Changing Trade Patterns and `Revealed` Selection

Changing Trade Patterns and ‘Revealed’ Selection
This version: March 28, 2015
Massimo Del Gatto1
‘G.d’Annunzio’ University and CRENoS
Abstract
We estimate potential selection effects, expressed in terms of long-run changes in country-sector average
marginal costs (i.e. cost competitiveness), associated with the recent trends international trade. Estimation
is based on bilateral trade flows (hence ‘trade-revealed’) and does not require firm-level data. Marginal costs
are predicted to decrease relatively more in high-income countries than in low-to-middle-income countries,
due to increasing competition from the latter. This aligns well with the logic of the selection-effect: firm
selection is more intense where competition is fiercer and marginal costs are higher.
Keywords: Selection effect, heterogeneous firms, gravity equation, trade costs.
J.E.L. Classification: F12 F14 F15 O47
1
Introduction
A significant reshuffle of international trade flows is currently underway. At the aggregate level, the share of
world trade involving developing and emerging countries as exporters increased between 1980 and 2011 from
34% to 47%; for importers the change was from 29% to 42% (WTO, 2013). At the firm level, considerable
resource and market share reallocation from less to more productive firms is documented in a number of
papers (beginning with Pavcnik, 2002). Compared to traditional trade theories, new trade models with
heterogeneous firms (Bernard et al., 2003; Melitz, 2003; Melitz and Ottaviano, 2008; Chaney, 2008) feature an
increased ability to explain these trends by ‘slicing up’ the within-country effects of international competition:
increasing competition fosters the survival of those firms which are able to produce at lower marginal costs, at
the expense of the less productive (i.e. selection effect). This process of firm-selection lowers marginal costs in
the aggregate. Through this channel, decreasing barriers to trade with low-cost countries are associated with
long run decreases in the marginal costs level above which firms are not able to survive in high-cost countries
(i.e. marginal cost-cutoff ) and, as a consequence, in aggregate production costs therein.
1 Contact details: ‘G.d’Annunzio’ University, Department of Economics (DEc); Viale Pindaro, 42; 65127 - PESCARA (ITALY);
tel: +39 085 453 7995; fax: +39 06 23311171; [email protected]
1
In this paper we estimate potential selection effects, expressed in terms of long-run changes in countrysector marginal costs, relative to a benchmark country, associated with the evolution of international trade
flows from the 1980s to the early 2000s.
Estimation hinges on taking advantage of the observability of international trade patterns to reveal information on cross-country differences in terms of firms’ marginal costs (hence ‘revealed’). The estimating equation
is consistent with both traditional trade models `a la Armington and new trade models with heterogeneous
firms, while the theoretical explanation differs depending on the reference framework adopted.
The analysis requires only country-sector information on bilateral trade flows (internal trade flows included)
and is carried out for a sample of 49 countries, covering about 85% of world trade, for which non-missing data
are available for all the years of interest.
As well as documenting substantial heterogeneity across countries and sectors, results show that marginal
costs in high-income countries are predicted to decrease comparatively more than in low-to-middle-income
countries, as a consequence of increasing imports from the latter.2 This apparently puzzling result is indeed
the essence of the selection effect: firm selection is more intense where competition is fiercer and marginal costs
are higher. While observed in trade data, this is not reflected by measures of marginal costs, like the GGDC EU KLEMS producer price index (PPI), recently used (e.g. Corcos et al., 2012) as a proxy for country-sector
cost cutoffs.
The paper proceeds as follows. Section 2 introduces the approach. Section 3 sketches the theoretical
framework and presents the equation to be estimated. Benchmark empirical analysis is presented in Section
4. Several robustness checks are then reported in Section 5. Section 6 concludes.
Additionally, in Appendix A and Appendix B we report the analytical derivation of the estimating equation
in, respectively, a variable (cfr. Melitz and Ottaviano, 2008; Corcos et al., 2012) and a constant (cfr. Melitz,
2003; Chaney, 2008) markup framework. In Appendix C we consider the extent to which the estimated
measure of marginal is correlated with some key indicators suggested by the literature. Appendix D reports
detailed country-sector results.
2
Intuition and related literature
The basic idea driving this analysis can be illustrated starting with a standard Anderson and Van Wincoop
(2003) setup, in which, using Tslh to refer to exports from country l to country h in sector s, the gravity
P lh
equation is ‘folded’ by dividing through, first, by total exports to country h (i.e.
l Ts ) in order to derive
country l’s export share in country h-sector s (i.e. Rslh =
lh
PTs lh ),
l Ts
and second by the export share of country
l in a benchmark country f in the same industry (Rslf ). Taking logs, the expression for the relative share of
2 When values are normalised with respect to the G20 average, large decreases in marginal costs emerge for Brazil, Korea and
Sweden, while substantial increases are estimated for the former USSR, Japan, Germany, South Africa, China, and Argentina.
Marginal costs increases are also predicted for the US, while cost competitiveness gains are predicted for the Euro area as a whole.
In general, large increases in marginal costs are estimated for countries that performed quite well in terms of export shares in
the period under examination (e.g. China, Mexico, and India in particular), while gains are estimated for countries, such as Italy
and Portugal, in which productivity progressively shrank after the nineties, arguably due to tougher competition associated with
importing increasing quantities and varieties at lower prices from low-cost countries, such as China or India.
2
˜ slh = Rslh /Rslf ) takes the form
expenditure in country h-sector s devoted to imports from country l (i.e. R
˜h
˜ slh = βs ln τ˜slh + lnP
ln R
s
(1)
˜ h = Ph /Pf is the importing country’s
where relative bilateral trade barriers are denoted by τ˜slh = τslh /τslf and P
s
s
s
relative multilateral resistance term.3
˜ h encapsulates crucial information on the ‘degree of competitiveness’ of the importing country
Arguably, P
s
in sector s. While several papers apply this ‘difference in differences’, or ‘ratios of ratios’, approach (e.g.
Romalis, 2007; Head et al., 2010; Fadinger and Fleiss, 2011; Costinot et al., 2012), the relative multilateral
resistance in Equation (1) can hardly be assigned a clear-cut interpretation on an aggregate basis, without
explicitly modelling the micro-foundations of trade shares. From this perspective, moving to a heterogeneous
firms framework is essential.
Therefore, we derive Equation (1) in a heterogeneous firm setup in which the multilateral resistance component is expressed in terms of country-sector marginal cost cutoffs (which, under standard model assumptions,
equal country-sector average marginal costs), relative to a benchmark country, and, insofar as it is inferred
from actual trade flows, is referred to as ‘Revealed Marginal Cost’ (RMC). The interpretation of the obtained
RMC varies only slightly depending on whether markups are modelled as variable (cfr. Melitz and Ottaviano,
2008; Corcos et al., 2012) or constant (cfr. Melitz, 2003; Chaney, 2008), the estimating equation being the
same.
Other papers have used a ‘trade-revealing approach’ to infer measures of productivity. In particular,
Fadinger and Fleiss (2011) rely on a monopolistic competition framework with CES preferences, while Finicelli et al. (2009) use a probabilistic Ricardian framework following Eaton and Kortum (2002). Waugh (2009)
adopts a variant of the latter, including traded intermediate goods and non-traded final goods. Other contributions include Hsieh and Ossa (2011), Levchenko and Zhang (2011), Shikher (2011), and Chor (2010). Our
work adds to this literature in two respects. First, with the exception of Levchenko and Zhang (2011), the
above studies take a cross-sectional perspective, while our focus is on changes (in marginal costs). Second, the
firm heterogeneity hypothesis allows us to distinguish between ‘exogenous’ and ‘endogenous’ cost cutoffs. This
is a crucial feature. While exogenous cutoffs are ‘first nature’ time invariant measures, endogenous cutoffs and
trade flows are endogenous to the model, being the long-run outcome of a process of firm selection, driven
by the degree of ‘accessibility’ (i.e. trade costs) and market size, as well as by other factors, such as entry
costs. As such, cost cutoffs, and thus our RMC measure, account for country-sector differences after the
firm-selection process. Previous literature is mainly concerned with the former, considering them the main
3 To
obtain Equation (1), start with a standard gravity equation (cfr. equation (5) in Anderson and Van Wincoop,
1−σ h
P (τslh /Ph
Ys
1−σs , in which τ lh denotes trade costs from l to h, Πl =
s)
2004) Tslh = (Ysl Ysh /Ysw )(τslh /Πls Ph
and
s)
s
s
h
Ysw
lh
l 1−σ l
P
(τ
/Π
)
Y
s
s
s
Ph
are the two multilateral resistance terms and the elasticity of substitution σs comes from the constant
s =
l
Yw
s
elasticity of substitution (CES) demand system. Aggregate output in countries l and h is denoted by Ysl and Ysh , respectively,
P lh
T lh
lh
P s lh
while Ysw denotes world output. Since Ysh equals the total imports of h (i.e.
l Ts ) in equilibrium, the ratio Rs =
T
l
1−σs , so that, dividing through by Rlf and taking logs yields (1), with β = 1 − σ .
amounts to (Ysl /Ysw )(τslh /Πls Ph
s
s
s
s)
3
s
driver of international specialisation. In contrast, we focus on the latter.
A more structural approach is followed by Bernard et al. (2003) and Corcos et al. (2012), who simulate the cost-cutoff changes associated with exogenous shocks to trade costs in specific theoretical contexts
(oligopolistic and monopolistic markets, respectively). In particular, the latter calibrate the model on the long
run equilibrium cost cutoff expression. In doing so, they rely on observed bilateral trade flows to estimate the
trade elasticity parameters and on PPIs to proxy for actual country-sector cost cutoffs. The latter choice is
a critical step in such an exercise. Indeed, the only country-sector measure compatible with the theoretical
implications of new trade models with heterogeneous firms is the GGDC - EU KLEMS PPI: a cross-section
of country-sector estimates for the year 1997.4 The measurement error associated with PPIs is likely to be
quite high and, in particular, higher than the error associated with trade flows. Moreover, trade flows and
PPIs are likely to be observed at different stages of the adjustment process. In fact, while trade adjustment
takes place along both the intensive (i.e. quantities traded by a given the number of firms) and extensive
margins (i.e. number of trading firms), the equilibrium cost cutoffs are the result of adjustments along the
extensive margin only. This clashes with the long run specification of the structural equation on which the
calibration-simulation exercise is based. Since it is fairly reasonable to consider the intensive adjustment as
more reactive (compared to the selection effect), to changes in the other structural variables (e.g. trade costs,
market size, and entry costs) and, in addition, data on international trade flows are much easier to obtain than
data on cost cutoffs, relying on the former to retrieve information on the latter offers a chance to estimate
model-based long run effects.
It is worth noting that our analysis is not concerned with welfare: our object of interest are the changes in
country-industry marginal costs induced by the process of firm selection. However, as discussed in Melitz and
Ottaviano (2008) and Corcos et al. (2012), consumer surplus is a function of the marginal cost cutoff: a lower
cutoff implies a greater number of varieties, a lower average price, and a higher average quantity. These effects
jointly imply a reduction in the deadweight loss due to imperfect competition.5 As a consequence, our work is
also related to a recent vein of literature that have revitalized the debate concerning the welfare gains from trade
associated with the globalisation process. In particular, Costinot and Rodr´ıguez-Clare (2014) and Ottaviano
(2014) have recently shown that cost cutoffs are not needed, in model calibration, if the ultimate scope consists
of simulating welfare gains from trade: when expressed in terms of real consumption, gains from trade turn
out to be a very simple function of the domestic share of expenditures and the trade elasticity parameter.6
4 The GGDC - EU KLEMS PPI is used also by Costinot et al. (2012) to study the impact of endogenous (‘observed’, in their
terminology) productivity differences on patterns of trade across countries and sectors in a Ricardian framework. In a robustness
section, Costinot et al. (2012) also adopt a ‘trade-revealing’ in which cross-country productivity differences are recovered as
exporter-industry fixed effects in a gravity equation estimation. It is worth noting that their exporter-industry fixed effects,
applied to our derivation, would reveal exogenous (‘fundamental’ in their terminology), and not endogenous, productivity (see
Sections below for the difference between exogenous and endogenous productivity/cost competitiveness.).
5 The welfare properties of the Melitz and Ottaviano (2008) model are extensively discussed in Nocco et al. (2014). Relative
to a hypothetical unconstrained optimum, they show that equilibrium firm selection is too weak, average firm size is too small,
low-cost firms are too small, and high-cost firms are too large. An insightful discussion of the relationship between productivity
and welfare is in Section 6.2 of Costinot and Rodr´ıguez-Clare (2014).
6 As shown by Arkolakis et al. (2012a), irrespective of the usage of an Armington-type model or a new trade model, aggregate
welfare gains from trade always depend on a few elements: relative expenditure on imported (rather than domestic) goods,
the elasticity of imports with respect to variable trade costs, and the elasticity of markups with respect to firm productivity,
if the framework is one of variable markups (Arkolakis et al., 2012b). Costinot and Rodr´ıguez-Clare (2014) take advantage
4
Although the unavailability of adequate cost cutoff measures is not an issue in that case, the reliability of
the simulated gains still relies on the quality of the gravity equation estimation through which the trade
elasticity parameters (the only parameters needed for simulation) are obtained. As recognised and discussed
by Costinot and Rodr´ıguez-Clare (2014), notwithstanding the tight connection between data and theory that
characterises recent literature on gravity model estimation, the current state of trade equation estimation
remains far from ideal. In particular, most of the concern arises about the fact that model calibration should
allow for cross-sector and cross-time variability in the trade elasticity parameter.
In our analysis, neither trade elasticities nor cost cutoffs are needed ex-ante. They are instead obtained
by estimating a single equation, as exporter-importer and importer fixed-effects, respectively. Moreover, since
the estimating equation is obtained by taking the ‘ratios of ratios’ of the predicted gravity equation, which
eliminates the theoretical discrepancies associated with different model structures, the resulting estimating
equation is consistent with both traditional Armington-type models (e.g. Equation (1)) and new trade models
with heterogeneous firms. Within the latter, the estimated equation is also robust to switching from a variable
to a constant markup framework with CES preferences. Although this weakens the need for model validation
(see however Section 2.4 in Corcos et al. (2012) for a validation analysis of their variable markup model, which
ours closely resembles), the fixed-effects estimation gives the obtained changes in cost competitiveness more of
a ‘catch-all’ flavour than those resulting from more thorough structural estimation exercises. This represents
an important drawback but is intrinsic to the idea of ‘trade-revealed’ analysis.
3
Theoretical setup
The relative multilateral resistance term in Equation (1) can be expressed in terms of real marginal costs
within a standard heterogenous firms setup. This expression, which represents the basis of our empirical
analysis, varies only slightly depending on whether the framework is one of variable or constant markups. The
analytical derivation of the expression in both cases is reported in Appendix A and Appendix B, respectively;
here we focus on its basic elements and interpretation.
The reference framework features S monopolistically competitive sectors (indexed s = 1, . . . , S), active in
M countries (indexed l = 1, . . . , h, . . . , M ).
Consumers maximise a ‘two-tiered’ utility function. In the first step they allocate a given fraction of their
income to goods produced in each sector; this share is then allocated among the different varieties in sector s.
The marginal cost faced by a generic firm is mls (c) = ωsl c, where ωsl denotes the unit input cost faced
by firms active in country l-sector s. The firm-specific unit input requirement c (i.e. inverse ‘total factor
h l
iγs h
iγs
ms (c)
c
productivity’) identifies the firm.7 Marginal costs are Pareto distributed: Gls (m) = max(m)
= max(c)
.
l
l
s
s
of these theoretical findings to calibrate different classes of models and obtain the simulated gains from trade associated with
counterfactual scenarios (i.e. moving to autarky or tariff reduction) in different theoretical contexts. See also Ottaviano (2014)
for a discussion of this approach and Balistrieri et al. (2011) for a disentangling between the Armington and Melitz models with
regard to the welfare consequences of certain thought experiments.
βx,s
Q
l
l
7 One might imagine ω l as equal to B
and βx,s referring to input x’s cost and share (in country
, with wx,s
s
x∈X wx,s /βx,s
P
l, sector s), respectively, (with
β
=
1)
and
B
denoting
the
bundle of parameters associated with the Cobb-Douglas
x,s
x∈X
5
l
max(m)ls is referred to as the ‘exogenous marginal cost cutoff’ in country l-sector s. The support [0, max (m)s ]
varies across sectors and countries.8 γs is the sector-specific shape parameter.
Firms independently maximise the profits earned in different destination countries. Exporting firms incur
a per-unit trade cost, encompassing quantity-related trade barriers. For each unit delivered from country l to
country h, τslh > 1 units have to be shipped. In the CES constant markup case (see Appendix B), exporting
firms also bear a fixed export cost Fslh ≡ ωsh ξslh fshh , with ξslh > 1.
A firm, wherever it is located, can serve market h provided that its cost including delivery does not exceed
the marginal cost, inclusive of trade frictions, faced by a producer in country h-sector s which is just indifferent
between serving its local market or not. Let mhh
s denote the level of such ‘domestic cutoff cost’ (in country
h-industry s).
Unlike max(m)ls , the cutoff level mhh
is endogenously determined by the model and varies over time.
s
Because our work focuses on such changes, we require a useful expression for recovering definitive information
on mhh
s from observable data. To this end, we use, as in Equation (1), the relative export share of country l
in country h (i.e. country h’s relative share of total expenditure on imported goods from country l). Letting
ρlh
s ∈ (0, 1] measure trade openness between country l and country h in sector s, and adopting a tilde to indicate
lh
lf
that a variable is expressed in relative terms with respect to the benchmark country f (i.e. ρ˜lh
s = ρs /ρs ,
˜ slh = Rslh /Rslf ), this yields:
R
where
P
l
lh
h
˜ slh ≡ ln PTs
ln R
= ln ρ˜lh
s + ln RM Cs
lh
T
l s
(2)
Tslh ≡ Ysh is the amount of national income that residents in country h devote to sector s. The term
ln RM Csh is analogous to the multilateral resistance term in Equation (1) and is still expressed in relative
terms with respect to the benchmark country f .
As shown in Appendixes A and B, the relation in Equation (2) can be obtained within a heterogeneous firms
framework irrespective of whether variable or constant markups are assumed. However, the interpretation of
its two components varies slightly as follows.
In a variable markup setup, we have
lh −γs
ρlh
s ≡ (τs )
˜
with m
¯ hh
s =
m
¯ hh
s
m
¯ hf
s
=
mhh
s
mhf
s
and RM Csh =
m
¯˜ hh
s
P˜sh
γs +2
,
(3)
denoting average 9 marginal costs in country h-sector s, and P˜sh = Psh /Psf , with Psh
denoting the exact price index in country h (see Appendix A).
production function.
8 Underlying (probabilistic) comparative advantages, encompassing both input costs and total factor productivity, can be easily
expressed in terms of marginal costs as
f
max(m)h
s /max(m)s
f
max(m)h
j /max(m)j
. While Costinot et al. (2012) build on this intuition to test for the
role of comparative advantages in shaping the geography of trade patterns, Chaney (2008) assumes an underlying unit input
requirement distribution G(c) = cγs , with support [0, max(c)s = 1], varying across sectors but not across countries. In this case,
the exogenous marginal cost max(m)ls reduces to ωsl and only comparative cost advantages exist across countries.
9 Under the hypothesis that firms’ marginal costs are Pareto distributed, average marginal costs mhh are a direct function of
s
s
the marginal cost cutoff: mhh
= γ γ+1
mhh
s
s . Being γs sector-specific (and not country-specific), this relationship simplifies to
s
˜
m
¯ hh
˜ hh
s =m
s when relativized respect to a benchmark country.
6
In a constant markup setup, instead, we have
γs +1−s
1−s
ρlh
˜s−γs ξ˜s
s ≡τ
h
i1−s
γs
˜h
˜¯ hh
and RM Csh = (m
,
s ) /(Ps )
(4)
where the RM C component differs from that in Equation (3) due to the presence of the elasticity of substitution
s (which comes from the CES utility function) and the fact that the trade openness parameter encompasses
both variable and fixed trade costs (see Appendix B).
Apart from the above difference, the RMC growth rate is in both cases given by:
˜ lh − ∆˜
∆RM Csh = ∆R
ρlh
s
s ,
(5)
where ∆V refers to the log-difference of the generic variable V .
Equation (5) represents the basis of our empirical analysis. It suggests that changes in country h’s RMC
can be traced back to the portion of the variation in h’s share of expenditure on imports from l that is not
counterbalanced by the variation in bilateral trade costs. Given total expenditure Y˜sh , an increase in imports
from country l (T˜slh ) entails long run increases in marginal costs in country h, when it is accompanied by a
less than proportional increase in the degree of trade openness with country l (e
ρlh
s ). Assume, for example,
an increase in trade openness between Mexico and the United States. The effect of this on average marginal
costs in the US depends on the effect on the share of US expenditure on imports from Mexico. An increase
ex,U S
˜ sM ex,U S < ∆˜
in the latter, such that ∆R
ρM
, leads to decreasing long run marginal costs in the US, viss
a-vis Mexico. In fact, when trade barriers shrink or market size (i.e. total expenditures) grows in country h,
`
fiercer competition in product markets reduces h’s marginal cost cutoff and makes it harder, for firms located
anywhere, to target consumers in h. This effect sets into motion a process of firm selection in country h. Less
efficient firms are forced to shut down and their market shares are reallocated in favour of more productive
firms, thereby reducing average marginal costs in h.10
In the next section, we exploit the simplicity of the relationships in Equations (2) and (5) to retrieve
information on long-run changes in the equilibrium marginal costs associated with observed changes in the
share of expenditure on imported goods. In this exercise, the exponents in the RM C term (i.e. γs + 2 or
1 − s , depending on the model) entail merely a sectoral re-scaling.
4
Benchmark Analysis
We use different specifications to estimate Equation (2) and quantify the growth rate in Equation (5).
In our benchmark analysis, we estimate, sector by sector and separately for the periods (1981-1990) and
10 In
turn, the reduction in country h’s share of expenditure on country l’s goods will foster firm selection in country l.
7
(1997-2006), the following log-linear version of Equation (2):
lh
˜ s,t
ln R
= δ˜slh +
|{z}
ln ρ˜lh
s
δ˜sh
|{z}
+ Yt + s,t ,
(6)
ln RM Csh
in which t is the year index (with t = 1980, . . . , 1990 in the earlier period and t = 1997, . . . , 2006 in the latter
period) and Yt is a year dummy. δ˜slh = δslh − δslf is an exporter-importer fixed effect (with δslh 6= δshl - i.e.
trade costs are not symmetric), meant to capture the relative trade openness from l to h (˜
ρlh
s in Equation
(2)), δ˜sh = δsh − δsf is an importer fixed effect capturing country h’s RMC in a given sector-period, and s,t is
an error term capturing measurement error.11
Data on bilateral flows are drawn from the TradeProd database maintained by Centre d’Etudes Prospectives et d’Informations Internationales (CEPII). Unlike other databases (e.g. NBER-UN), TradeProd reports
detailed information on internal trade flows from 1980-2006, i.e. just before the trade collapse associated with
the 2007 economic crisis. This enables us to properly measure Ysh using country h’s total imports in sector s,
inclusive of internal trade flows Tshh . Data are provided in nominal dollars at the three-digit level of the ISIC
Rev.2 classification. We truncate the data at $10,000 per annual bilateral flow.12 We set the United Kingdom
as the reference country, as it has the most observations as an importer-exporter.
To obtain comparable estimates, the sample is adjusted to ensure the same number of observations in each
period. In particular, we limit the analysis to the country pairs for which information is available for every
year in both periods.13 After data cleaning, the estimation is carried out on a sample of 49 countries, covering
about 85% of world trade. For expositional purposes, we concentrate on the G20 countries when reporting
the results.
The output of the analysis consists of detailed country-sector average RMCs and trade openness indicators
for the periods 1981-1990 and 1997-2006 for all countries analysed. Using sectoral import shares as weights, we
h
\
then obtain the average growth rates of the estimated RMC (∆RM
C s ) and trade openness (∆ρˆ˜lh
s ) reported
in Table 2 (see Specification 1); all values are normalised by setting the G20 sectoral average as the mean
(values are expressed in relative terms with respect to the G20 average, which is set to zero). While the full
set of country-sector RMC growth rates and rankings is reported in Appendix D (see Tables 6 and 7), it is
worth noting that the importer fixed effect, accounting for the RMC term, proved significant in almost all the
sectoral regressions. The very few country-sector combinations in which it was insignificant have been omitted
11 Compared to equation (18) in Costinot et al. (2012), our derivation features a key difference: here the ‘revealing term’ is the
importer fixed effect.
12 This has no notable effect on the results and avoids potential distortions due to errors in data units and implausibly small
trade values.
13 This is a conservative choice meant to maintain manageable computational intensity and ensure stability of the results, by
allowing the same set of countries to be observed in every year under consideration. We also experimented with a balanced (across
countries, sectors and years) panel including all the zero observations, which were replaced, as is standard, by negligible values.
This massively increases in computational time and also makes the country rankings more sensitive to the robustness controls.
However, it does not produce notable differences in our main results. The coefficient on H*grrate imp diff in Table 4 remains,
for instance, significative and negative (−0.4888). Instead of substituting zeros with small values, we also applied the PPML
estimator suggested by Silva and Tenreyro (2006). Again, the exponential increase in computational intensity, which sometimes
rendered it impossible to obtain the estimates, was not compensated for with appreciable differences in final results.
8
from the computation of the values reported in Table 2.14
The largest decreases in RMC are estimated for Brazil and Korea, followed by Sweden, and interestingly,
Portugal and Italy. Large potential increases in RMC are estimated for the former USSR, Japan, Germany,
South Africa, China, and Argentina. The US RMC suggests positive predicted growth; this is smaller than for
China, although the estimated reduction in trade barriers is higher in the latter. Overall, RMCs are predicted
to fall in the Euro area by 10.8%, notwithstanding a slight decrease (−3.7%) in trade openness.
A notable result in Table 2 is that countries that performed well in terms of export shares in the period
under examination (e.g. China, Mexico, and India) are characterised by large increases in RMC. This is not
surprising, as it is a consequence of the selection process: when competition becomes fiercer, firm selection
is less intense if average costs are already relatively low. This is the essence of firm selection. By the same
token, the relatively large drops in RMC obtained for countries such as Italy and Portugal, usually thought
of as having seen shrinking post-1990s productivity, must be understood in terms of long-run firm selection
associated to increasing competition from low-cost countries, such as China and India.15
To further elaborate on this, we classify trade flows according to the importing and exporting countries’
income. Based on the World Bank income classes16 , we compute, for each country, growth rates of imports
from two groups of countries, low-to-middle-income and high-income, which are then used as dependent
variables in Table 4. In column (1), the estimated RMC is regressed on a dummy variable (H), taking 1 if the
importing country belongs to the high-income group and 0 otherwise, as well as on the country’s average total
imports in period 1 (total imp). In column (2), we control for the income class of the country of origin by
disaggregating between growth of imports from countries in the same income class (grrate imp same), growth
of imports from countries in the other income class (grrate imp diff ), and growth of domestic trade (grrate
domestic). Finally interaction terms are included to control for high-income countries increasing their imports
from low-to-middle-income (H*grrate imp diff ) and high-income (H*grrate imp same) countries, as well as for
high-income countries increasing their share of domestic trade (H*grrate domestic).
The first coefficient of interest is that on the H dummy, according to which high-income countries are
predicted to gain the most, in terms of RMC reduction. Examining the coefficients on the interaction terms,
the negative and significant sign on the H*grrate imp diff coefficient captures the positive role that imports
from low-to-middle-income countries play in high-income countries, negatively affecting the estimated RMC.
This effect is particularly important for understanding the large decrease in RMC reported for countries such as
Italy and Portugal. Imports from other high-income countries (i.e. H*grrate imp same) exert opposite effects
on the estimated RMC, probably because the delivery costs in these cases are not low enough to generate
14 Detailed
regression output from the sectoral estimation of Equation (6) is available upon request.
reference model points to a new long-run equilibrium in which the increased import competition associated with the
availability of larger quantities at lower prices forces firms with marginal costs above mhh
s to leave the market, thereby reducing
aggregate marginal costs and increasing the overall competitiveness of the country.
16 According to the World Bank classification, the high-income group in our sample of countries consists of Canada, the United
Kingdom, Spain, Norway, Chile, Finland, Ireland, Austria, the United States, Israel, South Korea, Singapore, Belgium, the
Netherlands, Taiwan, Saudi Arabia, New Zealand, Japan, Denmark, Australia, Sweden, Greece, Hong Kong, Poland, France,
Italy, Czech Republic, Portugal, Germany, and Switzerland. Countries in the low-to-middle-income group include Iran, India,
Argentina, Colombia, Malaysia, Mexico, Nigeria, Hungary, Romania, Egypt, the Philippines, Brazil, China, Turkey, Thailand,
Indonesia, Venezuela, and South Africa.
15 The
9
sufficient competitive pressure to trigger the selection effect. Finally, domestic market growth is insignificant
both when considered alone (i.e. variable grrate domestic) and when interacted with the indicator for highincome countries (i.e. H*grrate domestic). ). Overall, these results corroborate the idea that the selection effect
is mostly driven by increasing competition from low-to-middle-income countries rather than by competition
associated with increasing domestic trade flows or imports from other high-income countries.
As shown in Appendix C, the estimated RMCs listed in Table 2 are not correlated with the GGDC - EU
KLEMS PPI, which is used (e.g. Corcos et al., 2012) as a proxy for country-sector cost cutoffs. This suggests
caution when relating PPP indexes to contemporaneous values of cost cutoff determinants in order to capture
the selection effect of international trade.
5
Robustness
To check the robustness of our results, we estimate alternative specifications of the benchmark equation and
change the reference data and period.
In Equation (6), we rely on importer-exporter fixed effects to capture the relative trade openness from l
to h (i.e. the term ρ˜lh
s in Equation (2)). Since these are likely to include other unmeasured and invariant
factors affecting bilateral trade, we estimate an alternative specification in which the exporter-importer fixed
effect is replaced by bilateral geographical distance and other controls, such as sharing common border and
language. In this case, the estimated parameter of distance reflects the distance elasticity of relative import
shares. On the one hand, this specification should arguably reduce the risk of labelling other types of effects
˜hl
as ‘trade costs’; on the other hand, it requires imposing reciprocal trade costs (˜
ρlh
s ). The estimated
s = ρ
equation (Specification 2) is in this case:
˜ lh =
ln R
s,t
δ˜sh
|{z}
ln RM Csh
+
˜ lh
βslh X
| {z s}
+ Yt + s,t ,
(7)
ln ρ˜lh
˜hl
s =ln ρ
s
˜ slh includes bilateral distance17 , a common border dummy, a common language dummy, and dummies
where X
controlling for belonging to the Euro Area, NAFTA, and the EU-15. The estimated RMC and trade openness
growth rates are reported in Table 2, while the regression output with the income country-groups is presented
in Table 4 (see Specification 2). Results are broadly consistent with those presented in the previous section,
with a notable difference arising for the domestic trade variable, which has an effect similar to that of imports
from countries in the same income group.
As a further control, we recognise that Specifications 1 and 2 might be influenced by the multiple counting
associated with the emergence of global value chains. Indeed, while countries’ exports increasingly rely on
intermediate imports, traditional trade metrics record gross flows of goods each and every time they cross
17 Bilateral geodesic distances between the most populous cities of country l and country h are used, with inter-city distances
calculated using the great circle formula and weighted by the share of the city in the overall population. Source: CEPII - GeoDist
database.
10
borders. This results in multiple counting, which is likely to inflate Rslh for specific partner countries l. To
control for this, we use cross-sectional information on ‘domestic value added embodied in gross exports’, made
available by the OECD-WTO Trade in Value Added (TiVA) database for 40 countries and 18 industries.
Trade flows are expressed in USD million and classified based on ISIC Rev.3.18 The estimating equation
(Specification 3) is:
˜ slh =
ln R
δ˜sh
|{z}
+
ln RM Csh
δ˜slh
|{z}
+ s .
(8)
ln ρ˜lh
˜hl
s =ln ρ
s
Where ρ˜hl
s is again an exporter-importer fixed effect. The advantage of Equation (8) over Specifications
1 and 2 is that it protects the estimates from being exaggerated by ‘multiple counting’ by allocating valueadded figures to the source industries and countries. However, this does not come without costs. First, the
panel dimension is lost and we must estimate Equation (8) separately for 1995 and 2005, without year effects.
Second, because the information is cross-sectional, we have to set ρ˜lh
˜hl
s = ρ
s , as in Specification 2. Third, the
TiVA dataset does not include figures on internal trade (i.e. domestic value added embodied in country h’s
gross internal trade flows).
Results, reported in column (5) of Table 2 (see Specification 3), broadly confirm our benchmark results,
though interesting differences emerge. The range is smaller, and the formerly huge RMC decrease enjoyed
by Brazil is considerably reduced. In addition, the competitiveness losses estimated for China and Germany
shrink, while RMC decreases are now predicted for Japan and Argentina. The regression output presented
in Table 4 is in line with the benchmark analysis, with the notable difference that the coefficient of H is no
longer significative. These changes confirm the importance of controlling for multiple counting, although the
differences are likely to be driven by the absence of domestic trade.
Finally, to check to what extent the benchmark results depend on time splitting, we consider two alternative
periods. In the first one (Specification 4), we use years 1985 − 1989 and 1995 − 1999 as the early and the late
periods. The latter period lies between the implementation of the EU Single Market Program and China’s
WTO accession. In the second check (Specification 5), we again use shorter periods than under the benchmark
approach and now center the periods in 1983 and 2003 (i.e. 1981-1985 and 2001-2005), thus grouping Euro
adoption and Chinese WTO membership in the second period and remaining somewhat removed from both
the implementation of the EU Single Market Program and the economic crisis. The estimated growth rates
of RMC and trade openness are reported in Table 3, while the regression output with the income countrygroups is provided in Table 4. Income regression results are robust to these changes and confirm the strength
of the selection effect associated with increasing import competition from low-to-middle-income countries.
Some differences can be noted, however, in terms of RMC growth rates. Interestingly, comparing the last
rows of Tables 2 and 3, one might be tempted to attribute the RMC increase in Specification 5 to the trade
(creation and diversion) effects of the first phase of Euro adoption, in which the appreciation of the Euro likely
18 The
TiVA database (http://oe.cd/tiva) is a joint OECD-WTO initiative. Values are derived from OECD Input Output
Tables linked together in terms of goods by industry using the Bilateral Trade Database. The name of the variable used is ‘EXGR
DVAIND’.
11
weakened import competition from more competitive countries. This is supported by the smaller coefficient
on the H*grrate imp diff variable in Table 4.
6
Conclusions
We used data on actual trade flows to estimate theory-based changes in the degree of cost competitiveness,
expressed in terms of long-run growth in country-sector average marginal costs. The key variables of the
estimating equation countries’ expenditure shares on imports and the degree of trade openness.
The empirical analysis, which required only data on country-sector bilateral trade flows (including internal
trade flows), was carried out using CEPII-TradeProd data for 49 countries, accounting for about 85% of
world trade, to estimate a modified gravity equation in which the country-sector specific RMCs were identified
through fixed effects.
Bearing in mind all the limitations of the fixed-effect specification, one robust result arising from this
analysis is that marginal costs are predicted to decrease relatively more in high-income countries than in
low-to-middle-income countries, due to increasing competition from the latter group.
This is in line with the logic of the selection effect: firm selection is more intense where competition is fiercer
and marginal costs are higher. Accordingly, aggregate marginal costs decrease more in high-cost countries,
due to the increasing competition associated with importing increasing quantities and varieties of goods at
low prices from low-cost countries.
12
References
[1] Anderson, J.E. and van Wincoop, E. (2003). ‘Gravity with gravitas: a solution to the border puzzle’,
American Economic Review, vol. 93, pp. 170-192.
[2] Anderson, J.E. and van Wincoop, E. (2004). ‘Trade Costs’, Journal of Economic Literature, vol. XLII,
pp. 691751.
[3] Arkolakis C., Costinot, A. and Rodr´ıguez-Clare, A. (2012a). ‘New trade models, same old gains?’, American Economic Review, vol. 102, pp. 94-130.
[4] Arkolakis C., Costinot, A., Donaldson, D. and Rodr´ıguez-Clare, A. (2012b). ‘The elusive pro-competitive
effects of trade’, Mimeo.
[5] Balistrieri, E., R. Hillberry, and T. Rutherford (2011). ‘Structural estimation and solution of international
trade model with heterogeneous firms’, Journal of Intermational Economics, vol. 83(1), pp. 95108.
[6] Bas, M., Mayer, T. Thoenig, M. (2014). ‘From micro to macro: Demand and supply-side determinants of
the trade elasticity’. Mimeo.
[7] Bernard, A., Eaton, J., Jensen, J. and Kortum, S. (2003). ‘Plants and productivity in international trade’,
American Economic Review vol. 93, pp. 1268-1290.
[8] Carlin, W., Glyn, A. and Van Reenen, J. (2001). ‘Export market performance of OECD countries: an
empirical examination of the role of cost competitiveness’, The Economic Journal, vol. 111, pp. 128-162.
[9] Chaney, T. (2008). ‘Distorted gravity: the intensive and extensive margins of international trade’, American Economic Review, vol. 98(4), pp. 1707-1721.
[10] Chor, D. (2010). ‘Unpacking sources of comparative advantage: a quantitative approach’, Journal of
International Economics, vol. 82(2), pp. 152-167.
[11] Corcos, G., Del Gatto, M., Mion, G. and Ottaviano, G.I.P. (2012). ‘Productivity and firm-selection:
quantifying the new gains from trade’, The Economic Journal, vol. 122(561), pp. 754-798.
[12] Costinot, A., Donaldson, D. and Komunjer, I. (2012). ‘What goods do countries trade? A quantitative
exploration of Ricardo’s ideas’, Review of Economic Studies, vol. 79, pp. 581-608.
[13] Costinot, A. and Rodr´ıguez-Clare, A. (2014). ‘Trade theory with numbers: quantifying the consequences
of globalization’, in (Helpman E., Rogoff K. and Gita G. eds.), Handbook of International Economics,
vol. 4, Elsevier.
[14] Eaton, J. and Kortum, S. (2002). ‘Technology, geography, and trade’, Econometrica, vol. 70(5), pp.
1741-79.
13
[15] Fadinger, H. and Fleiss, P. (2011). ‘Trade and sectoral productivity’, The Economic Journal, vol. 121(555),
pp. 958-989.
[16] Finicelli, A., Pagano, P. and Sbracia, M. (2009). ‘Trade-revealed TFP’, Banca d’Italia Working Paper,
vol. 729.
[17] Head, K., Mayer, T., and Thoenig, M. (2014). ‘Welfare and trade without pareto’. American Economic
Review, vol. 104(5), pp. 310316.
[18] Hsieh, C. and Ossa, R. (2011). ‘A global view of productivity growth in China’, NBER Working Paper
Series, 16778.
[19] Levchenko, A. and Zhang, J. (2011). ‘The evolution of comparative advantage: measurement and welfare
implications’, NBER Working Paper Series, 16806.
[20] Melitz, M. (2003). ‘The impact of trade on intra-industry reallocations and aggregate industry productivity’, Econometrica, vol. 71, pp. 1695-1725.
[21] Melitz, M. and Ottaviano, G.I.P. (2008). ‘Market size, trade, and productivity’, Review of Economic
Studies, vol. 75, pp. 295-316.
[22] Nocco, A., Ottaviano, G.I.P. and Salto, M. (2014). ‘Monopolistic competition and optimum product
selection’, American Economic Review, vol. 104(5), pp. 304-09.
[23] Ottaviano, G.I.P., Tabuchi, T. and Thisse, J.F. (2002). ‘Agglomeration and trade revisited’, International
Economic Review, vol. 43, pp. 409-436.
[24] Ottaviano, G. I. (2014). ‘European integration and the gains from trade’, in (H. Badinger and V. Nitsch,
eds.), Handbook of the Economics of European Integration, Routledge.
[25] Pavcnik, N. (2002). ‘Trade liberalization, exit, and productivity improvements: evidence from Chilean
plants’, Review of Economic Studies, vol. 69, pp. 245-276.
[26] Piermartini, R. and Teh, R. (2005). ‘Demystifying modelling methods for trade policy’, WTO Discussion
Paper, vol. 10.
[27] Romalis, J. (2007). ‘NAFTAs and CUSFTAs impact on international trade’, Review of Economics and
Statistics, vol. 89(3), pp. 416435.
[28] Silva, J. S., Tenreyro, S. (2006). ‘The log of gravity’, The Review of Economics and statistics, 88(4),
641-658.
[29] Shikher, S. (2011). ‘Capital, technology, and specialization in the neoclassical model’, Journal of International Economics, vol. 83(2), pp. 229-242.
14
[30] Waugh, M.E. (2009). ‘International trade and income differences’, Staff Report 435, Federal Reserve Bank
of Minneapolis.
[31] World Trade Organization (2013). World Trade Report 2013: Factors shaping the future of world trade,
Geneva: WTO.
15
16
Tables
17
Food products
Beverages
Tobacco
Textiles
Apparel
Leather products
Footwear
Wood products except furniture
Non-metallic furniture
Paper products
Printing and publishing
Industrial chemicals
Other chemicals
Petroleum refineries
Rubber products
Plastic products
Pottery china earthenware
Glass products
Other non-metal min. prod.
Iron and steel
Non-ferrous metals
Fabricated metal products
Machinery except electrical
Electric machinery
Transport equipment
Prof. and scient. equipment
SECTOR
SECTOR
ABBREV.
FD
BV
TB
TX
AP
LT
FT
WO
FU
PA
PP
IC
OC
PE
RU
PL
PT
GL
NM
ST
NF
MP
MA
EM
TR
PS
13790
7120
1960
13650
9890
8600
6070
8420
6680
9790
9090
13570
11730
5180
9880
9110
6010
8650
8440
9640
9790
12560
13730
12850
10800
11580
# obs
% of tot. trade
(1980-1990)
92.0
82.2
82.6
94.2
96.1
88.4
95.6
97.1
92.2
95.4
86.4
90.7
88.0
71.8
91.8
93.1
79.6
92.6
93.0
90.6
88.9
93.3
92.2
94.7
94.5
90.5
Table 1: Data description.
% of tot. trade
(1997-2006)
95.1
86.8
79.3
93.8
96.9
82.1
86.4
95.8
70.7
94.3
88.3
90.6
95.5
65.0
86.5
92.1
76.2
88.7
88.4
79.4
89.5
94.8
95.5
92.9
97.2
93.6
18
COUNTRY
Brazil
Korea
Portugal
Sweden
Italy
Denmark
Ireland
Indonesia
Turkey
Finland
UK
Netherlands
Belgium
Greece
France
India
Austria
Mexico
Spain
USA
Australia
Canada
Argentina
China
South Africa
Germany
Japan
Fm USSR
Euro Area
Specification 1 (benchmark)
h
\
CODE
∆RM
Cs
∆ρˆ
˜lh
s
BRA
-2.456
0.128
KOR
-1.069
0.499
PRT
-1.045
0.723
SWE
-0.790
0.370
ITA
-0.660
0.019
DNK
-0.554
0.091
IRL
-0.420
0.132
IND
-0.382
-0.070
TUR
-0.321
0.628
FIN
-0.267
0.036
GBR
-0.199
0.024
NLD
-0.084
-0.137
BEL
-0.008
-0.084
GRC
0.032
0.346
FRA
0.140
-0.422
IND
0.150
-0.130
AUT
0.192
-0.247
MEX
0.202
-0.290
ESP
0.229
0.001
USA
0.262
-0.124
AUS
0.400
-0.064
CAN
0.562
-0.130
ARG
0.614
-0.052
CHN
0.663
0.053
ZAF
0.673
-0.189
DEU
0.701
-0.778
JPN
0.905
-0.433
SUN
2.872
-0.362
EUR
-0.108
-0.037
Specification 2
h
ˆ
\
COUNTRY
∆RM
Cs
∆ρˆ
˜lh
˜hl
s ≡ ∆ρ
s
Brazil
-2.300
-0.028
Italy
-0.649
0.008
Korea
-0.528
-0.042
Denmark
-0.515
0.052
Sweden
-0.469
0.050
Ireland
-0.407
0.119
Portugal
-0.398
0.076
Indonesia
-0.369
-0.083
France
-0.295
0.013
Finland
-0.281
0.050
Netherlands
-0.240
0.019
UK
-0.229
0.054
Belgium
-0.125
0.033
Austria
-0.080
0.026
Mexico
-0.075
-0.012
Germany
-0.067
-0.009
India
0.047
-0.027
USA
0.129
0.009
Spain
0.190
0.039
Greece
0.311
0.067
Australia
0.329
0.007
Turkey
0.372
-0.065
Canada
0.434
-0.001
South Africa
0.481
0.003
Japan
0.535
-0.063
Argentina
0.611
-0.049
China
0.786
-0.069
Fm USSR
2.534
-0.023
Euro Area
-0.186
0.040
Specification 3
h
ˆ
\
COUNTRY
∆RM
Cs
∆ρˆ
˜lh
˜hl
s ≡ ∆ρ
s
Brazil
-0.806
0.042
Indonesia
-0.684
0.105
Portugal
-0.306
-0.376
Argentina
-0.283
0.056
Japan
-0.253
0.128
Italy
-0.187
0.054
Netherlands
-0.172
-0.028
Korea
-0.168
-0.171
Denmark
-0.163
-0.044
Ireland
-0.043
-0.533
France
0.010
-0.020
Greece
0.011
-0.044
Belgium
0.020
0.310
UK
0.029
-0.003
Germany
0.034
0.130
China
0.045
0.044
Finland
0.050
0.167
USA
0.143
0.007
India
0.148
0.027
Spain
0.154
0.053
Sweden
0.173
0.073
Australia
0.266
0.122
South Africa
0.282
-0.222
Canada
0.302
0.185
Austria
0.315
-0.258
Mexico
0.325
-0.126
Turkey
0.411
-0.036
Fm USSR
1.117
0.195
Euro Area
-0.010
-0.050
Table 2: Growth Rates of RMC and Trade Openness from the Early Period (1981-1990) to the Late Period (1997-2006).
19
COUNTRY
Brazil
Korea
Portugal
Sweden
Italy
Denmark
Ireland
Indonesia
Turkey
Finland
UK
Netherlands
Belgium
Greece
France
India
Austria
Mexico
Spain
USA
Australia
Canada
Argentina
China
South Africa
Germany
Japan
Fm USSR
Euro Area
CODE
BRA
KOR
PRT
SWE
ITA
DNK
IRL
IND
TUR
FIN
GBR
NLD
BEL
GRC
FRA
IND
AUT
MEX
ESP
USA
AUS
CAN
ARG
CHN
ZAF
DEU
JPN
SUN
EUR
Specification 1 (benchmark)
h
\
∆RM
Cs
∆ρˆ
˜lh
s
-2.456
0.128
-1.069
0.499
-1.045
0.723
-0.790
0.370
-0.660
0.019
-0.554
0.091
-0.420
0.132
-0.382
-0.070
-0.321
0.628
-0.267
0.036
-0.199
0.024
-0.084
-0.137
-0.008
-0.084
0.032
0.346
0.140
-0.422
0.150
-0.130
0.192
-0.247
0.202
-0.290
0.229
0.001
0.262
-0.124
0.400
-0.064
0.562
-0.130
0.614
-0.052
0.663
0.053
0.673
-0.189
0.701
-0.778
0.905
-0.433
2.872
-0.362
-0.108
-0.037
Specification 4 (85-89 vs 95-99)
h
\
∆RM
Cs
∆ρˆ
˜lh
s
-2.501
0.064
-0.909
0.306
-0.546
-0.050
-0.533
0.259
-0.443
0.022
-0.413
0.553
-0.281
0.305
-0.254
-0.032
-0.239
-0.056
-0.221
-0.068
-0.216
0.065
-0.200
-0.228
-0.199
0.305
-0.079
-0.107
-0.052
0.156
-0.031
-0.064
-0.010
-0.062
0.102
0.034
0.130
-0.601
0.144
-0.108
0.254
0.028
0.302
-0.428
0.334
-0.480
0.381
-0.442
0.498
0.207
1.123
0.161
1.197
-0.017
2.716
0.382
-0.211
0.009
COUNTRY
Brazil
Portugal
Italy
Finland
Sweden
Turkey
Mexico
UK
Ireland
France
USA
Denmark
Spain
Netherlands
Greece
Germany
Austria
Japan
Korea
Belgium
Australia
Indonesia
India
Canada
China
South Africa
Argentina
Fm USSR
Euro Area
Specification 5 (81-85 vs 01-05)
h
\
∆RM
Cs
∆ρˆ
˜lh
s
-2.750
0.109
-1.015
0.360
-0.961
0.302
-0.925
0.643
-0.859
0.489
-0.712
0.409
-0.676
0.432
-0.635
0.138
-0.564
-0.081
-0.470
0.252
0.032
0.424
0.090
0.372
0.144
-0.268
0.165
-0.180
0.199
-0.102
0.225
-0.401
0.305
-0.360
0.313
0.295
0.403
-0.618
0.426
0.376
0.574
-0.860
0.622
-0.667
0.666
-0.489
0.718
-0.150
0.775
-0.347
0.795
-0.426
1.265
-0.271
4.021
-1.371
0.035
-0.127
COUNTRY
Brazil
Indonesia
Korea
Argentina
Sweden
Ireland
Portugal
Denmark
Italy
UK
Spain
Greece
Belgium
Austria
South Africa
India
Finland
Turkey
Netherlands
Canada
France
Germany
Mexico
Australia
USA
Japan
China
Fm USSR
Euro Area
Table 3: Growth Rates of RMC and Trade Openness in different periods.
20
h
-0.124*
(0.0583)
653
0.290
0.214**
(0.0787)
0.215***
(0.0235)
-0.184***
(0.0206)
-0.220**
(0.0785)
0.000244
(0.000310)
0.184***
(0.0206)
1.78e-09*
(7.02e-10)
-0.225**
(0.0700)
(4)
Specification 2
0.241*
(0.114)
240
0.0714
-0.753
(0.638)
-1.733**
(0.643)
0.843**
(0.267)
1.732**
(0.643)
-0.00000134
(0.00000490)
-0.207
(0.127)
(5)
Specification 3
\
Dep. Variable: ∆RM
C s . Estimation method: OLS. Robust standard errors in parentheses.
* p < 0.05, ** p < 0.01, *** p < 0.001
N
R2
-0.0554
(0.0989)
653
0.143
constant
0.124
(0.0969)
653
0.0453
0.142
(0.133)
H*grrate domestic
0.192*
(0.0950)
653
0.0283
0.279***
(0.0398)
H*grrate imp same
-0.142
(0.133)
-0.166***
(0.0350)
-0.00564
(0.0113)
-0.0000121
(0.000526)
0.166***
(0.0349)
4.00e-09***
(1.19e-09)
-0.381**
(0.119)
(3)
H*grrate imp diff
grrate domestic
0.00135**
(0.000469)
grrate imp same
3.63e-09**
(1.25e-09)
0.00237
(0.00145)
3.54e-09**
(1.26e-09)
total imp
-0.395***
(0.116)
(2)
grrate imp diff
-0.434***
(0.114)
H
(1)
Specification 1 (benchmark)
Table 4: RMC and Income Country-Groups.
0.258*
(0.115)
592
0.0882
-0.462
(0.275)
0.111***
(0.0314)
-0.0390*
(0.0174)
0.389
(0.271)
0.000358*
(0.000172)
0.0384*
(0.0174)
3.03e-09
(3.77e-09)
-0.601***
(0.134)
(6)
Specification 4
0.0234
(0.138)
592
0.158
0.528
(0.330)
0.186***
(0.0377)
-0.123***
(0.0209)
-0.626
(0.325)
0.000134
(0.000206)
0.122***
(0.0209)
1.63e-08***
(4.53e-09)
-0.468**
(0.161)
(7)
Specification 5
Appendix
A
Derivation of the RMC index with variable markups
The derivation in this section builds on the variable markup model of Melitz and Ottaviano (2008), along
the lines developed by Corcos et al. (2012). Unlike the latter, however, we do not include an outside good,
whereas we do assume a two-tiered utility function. These changes enable us to obtain a gravity-type equation
in which total expenditure, not population as in the above models, is used to account for the size of the
importing country.
Consider S sectors (indexed s = 1, . . . , S) active in M countries (indexed l = 1, . . . , h, . . . , M ). Each
country-industry is endowed with a given amount of labour Lls and capital Ksl , with the output of each
industry being horizontally differentiated in a large (continuum) set of varieties (indexed by i ∈ Θs ).
Firms compete in a monopolistic market and each variety is supplied by one and only one firm. The
marginal cost faced by a generic firm is
mls (c) = ωsl c.
(9)
where ωsl denotes the unit input cost faced by firms active in country l-sector s. The unit input requirement
c (i.e. inverse ‘total factor productivity’) is firm-specific and identifies the firm.19
National markets are segmented but firms can export, Because production functions have constant returns
to scale, firms independently maximise the profits earned in different destination countries. Exporting firms
incur a per-unit trade cost, encompassing quantity-related trade barriers. For each unit delivered from country
l to country h, τslh > 1 units must be shipped. Moreover, we also allow for costly trade within a country, with
τslh > τsll ≥ 1.
To start producing, each firm has to bear a sunk cost Fsl = ωsl fsl . At this stage, firms are only partially aware
of their marginal costs: while the exogenous country-sector specific cost ωsl is known ex-ante, c is revealed
only once hthe sunki costs hare paid.
iγs This phase is modelled as a draw from a known Pareto distribution
γs
mls (c)
l
c
Gls (m) = max(m)
, with the support [0, max (m)s ] varying across sectors and countries,
= max(c)
l
l
s
s
where max(m)ls is referred to as the ‘exogenous’ marginal cost cutoff in country l-sector s.
Consumers maximise a ‘two-tiered’ utility function. In the first step, they allocate a fraction σsl of their
income Y (i)l to goods produced in each sector according to
U (i)ls =
Y
σl
u(i)ls s
with
s
X
σsl = 1.
(10)
s
In the second step, consumers allocate σsl Y (i)l among the different varieties in sector s by maximising the
following quasi-linear utility function with quadratic sub-utility (Ottaviano et al., 2002):
u(i)ls
Z
=α
dls (i)di
i∈Θs
1
− υs
2
Z
i∈Θs
l 2
1
ds (i) di − η
2
Z
!2
dls (i)di
(11)
i∈Θs
subject to
Z
pls (i) qsl (i)di = σsl Y (i)l
(12)
i∈Θs
where dls (i) represents the individual consumption level of variety i of good s. The demand parameters α,
βx,s
Q
l /β
l
and βx,s referring to input x’s cost and share
might imagine ωsl being equal to BP x∈X wx,s
, with wx,s
x,s
(in country l, sector s), respectively, (with
β
=
1)
and
B
denoting
the bundle of parameters associated with the
x∈X x,s
Cobb-Douglas production function.
19 One
i
η, and υs are all positive. The parameter υs indexes the degree of product differentiation among different
varieties of good s.20
With this preference structure, marginal utilities are bounded, and utility maximisation yields the following
expression for the individual demand of a generic variety i
λls [max(p)ls − pls (i)]
υs
(13)
σsl Y (i)l
,
Nsl max(p)ls p¯l1,s − p¯l2,s
(14)
dls (i) =
where the Lagrange multiplier λls amounts to21
λls = υs
and where max(p)ls = (α − υs Dsl )/λls denotes the price level, above which the demand for a generic variety in a
given country-sector is positive. p¯l1,s and p¯l2,s represent the first and second moments of the price distribution
of the Nsl varieties consumed in the country, respectively . As well as the number of varieties actually consumed
in country l, Nsl also represents the number of firms serving country l-sector s.22
Let us now use h to refer to the destination market. Only those firms with sufficiently advantageous cost
draws to sell to market h at a price below max(p)hs earn non-negative profits and can afford to serve that
h
hh
market. Let mhh
denote the marginal cost, inclusive of trade friction, faced by a producer in
s = ms (c)τs
country h-sector s that is indifferent between serving the local market and not. Then, the zero profit condition
h
mhh
s = max(p)s holds true. As a consequence, a firm, wherever located, can serve market h provided that its
delivered cost does not exceed mhh
s . In other words, firm c producing in country l is able to target market h
hh
lh l
when τs ms < ms and cannot do so when τslh mls (c) > mhh
s . It is indifferent between serving and not serving
hh
measures
the
domestic ‘cutoff cost’ in country h-industry s. The
.
Thus,
m
market h when τslh mls (c) = mhh
s
s
lh
hh
lh
export cutoff is measured by ms = ms /τs .
From profit maximisation, the aggregate demand and aggregate price for the variety sold in country h by
firm c, which has its production based in country l, are respectively given by
qslh (c) =
λhs Lh hh
[ms − mlh
s (c)]
2υs
and
plh
s (c) =
1 hh
[m + mlh
s (c)],
2 s
(15)
lh l
h
where mlh
s (c) = τs ms (c) and L is the population of the destination country.
l
to denote the number of entrants in country l-sector s, aggregate exports of
Given this and using NE,s
good s from country l to country h can be obtained by solving:
l
Tslh = NE,s
Z
mlh
s
lh
l
l
plh
s (c) qs (c) d ms (c)/ max(m)s
γs
,
(16)
0
lh
hh
lh
Using the above equations for qslh (c) and plh
s (c) (and considering that ms = ms /τs ), the solution yields
hh h γs +2 h
l
Tslh = Υ1,s NE,s
[max(m)ls ]−γs ρlh
ms /Ps
Ys
s
20 The
(17)
degree of product differentiation increases with υs , as consumers give increasing weight to the distribution of consumption
levels across varieties. In contrast to the standard Melitz-Ottaviano framework, α and η are assumed sector invariant to ensure
that consumers always allocate a given expenditure share σsl to each sector. Note that assuming exogenously given σsl has no
effect on our final goal, as the assumption does not imply that σsl is also time-invariant.
21 To derive the expression for λl , substitute (13) into the budget constraint (12).
s
R
22 Utility maximisation yields dl (i) = (α − λl pl (i) − ηD l )/υ , where D l =
l
s
s
s s
s
s
i∈Θs ds (i)di is the total consumption of good s.
l
l
ds (i) drops to zero at any price level below max(p)s . In this setting, each firm has a negligible impact on the market and does not
compete directly with other firms. However, given the demand structure, firms interact indirectly through an aggregate demand
effect, as the total output of the industry influences firm profits.
ii
in which
1
Psh ≡ Nsh max(p)hs p¯h1,s − p¯h2,s γs +2
and
Ysh ≡ σsh Lh Y (i)h ,
(18)
lh −γs
and where Υ1,s ≡ 2(γs1+2) is a bundling sectoral parameter, ρlh
∈ (0, 1] is a measure of trade
s ≡ (τs )
h
openness between country l and country h-sector s, Ps is the exact price index in country h, and Ysh is the
amount of national income (Y h = Lh Y (i)h ) that residents of country h spend in sector s.
Exogenous (i.e. max(m)ls ) and endogenous (i.e. mhh
s ) marginal cost cutoffs play different roles in Equation
(17). On the one hand, a high max(m)ls (i.e. high input costs and/or low tfp) reduces the export performance
of country l by weakening its comparative advantage, as in traditional trade models. On the other hand, a
l
relatively high mhh
s in the destination country facilitates exports from l. Unlike max(m)s , the endogenous
hh
cutoff ms varies over time.
The aggregate export share of country l in country h-sector s equals the expenditure share of country h in
P
imported goods from country l-sector s. Since Ysh = l Tslh , this can be expressed as:
hh h γs +2
T lh
l
[max(m)ls ]−γs ρlh
ms /Ps
Rslh = P s lh = Υ1,s NE,s
s
l Ts
(19)
Equation (19) suggests that country l’s export share in country h can be used to retrieve definitive information on the marginal cost cutoff mhh
s . To this end, note that the terms in Equation (19) are specific to both the
h γs +2
l
lh
)
), or the latter ( mhh
origin and destination country (ρs ), to the former only ([max(m)ls ]−γs NE,s
s /Ps
only. Thus, we can use country l’s export share in country f (i.e. Rslf ≡ Tslf /Ysf ) to rid Equation (19) of
l
. We thus obtain the following prediction for the relative export
the hard to observe term [max(m)ls ]−γs NE,s
share of country l in country h (i.e. country h’s relative share of total expenditure on imported goods from
country l):
hh γs +2
˜¯ s
m
h
˜ slh = RM Csh ρ˜lh
R
with
RM
C
=
,
(20)
s
s
P˜sh
where a tilde has been used to indicate that a variable is expressed in relative terms with respect to the
f
h
f f ˜h
hh
lf
lh
lf ˜ lh
lh
˜
¯ hh
˜ hh
benchmark country (i.e. ρ˜lh
s =
s = ms /ms , Ps = ps /ps ). The term m
s = ρs /ρs , Rs = Rs /Rs , m
hh
hh
m
¯s
ms
hh
= mhf = m
˜ s denotes average marginal costs in country h-sector s. As described in footnote 9, if firms’
m
¯ hf
s
s
γs
hh
hh
marginal costs are Pareto distributed, average marginal costs mhh
s are in fact a function ms = γs +1 ms of the
˜¯ hh
˜ hh
marginal cost cutoff. Because γs varies across sectors but not across countries, this implies that m
s .
s =m
Equation (20) expresses the (observable) export share as a function of trade openness and marginal cost
cutoffs, both relative to the benchmark country f . Note that the (endogenous) number of firms entering the
market (i.e. the size of the exporting country) cancels out, as does the unmeasurable exogenous marginal cost
component.
To highlight the determinants of the RMC measure, note that the export share of country l in country h
iii
can be expressed as23
˜ slh
R
P
ρ˜lh
= s
˜ hs
Λ
˜ hs
Λ
with
=P
j
j
NP,s
j
j NP,s
jj
ρjh
s /ρs
−γs
[mjj
s ]
jj
ρjf
s /ρs
−γ
[mjj
s ] s
.
(24)
Thus, whether country l is more competitive in market h than in country f -sector s depends on two factors.
First, it depends positively on country l’s relative degree of trade openness with the two countries, the export
lf
˜h
share being higher in country h if ρlh
s > ρs . Second, it depends negatively on the term Λs , which dictates
that the export share in h is higher if the degree of competition is in market h is lower, compared to market
f.
From (20) and (24) it follows that
hh γs +2
˜¯ s
m
1
=
.
(25)
h
˜
˜
Ps
Λhs
Equation (25) highlights that the ultimate determinant of cross-country marginal cost variability in the
model is market competition. Marginal costs are lower in country-sectors in which relatively more producers
j
jh
(higher NP,s
), with lower marginal costs (lower mjj
s ), compete in a more open (higher ρs ) market.
B
RMC with constant markups
In this section, we show that an expression similar to Equation (20) can be derived within a framework with
constant (and thus exogenous) markups. This framework is identical to that of the previous section except
for the way in which consumers allocate their income across varieties (i.e. the second stage of the utility
maximisation process), which is now based on a CES utility function, and the presence of a fixed cost for
exporting. The resulting model is a multi-country, multi-sector version of Melitz (2003), in the spirit of Chaney
(2008), which is essentially a heterogeneous firm version of the Fadinger and Fleiss (2011) representative firm
setup.
While the first step of utility maximisation is still described by (10), let us assume, for the second stage,
that consumers maximise sub-utility
u(i)ls =
23 To
Z
i∈Θls
dls (i)
s −1
s
s
! −1
s
,
(26)
obtain Equation (24), first solve
p¯h
1,s =
X
l
l
NE,s
Z
mlh
s
γs
l
l
plh
s (c) d ms (c)/ max(m)s
and
p¯h
2,s =
0
l
l
remembering that NE,s
= NP,s
X
l
NE,s
l
ll
mll
s /τs
max(m)ls
−γs
Z
mlh
s
γs
2
l
l
[plh
s (c)] d ms (c)/ max(m)s
(21)
0
l
l
, with NE,s
and NP,s
denoting the number of entrants and producers in country
l-sector s, respectively. The resulting expression for the price index in (18) is

Psh
X j
1
γs +2
=
[mhh
NP,s
s ]
Υ1,s
j
ρjh
s
ρjj
s
!

−γs 
[mjj
s ]
1
γs +2
.
(22)
Using (22) to substitute for Psh in Equation (19), the aggregate export share of country l in country h-sector s can be rewritten
as:
!
lh 1
X j
ρjh
s
lh
l
h
−γs
ll −γs ρs
Rs = NP,s [ms ]
with
Λs =
NP,s
[mjj
.
(23)
s ]
h
ρll
ρjj
s Λs
s
j
Dividing by the equivalent expression for Rslf yields Equation (24).
iv
subject to
Z
pls (i) qsl (i)di = σsl Y (i)l ,
(27)
i∈Θs
where Θls is the set of available varieties in sector s-country l and s is the elasticity of substitution among
varieties. The associated demand function can be written as
dls (i)
[pl (i)]−s σsl Y (i)l
= s
[Psl ]1−s
with
Psl
1
! 1−
s
Z
=
i∈Θls
[pls (i)]1−s di
.
(28)
From profit maximisation, we know that
l
pll
s (c) = ms (c)
s
s − 1
ll
lh
and, since plh
s (c) = ps (c) τs ,
l
lh
plh
s (c) = ms (c) τs
(29)
s
.
s − 1
(30)
In contrast to in the variable markup case, let us assume that firms in country l face a fixed cost Fslh ≡
ωsh ξslh fshh , with ξslh > 1, when targeting consumers in country h. The following condition of zero operating
profits must be satisfied
πslh (c) =
1
s
s
s − 1
1−s mls (c) τslh
Psh
1−s
σsh Y h − Fslh = 0,
(31)
where Y h = Y (i)h Lh is country h’s national income.
From Equation (31) together with the analogous zero profit condition for domestic sales in country h (i.e.
hh
πi,s ) it is possible to derive a relationship between the export cutoff in the exporting country (mlh
s ) and the
hh
domestic cutoff in the importing country (ms ) as
lh
mlh
s = ξs
1
1−
s
mhh
s .
l
Aggregate exports from country l to country h are given by NP,s
(32)
R mlh
s
0
lh
plh
s (c) qs (c) d
mls (c)
max(m)ls
γs
, where
l
NP,s
denotes the number of firms producing in country l-sector s. Using Equations (30) and (32) to solve the
integral, the export share of country l in country h-sector s can be written as:
γs −s +1
T lh
γs −s +1
l
lh 1−lh
s
Rslh = P s lh = Υ2,s NP,s
[max(m)ls ]−γs ρlh
[mhh
[Psh ]s −1 ,
s ]
s [ξs ]
l Ts
(33)
1−s
γs
s
with Υ2,s ≡ γs −
. Under the regularity condition that γs > 1 − s , Equation (33) can be used
s −1
s +1
to derive an expression that is estimationally equivalent to Equation (20). To this end, comparing country l’s
exports to a reference country f , we obtain
h
i1−s
hh γs
h
˜ lh = (m
˜
˜
R
¯
)
/(
P
)
%˜lh
s
s
s
s ,
|
{z
}
(34)
RM Csh
γs +1−s
where %˜lh
˜s−γs ξ˜s 1−s is a measure of trade openness, encompassing variable and fixed frictions.
s =τ
Equation (34) shows that the estimation strategy in Equation (6) is broadly consistent with a constant
markup setup, although the interpretation is slightly different, due to the presence of fixed export costs.
v
To highlight the marginal cost determinants, we proceed, as in Section A, by noting that the export share
of country l in country h can be expressed as24
P
j
jh
j −γs
j NP,s %s [max(m)s ]
h
˜
.
with Ψs = P
j
jf
j −γs
j NP,s %s [max(m)s ]
lh
˜ slh = %˜s
R
˜ hs
Ψ
(38)
This formulation is similar to Equation (24) and can be used together with Equation (34) to show that
γs
˜¯ hh
(m
s )
P˜ h
1−s
=
s
1
.
˜h
Ψ
s
(39)
Thus, the interpretation of the estimated RMC is very similar to in a variable-markup framework. Crosscountry differences in marginal costs are still driven by the level of market competition. However, unlike the
variable-markup case, the degree of trade openness is decreased by the presence of fixed export costs and it is
ultimately the exogenous, not endogenous, marginal cost cutoff that is important in Equation (39).
C
Correlations
In this section we analyse the correlation of our estimated RMC with some key indicators suggested by the
literature.
Producer prices. Ricardian frameworks with intra-industry heterogeneity feature a close relationship between productivity (or marginal costs) and producer prices. In fact, under the Pareto distributional assumption, average prices in country h are in fact a very simple function of the marginal cost cutoff: p¯hs =
2γs +1
hh 25
Corcos et al. (2012) take advantage of this relationship using country-sectoral PPIs, drawn
2(γs +1) ms .
from the GGDC - EU KLEMS database, to proxy for the cutoff.26 Under the hypothesis that economies
are observed at their long-run equilibrium, the expected correlation between our RMC measure and relative
producer prices in the GGDC - EU KLEMS database is positive. However, as already highlighted in Section
2, the measurement error associated with PPIs is reasonably high and the adjustment process due to the selection effect is likely slow. Therefore, it is not surprising that we find no correlation between contemporaneous
values of RMC and Producer Prices in Table 5 where, since the PPI is available only for 1997, we estimate
and include the 1996 − 1998 average of our RMCs to examine contemporaneous correlations (Spearman’s rank
24 To
obtain Equation (38) solve

Psh
=
N
X
j=1
j
NP,s
Z
mjh
s
1−s
[pjh
d
s (c)]
0
mjs (c)
max(m)js
1
!γs  1−
s

(35)
to obtain the following expression for the price index:
1
1−s
γs −s +1
Psh = Υ2,s [mhh
Ψh
s
s ]
with
Ψh
s =
X
j
j −γs
NP,s
[%jh
.
s ] [max(m)s ]
(36)
j
This equation can be used in (33) to express the aggregate export share of country l in country h-sector s as
l
Rslh = NP,s
[max(m)ls ]−γs %lh
s
1
.
Ψh
s
(37)
Dividing by Rslf yields Equation (38).
25 The
R mlh
l
lh
l
l
s
expression for p¯h
¯h
plh
s is obtained by solving p
s ≡ 1/Gs (ms ) 0
s (m)dGs (m) under the hypothesis that Gs (m) is
Pareto.
26 The same measure is also used by Costinot et al. (2012) in a robustness check.
vi
correlations are reported).27 . This result suggests caution when relating PPP indexes to contemporaneous
values of cost cutoff determinants in order to capture the selection effect of international trade: economies are
hardly unlikely to be observed at their true long-run equilibrium.
Unit labour costs, labour productivity, and export shares. Countries’ export performance is crucially
affected by ‘cost-competitiveness’ (Carlin et al., 2001). This makes it interesting to examine the extent to
which our country-sector rankings relate to those that can be obtained through conventional measures of cost
competitiveness, on the one hand, and to countries’ export shares, on the other.
We therefore consider the correlation of the RMC with labour productivity and with a measure of Relative Unit Labour Costs (RULC) obtained (Carlin et al., 2001) by applying country-sectoral wages to labour
productivity and dividing by the G20 average. Being derived under the hypothesis that marginal costs are
ωsl c, our estimated RMC should generally be correlated with both these measures. Correlations are reported
in Table 5, where we also consider real measures of RULC and labour productivity, obtained using the above
country-sectoral (GGDC - EU KLEMS) PPIs as deflators. While we find the export share to be negatively
correlated with the contemporaneous RMC (all variables other than RMC 1997 refer to the earlier period) and
not with the subsequent RMC (RMC in 1997), the correlations with RULC and (inverse) labour productivity
are slightly positive and significant, as expected.
D
Country-Sector Results
The average values presented in Tables 2 and 3 hide substantial heterogeneity across sectors. This is shown
in Table 6, where the full set of country-sector RMC growth rates is reported, and Table 7, which reports
the country ranking of each sector in the two periods under consideration. In the US, for example, predicted
changes in RMC range from −1.87 points in the plastic products industry to 2.43 points in the rubber products
industry. Additionally, while the former USSR countries are estimated to lose the most in aggregate terms,
they report strong RMC decreases in some industries, such as tobacco and printing and publishing.
27 We
also considered the correlation with the RMC estimated for the first period, again finding no correlation.
vii
viii
RULC
# obs.
Stars denote 5% significance level.
0.1454*
340
0.1562*
340
Producer Prices 1997
# obs.
Inverse Labour Productivity (real)
# obs.
0.0736
340
Export Share
# obs.
0.1550*
340
-0.1220*
340
RMC (Specification 2)
# obs.
Inverse Labour Productivity
# obs.
0.4323*
340
RMC 1997 (benchmark)
# obs.
0.1427*
340
0.1070*
340
RMC (benchmark)
# obs.
RULC (real)
# obs.
RMC
(bench.)
1
340
-0.0939
340
-0.1103*
340
-0.0748
340
-0.0912
340
-0.0942
340
0.0303
340
0.1309*
340
1
340
RMC 1997
(bench.)
0.3486*
340
0.3445*
340
0.3285*
340
0.3215*
340
0.0116
340
-0.4454*
340
1
340
RMC
(Spec. 2)
-0.7701*
340
-0.7591*
340
-0.6694*
340
-0.6491*
340
-0.0225
340
1
340
Export
Share
0.0099
340
0.1850*
340
0.0769
340
0.2430*
340
1
340
ppp
1997
0.8991*
340
0.9344*
340
0.9814*
340
1
340
RULC
Table 5: Spearman’s Rank Correlations (Levels).
0.9313*
340
0.9308*
340
1
340
RULC
(real)
0.9804*
340
1
340
Lab. Prod.
1
340
Lab. Prod.
(real)
ix
avg
0.61
0.40
0.19
-0.01
-2.46
0.56
0.66
-0.56
-0.26
2.52
0.14
0.70
0.03
0.15
-0.38
-0.42
-0.66
0.90
-1.07
0.20
-0.08
-1.04
0.68
0.23
-0.79
-0.32
-0.20
0.26
avg
0.61
0.40
0.19
-0.01
-2.46
0.56
0.66
-0.56
-0.26
2.52
0.14
0.70
0.03
0.15
-0.38
-0.42
-0.66
0.90
-1.07
0.20
-0.08
-1.04
0.68
0.23
-0.79
-0.32
-0.20
0.26
COUNTRY
Argentina
Australia
Austria
Belgium
Brazil
Canada
China
Denmark
Finland
Fm USSR
France
Germany
Greece
India
Indonesia
Ireland
Italy
Japan
Korea
Mexico
Netherlands
Portugal
South Africa
Spain
Sweden
Turkey
UK
USA
COUNTRY
Argentina
Australia
Austria
Belgium
Brazil
Canada
China
Denmark
Finland
Fm USSR
France
Germany
Greece
India
Indonesia
Ireland
Italy
Japan
Korea
Mexico
Netherlands
Portugal
South Africa
Spain
Sweden
Turkey
UK
USA
0.67
-0.31
-0.06
-0.89
-3.23
1.56
2.26
-1.25
1.34
1.82
-0.14
0.93
0.40
0.43
-1.79
-0.40
-0.23
0.16
-2.49
-0.54
1.10
0.06
2.91
-1.66
-0.69
0.92
-1.01
-0.51
PE
1.27
-0.95
-0.31
0.21
-2.64
0.07
1.38
0.25
-0.25
2.82
-1.47
0.54
-0.70
0.92
0.56
-0.19
-0.77
0.38
-0.41
2.42
0.37
-2.97
1.28
-0.49
-1.84
-0.66
0.10
-0.43
FD
-0.09
0.08
0.77
-1.16
-2.96
0.18
1.71
-1.63
-0.93
5.31
0.42
2.61
-1.21
0.65
0.25
-0.69
-1.08
-0.32
-0.85
1.04
0.91
-1.78
0.11
-0.91
-1.57
-0.91
-0.49
2.43
RU
1.32
-0.62
1.21
-0.65
-2.14
1.08
0.15
-0.34
-0.24
1.77
-0.10
0.96
1.19
1.06
0.23
-0.20
-0.29
-3.76
1.21
1.50
-0.61
0.97
1.72
-1.34
-1.78
0.07
0.01
-1.29
BV
-0.35
2.58
0.91
0.34
-2.97
2.42
0.71
-2.40
-1.20
3.53
-1.37
-1.22
0.10
0.32
-0.17
-0.82
-2.11
-0.15
-1.18
2.38
-0.41
-1.76
1.32
-0.01
-1.86
0.67
-1.02
-1.87
PL
1.16
-3.44
1.51
-0.57
-6.28
0.56
1.05
-0.10
0.51
-5.18
-2.08
-0.03
0.90
1.21
-1.31
-0.08
-2.12
-2.26
1.49
1.04
-1.51
3.02
4.84
2.59
0.88
0.19
-1.88
-0.25
TB
1.35
-1.00
-0.70
0.94
-0.61
0.34
3.07
-1.86
-1.97
3.53
-2.40
-1.04
0.72
0.73
-0.09
-2.12
1.59
-0.38
0.61
0.77
-2.40
-1.00
2.59
0.40
1.76
1.43
-0.19
-1.02
PT
0.84
0.12
-1.28
-0.17
-3.57
0.50
0.04
0.07
-0.47
4.98
-1.03
0.42
-0.46
0.26
0.14
-0.13
-0.60
-1.42
0.13
1.44
0.44
-2.06
-0.09
-0.01
0.11
-1.05
-0.01
0.51
TX
-0.24
-0.57
1.15
0.27
-3.28
0.62
3.22
0.23
-0.73
3.94
-2.93
0.47
0.60
0.47
-0.32
-0.11
0.18
0.40
-1.07
1.02
1.60
-1.20
0.22
-0.51
0.12
0.68
-0.57
-0.01
GL
0.27
0.05
-0.27
0.09
-2.91
0.13
2.10
-0.08
-0.10
3.71
-0.84
0.36
0.91
-0.32
-1.63
-1.54
-1.58
-1.38
-1.33
2.84
-0.83
-1.84
0.02
2.22
-0.26
-1.55
0.40
0.23
AP
-1.70
1.04
0.71
-1.25
-3.47
0.81
2.73
-0.02
-0.02
1.95
-0.87
0.05
0.23
0.21
-0.60
-0.78
-1.44
1.81
0.27
2.07
-0.77
-1.80
0.42
-0.63
-1.34
-0.76
0.85
0.40
NM
0.06
-1.06
-0.10
-0.61
-2.04
-1.19
2.05
-1.35
0.10
4.49
-0.76
1.07
-0.59
-0.10
1.56
0.32
-0.57
0.38
-1.00
1.05
-2.40
1.23
2.33
-0.34
-2.56
-0.21
-0.03
0.07
LT
-0.43
2.70
0.55
0.22
-2.00
1.55
-1.27
0.28
0.70
-0.23
1.30
0.67
0.07
-0.55
-1.81
-0.03
0.54
-1.02
-0.35
1.38
0.45
0.42
0.25
0.92
-1.15
0.41
1.06
-1.12
ST
1.71
-1.29
0.36
-0.08
-3.76
0.41
-3.31
-1.01
-0.13
2.76
0.29
0.07
0.99
4.01
-0.17
0.73
-3.71
4.23
-3.51
0.44
0.76
1.75
0.07
1.93
-0.48
1.42
0.57
-1.20
FT
1.42
1.76
-4.03
-1.35
-2.38
0.90
0.69
1.48
-1.94
4.32
-0.42
-0.01
-1.02
-1.16
-0.88
-0.53
0.01
1.47
-0.95
0.06
0.35
-1.38
-0.13
-0.54
-0.88
0.75
0.88
1.76
NF
1.06
-0.30
-1.87
-2.83
-5.14
4.48
2.86
-1.20
-0.66
3.74
-2.28
-2.20
0.62
-1.46
0.06
0.23
-1.33
5.14
0.56
-0.65
-1.94
-1.83
-0.25
2.40
-0.36
1.60
-0.77
0.54
WO
0.39
0.31
-0.78
1.23
-3.08
0.18
0.86
0.10
-0.34
0.96
-0.13
0.09
1.24
-0.25
0.91
-0.03
-1.57
0.07
-0.41
1.42
-0.31
-0.52
0.90
0.17
-0.22
-0.18
-0.13
0.11
MP
3.56
-1.03
1.15
0.61
-2.93
-1.66
4.15
-1.36
-0.75
5.31
-1.22
0.34
3.42
-1.16
-2.61
-0.09
-1.31
-0.42
-2.05
-0.12
0.17
2.11
1.35
-0.64
-1.31
-0.98
-0.13
-0.98
FU
-0.96
-0.11
1.28
1.04
-1.24
0.02
1.42
-1.11
-1.04
3.24
1.08
1.47
-1.34
-0.51
-0.93
-0.23
0.26
2.18
-1.08
-0.34
-1.70
0.22
0.04
2.04
-1.56
-0.55
-0.44
1.69
MA
1.31
0.49
-0.69
2.15
-1.53
1.41
0.47
0.15
-0.02
2.41
-0.82
0.56
-1.99
-0.05
-0.29
-0.28
-0.83
-2.09
-1.68
0.52
-0.58
-0.28
2.00
0.67
0.22
0.76
0.17
-1.51
PA
0.87
0.93
0.95
0.55
-1.01
0.39
-1.03
0.26
0.06
3.16
0.23
1.74
1.13
0.23
-0.22
-0.38
-0.47
0.35
-1.81
-2.78
-0.91
-0.75
1.24
-0.05
-0.74
-0.39
-0.19
0.45
EM
2.28
1.67
0.25
-0.18
-5.25
1.60
0.45
-2.11
0.24
-4.08
-0.24
0.61
0.78
1.39
-1.64
-0.01
-0.51
-0.76
-0.44
0.13
0.32
-1.04
1.12
-0.54
-0.56
1.20
-0.28
0.78
PP
0.23
-0.13
1.19
-0.08
-2.84
-0.44
-0.85
-2.09
-0.33
5.76
2.10
1.19
0.20
-0.37
-1.77
-0.45
0.19
3.74
-1.92
-0.27
0.69
-2.31
-0.20
1.38
-0.13
-1.01
-0.24
0.60
TR
0.04
0.38
0.25
0.58
-1.19
0.63
-0.02
1.14
-0.33
2.70
0.92
-0.69
-0.98
0.06
-0.83
-1.08
-0.81
-0.94
0.24
0.21
-0.02
1.43
0.09
-0.25
-1.56
-0.38
0.55
0.57
IC
-0.60
-0.26
-0.11
-0.39
-0.17
-0.77
1.91
-1.14
-0.83
2.73
0.03
-0.75
-0.84
0.21
-0.98
-0.71
-0.50
1.59
0.42
-0.26
-0.04
0.34
0.24
0.33
-1.06
-0.11
-1.00
0.18
PS
-0.40
-0.08
-0.61
-2.51
-3.07
0.24
4.03
-1.40
-0.74
2.20
0.02
1.51
-0.73
0.23
1.76
-0.08
-0.52
1.68
-0.47
-0.19
1.41
-1.46
-0.54
0.77
-1.04
0.13
-0.18
1.32
OC
Table 6: RMC Changes from the Early Period (1981-1990) to the Late Period (1997-2006) by Country-Sector.
early
ARG
SUN
MEX
IRL
IND
TUR
AUT
AUS
DNK
BRA
FIN
BEL
CHN
PRT
NLD
GBR
SWE
ZAF
GRC
FRA
ESP
USA
CAN
DEU
JPN
IDN
KOR
ITA
FD
late
PRT
BRA
TUR
IRL
ARG
AUS
SWE
AUT
FRA
IND
FIN
GRC
DNK
BEL
ESP
SUN
NLD
GBR
MEX
USA
CHN
ZAF
CAN
JPN
ITA
DEU
KOR
IDN
gr.rate*
‐2.97
‐2.64
‐0.66
‐0.19
1.27
‐0.95
‐1.84
‐0.31
‐1.47
0.92
‐0.25
‐0.7
0.25
0.21
‐0.49
2.82
0.37
0.1
2.42
‐0.43
1.38
1.28
0.07
0.38
‐0.77
0.54
‐0.41
0.56
early
IND
DEU
TUR
ZAF
ARG
FRA
KOR
MEX
PRT
AUT
GRC
BRA
ITA
NLD
USA
CHN
BEL
GBR
FIN
IDN
AUS
IRL
CAN
ESP
DNK
SUN
SWE
JPN
*growth rates refer to the countries in the second column.
rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
BV
late
TUR
BRA
IND
DEU
JPN
FRA
USA
NLD
ITA
ZAF
ARG
BEL
ESP
SWE
PRT
KOR
AUT
GRC
AUS
CHN
MEX
FIN
GBR
IRL
IDN
DNK
CAN
SUN
gr.rate*
0.07
‐2.14
1.06
0.96
‐3.76
‐0.1
‐1.29
‐0.61
‐0.29
1.72
1.32
‐0.65
‐1.34
‐1.78
0.97
1.21
1.21
1.19
‐0.62
0.15
1.5
‐0.24
0.01
‐0.2
0.23
‐0.34
1.08
1.77
early
USA
PRT
IND
CHN
JPN
SWE
IRL
ESP
IDN
AUT
DNK
MEX
CAN
KOR
DEU
ARG
GRC
ZAF
BEL
GBR
AUS
FIN
FRA
ITA
TUR
SUN
NLD
BRA
TB
late
USA
JPN
IDN
AUS
IND
CHN
IRL
GBR
SUN
DNK
SWE
BRA
PRT
BEL
DEU
FRA
ITA
AUT
CAN
MEX
ESP
GRC
ARG
KOR
FIN
NLD
TUR
ZAF
gr.rate*
‐0.25
‐2.26
‐1.31
‐3.44
1.21
1.05
‐0.08
‐1.88
‐5.18
‐0.1
0.88
‐6.28
3.02
‐0.57
‐0.03
‐2.08
‐2.12
1.51
0.56
1.04
2.59
0.9
1.16
1.49
0.51
‐1.51
0.19
4.84
early
SUN
FIN
IND
NLD
CHN
ESP
SWE
BEL
DEU
IRL
KOR
AUT
ARG
USA
FRA
IDN
GRC
CAN
TUR
ZAF
PRT
GBR
AUS
DNK
ITA
JPN
MEX
BRA
TX
late
FIN
AUT
PRT
IND
CHN
NLD
FRA
BEL
ESP
SWE
TUR
IRL
GRC
DEU
KOR
BRA
IDN
SUN
USA
ZAF
JPN
CAN
ARG
GBR
ITA
AUS
DNK
MEX
gr.rate*
‐0.47
‐1.28
‐2.06
0.26
0.04
0.44
‐1.03
‐0.17
‐0.01
0.11
‐1.05
‐0.13
‐0.46
0.42
0.13
‐3.57
0.14
4.98
0.51
‐0.09
‐1.42
0.5
0.84
‐0.01
‐0.6
0.12
0.07
1.44
early
CHN
SUN
ZAF
USA
ARG
CAN
AUS
ESP
BEL
FRA
AUT
SWE
ITA
MEX
GBR
JPN
IND
BRA
KOR
GRC
NLD
DEU
TUR
IRL
FIN
PRT
DNK
IDN
AP
late
ZAF
USA
ARG
ITA
BRA
AUS
CAN
CHN
FRA
JPN
AUT
SWE
BEL
TUR
IRL
KOR
SUN
NLD
PRT
IND
GBR
ESP
FIN
DEU
GRC
DNK
MEX
IDN
gr.rate*
0.02
0.23
0.27
‐1.58
‐2.91
0.05
0.13
2.1
‐0.84
‐1.38
‐0.27
‐0.26
0.09
‐1.55
‐1.54
‐1.33
3.71
‐0.83
‐1.84
‐0.32
0.4
2.22
‐0.1
0.36
0.91
‐0.08
2.84
‐1.63
early
CHN
SUN
AUS
MEX
IDN
PRT
ARG
IRL
USA
DEU
ESP
GRC
FRA
AUT
BRA
TUR
JPN
GBR
FIN
IND
ZAF
DNK
CAN
NLD
BEL
KOR
SWE
ITA
Table 7: RMC Country Rankings and Gowth Rates from the Early Period (1981‐1990) to the Late Period (1997‐2006).
LT
late
AUS
CHN
BRA
MEX
ARG
FRA
GRC
USA
ESP
IDN
IRL
NLD
PRT
AUT
TUR
SWE
DNK
GBR
DEU
CAN
IND
FIN
JPN
SUN
BEL
KOR
ITA
ZAF
gr.rate*
‐1.06
2.05
‐2.04
1.05
0.06
‐0.76
‐0.59
0.07
‐0.34
1.56
0.32
‐2.4
1.23
‐0.1
‐0.21
‐2.56
‐1.35
‐0.03
1.07
‐1.19
‐0.1
0.1
0.38
4.49
‐0.61
‐1
‐0.57
2.33
early
IND
ARG
JPN
FRA
ZAF
ESP
TUR
MEX
AUT
GRC
DEU
IRL
SUN
FIN
PRT
BRA
GBR
BEL
NLD
DNK
IDN
AUS
SWE
USA
CAN
ITA
KOR
CHN
FT
late
BRA
ARG
ZAF
FRA
ITA
AUT
MEX
IND
DEU
KOR
FIN
TUR
DNK
AUS
ESP
JPN
GRC
BEL
IRL
USA
GBR
SWE
IDN
CHN
NLD
PRT
SUN
CAN
gr.rate*
‐3.76
1.71
0.07
0.29
‐3.71
0.36
0.44
4.01
0.07
‐3.51
‐0.13
1.42
‐1.01
‐1.29
1.93
4.23
0.99
‐0.08
0.73
‐1.2
0.57
‐0.48
‐0.17
‐3.31
0.76
1.75
2.76
0.41
early
CAN
CHN
ESP
AUT
SUN
KOR
JPN
USA
DEU
AUS
TUR
FRA
DNK
PRT
IND
FIN
MEX
SWE
IRL
ZAF
NLD
GBR
ARG
GRC
BEL
BRA
IDN
ITA
WO
late
AUT
DEU
BRA
FRA
PRT
AUS
KOR
IND
DNK
USA
NLD
CHN
FIN
MEX
BEL
SWE
ESP
GBR
ZAF
CAN
IRL
TUR
SUN
GRC
IDN
ARG
JPN
ITA
gr.rate*
‐1.87
‐2.2
‐5.14
‐2.28
‐1.83
‐0.3
0.56
‐1.46
‐1.2
0.54
‐1.94
2.86
‐0.66
‐0.65
‐2.83
‐0.36
2.4
‐0.77
‐0.25
4.48
0.23
1.6
3.74
0.62
0.06
1.06
5.14
‐1.33
early
CHN
SUN
AUT
ARG
DEU
PRT
GRC
ITA
USA
NLD
BEL
ZAF
ESP
AUS
FIN
JPN
FRA
CAN
SWE
TUR
GBR
IRL
BRA
KOR
MEX
IDN
IND
DNK
*growth rates refer to the countries in the second column.
rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
FU
late
ITA
USA
AUS
BRA
ESP
AUT
CAN
DEU
FRA
FIN
NLD
SWE
JPN
BEL
CHN
TUR
KOR
IDN
PRT
ZAF
GBR
IRL
SUN
ARG
MEX
IND
DNK
GRC
gr.rate*
‐1.31
‐0.98
‐1.03
‐2.93
‐0.64
1.15
‐1.66
0.34
‐1.22
‐0.75
0.17
‐1.31
‐0.42
0.61
4.15
‐0.98
‐2.05
‐2.61
2.11
1.35
‐0.13
‐0.09
5.31
3.56
‐0.12
‐1.16
‐1.36
3.42
early
SUN
ZAF
FIN
IND
AUS
SWE
BEL
PRT
NLD
DEU
USA
ESP
AUT
TUR
CHN
ARG
FRA
GBR
JPN
IRL
ITA
KOR
GRC
DNK
IDN
CAN
BRA
MEX
PA
late
FIN
USA
JPN
IND
ZAF
NLD
PRT
SUN
AUS
AUT
GRC
SWE
FRA
KOR
ITA
DEU
IRL
ESP
GBR
BRA
CHN
TUR
BEL
ARG
IDN
DNK
MEX
CAN
gr.rate*
‐0.02
‐1.51
‐2.09
‐0.05
2
‐0.58
‐0.28
2.41
0.49
‐0.69
‐1.99
0.22
‐0.82
‐1.68
‐0.83
0.56
‐0.28
0.67
0.17
‐1.53
0.47
0.76
2.15
1.31
‐0.29
0.15
0.52
1.41
early
JPN
CHN
KOR
NLD
ITA
FIN
ESP
AUT
FRA
GRC
PRT
SWE
BEL
AUS
MEX
USA
TUR
GBR
IND
DEU
IDN
ZAF
DNK
IRL
CAN
BRA
SUN
ARG
late
JPN
BRA
KOR
ITA
CHN
PRT
ESP
NLD
DNK
FRA
SWE
IDN
FIN
AUT
BEL
SUN
MEX
GRC
GBR
USA
DEU
IRL
AUS
TUR
IND
ZAF
CAN
ARG
PP
gr.rate*
‐0.76
‐5.25
‐0.44
‐0.51
0.45
‐1.04
‐0.54
0.32
‐2.11
‐0.24
‐0.56
‐1.64
0.24
0.25
‐0.18
‐4.08
0.13
0.78
‐0.28
0.78
0.61
‐0.01
1.67
1.2
1.39
1.12
1.6
2.28
early
SUN
DNK
ARG
PRT
FRA
AUT
USA
IRL
CHN
DEU
FIN
GBR
ITA
IND
ESP
JPN
TUR
GRC
ZAF
BEL
IDN
AUS
NLD
KOR
SWE
BRA
CAN
MEX
late
IRL
ARG
DEU
ITA
GRC
JPN
CHN
FIN
AUT
DNK
ESP
SUN
USA
TUR
IND
PRT
FRA
IDN
SWE
GBR
ZAF
BRA
BEL
NLD
AUS
KOR
MEX
CAN
IC
gr.rate*
‐1.08
0.04
‐0.69
‐0.81
‐0.98
‐0.94
‐0.02
‐0.33
0.25
1.14
‐0.25
2.7
0.57
‐0.38
0.06
1.43
0.92
‐0.83
‐1.56
0.55
0.09
‐1.19
0.58
‐0.02
0.38
0.24
0.21
0.63
early
CHN
JPN
USA
SUN
ITA
DEU
IND
ESP
FRA
IDN
NLD
TUR
ZAF
GRC
FIN
KOR
AUT
SWE
GBR
PRT
ARG
IRL
AUS
MEX
CAN
DNK
BRA
BEL
Table 7: RMC Country Rankings and Gowth Rates from the Early Period (1981‐1990) to the Late Period (1997‐2006).
OC
late
ITA
USA
JPN
IND
CHN
BRA
FRA
PRT
GRC
SWE
ZAF
FIN
BEL
ESP
AUT
KOR
TUR
DEU
SUN
DNK
GBR
ARG
IRL
NLD
AUS
MEX
IDN
CAN
gr.rate*
‐0.52
1.32
1.68
0.23
4.03
‐3.07
0.02
‐1.46
‐0.73
‐1.04
‐0.54
‐0.74
‐2.51
0.77
‐0.61
‐0.47
0.13
1.51
2.2
‐1.4
‐0.18
‐0.4
‐0.08
1.41
‐0.08
‐0.19
1.76
0.24
early
CHN
SUN
CAN
AUT
IND
FIN
TUR
PRT
JPN
DEU
NLD
BEL
USA
ITA
FRA
IRL
GRC
SWE
ZAF
ARG
GBR
ESP
KOR
BRA
DNK
AUS
MEX
IDN
PE
late
AUT
SUN
CHN
CAN
IND
BEL
KOR
PRT
GBR
USA
JPN
ESP
SWE
IRL
ITA
TUR
FRA
BRA
FIN
GRC
DEU
NLD
ARG
DNK
ZAF
AUS
MEX
IDN
gr.rate*
‐0.06
1.82
2.26
1.56
0.43
‐0.89
‐2.49
0.06
‐1.01
‐0.51
0.16
‐1.66
‐0.69
‐0.4
‐0.23
0.92
‐0.14
‐3.23
1.34
0.4
0.93
1.1
0.67
‐1.25
2.91
‐0.31
‐0.54
‐1.79
early
SUN
USA
CHN
DEU
FRA
IND
TUR
IDN
ESP
ITA
AUT
JPN
ZAF
KOR
PRT
GRC
NLD
FIN
GBR
ARG
DNK
SWE
AUS
IRL
BEL
MEX
CAN
BRA
*growth rates refer to the countries in the second column.
rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
RU
late
USA
CHN
FRA
PRT
TUR
ITA
ESP
IND
GRC
DNK
KOR
SWE
JPN
DEU
FIN
IDN
BRA
ZAF
GBR
SUN
AUT
ARG
BEL
IRL
AUS
NLD
CAN
MEX
gr.rate*
2.43
1.71
0.42
‐1.78
‐0.91
‐1.08
‐0.91
0.65
‐1.21
‐1.63
‐0.85
‐1.57
‐0.32
2.61
‐0.93
0.25
‐2.96
0.11
‐0.49
5.31
0.77
‐0.09
‐1.16
‐0.69
0.08
0.91
0.18
1.04
early
SUN
AUS
AUT
ZAF
TUR
ESP
JPN
ITA
PRT
BEL
GRC
CAN
IND
FRA
KOR
FIN
DEU
MEX
GBR
NLD
CHN
SWE
DNK
ARG
IDN
IRL
USA
BRA
PL
late
ITA
PRT
DNK
JPN
ESP
SWE
FRA
AUT
FIN
KOR
DEU
TUR
GBR
BRA
ZAF
USA
GRC
NLD
BEL
AUS
IND
ARG
IRL
SUN
IDN
CHN
CAN
MEX
gr.rate*
‐2.11
‐1.76
‐2.4
‐0.15
‐0.01
‐1.86
‐1.37
0.91
‐1.2
‐1.18
‐1.22
0.67
‐1.02
‐2.97
1.32
‐1.87
0.1
‐0.41
0.34
2.58
0.32
‐0.35
‐0.82
3.53
‐0.17
0.71
2.42
2.38
early
CHN
GRC
ARG
MEX
SUN
TUR
KOR
JPN
IND
ZAF
IRL
AUS
CAN
PRT
USA
GBR
ESP
AUT
ITA
IDN
BEL
SWE
FIN
BRA
NLD
FRA
DNK
DEU
PT
late
IRL
CHN
JPN
AUS
PRT
GRC
USA
MEX
KOR
AUT
IND
FIN
ARG
GBR
TUR
CAN
NLD
IDN
ESP
FRA
BRA
DNK
ZAF
BEL
SUN
ITA
DEU
SWE
gr.rate*
‐2.12
3.07
‐0.38
‐1
‐1
0.72
‐1.02
0.77
0.61
‐0.7
0.73
‐1.97
1.35
‐0.19
1.43
0.34
‐2.4
‐0.09
0.4
‐2.4
‐0.61
‐1.86
2.59
0.94
3.53
1.59
‐1.04
1.76
early
CHN
KOR
SUN
TUR
ZAF
USA
AUT
NLD
ITA
ESP
GRC
IND
DNK
ARG
PRT
DEU
JPN
FIN
IDN
BEL
IRL
GBR
SWE
AUS
FRA
MEX
CAN
BRA
GL
late
KOR
FRA
PRT
TUR
ZAF
ESP
USA
ITA
ARG
FIN
CHN
DNK
IND
GRC
GBR
BRA
AUT
IDN
JPN
DEU
AUS
IRL
NLD
BEL
SWE
SUN
MEX
CAN
gr.rate*
‐1.07
‐2.93
‐1.2
0.68
0.22
‐0.51
‐0.01
0.18
‐0.24
‐0.73
3.22
0.23
0.47
0.6
‐0.57
‐3.28
1.15
‐0.32
0.4
0.47
‐0.57
‐0.11
1.6
0.27
0.12
3.94
1.02
0.62
early
CHN
JPN
AUS
SUN
USA
MEX
AUT
KOR
CAN
IND
GBR
IRL
ESP
GRC
ZAF
FRA
PRT
NLD
DNK
FIN
TUR
DEU
ARG
ITA
IDN
BRA
BEL
SWE
NM
late
CHN
AUS
JPN
USA
KOR
BRA
IRL
AUT
PRT
ESP
IND
SUN
MEX
CAN
FRA
GBR
ARG
NLD
ITA
TUR
GRC
ZAF
SWE
DNK
BEL
FIN
DEU
IDN
gr.rate*
2.73
1.04
1.81
0.4
0.27
‐3.47
‐0.78
0.71
‐1.8
‐0.63
0.21
1.95
2.07
0.81
‐0.87
0.85
‐1.7
‐0.77
‐1.44
‐0.76
0.23
0.42
‐1.34
‐0.02
‐1.25
‐0.02
0.05
‐0.6
Table 7: RMC Country Rankings and Gowth Rates from the Early Period (1981‐1990) to the Late Period (1997‐2006).
early
MEX
AUS
KOR
FRA
SUN
CAN
SWE
ESP
ARG
DEU
TUR
BRA
NLD
AUT
GBR
IRL
ITA
PRT
GRC
FIN
DNK
BEL
JPN
ZAF
IND
CHN
USA
IDN
ST
late
KOR
SWE
BRA
SUN
MEX
ARG
FRA
AUS
CAN
DEU
IRL
TUR
ESP
JPN
NLD
AUT
GRC
PRT
ITA
CHN
GBR
FIN
IND
BEL
DNK
USA
ZAF
IDN
gr.rate*
‐0.35
‐1.15
‐2
‐0.23
1.38
‐0.43
1.3
2.7
1.55
0.67
‐0.03
0.41
0.92
‐1.02
0.45
0.55
0.07
0.42
0.54
‐1.27
1.06
0.7
‐0.55
0.22
0.28
‐1.12
0.25
‐1.81
early
SUN
ARG
TUR
CHN
DNK
AUS
ITA
USA
GRC
DEU
ESP
CAN
IND
NLD
FRA
PRT
IDN
SWE
JPN
GBR
ZAF
IRL
KOR
AUT
BEL
MEX
BRA
FIN
NF
late
AUT
GRC
IND
TUR
PRT
ARG
ITA
CHN
ESP
IDN
SWE
DEU
FRA
BRA
KOR
NLD
DNK
IRL
BEL
ZAF
CAN
AUS
SUN
USA
GBR
JPN
FIN
MEX
gr.rate*
‐4.03
‐1.02
‐1.16
0.75
‐1.38
1.42
0.01
0.69
‐0.54
‐0.88
‐0.88
‐0.01
‐0.42
‐2.38
‐0.95
0.35
1.48
‐0.53
‐1.35
‐0.13
0.9
1.76
4.32
1.76
0.88
1.47
‐1.94
0.06
early
CHN
AUS
ESP
BEL
FRA
FIN
MEX
ARG
GRC
ITA
DNK
AUT
JPN
CAN
USA
IND
SWE
NLD
SUN
IRL
GBR
PRT
KOR
ZAF
TUR
DEU
BRA
IDN
*growth rates refer to the countries in the second column.
rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
MP
late
ITA
AUS
ESP
FIN
AUT
FRA
CHN
BRA
ARG
DNK
IND
JPN
SWE
USA
NLD
CAN
BEL
PRT
MEX
GRC
IRL
GBR
KOR
TUR
SUN
DEU
ZAF
IDN
gr.rate*
‐1.57
0.31
0.17
‐0.34
‐0.78
‐0.13
0.86
‐3.08
0.39
0.1
‐0.25
0.07
‐0.22
0.11
‐0.31
0.18
1.23
‐0.52
1.42
1.24
‐0.03
‐0.13
‐0.41
‐0.18
0.96
0.09
0.9
0.91
early
JPN
USA
AUT
SUN
ESP
CHN
FRA
DEU
ITA
IRL
PRT
BEL
TUR
CAN
GRC
ZAF
FIN
IND
DNK
GBR
SWE
NLD
AUS
KOR
IDN
BRA
MEX
ARG
MA
late
JPN
AUT
IRL
GRC
USA
ITA
PRT
NLD
FIN
SWE
TUR
DNK
CHN
FRA
KOR
CAN
ESP
IND
DEU
ZAF
GBR
BRA
IDN
SUN
BEL
AUS
ARG
MEX
gr.rate*
2.18
1.28
‐0.23
‐1.34
1.69
0.26
0.22
‐1.7
‐1.04
‐1.56
‐0.55
‐1.11
1.42
1.08
‐1.08
0.02
2.04
‐0.51
1.47
0.04
‐0.44
‐1.24
‐0.93
3.24
1.04
‐0.11
‐0.96
‐0.34
early
SUN
JPN
AUT
CAN
FIN
DEU
IND
CHN
FRA
AUS
USA
GRC
ARG
ESP
BEL
KOR
BRA
DNK
IRL
ZAF
GBR
MEX
SWE
TUR
ITA
PRT
NLD
IDN
EM
late
CHN
KOR
MEX
JPN
FIN
CAN
IND
FRA
BRA
ESP
AUT
USA
SUN
BEL
AUS
IRL
ARG
GRC
SWE
DEU
DNK
GBR
ITA
TUR
NLD
PRT
ZAF
IDN
gr.rate*
‐1.03
‐1.81
‐2.78
0.35
0.06
0.39
0.23
0.23
‐1.01
‐0.05
0.95
0.45
3.16
0.55
0.93
‐0.38
0.87
1.13
‐0.74
1.74
0.26
‐0.19
‐0.47
‐0.39
‐0.91
‐0.75
1.24
‐0.22
early
SUN
FRA
JPN
ITA
MEX
ESP
DEU
CHN
SWE
BEL
NLD
AUT
USA
IND
ARG
PRT
GRC
GBR
ZAF
FIN
TUR
AUS
KOR
DNK
IRL
BRA
CAN
IDN
TR
late
MEX
ITA
FRA
CHN
PRT
SWE
ESP
BEL
IND
DNK
KOR
SUN
DEU
TUR
BRA
NLD
GBR
ZAF
USA
FIN
ARG
GRC
AUS
AUT
JPN
IRL
CAN
IDN
gr.rate*
‐0.27
0.19
2.1
‐0.85
‐2.31
‐0.13
1.38
‐0.08
‐0.37
‐2.09
‐1.92
5.76
1.19
‐1.01
‐2.84
0.69
‐0.24
‐0.2
0.6
‐0.33
0.23
0.2
‐0.13
1.19
3.74
‐0.45
‐0.44
‐1.77
early
JPN
CHN
SUN
USA
FRA
ITA
ESP
BEL
PRT
KOR
NLD
AUT
IND
MEX
GRC
BRA
IRL
FIN
CAN
DNK
ARG
AUS
ZAF
SWE
TUR
GBR
IDN
DEU
Table 7: RMC Country Rankings and Gowth Rates from the Early Period (1981‐1990) to the Late Period (1997‐2006).
PS
late
JPN
ITA
BEL
FRA
USA
ESP
GRC
DNK
NLD
FIN
IRL
MEX
PRT
AUT
CAN
CHN
KOR
SWE
BRA
ARG
IND
GBR
IDN
AUS
ZAF
TUR
SUN
DEU
gr.rate*
1.59
‐0.5
‐0.39
0.03
0.18
0.33
‐0.84
‐1.14
‐0.04
‐0.83
‐0.71
‐0.26
0.34
‐0.11
‐0.77
1.91
0.42
‐1.06
‐0.17
‐0.6
0.21
‐1
‐0.98
‐0.26
0.24
‐0.11
2.73
‐0.75