U.S. affiliates of
foreign multinationalsin manufacturing, 7 countries, 1989-93*
Percentage change in foreign content levels in sales
made by the U.S. affiliates of foreign multinationals
-
40
-1
-7
--
Il
i mili i
iii...,1
i
30
20
Kingdom
_.......... ........._......:2...... ...... ......RtUnited
_
Canada
Netherlands
~~~~I.·~~~~~~~~~~~~
10
0
:
wi
·
-erland
....................
:.........
.....................
..... ...... ........
Germany
-10
-20
Japan
Japan~.-~~~-~.-~~~-~·-~
-30
-40
_P___
L
-10
-20
0
10
20
Percentage change in home-country real exchange rates
Changes > 0 are depreciations in home country currency between 1979 and 1985
Changes < 0 are appreciations in home country currency between 1985 and 1989
Sources: Author's estimates based on data obtained from the U.S. Bureau of Economic
Analysis, U.S. Direct Investment Abroad; IMF, InternationalFinancialStatistics;OECD,
Indicatorsof IndustrialActivity; U.S. Bureau of Labor Statistics. U.S. Export Prices.
*: For Japan, the exchange rate change episode begins in 1990 and goes through 1993.
fdius: maf onty - 1989-93
3
·IFI*l II^-·(
I^1L..L*- ^--I-..I ·.
)
·--·1
I.
· · · ·· ·-yl----_ll__l^1·_·1111·111(·1111
I
I
Table 1. Regressions explaining year-to-year changes in the foreign-content levels in sales
made by the U.S. affiliates of Japanese, British, French, and German multinationals,
aggregatemanufacturing, 1977-93
Countryof Headquarters
All
Foreign
Japan
United
Kingdom
Constant
2.24
(1.47)
3.24
(1.17)
0.17
(0.06)
4.67
(0.80)
3.0
(1.03)
Changes in real exchange
rates
1.06
(7.25)
1.31
(5.34)
0.89
(3.19)
1.31
(2.35)
0.96
(3.91)
Summary statistics
Adjusted R-squared
.32
.65
.38
.23
.49
Number of observations
112
16
16
16
16
Independent variables
-
Note: T-stats in parentheses.
resulu: an. uk france.germany - annul changs
33+
x'--
France Germany
Table 2. Regressions explaining changes in the home-content levels in sales made by foreign
affiliates of U.S. multinationalsand the U.S. affiliates offoreign multinationals,aggregate
manufacturing, 1979-93
ForeignMultinationals'
U.S. Affiliates
U.S. Multinationals'
ForeignAffiliates
Home
Currencies
Home
Home
Appreciating
Currencies
Currencies
and
Depreciating Appreciating Depreciating
1979-85*
1985-87
1989-93
Independent variables
Constant
Changes in realiexchange
rates
Summary statistics
Adjusted R-squared
Number of observations
Dollar
Appreciating
Dollar
and
Depreciating Depreciating
1985-89
1989-93 **
-60.80
(-0.99)
3.3
(0.25)
8.95
(1.96)
-28.51
(-1.65)
9.345
(2.13)
2.80
(2.46)
1.17
(2.61)
1.39
(3.07)
1.49
(3.19)
1.88
(3.83)
.46
.49
.58
.51
.67
7
7
7
9
8
Note: T-stats in parentheses.
*: For Japan, the yen depreciation is measured from 1978.
**: Results shown are without Japan. With Japan, the coefficient on exchange rates is 0.45 with a t-statistic of
0.66.
resuls: us.t - fortimn msson restl
35
-
-----
·-----------------··--
-- ·
---------
--------
---
A PENID
3b
______I
__l___l·_·___CI)____1_111__-·--_-_---__
l
e-"
*
-
"-
v:
CZ
V:
'0
Be'B~
8~~ ee
-QrOc~
5U
5-4r;-
0S
(d
-
U-
4
0
N
-
-
'-
m
oIn
ofi
0%
,0%
c
J.
0%
g~d~Ei2ep~~d a
~~"~`0"~~~ 4
a
C
@0
4
aa
0e
a,
00
r
15
E
4.)
V.
EL
B
C
· e 'r
cc
a
s
'
r0
_
00
(J
v:
4;
tt,
5os~~~~~~~~~~~~~~~
-h
go
'-
B-
0*
I
a
E:
.D, o
CC 0
au..
a~~~~
>
"
_ I..0
_
a
ay
3~~~~
.2
Ir-
2
10
9
se
0.
0
0%
4)
0
_0
a
oc
C
_
I00
-EL
a
0%
00
C.)
00
r
0
(U
g0.
0%
co
-S
4-'
U
(0
@0
00
c)
vco
C
ab
I
.0
0
0
-
>
0
5-
g
.2
_
a
a
4)
_·
2
L0
to,
- .. ,
00
.a
O
-
I-
Z
(c
'U=
(14
(.
-
-
V-60
-
o
I
I
M
co
I
I
9
U~~~
10
.0
00
00
fn
4-I
a,
t-
I
V=
.s
0
U.
0
00
a8
_
_c
fn
_
C
la
a,E
0%
00
k-
4.-
cr
00
C
wt
00
°o
r~~~~r
.)
00
5%.
0
U
(U
.2
v,
4.5--
,
,
.
.
a
r~~-
rr
13
(-4
u
c
.90
-
5
-
ao
"
I
8(C14
C
0%
t 0%S
co
Im
00
@0
c
U
a
0
O
42
C<
00
5~s
00
0
C
(.
-_
8_
'C
0
M-I v 4)
;a
V
_;
._
-
·------Ull-----*rre
C
t-
,
e
CI
<
-
=
C
m
0
-
0
U.
··L---;-
----- c··---------·I-·------
VE
00
00
0%
Kt
e
.
.
C
o
Ifn
0%
II
0%
It-00
.1
tgo
.S!
0v.,
00
(0
U
c
5-
5-l
O
I
b
I
0%
II
I
00
00
5WI
U
So
m
0%
C)
C
._
'Ud
c :2
tui
o
<ca
a
r~~~~--
-
5
a
o
_
...
h
U
4)
C:r
t
0
Et
0j
W
.
"a
.Q
O X 0_
'E
O a ' '.0
,.
.
r
r
j
rov
.)c
0%
Q)
-
bc
0
4.)
4)
.0
8
n
o
a
0
Q
(2
~~~~~r
3
S
Ir
00
0
0
I
aa
0.
m
G
co
aU
(u
@0
U
4)0b.
* 3
00~
Xoc
co
a
00
Si--
0%
ON
00
e
.!P
J0
4)
-o
-4)
n
..
GI'
2
r
-
0%
I
4)
r
-.
4
-
=
en
j
o
o
o
0%
-
4)
c
5
j~5U U
-
0O
0
Sco
00
I ¾,
00
00a
4)p
'a
x
4-I
61
.0%
U
(J
.0
(m
4-.
.C-
O *;
f-
CD
C.:
8-
a
--
2
-
0..*
a
I
--
-
*
"oo
O
en
li
*
C
x
=
_
.E
-
'n
r
_
0o
o'
r>
8 c 8
_
r_
8
(
<n
8
o
.
'n
8
·" o
8
n
-
j
0%
U
U
0%
o\
CSo
0%
'o
I.Co
a
Co
s~
-
(U
-2
Vs.'
U
.-
co
0
m
-c
~
o
0%
0
0r
E
V
8a
0
8:8
iU
4$
O
I
a
%
w
4
(C
C
.2.
a
o
t-
0%
rI
I. F
(U
4-
E
*0
Co
I
0oo
o
Q
U
0%
.E
11
cc
U
co
-C
c
Ue
4$
-
_
3S
zV0. c
o -
(C
:2 -
Oa
I
(U
,0
.
'Sa
Co
all
I'Wo
..
U
.o
C
VI
C
E
c
-d
o
_
en
0
4$
j
.;a
0
0
-
O
.2
"
0
8N
00
a10
c
'I
._
C
<2
U
O
.0
o
la
c
_
U
-C
.
z coi
4.-Xc
0
co
0 C4
(
(._
U
4$
c
04$
8 ,-0
-
8M
"n
C1
co
I
0
n
84n , 8
o
-Y
8,, ·
.}
o
h.
-4$c
c
c
t
~~~~
ine
0%
<0
0%
C4
i7
00
0 la
'S1
0·
P
C.
.2
0
cr
0
co
go
zz
co
CA
(U
U
_,
t.oo
g
_r _
-
00
WI
00
'n
Co
a
U
0%
C
U
.Q
10
CI.-
a
8
0
2
u
SS
ci
.2
I
9
8
!
m 8
*UcmC
U'
(C
.0
Z
-0
z
co
U
C
0
V
,U
o
V
co
.;-3
ao
c
co
o jc
.VC
C
o
0%
U,
s$
V1
%
*E 0
"
'I
~
8_
a-
I
4$
"s
.,
((-I
0%
,,
_j
,v,
-_
_
O
,
-
001
,
C
C
C
Uc
ooa
o.
_:
's-3
S
.SC
00,
0
Oe
9_
0°C
C
> a
4
4$,z
'
._
L
4U
.'
U *O
_
0 .. .
Co .
*
38
---^I-------^-------r^----l--c\i
_______
_
l
0 r
_' Cl
%
O V'n
-0
e'"
C
*
f*4
14
-
tM~
Si
0,
0
r.m
_
-_
t-
n
o
cW
0%
0%
_
'o
18
C
f00
n
WI
C
U
0
E
oo
C%
00
0
4'00
X
CC
4
U
0
U
1^
E
S
I'
oO
en
U
0%
'V
U
c
'C.
.C
CC
1 oe·
8
I-
-·
(j
4.
'u
018
SM
o
e,.
C
V
ao
m
'
o
4.)
0U
~OI
i~cj~
Y)
r-
-
E
a.
-
5*
0,
0
-
;..
>
0%
4-
12
UC
.001
LG
a
0%
v
a
00
z
00
.)
00
la
q
c
a
O
M
ao
---
C
I
s
Ir c
00
I 4'
II
0
4. 0C .
._
_
~~~~
o U
_~ J
C
:
U
oi
._
09
Z4
C
a
oiC
q'C
O
'
|
--
"
--
--
0 00
1.1
C
-
0
*_
U
2=
CC
I-
-
U
-0
4
c
I;
n-
0
v,
m
t
0%
f.O
t4
r3
nto
r
m
--
0%
_a
o
-
|
E'4
o
at
q
0t
F-
'.-e c
2
Q
o
0
r0
CIN
C%
o
2U
Im
-n
00
'V
C4
f"
co
4.
CC
1=
C
,
0o
00
C'
'E
a
00
aC.2' ,s
.r
a"
e
0'
U
Q
I
0
00
o
a
Y
0
*I
0
0c
6
%n
t,
cc
eo
00
w
..
3
3
0%
CO
S-
O, 4o
oo
ra
zR
S.
Ca
00
cr
U
-
F
Ch
II
·S9
Y
-%q
0%
a
I
00
C
jo
-
I',
i
'a
Q1
r-
U
L..
C
U
4-.
t-
08
-
aco
PS
I
-
.
3001
.)
W
D' .
-
_
.
e0
3 9-
oo
O
--
mr mImm
-
EC
. C
s
_
0%
I
g
,
C4
8
14
8-
,
o oCjC
._
_,
e"
0%
C
C
cv
_
O
,Im
e
0%
0%
.,
00
co
I-.
2C
ro0
!2
-o4
:
0
00
,z
4o
04
en
o-
a
t,
C
Sa:
Iro
WI
00
b
r,
4.
)
211
0%
4oo
I-
Cb-
C-
a.
a s
3
::
P.
S
C
c
s
cc
U
C
01
0
C
cc
&
3q
~-
1----·
<i:
U
'I
C
t
00
oo
oo
. C
S
04
Ca00
la
U
U
.)
00
2C
C
-------
.>
j £
C -
2
M,U
.M
00
.)
on
0%
00
CO
I-
o
g
"I
~
--
-- 1-1-~--
-
-
0
8
0
2
:1._..
O0 ..
-
l
eCt
*
*
e;
8
.e0
V:
C
rNf
- r
-ns
'n
4cZ4
z;
~n-
2O
0
31
0%
CC
.)
o
os;
C U
r
C
V.
U
;C
~~U
._
C.
S .!0°
U
a
_·I
C
Cu
c
I
C
C:
U
r
E
I.r-
C
4Cu
CC)
aI
0%
C
C
m
U)
U
_
a
._
U
I-
o
cc
0%
U
.
o
I
0
0
E
_i
0%
.C
el
10
co
cc
r·
U
C
*C
1.s
c
.In.
u
C~
_
_
0\
ao
.E
CZ
E
W
2
0
U
C
.)
f:
cc
<:s
_
az
oo
z
n
o
C
0
Cl
ar
8
D
0%
a,
co
C
_
a
0%
c~
C
C
S
-O~~~O
In
E
'-. e4
< O
-
_"
<
00C
0
IC
0.
C
It
)*0
.0
gC
o
U
U-
= Ua-
C
OUC
04
C
1
C
U
o
_
~~
*R
Cu
o
_
a
ao
_
0
n
0%
U)
11
00
co
rr
*P *C
cc
-
cc
C"
co
xC.
m
CP
C
I
z
0%
I-
ci
.)
U
0_
0:1
0
en
0%
0%
co
U
Ccc
0
U,
_-
.9
C
U
U
.C
9
2
(04
U
10
0
C
co
o
Icc
U
C
o
aO
cn
4.)
0
.
m
(04
C,
6
0%
m
'
!
CU
Cr
-
o
ao
4
,,oo0
a
0
C%
-'
r U
u
v
0%
0%
0%
8
C
U
C
c
C
2
°U-
0%
I-
4.)
1
Cu
W
U
U
ICRa
F
9
I.
C-
P
4.
C
_
.2
>
l
xsr8
2 -o
e
< n CC
'~
.0
r
C
-.
Gj 'Cow
i
I
I
_a
a_
U)EC*~
0...
C
C
U
bC
*)-C C
-
Cu
E
U
C
-e
aU CotC 0
U
I
1
I
-
c
U
0
Q
-
Cu
0o
p C
C
cc.
oo
,
_
-0
U
U
m
6'
C-
U
2
4.)
00
10
C
cc
C-
~U U
C-
C
.I
.i
10
t
LL.
19
C
0
.)
3
C
0:
-a
(11
C
:n
o
= a
8ar
s
_
·
--------------------
_ ..
.
0%
K
c,
-,
t
e
* ,e ?e
-
-r
WIC-
\ao000 0
*
0
a,
-
00
8
4d1
a
8
I , C14 n
1
-
0
,
O~~~~~~~~~~~~~~<
U
0
Z
C
0
00
C
C-
h
0%
E
C
V
9e
'U
Co
4-)
0
0
to
Q
43
a,
U
"U
v
c
.)
C
cX
.
Oc
0
_
r0
co
.)
E
.
C:4
S0
4,
m
0%
V
-
-
o
0%
U
*c
-u
--
i
U
~
SOY
a
o
40a
4-.
*~
.w
C
Z4
U
U~~~
Be~~~~~~~~~~~~4
I-
,.%
0\
0
0
a
C
.e.
~30
II
3
00
ou0
C
0-
·· O
to
o
,u
1=
c.
8
-
-
-
-
.
-
-
-
(1,
0
'
-
_
n
a
-G
._
00
0
_j
e
e
I
4c
Z
0
Lt I a
<r
XJa
8
w
_
I
a
C.-
UI
.)
A
wj
_
0
C.-
C.)
!
o
0
a
u
C.rr
F0
U
Z
0
=
0
4.-
cu
a.
o
C)
0%
.'2
C
en
00
4
00
la
,=1
10
LL
L
-
or
0
II
4)
.J4
C.e
/
.
.9
WI
00
._
3
0%
oo
a
5
II
E
S
t.-
3
G-
la
'
11
00
K
%
0
c
0%
0%
00
0
a
0%
a
4
.!
-
C
-
-3
%
.
.-
-
en
&i
43
3
a.
;
8
_
0
3C4
U 00
40
U
0...'U
o
8
__
'4
a
C >
4.e
0
n 8-
8
0
C4
co
C
0\
_
c
tq
n
e
4,
C
6
=
00
U
Cu
4
2
=1
C-
00
°
°4
a
40
en
(-4
O
0I
0%
0%
o
0%
00
CZ
-2
00
oo
("I
=.
'V
C
c
4.)
co
to
co
00
8
,U
C
._4
h
3
_
a-
Cu
,U
Z
00
w
U
C
0%
c
r.
b-
4.)
C..
C-
2!s
C
(S.
a
L
(Z 8
=
0
2
ZC
"
--
41
-
-·
--
··-
-
-I·
.
C
.
vjj
O;
a-
a
a
4)
00
.)
CI
UN
C-
:~~~~~~~O
as
U ,
CU o
0
U
2
Ci
X
Cco
a-
U
00
4=
43
o
r,
c
-
0%
'. C
O
0 4
(4
Cr
.a
-3
{N h_
-
"
_
..
.
* * 91 tp,
~ru,
f-
a)
=t
-
- esooO
c
.E
LO
t~-
1, S~l
n4
.s
wtto
Z
'aa
'o
c
to
o
p
~~of
o
a
u
c
v;
1
E~~~
no
to
-E
c
-Ni
a
E
Uo,
o
.-
s
t
-
V
_
.E
_
Sa EY4
-cc
O
UO3 C
e
,
s
-Z
r a
e
.
-
:,,Ur .n -10
.- _
._
r
-U
~~~~~~~~~~~~
-- r ~U
C j
oX
- U
a -2
'
O
.
--
ne
°
8.
,
Oj_
'
C
c
>.
,;Z U
,,
I
°8urqq_
I
(YN.
I
!
':
Ui O
U)
I
I
.-o
Z
a~,
C
a,
_R
*C m
°
oso
I00
Ie
Q,
E
:Z
.Q 1
.~
.",·
.e
I0
10
s
r.
C
U
o
o
U
U
C
~~~~~~~~~~~~
._
Z
LS0
Z
(U
C
Ex
.b
.C6
V
c/O
O
aE
_,
-
.
1
.
_
8
_8
8
!i
Z
.9 C~
;
;6
2
V
_
:w
,3
-Z.0
a.
.....
E3'
~~~~~~~~8
Z,
CMn
.r-OI
o
uj
_
"
0c -0 o tO
.0
·c
p
"^I '
4 0
.
0
0
C)
;.
g d~~~~~~~~~~~~
HS%~~~~~~~
%).
_
.;-
=.'^
c
Z U
w <)
_
., !
Sa
< 'o
So
_
On
Q
C
.
K
On*U
_
1C
oa
I
O
0
')
U
.
mVJ
4-2.
---------
_
..
*
r
*
I
"
V
4
vl
-
-
n
o
e-
0%
._
C-1
.E
0
l
.,
>e
-
>
-
(0-
0%
'5
2c
-
0t
0
U
0
0)
0%
0%
ctru
vl
_J
.
Sc
0.
cr
o
.n
00
co
4c
c
.
os?
a
U
C
CO -4.2
_
0%
0
0
u:
4co
0
Ch
-
C
Q
0
00
II
S
.
a~
.
.4
_
"
.0
U
U
Q
9~
o
.
u:
C.
a
Ii
0
E
cR
0
.,IT
N
co
. 0e
*U
ao(U
cIu
Ot
U
-i O
8
I
Occ,,
;a
,0
o
.
-
.,
Oo
ca
~~~~~~
--
--
._
m
-
,
-
,
-
-
8
S,--,-
t'4
2
:
0%
e
0%
-.
.>
0
cc
0
8
o
0 .
· O*0
GV
e
U
'C
_O c
O
0%
Co
I-
E
0
0%
a
0\
I
M
U
a
60
C
L:.
0
C,
0
_C.'
J
._
Q
U
U
0
cor
0
.-
0
0
0%
.
!----
I --
.--
I
'S
.
CO
; -0
-~~
---:-
-
-
--
C
cm
: :_
C,
0%
.
I
C%
o
0
co
C
0
nn
0
00
4
-
0
e
0
eo
co
za,
CF
W
0%
14
0
c
r-
U
.0
Co
U
0
c
0
V
U
V
0.
.b
I-
0
c0
j,0
-
43
·--
·I-----
CE
e>
a
o *a
C
Oi
y
C
C
eo
O
e
o
C
*
a
-
M
· :
°°
Oc
ft
go
co
VI
00
C--
0
-
6~
4-
Co
U
U
_
Co
c
.c4U
r
.
-
--
o
v
c
C
4.
C.-
:e
Q
.E
oa
c
0
.4
U
-
v
4W
E
.
0r
C
C
m
CE
O
o
c
LLQ
U
O
,0
00
U
.0
.
9a
c.,
0
CT
o
0
,
(U
2
C.
<5r
0
Cu
U,
=
a
;j0
a
0
cr
-n
0
0
7aO0
0
0
U
C-.
4'
0%
en
co
t0
( _
o.
o .
l
tR tR e_ e
I
I
-00
tBl e
N
o
C
-r< 0
0x
'r
06
o\O?
0
- C
o0%
C~\r)\V)
I
-o
-O
13
0\
m
1-1
os
o
00
~
c
_,
'"i00
W
ex
7
o.
oo
.s
-
"
-C
I
_
A
A
C
_
0
U
0
zi,
I'
0%
A
S ·*
cs
r0
00
In
c
I.-
>0
'3
26
'1,ul 22
g
-
CO
*
0
-
*;
0
00
00
,0%EU
PC
Q
z
0
00
0%
a,
E
E
7-
-
-c
a
-
0,
u0
a
C
a,
cc~
x
aCU
o00
00
Z-
li
U
se
dtdeeeetet
r
.
O
_
~
8
_
~
a
(
O
0
>
U
0i
:
u
!
_
_
Ce
C
oF~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-
-~
r
U.
3
\
C:.2 >
·-.
.
.I"Z
0
IC
w
=r
u
C
U
:0
Q
_
U.0
UO
tI.
a
000
co
._-
_
u
.
-
(2
f
_
i
I
s
C
z
tmB
I
oa
84
. .. .
I-,
0\
I
°
. .
8
.
cm
en
C.,
0r%
Uw-,F
.
_
s
0%
co
co
00%
I,
o
0\
a
0
!,
0
65 zo
C
2M I
s
.s: IC
a
co
00
0
00
°
I 030,u GV
g
C
I
a
0%
C-
0
'W
u,
U
0o
0%
00
t 00
m
0u
00
00
8
I
9.
a,
I_
W
In
to
cr
.2
.2
6b
0
00
00
.Ca0 a-0
_; 1's
o
00
'A
,,
0
'E
a
0
0
II
II
I
S
-
0c
r-
u
S
Z
.S
C
Vs
_
GI'
o;
ci
_"
_
T
C
m
r.
.t.
o
CWI
s
U
U
F_
f-
P
C
t
U
U
u
_0
0%
laI
0
_
_
8
o
O
-a
r
.r
c
s
O
W
C00
_
_
.
_
_
i
a~
.. u> r--
Z g
,
''
_`
"
oo
S:
m
c
.s
C%
z
.s
,
Z C
-i V.
_
-0>
I
4.
U
._
'
P
U
0%
*u
- .2C Q
_
co
C~CO
I
I
0%
I
0
00
.
0%
00
tj.;
C
._
00
0%
co
Qr
C .t; C
o cc
ZU
cOn
oSU
X
O 0...:
I
I.tt
IO
-.
1c
8
>
O
C
l
-
,v
=l
4
lo
U
I.oo
.=
COC
c
10
*,
_
_
00
1
1
_
-r
44
.-^`.I^--------I
f
o
-
*
so
as_ .f4
(N 10_ "
'0_% 0%
-i (N u -00 (N
*
8
r;
- -
g m8
8
_
-
00 ..
0%
e-C
I
c
pr
'
\
C
C
0%
c
C
V.
I.
0s
c
U-
2
0
C
ic
C,
-'U-
0%
0%
U
e
.
C
cc
v
.C
V.
I
00
"
0
00
r-
00
E
0.
Z
0%
00
co
00
s!g
U9T
(0
U
(I,
cc
C
I
0'.
..0~
"U
A
S
v
a--
c
.-
(U
~~~sr,~
· ~
E;
U,
B
to
'U
U
r
* "-
o\
0o
P
V' 0 .
4Z
I)
0
S
I
-Z
r-
0%
00
"I
v,
0
0%hh
i
I
1v
U
4 ,
au~~~~~~~~(
0%
CI
.Y
0
C
--
(I-
a 8C a
no s 'r.
- -- -
~
0
L·
(U
O
0
00
~
*C
'V
C
V
1ta~o
r
P
O
(U
C
-
-
--
-
o e
°
,,
W_
·C
II
.
>
C
to
C
0
_
.
C
w
Q
I;/, .s~
.x
e
4s,
Q
8
I
c~
(N
80
(-
_
-
-C
o
(ff
0
G%
to
:
J)
--6
Sr2
-(I
4,
0)
0
8
°
0%
'3'-i
oC
' E'O,3i
4,o
cv
cc
.
I
00
co
Q·
0)
a*
I.0c
Ws
Q)
(-
'0
00oo
co
0
en
;:j
c G
U
10
"U
SO
-
0)
0 0
0
4)
o
0
88
0
O
'a
c
*
t-
-4 ,
..
C(
O
C
4
c
'"
t:
- .e
I
~4,. o
4,0g
el
0%
,n
C-4 & -
;-- Qn-0
4,
4C
en
Co
0
0%
a
o
ao
00
-
e
s.
O
o
I
co
r.,
o
flj
04
2
C-,
ai
2
oo
_
r
a
en
C14
0O
_
8
0CO
o
co
0%
0)
OC
*::.Q.
'-
U
0cCn
0)(
C0
o\
rC--
m
C
c
._
0
0.)
0
8
0 ...
U, *
Ul
45,
--------
. I-r
rf.
"U
1
.5U
0)
.
o
:)Co
._,
<
00
0r
4,
0%
co
Cco
%
00
00
C
,C t-:
_
0%
0%
a,
~a
Q' U
_
-
---
c
0
1t-
co
(U
cg '0
_
00
a
U
.)
--
_
_
O
C%
0%
2L
0
o
0%
.co
e
F'
FI,
I
U,
_-
2
. tt
10
o
(N
Z
w
0)
ih
C
4.)
e(N
° °
n
C
.0
-.
'iC
_
-.
co'0
I.._
0
M
a
.m
=
'U
-
,
('4
- ('4J
(' 4 *
%Q
0o
- (<
-
0
0
00
es
Q
C.;
U
OC
.,
(n
00
.
00
u
4)
cc
.X
CC
-(U U
(U
cn
t
Q
.
n
'u *G
U
U
>
C4
ccU
.co
-5
4)
-_
Q
C'
E
T
- >
0
,
4i
:
c
r
Z
_
O
e
C
co
O
rf
(U
_r
_
C
cC
tjC .
(U
X
._-
C
O
.r
c
l
U
c
0
I
0
8
a0
R-
on
;
0
o0
_
.n
To4)
-
-U, a40
4)
O=
ho0ci
C
r_
c
2.a
C
m
CI
0%
0%
co
0%
o0
14
ao
C.
2
co
0%
t
co
C00
)cn
Co
U,
e,
co
r00
,
In
.3
Z,
(U
O
co
4.-
9,
0%
4)
E
w
u
0
:=
-
00
.L4'
4-
-
I
.
80
2
b
8
C
-~C
00
wo
a Q
0%
Z
4)
'_
C
'
2
00
cr
4)
.)
CI
14
0%
C-.
c
00
00
c.
4 oco
a
0%
00
co
00
a
0
ds
*U
C
S
CC
.
fn
0%
la
C-.
9
'ES
u
.)
~~8
°~8~g
f.4-.
c
C
za
,,
O
C
u
U
oo
.CI
e C° C
U4
4)
o
4)
O
*C
,
-
C
C
5
J
'
_~
O
_.
or
tC
oCo .e
c,-0.- as
4)
t~c
..
=
(U
O
U
=
r
Cfn
0%
· fO
U
_
a,
co
IIa
.O
4)
.-
(U°
r
4)&
4.:e
8
aC
F
'G
_
oo
co
X
C.
,.
u
u
4)
m
00
Z
4)C
0%
t4C-
C.)
2C
w
r_
L o
_.
-------
.
-ur
_... ._..______
8
°
C4
_
8
---"II"
-------
°
U
X
C_
0.. .
ci
C
..
.
-------------------~""~
*U
@0
'10 C' f-_
'0
-
(
W
1
C
-
-
'-4
=
x
,-4
84
0
-
ty'j
n
8-
a"
0
-
'n
0
8
0%
a
@0
W)
In
C
115Vj
c0
0%
C~~~~~~~~~~~~~~~(
_
*c
Ps
N 00
-.-
0
@0
0
cU
_
ECc
E
4)
_
So
co
@
@0
co
_
'n
I-
c/
O.
8-
" n
o
@0
00
C
U
E
I-.
a
t-I
a
cr
W
10
c
0
Q
4.)
0
C
44
0
m
a,
Z
cu
W)I
2
CA
I
14
11:1
c
cl
EEU
7
a Te
E
'
.e
~~~~~
.
Q
ja
'
-O
*
0
4.
4.
_
I-
_6.
tC14
8 !2
8
'U
"C
.I
C
2
aa
a
.rZ;0
C
.a
+
I
W
C48
I
-"
.E
>
°
0
O
I
t
O
,j
a
·f CQ a
C
U
c0
U
W.O
.a
_a
C
't
C)
_
"a
4)
,- -- Z
U
C
Z
4
-
8
0
8
,.
C
00
c,
co
UU
C. .5
o
trj
- r;O^a
r
-
·O
I
:
laJI
9
cn
-0
Ji
10
.s
-
i
b
--
z
4"~
Ch
az
C%
@0
c
C
0%
o0
a'
C,
-oo
|
m
0%
S
Z
.@
0
a
I
c"
C
C".
c0
@0
o2
0>lo
E
4J
U
U
C
eo
0
ra
t@0
@0
.
C
0%
C"
I
(A
4
c-
U
la.'
~a
4)"
4-
C".
_0 _
WI
8
a
1
W
°0n
8~88
0%t
8
jn
I
_J'
sa
Z
.*'
a
43
co
.
a
0%
.S.
C
2
a
C.
c-
a
-
0
8
i
t>
,a 8
2
1
U
c
c
)O
a
t=4)
O It
O~
a'0
-
CC
;:a
+ .~o04C
03
a - ai
a
'I
E
CX
0j
"C
O
C
.. _
O
0
.I
- -- ·---
·--
:-·-·----------3-----
~
~
I~~"'i,,
..
*
*
*
NR k
*:
c
-
~I
?PI
P99
%8
:2 a
00 ao c('
00 1e
-
I
a ('4
2
o
0%
o
-
r-
o
I
a"
00
C
o0
c
0
7
v
-
R
0%oo
00
C.-
_
r
E
U
0
4.
.
U
oC
00
o
a,
o
0
0C
CE
t5
oo
*-
CK
b
4) e
c
._
"
zr
OUUU
!:
=;~
0~
.~4i
._
a
l
0
=
4.-
a
0
O
c
c
.Z
2
7.
_
-;
00
.0
S
3*'~
(8
'
C4
8
_
a
fa
.
I
Y
x
10
)
>
:
O
Q
C
, 1.
,S'0
III
_,S
c
00t
04m
;l
- >
-U~~~~~~~~~~~~~
!
Ua
!n U
00
-
0
b: 'E
0
$~0
0
-
-
,
I
-
Cd,
t:Z
I
I
m
0%
a
S
.Qa
=
B
ao
U
00
i"1aB
I
.2
U
co
_" . a
V
8
a->
-10>j,.
a00
00
0
V0
ts
_ z
E
C
e
0%
oo
ac
IM
._
U
U
-
a
I
.-00
4)
16
E
" l '
Co
a%
o
E
10.4
iW
_
00
ii; - 0
va a
cnl
-a i.0
C
.
._
-.
('4
4)
.0
.
_ 0*C
S
O
."UU
·e
la
t.
'58
.5
,o
.a
W
U
(
-
'aa
-
...
Ws_
..
Z! ,-"
§
r
;;
-151
.e0
-·-·
e
a~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I-
1-
o0
o
('4
o
0%
o
o
l~O
6
'-.4
V
E
00~5
(0
" 'd.
O
S a"
U
0%
o
a
I-
-
001
m
0
U
co
d
4.SO
co
.! I, u
0%
o
=
(U
U
Q:
'3
eU
U
U
00
co
_-
4
013
Ut'=
t
co
00
0
0
0
V-
00
ccc
to
;-o
U
U
U
pp
co
0%
.-
U
U
C
x
w.
a'
9
Io
1
1
0
Ca.
C-QI
c
.
Is
4.tU
_o_
Z'
x
i~4)
____
8
0
8
(D
11
(1
3
___
('4
-
-
I
__II
U
(
1
---------------------
______
_____
8C1c
t
0
4-4
-
8
a
-
(i----·
.
!
,.
0 ..
cn
_
.
·---------`)··-s113aaol
,*
..
N11- Q
-Q
,
-F
-Q
4
Pol
-C
c
vl
.
c
"
.
0~~~0
o
C
0'
C
4
wU .C0 '3C
o
o
_em
_ S
eiar
00
t-`" ~
_
04
)
tsacUO,
L_
a
0 4)0rj
O
C
r
b
c~ebt~o
00r01O(IQ
c
*
4o
\00\~0r~g\00
=
c
0
-
.t0=3
(
O
MI
0
d
_
fi
_
_
_
__
Z
..
w
-n;
II
Z
C
'
4
I-.-'
a;~~~~~~~~~~~~~~--(0
U M
R
to)0
E
a0
C- U
.
>1;
8"
*U~~~~~~~~~~~~~
.
.o
4)a
-
V
.
-00
_t3 .2
1 ._
jO
8
C4=
_~
0
8
-
,
8C4
,
8-
0n
",
.=
.U
d5
O
"
06
-
c
_S
S
-%
0%
)
0
!2
Q
.
El
CZa
0%
br
~S
0
2
i
.
aD
VN
co
I.,. I 00
c
00
.n
cI
'4
0
C
43
-C
V
' _I
-.
O
*0
C
'U
t
-,.
00
- -a
_: U
W aa V
.
10
n
o0
0
oO
2
Q,
m
c
T-
U
U
C_
.00a
U
2
(*o4
a-
0%
E
V.'
.
8
co
.t-
U
U
C
0
i
b
UI
00
U
oo
Q
I
V
I
UI
.C
wI
O
O
8
C
la
V
ar
]
(-.·
oC
o
(n
ac
IC-
82(.4
a
o-
*-
a
1ca-
_
B~
(0'2
n
R8
Z!
-
1"
SU
0c
.
o00oo
II-',*
J
0
>)
.-
p(J
'
--
0000
Q 5-
Q
fi
Cs
(.4
-
-
Z-.
a
4)
Ta
'Jo
_
Q
_ . oS
14.
1
%)
C
00
00
00
00
0
04
4)
00
(
co
ro
Q
co
<
0%
r.
I.°S
9;
rt
00
we
0
V
C
-9
CO
oo
V
C.
2
C
oW
O0.-
C
p
o=m
t
00
c
4)0=
_
_
8
:D
n
00
'P
,.n
*e;
o
oo
4
r-
P
,t
F%
mlE
oU
0
.)
MVI
-'
00
3 csSV
a
C
I
m
0%
0%
2s
°
8
2
(
n
0
0
8
e,4
C
-
I C Z'_
o
_
U.
Jg
.
4c
-
4011.
-----c-c
·l·---·l*alpCllr·lslr-c·s^-·--·-----I--------·-
--
---
ii
Download