Received: 04 May 2018 

|
  Revised: 16 May 2019

DOI: 10.1111/meca.12265

ORIGINAL ARTICLE

New‐structuralist exchange‐rate policy and the

pattern of specialization in Latin American
countries

Ariel Dvoskin1,2  | Germán David Feldman1,2  | Guido Ianni1,2

1
Institute for High Social Studies
(IDAES), National University of General
Abstract
San Martín (UNSAM), Buenos Aires, The present article critically examines the transmission
Argentina channels from real exchange rate undervaluation to output
2
National Scientific and Technical
growth adduced by New‐Structuralism (NS). To this end, we
Research Council of Argentina
(CONICET), Buenos Aires, Argentina slightly modify new‐structuralist canonical model for small
open peripheral economies, and make the underlying pro-
Correspondence
ductive structure depend on income distribution. This allows
Germán David Feldman, National
Scientific and Technical Research Council us both to discuss: (a) the limits of the transmission channels
of Argentina and Institute for High Social adduced by NS in their own terms, that is, under the assump-
Studies, National University of General
San Martín, Roque Sáenz Peña Av. 832,
tion that the tradable sector is the modern sector; and (b) the
2nd floor (C1035AAQ), Buenos Aires, further limitations of these channels when the pattern of trade
Argentina. is more suitable to characterize Latin American countries,
Email: gfeldman@unsam.edu.ar
whose tradable sector mainly exports primary goods under
conditions of differential rent. We conclude that the channels
work under highly restrictive assumptions, and therefore the
alleged positive relationship between real exchange rate and
growth cannot be generally ascertained.

1  |   IN T RO D U C T ION
In recent times, we have witnessed the emergence of a vast literature on development models inspired,
fundamentally, by the experiences of several Southeast Asian countries, but whose conclusions have
been extended to formulate policy prescriptions for Latin American countries. This standpoint, known
as “New‐Structuralism”1 (henceforth, NS), views the real exchange rate as the key variable to achieve
1
 This term has been employed in the literature in a broader sense: it refers to the contributions of the Economic Commission for Latin
American and the Caribbean (ECLAC) in different fields of economics, from the 1990 onwards (structural reforms, globalization,
innovation, macroeconomic volatility, etc.; for a revision of this literature, see Bielschowsky, 2009). Here, however, we will only
discuss those works that, following ECLAC's tradition, explore the link between exchange rate policy and economic growth.

Metroeconomica. 2019;00:1–27. wileyonlinelibrary.com/journal/meca |

© 2019 John Wiley & Sons Ltd     1
 |
2       DVOSKIN et al.

sustained economic growth. Although the transmission channels from the former to the latter are dis-
cussed in a wide variety of models, here we will focus on the dual‐economy models for small, open
countries with persistent unemployment, firstly presented in Frenkel and Ros (2006), and then more
rigorously developed, among others, by Razmi, Rapetti, and Skott (2012), Rapetti (2013), and Neto
and Lima (2017).2
Inspired by Lewis' (1954) seminal contribution, these models identify two strikingly different pro-
ductive sectors within a peripheral economy: a sector that profitably exports part of its production
at currently given international prices (sector T, or the “tradable sector”) and a second sector of low
productivity, which sells its entire production in the domestic market (sector NT, or the “non‐tradable”
sector). A central feature behind the new‐structuralist argument is that sector T, typically a modern
industrial sector, is envisaged as the most dynamic sector in terms of innovation and productivity
growth, with the implication that it should be the sector heavily promoted by the government.
There are three main channels discussed by this literature through which depreciations may foster
growth. In the “short run”, a rise in the exchange rate would have expansionary effects on employ-
ment and exports of sector T. In the medium run, the increase in the relative profitability of industry
T would induce a rise of its share in output, and thereby, being the most dynamic sector, average pro-
ductivity would be raised too. And finally, in the long run, the higher average profit rate would trigger
a profit‐led‐growth path.
To these three main mechanisms, we can add a fourth expansionary channel, which is assumed to
work both in the medium and long runs. It rests on the possibility to diversify the existing productive
structure by incorporating new sectors that were not profitable before devaluation.3
Now, while the results of these models are supported by an important empirical literature, the
fact is that the evidence is far from conclusive. In the first place, there are difficulties with the iden-
tification of the “equilibrium” or “long‐run” level of the exchange rate, against which the observed
misalignments are computed. The usual procedures, to rely either on some “adjusted purchasing‐
power‐parity” index (see e.g., Missio, Jayme, Britto, & Luis Oreiro, 2015; Razmi et al., 2012; Rodrik,
2008) or on some other measure that depends on “fundamentals” (e.g., Razin & Collins, 1997) are far
from being universally accepted (see Nouira & Sekkat, 2012; Smithin, 2002, Woodford, 2008). Other
contributions adopt not less questionable methods. In Levy‐Yeyati and Sturzenegger (2007), just to
mention another empirical work usually quoted in defense of the positive relationship between the real
exchange rate and economic growth, undervaluation is proxied counterfactually by the notion of “fear
of floating”. It denotes the idea that the exchange rate is artificially kept above its market equilibrium
level through the interventions of the Central Bank in the foreign exchange market. In this case, the
authors are unable to capture the effective impact of devaluation. They can only measure the effect
of an intervention that attempts to avoid the appreciation of the domestic currency, while, at the same
time, they use the change in the level of international reserves as a proxy of the hypothetical move-
ment of the exchange rate. Clearly, however, the rise in the level of international reserves can be the

2
 Ros and Skott (1998) could be also included in this list, but their model is employed to examine the effects of trade
liberalization on growth, and not of devaluation per se. For a different kind of model used to support a positive relation
between the real exchange and growth, which will not be examined here, see for example, La Marca (2010), Razmi (2012)
and Tadeu Lima and Porcile (2013).
3
 There is a fifth channel, not discussed by these authors, that examines the impact of devaluation on net exports. This
channel, certainly popular within mainstream circles, assumes that price elasticities of both exports and imports satisfy the
Marshall–Lerner condition. Differently from the above‐mentioned channels, this mechanism assumes that the economy is
“big” (i.e., it determines the international price of its exports); moreover, it is also assumed that the action of the channel is
limited to the short run, since only within this time span, it is further argued, there is involuntary unemployment. While in the
long run, the exchange rate is assumed to adjust to its general equilibrium level.
DVOSKIN et al.   
|
   3

outcome of multiple reasons, such us the inflow of capitals due to the very same dynamic of growth,
which may, therefore, be a source of endogeneity bias.
Moreover, the empirical evidence provided by the “favourable” works should be qualified, since
the relation is subject to non‐linearities that either show that (a) only moderate depreciations acceler-
ate growth, while large ones tend to reduce it (Aguirre & Calderón, 2005; Couharde & Sallenave,
2013); or (b) the effect of changes in the exchange rate depends on the kind of regime considered
(Mejía‐Reyes, Osborn, & Sensier, 2010): while appreciation regimes are detrimental to growth, the
effect of undervaluation is indeterminate.4
Lastly, beyond the seminal contributions that document the contractionary effects of devaluation
in developing countries (Braun & Joy, 1968; Diaz Alejandro, 1963; Krugman & Taylor, 1978), there
is also a more recent literature that finds evidence in this direction, both in the short and long runs
(Ahmed, 2003; Kim, An, & Kim, 2015; López, Sanchez, & Spanos, 2011; Mitchell & Pentecost,
2001; Miteza, 2006; Oskooee & Miteza, 2006; Serena & Sousa, 2017), that obtains indeterminate re-
sults, that depend on the groups of countries and/or the period of time considered (Oskooee & Hajilee,
2010), or simply that, beyond the observed correlation in the short run, finds no relation whatsoever in
the long run (Acar, 2000; Chou & Chao, 2001; Goncalves & Rodrigues, 2017; Kamin & Klau, 1997;
Nouira & Sekkat, 2012; Oskooee & Kandil, 2011; Upadhyaya, 1999).
The indeterminacy of empirical results should not, however, be a cause of surprise. Since outside
the neoclassical (or marginal) approach, it does not seem possible to obtain general relations between
relative prices and the dynamics of quantities, which is, on the other hand, precisely what NS is trying
to establish through its transmission channels.
On the other hand, these models do not seem to fit very well with Latin American countries, whose
most profitable sector produces primary commodities or manufactures based on primary inputs; and
under conditions that, even some authors that broadly speaking also belong to NS, qualify as “extremely
favourable”, and hence, “yield Ricardian [differential] rents” (Bresser Pereira, 2008, pp. 53–55).5
While, due to an incomplete process of import substitution inherited from the past, the industrial sec-
tor is a backward sector that, if any, can only produce for the domestic market.6
To have just a quick glance at this feature, we can examine the composition of exports of Latin
American countries and the Caribbean. As it can be seen in Chart 1, currently 52% of total exports of
the region to the rest of the world are explained by primary commodities or manufactures based on
primary goods. Moreover, if the case of Mexico is excluded due to the peculiarities of the “Maquila”,
and we focus on South America, which concentrates the remaining major countries of the region, this
percentage rises to 86. Far from being a circumstantial feature, the predominance of primary goods
in the export baskets of Latin American countries is a persistent structural characteristic that has
survived over the last decades. In the period 1985–2002, 59% of total exports of the main economies
of the region (Argentina, Bolivia, Brazil, Chile, Colombia, Mexico, Peru and Uruguay) were based
on natural resources, while, for instance, it was only 30% in the developing Asian countries that has
inspired NS.

4
 That non‐linearities are sufficiently damaging for the approach is even accepted by some of its proponents (see the
contribution by Tadeu Lima jointly with Marques Ribeiro and McCombie (2017, pp. 2, 3).
5
 See also Guzmán, Ocampo, and Stiglitz (2018, p. 53), who accept that “[m]any developing economies, especially in Africa
and South America, are highly dependent on agricultural and/or nonrenewable natural resource exports (fuels and minerals)”.
6
 Actually, this kind of pattern of specialization seems to have also been recently accepted even by some representative
scholars of NS examined here: “The region [Latin America] has experienced a reduction in its capacity to produce tradable
goods other than commodities … The region was de‐industrialized, and to reverse this process will take time” (Damill &
Frenkel, 2017, p. 5).
 |
4       DVOSKIN et al.

ϭϬϬй 2 6 3
7 12
ϵϬй 15
5 23
ϴϬй 23
23 27
ϳϬй 24
ϲϬй 20
9
ϱϬй
48
22 15
ϰϬй
30
ϯϬй 63
11
ϮϬй
36 37
7
ϭϬй 21
11
Ϭй
South America Caribbean Central America Mexico Latin America and
the Caribbean
Manufactures with high technological content Manufactures with medium technological content
Manufactures with low technological content Manufactures based on natural resources
Primary Goods

C H A R T 1   Structure of exports of Latin American and the Caribbean countries, 2017. Source: ECLAC (2018)

If the mixed empirical evidence suggests that the transmission channels adduced by NS are not
sufficiently robust, and calls for a thorough revision of the issue, the inadequacy of these models
to represent Latin American countries suggests that the analytical framework employed, although it
must respect the main features of the canonical model used by NS, should be also sufficiently flexible
to accommodate different productive structures. To this end, after this introduction, in Section 2 we
present a model that determines the pattern of trade as a problem of technical choices, and of which
the productive structure assumed by NS emerges only as a special case. On this basis, in Section 3 we
critically examine the transmission channels adduced by NS “in their own terms”, namely assuming
the same productive structure as the authors of the approach. While in Section 4, we discuss further
limits of these channels when the pattern of trade is more suitable to represent Latin American coun-
tries. Section 5 resumes the argument and presents the main conclusions of the article, which can
be very briefly summarized in the following statement: new‐structuralist channels work only under
highly restrictive assumptions, and therefore the kind of relationship between real exchange rate and
growth cannot be generally ascertained.

2  |  A NA LY T ICA L F R A ME WO R K

In this section, we introduce a formal framework that captures the main features of the dual‐economy
model firstly presented in Frenkel and Ros (2006), and then developed, among others, by Razmi et al.
(2012), Rapetti (2013), and Neto and Lima (2017). Therefore, we conceive a small (i.e., price taking)
peripheral economy opened to trade and capital flows, which also has persistent unemployment.
The productive structure is characterized by the following features: first, there is an intrinsically
non‐tradable good, NT, typically subsistence agriculture or services; second, there is one tradable
commodity, T. However, to explore the robustness of the transmission channels from the real exchange
rate to economic growth adduced by NS on different productive structures, we assume, differently
from the above‐mentioned contributions, that the nature of good T is not predetermined, but rather
DVOSKIN et al.   
   5
|
is, as we shall see, the outcome of a problem of technical choices. As a result, commodity T can be
either a necessary primary consumption good (such as corn)—eventually produced under conditions
of differential rent—or a more dynamic industrial good. Let us call the former commodity C and the
latter commodity I.
We also follow NS in assuming that commodity T is produced by direct labour and an imported
capital good.7,8 To avoid the problems raised by fixed capital, which are not relevant for the present
analysis, for simplicity we further assume the capital good to be entirely circulating. Commodity NT,
on the other hand, requires unassisted labour alone.
If w stands for the uniform nominal wage rate across sectors,9 r for the normal rate of profits, lNT
and lT (with T = C, I) are the unitary labour requirements of sectors NT and T , kT is the unitary require-
ment of the capital good K10 in sector T , p∗K is its exogenously given price, and E is the nominal ex-
change rate, then the costs of production or the supply prices of commodities NT (psNT ) and T (psT ) can
be represented by the following equations11:
psNT = wlNT (1 + r) (1)
( )
pST = wlT + kT Ep∗K (1 + r) (T = C, I) (2)

These supply prices represent the minimum amount of money per unit of output to regularly (under
“normal conditions”) deliver each commodity on the market.
For reasons that will be clear below, here we slightly depart from NS and explicitly introduce a sec-
ond notion of price, which we shall denominate demand or selling price. It represents the maximum
amount of money that consumers are willing to pay for a certain commodity. If, like the authors of the
approach, we abstract from transport costs, import tariffs and other expenses implied by international
trade, we should notice that, since the domestic economy takes the international price of the tradable
good ( p∗T ) as given, once the level of the exchange rate is fixed, demand prices for tradable goods
T = C, I are univocally determined. These prices are:

pdT = Ep∗T (T = C, I) (3)

7
 Except for Razmi et al. (2012, p. 152), who explicitly argue that “all capital goods are imported”, the remaining papers
examined here do not explicitly mention where capital goods are produced. At any rate, Frenkel and Ros (2006, p. 635)
eventually accept that domestic production requires “an important portion of imported capital goods”, and Rapetti (2013,
p. 2) similarly points out that in development countries, imports of capital goods may be “significant”. Anyway, the domestic
production of capital goods will be considered in Section 3.
8
 When good T is a primary commodity, it will also use land as an input. However, throughout the analysis we will abstract
from absolute rent, while the possibility and implications of differential rent will be explored in Section 4.
9
 Differently from Lewis' (1954) model that has inspired it, NS assumes the uniformity of money wage rates across sectors,
and that is why we adopt the same assumption here. We thank a comment made by an anonymous referee that allowed us to
clarify the issue.
10
 The assumption that both tradable sectors employ the same capital good is no doubt restrictive. However, this is immaterial
to the results reached in the paper and can be anyway easily relaxed when necessary.
11
 Since the scope of this section is only to determine prices and distribution within new‐structuralist “canonical” model, for
the moment we assume given sectorial outputs, and therefore the technical coefficients are expressed per unit of output. The
discussion of the effects of income distribution on the change in the level of outputs, and the related question of the kind of
returns assumed by NS (both of scale and marginal ones) will be postponed to the following section, when the alleged
channels that link the exchange rate and economic growth are examined in detail.
 |
6       DVOSKIN et al.

While for the non‐tradable commodity, the demand price is determined by its respective supply
price, as it is the case with any commodity produced in a closed economy12:

pdNT = pSNT (4)

The six Equation (1)–(2C, I)–(3C, I)–(4) have nine unknowns: E,r,w, psNT , psT, I , psT,C , pdNT , pdT, I , pdT,C.
Therefore, there are three degrees of freedom left. To eliminate the first two of them, consider first that
NS assumes that, due to “nominal stickiness” (e.g., wage contracts are the result of previous negotia-
tions between workers and capitalists), the level of money wages is given, at least in what NS calls the
“short run” (see Rapetti, 2013, p. 11)13:

w = w̄ (5)

Second, that the nominal exchange rate is determined by the monetary authority (Rapetti, 2013,
p. 11)14:

E = Ē (6)

This implies that the wage rate expressed in foreign currency (which is none other than the inverse
of the E∕w ratio) is independent from market conditions and can therefore be known before prices and
distribution are determined. If we define the variable e ≡ E∕w, then (5) and (6) imply:

Ē
e = ē = (7)
w̄

To eliminate the last degree of freedom, one must determine the pattern of specialization in the
tradable sector. Notice however that in our model, differently from the one endorsed by NS, it is
not possible to ascertain which of the tradable commodities will be effectively produced before the
relationship between demand and supply prices of each of these commodities is established. Hence,
before income distribution is known.
The pattern of specialization of the economy will be regulated by the following conditions:

pdT ≤ psT (T = C, I) (8)

12
 In Rapetti's (2013) model, for instance, the supply price Equations (1) and (2) are not explicitly formalized, but one can
derive them from his Equations (3) and (6). While our Equation (3) corresponds to Rapetti's Equation (2).
13
 This fact does not imply that for NS, over longer time spans the nominal wage adjusts to clear the labor market. It simply
means that eventually the level of money wages change because they are renegotiated according to new market and
institutional conditions. This view of the institutional determinants of the level of wages is clearly stated when Rapetti (2013,
pp. 16, 17) recalls that wage determination is influenced by two factors: besides workers' bargaining power vis‐à‐vis
capitalists, “wages are also influenced by institutional, political and moral elements which are context dependent and can be
influenced by government intervention”. This implies that, even over longer periods, we are authorized to take w as given
when determining prices and distribution.
14
 E is defined as the amount of domestic currency per unit of foreign currency, which means that E goes up with a deprecia-
tion of the local currency.
DVOSKIN et al.   
   7
|
The generic tradable commodity T will be produced and (potentially) exported only if pdT = psT . In
contrast, when pdT < psT , the sector will not be viable because its normal costs of production exceed its
demand price.15
We can now derive for each commodity T , a function that gives, for each level of e, the maximum
affordable profit rate by each tradable sector under given technical conditions and international prices.
This is obtained by equalizing supply and demand prices for each commodity T . Hence, from condi-
tions (2) and (4), one obtains:
1
rT (e) = e ( ) −1 (T = C, I) (9)
l T + kT e

Where for convenience, we have normalized the given international prices to one. Figure 116 rep-
resents a possible shape of these curves:
As can be immediately seen from the figure, the e − r curves cut the vertical axis at r = −1, and are
positively sloped. Moreover, given the input–output relations assumed by NS for the tradable sector,
it can be shown that the curves intersect only once, at the value of e given by:

l −l
ê = ( I c ) (10)
kc − kI

Economically meaningful values of (10) require that the following condition is fulfilled:

( )( )
lI − lc kc − kI > 0 (11)

whose interpretation will be given in the next paragraph.

These curves can be used to determine the pattern of specialization of the peripheral economy as a
problem of technical choices. In effect, due to the action of competition among capitals, the tradable
sector of the economy will specialize in the production of that commodity which, for the given value
of e, can afford the highest profit rate. If we assume, without loss of generality, that sector I operates
with higher “capital intensity”, that is, if l I > l c , then, for any level e < ê there will be full specialization
k k
I c

in the production of commodity I .17 In effect, if for instance: e = ē (< ê ), then rI (̄e) > rC (̄e) and there
will be no incentive to invest in sector C. The opposite occurs when e > ê . An only by a fluke e = ê ,
which is the level that allows the coexistence of the two tradable sectors in the economy.
It is now easy to interpret the necessary condition (11) for the existence of an intersection of the
curves: for income distribution to eventually cause a shift in the pattern of specialization, it is neces-
sary that the tradable sector with the highest coefficient of labour per unit of output is not also the one
with the highest capital output coefficient.

15
 As rightly pointed out by an anonymous referee, there are of course other non‐price factors that may allow a country to
export specific commodities (e.g., product differentiation, etc.). The implications of this point are beyond the scope of this
paper, since this issue is not considered by the models under examination.
16
 The technical coefficients of production that support the shape of these curves are: lC = 3∕8, kC = 1∕8, lI = 1∕4, kI = 1∕4.
Additionally, ê = r̂ = 1, ē = 1∕2, rc (̄e) = 1∕7 and rI (̄e) = 1∕3.
17
 If one assumes that the second factor in (11) is positive, then kc > kI . And, since the first factor must be also positive for ê
to be positive, one obtains the factor intensity condition documented in the main text. The opposite sectorial order would arise
if both factors in (11) were negative.
 |
8       DVOSKIN et al.

F I G U R E 1   e − r Relation in the small open economy

The outer envelope of the curve (thick black line) illustrates the economically relevant pairs of e
and r.18 In analytical terms, the e − r relationship is given by:

⎧ rI (e) if e ≤ ê
⎪
r=⎨ (12)
⎪ r (e) if e > ê
⎩ C

As previously mentioned, the models under consideration do not discuss the dependence of the
pattern of specialization on income distribution. Instead, they directly assume that the modern sector
I profitably competes in the international markets at the given prices (in other words, the tradable good
is always produced by sector I),19 while if consumed, commodity C is imported from abroad.20

18
 Note that, with more than two commodities, in general there will be no value of e that will allow the coexistence of all
tradable sectors, unless one admits heterogenous remunerations within social classes (see Baldone, 2001; Steedman, 1999,
p. 272), or the existence of differential rent (see Section 4.2 below).
19
 The equality between supply and demand prices of commodity I is ensured in Rapetti's model by Equation (3).
20
 Therefore, the small peripheral economy exports commodity I in exchange for imports of the capital good, K and, if
consumed, of the primary commodity C. No particular result of the current account is assumed by NS (see e.g., Neto & Lima,
2017, pp. 4, 7). If there is a current account surplus, the implicit assumption is that the country accumulates foreign assets.
While in case of current account deficits, since the economy is also small in financial terms, it is argued that it can finance
this imbalance through capital inflows (Neto & Lima, 2017, p. 4). Hence, this kind of models does not deal with the potential
problems caused by the external constraint to growth usually faced by peripheral economies, and therefore, the issue,
although extremely relevant for any growth strategy, will not be discussed in the present contribution. We thank the comment
by an anonymous referee that allowed us to clarify this point.
DVOSKIN et al.   
   9
|
In the light of our model, the implied pattern of specialization is such that e = ē (< ê ) in Figure 1,
and hence, the model is closed by considering the first branch of Equation (12), namely:

r = rI (̄e) (13)

In sum, the eight Equation (1)–(2C, I)–(3C, I)–(4)‐(7)‐(13) determine the following unknowns:

{ }
e,r, psNT , psI , psC , pdNT , pdI , pdC

The model presents the following three features that are worth mentioning. The first two, because
they differ from the standard results for a closed economy, while the third one because it may give the
wrong impression that some of the neoclassical “parables” about factor intensities and their respective
remunerations may generally hold.
First, notice that, not only does condition (13) determine the profit rate in the tradable sector, but it
also determines, due to the action of free competition, the normal level of the profit rate for the whole
economy; even when the tradable commodity is not used as an input in the production of any
commodity. While the price equation of the non‐tradable sector only determines the supply price of
this good.21
Second, note that in the model both the real wage and the profit rate are endogenously determined.
This is because the existence of intrinsically non‐tradable goods disentangles the money wage ex-
pressed in foreign currency (i.e., the inverse of e), which is a given variable, from the real wage.
However, both variables necessarily move in the same direction.22 To see both features, let us define
a given consumption basket for the representative worker as c = (cC , cI , cNT ). Thus, the real wage 𝜔 can
∑
be expressed as 𝜔 ≡ wP, with P = cNT pNT + E T=C, I cT p∗T . Notice then that, even if the value of e were
known, the price level P, and therefore, the level of 𝜔, would not be not known yet because pNT is not
determined (this would not happen if the three commodities were tradable goods, since in this case P
, and hence, the number of consumption baskets 𝜔 affordable by the representative worker, would be
univocally determined by e). But also note that, since a rise in e implies a rise in the normal profit rate,
and therefore, for a given money wage, a rise in pNT , then it follows that the wage in foreign currency
and the real wage still move in the same direction as e varies.23,24
The third feature is the following. The so‐called neoclassical parables about the relationship be-
tween factor intensities and their respective remunerations are verified in the model: as depicted
in Figure 1, the curve with the highest capital/labour ratio intersects the other curve from above,
namely a lower real wage (a higher e) eventually shifts the pattern of specialization towards the more

21
 The point has been raised by Steedman (1999, section 2; cf. also Siqueira Machado's, 2017, unpublished dissertation). As he
correctly argues, this pure consumption tradable commodity behaves “as only a basic commodity is said to do” (Steedman,
1999, p. 267). All this suggests, as Steedman himself concludes, “that in the context of analysing the small open economy,
the concept of a basic commodity must be set aside or at least significantly modified”.
22
 We will see in Section 4 that this condition may cease to hold when differential rent is considered.
23
 We thank a comment made by an anonymous referee, which gave as the opportunity to clarify this point.
24
 That both the real wage and the profit rate are endogenously determined variables, does not mean that class conflict is
absent from the analysis. Given the money wage, distributive conflict is expressed in the level of the nominal exchange rate.
And the latter, being exogenously determined by the monetary authority, ultimately reflects the relative power between
capital and labour.
 |
10       DVOSKIN et al.

“labour‐intensive” sector (in this case, sector C). However, we shall show in Section 4.1 that this is
a strict consequence of the very particular conditions of production assumed by NS. Indeed, we will
also show that if one slightly modifies the input–output relations and, for instance, assumes that the
non‐tradable commodity is employed as an input in the tradable sectors, the two e − r curves could
easily intersect more than once and, therefore, the conditions required for the validity of these parables
would no longer hold.
Before we move to the next section, it is convenient to remark a second limitation of the ca-
nonical model presented so far (the first one being the alleged independence of the productive
structure with respect to changes in income distribution): the distributive closure assumed by NS
(a given e and an endogenously determined r ) will be generally incompatible with the equalization
of profits rates across countries. A result, on the other hand, that one would have expected in the
current era of “globalization”, under which capital can very quickly migrate from one country to
another. This, however, would not be in itself cause of big surprise if it not were the case that,
additionally, and due to the assumption of free financial capital mobility across countries, the
tendency towards the equalization of returns on financial assets is accepted (see Rapetti, 2013, p.
11). This is indeed surprising because, due to the later tendency, one would have also expected the
arbitrage between the domestic and the international profit rates to occur. After all, once capital
flows from one country to another in the form of financial assets, unless purely institutional rea-
sons are advanced (e.g., capital controls, minimum stay requirements, etc.), there does not seem to
be economic reasons that may prevent financial capital from taking the form of productive capital.
In other words, unless one highlights motives of purely extra‐economic nature, which in any case
are neglected by the canonical model, it seems arbitrary to assume the free mobility of financial
capital and not that of productive capital.
Now, let us assume for the sake of argument that those institutional reasons exist and express the
gross profit rate, r, as the sum of two components: the domestic interest rate, i , that due to financial
capital mobility is equal to the international rate, i∗, and a normal “profit of enterprise”, 𝜎, that some-
how reflects the “risks and troubles” of investing in the productive sector:

r = i∗ + 𝜎 (14)

It should be then clear that the difference among countries' profit rates is entirely explained by the
differences in their respective 𝜎. Differences that, as we have already explained, under the conditions
assumed by NS can be only plausibly justified by reasons that are independent of the market, and
hence susceptible of being known before relative prices are determined. But if this were effectively the
case, and hence 𝜎 were given for the price system, there would be two equations to determine r—(13)
and (14)—and the system would be evidently overdetermined.
The formal consistency of the distributive closure assumed by NS requires, therefore, two highly
restrictive assumptions, which are never explicitly stated by their proponents: (a) to arbitrarily assume
that the differences in the normal profits rates across countries are explained, contrarily to what we
have just argued, entirely by economic forces; (b) that these differences rise with e, namely that there
is a positive and systematic relationship between the normal profit of enterprise and the real exchange
rate. This relationship could be formalized as:

�
𝜎 = 𝜎 (e) , with 𝜎e > 0 (15)
DVOSKIN et al.   
   11
|
3  |   T R A N S M ISS ION CHA N N E L S FROM THE EXCHANGE
RAT E TO ECO NOMIC GROW T H

In this section, we discuss the scopes and limits of the transmission channels from the exchange rate
to output growth adduced by NS in their own terms, namely, under the assumption that sector T is the
industrial sector (i.e., e < ê ). While some further limitations of these channels under a different pattern
of specialization will be addressed in Section 4.

3.1  |  Short‐term mechanisms

Let us begin with those mechanisms that are assumed to operate in what NS calls the “short run”,
namely under a given level of money wages and a given rate of capital accumulation. According to
NS, the rise in e exerts an expansionary effect through three different mechanisms, all induced by the
subsequent fall in the real wage in terms of the tradable good, 𝜔T = w∕pdI. The first effect is to induce
producers in sector I to adopt more “labour intensive” methods of production, and therefore, to raise
labour demand (LI ) (this is denominated by this literature as “labour intensity channel”, see Frenkel &
Ros, 2006). Moreover, since 𝜔T is now lower than the marginal product of labour, profit‐maximizing
firms will be induced to expand industrial output yI 25:

↑ e → ↑ pdI → ↓ 𝜔T → ↑ LI →↑ yI (A1)

These expansionary effects also cause a rise in the production of good NT . Firstly, because the
expansion of yI increases aggregate income and, therefore, the demand for NT (DNT ) (this is called
“income effect” by Rapetti, 2013, p. 7).

↑ yI →↑ DNT → ↑ LNT →↑ yNT (A2)

pdI
Second, because the rise in the relative price pdNT
increases the demand for—the now relatively
cheaper—good NT (called “substitution effect” by Rapetti, 2013, p. 7).

↑ e →↑ DNT → ↑ LNT → ↑ yNT (A3)

25
 Of course, the labour intensity channel assumes that there are several methods of production to produce commodity I . While
this aspect of the productive structure has been neglected in the presentation of the canonical model in Section 2, we will see
below that the model can be easily reinterpreted to account for the existence of several methods of production in sector T . In
this respect, it is worth pointing out that the technology presented by NS for the tradable sector not only does assume
decreasing marginal returns to both capital and labour (this is formalized by assuming a Cobb–Douglas production function
that uses labour and capital as inputs, but this is immaterial for our analysis), but also increasing returns to scale. Moreover, to
be compatible with free competition, the returns to scale are assumed to be external to the firm and internal to the industry.
This is allegedly due to “externalities in the learning process” (Rapetti, 2013, p. 12). Note that these conditions are highly
restrictive since “…the economies of production on a large scale can seldom be allocated exactly to any one industry: they are
in great measure attached to groups, often large groups, of correlated industries” (Marshall, cited by Sraffa, 1926, p. 540). It
seems therefore arbitrary to assume that the externalities that affect sector T do not have any influence on sector NT as well.
The assumption is tantamount to assuming the technical disconnection among productive sectors, ignoring the possible
input–output relations that generally emerge among them, and would immediately cease to hold if, for instance, one slightly
modifies the canonical model and considers the possibility that commodity T were used as input of commodity NT .
 |
12       DVOSKIN et al.

To ascertain the validity of these three mechanisms, it is necessary to make two observations: first,
that the negative relationship between labour demand and e is a necessary yet not a sufficient condition
to justify the initial increase in LI . Without any additional consideration, and under the plausible as-
sumption that workers' propensity to consume is higher than capitalists' propensity (see, for instance,
Rapetti, 2013, p. 14), one should expect that the incentive to expand the production of commodity I
will be sooner or later counterbalanced by a fall in DI that, ceteris paribus, is caused by the fall in the
real wage (recall that e and 𝜔 move in the opposite direction). Hence, to ascertain that the rise in e will
anyway exert a favourable effect on employment, it is further necessary to justify how the excess of
production of I over internal consumption will be absorbed. For instance, the marginal approach ar-
gues that the higher level of savings caused by the increased levels of output are eventually absorbed
by higher levels of investment through movements in the interest rate. But NS does not envisage this
mechanism, at least explicitly.26 And while it recognizes the possible recessive effects of the fall in the
real wage,27 the issue is virtually neglected because every excess of production over internal consump-
tion is assumed to be passively absorbed by the demand for exports (XI ). Hence:
XI = yI − DI (16)

Therefore, this adjustment presumes the absence of effective demand problems. Now, behind the
idea that the demand for exports is infinitely elastic at current prices we find NS arguing the only
apparently equivalent notion that the economy is sufficiently “small”, that is, an economy that takes
the selling price of commodity I as given. Note however that for this last condition to hold, it is only
necessary that neither the domestic method of production nor domestic income distribution are rele-
vant magnitudes to determine the international price of this commodity; and hence, it does not need
assume that the share of exports in the global market is of any particular magnitude (as long as this
proportion is smaller than 1).28
The second consideration refers to the bases that justify the negative (and sufficiently elastic)
relationship between employment and the level of the real wage itself. As seen above, here NS does
closely follow the marginal approach and argues that the fall in 𝜔T increases employment because
productive processes that employ more labour per unit of output become more profitable.
The fact is, however, that the labour intensity channel works in the direction adduced by NS only
because all the produced means of production are imported at given international prices.29 To see this
more clearly, we must reconsider our model in the following way: instead of assuming that there is one
single method to produce each tradable commodity, let us assume that I and C are now two different
methods to produce the specific tradable good T .30 Because international prices are given, it makes no

26
 For reasons of space, it is not possible to deal with this issue here, but labour‐capital substitution, accepted by NS, implies
the existence of a negative relationship between aggregate investment and the interest rate (see Dvoskin & Petri, 2017).
27
 See Frenkel and Ros (2006, p. 635), Razmi (2007) and Rapetti (2013, p. 6, fn. 9).
28
 To link the assumption of given prices with particular conditions of demand, it is further necessary to establish general
price–quantity relations that univocally determine the way in which demand influences technical conditions of production
and/or distribution (see Garegnani, 1983). But these relations do not seem to exist outside the very restrictive world of
“well‐behaved”, neoclassical factor demand curves.
29
 On this point see also Steedman (1999, p. 262).
30
 This reinterpretation of the model is not actually necessary, since as Frenkel and Ros (2006, p. 635) argue, the labour
intensity channel takes place, either through “the adoption of more labour intensive techniques”, as explained in the main
text, or through “the reallocation of labour and investment towards labour intensive tradables”, which is our formalization in
Section 2 can capture.
DVOSKIN et al.   
|
   13

difference whether we assume that each method uses a specific imported capital good, or not. If we
normalize foreign prices to one, the relative cost of production of the two methods is:
( )
pSC (1 + r) lC w + EkC lC + ekC
S
= ( ) = (17)
pI (1 + r) lI w + EkI lI + ekI

It is evident that a sufficient increase in e decreases the cost of production of the method with a
lower capital/labour ratio and, therefore, profit maximizing firms will adopt it. If without loss of gen-
erality, we assume that l C < l I , then31:
k k
C I

pS
d pCS
I (18)
<0
de

The reason is evidently that, although there is production of capital goods, since their prices are
given for the domestic economy, changes in income distribution do not affect, by assumption, capital
good prices and, therefore, the “paradoxes” highlighted during the so‐called “Capital Debates” of the
1960s and 1970s cannot occur.
As seen in the previous section, in terms of Figure 2 all this means that the curve that represents the
conditions of production of the method with the highest capital/labour ratio (i.e., the highest kT ∕lT ) is
employed for lower levels of e and eventually cuts the less capitalistic method from above. However,
the moment one slightly departs from the particular conditions of production assumed in the canoni-
cal model, and considers instead the possibility that, for instance, the non‐tradable good is used as an
input in the production of the tradable commodity (a possibility that is even accepted by some authors
of the approach, see e.g., Ros, 2016, p. 234), changes in income distribution will affect capital goods
prices and therefore, the labour intensity channel may not work in the direction predicted by NS.
To see this, consider the case that, besides labour and the imported capital good, each method of
production employs a quantity bT (T = C, I) per unit of output of commodity NT . Now the relative
pS 32
price of commodity T under both methods, pT,C S , is :
T, I

pSC [lC + bC lNT (1 + r) ]e + kC

= (19)
pSI [lI + bI lNT (1 + r) ]e + kI

Condition (19) shows that the relative price is now a function of the profit rate r, and this will gen-
pS
erally mean that pCS may move in any direction when e varies.
I

pS
𝜕 pCS
I (20)
≷0
𝜕e

31 pSC kC (lI +ekI )−kI (lC +ekI )

 Since ded pSI
= 2 , the sign of this derivative is determined by the sign of the numerator, which can be
(lI e+kI )
pSC
re‐expressed as: lI kC − lC kI . Hence, ded pSI
<0⇔
kC
lC
< lI .
k
I

32 ( )
 The supply price of the tradable good under method T = C, I is pST = lT w + EkT + bT pNT (1 + r). If we replace pNT by
( )
Equation (1), we obtain pST = lT w + EkT (1 + r) + bT wlNT (1 + r)2. Finally, dividing both supply prices yields (19).
 |
14       DVOSKIN et al.

F I G U R E 2   Reswitching of techniques in the tradable sector

Graphically, we could easily observe that the two e − r curves intersect more than once, namely that
the convenience of a particular
( method
) does not (show any
) relation whatsoever with e: the same method
could be adopted for low e < e�1 and then high e > e�2 values of the wage in foreign currency, show-
ing a sort of “reswitching” in methods.33 This is shown in Figure 2.34

33
 Besides the difficulty addressed in the main text, it is worth mentioning a second difficulty with the labour intensity channel
when it is used within NS: it is unclear what is assumed to be given with respect to the factor “capital” when the marginal
product of labour in sector T is determined. The neoclassical mechanisms of substitution duly consider that capital goods are
specific to each technique, and hence plausibly presume that the “form” of the cooperating capital generally changes with the
change in the cost minimizing method. To proceed in this way, neoclassical theory is forced to assume a given “quantity” of
capital, measured in value terms. A magnitude that, therefore, does not change when income distribution changes, something
that is well known to be unacceptable. On the contrary, NS seems to take as given the physical stock of capital goods existing
at any given moment. But if the available capital goods before and after devaluation are the same, it may well happen that the
labour demand curve is highly inelastic, with the implication that a drastic—socially unbearable—fall in the real wage is
needed to achieve a certain target level of employment. The problem is somehow hidden in the new‐structuralist versions of
the canonical model because the conditions of production assumed presuppose that the same kind of capital good is used in
different proportions when the quantity of labour employed changes. But it would become immediately evident once a more
general specification of the available techniques is considered (for a detailed analysis of this problem, see Dvoskin & Fratini,
2016). It could be finally argued that the higher level of employment can be accommodated without any change in form of
the existing capital stock because, in the short run, there is idle capacity. The problem here is that the marginal product of
labour should be constant, and therefore, the positive relationship between LI and e, would be lost. If before devaluation the
marginal product of labour in the tradable sector equals w∕pI , one would have to accept that the rise in e would indefinitely
expand output and employment in sector I , at least until full capacity is reached. Anyway, once full capacity is restored, for
the above‐mentioned reasons, the marginal product of labour would still be zero (or very closed to zero). Finally, as long as
there is idle capacity, NS would also have to accept that the marginal product of capital would be zero, contradicting the
assumptions of the canonical model.
34
 The technical coefficients of production that support the shapes of these curves are: lI = 1∕15; bI = 7∕30;
kI = 1∕20; lC = 7∕60; bC = 17∕60; kC = 1∕10; lNT = 1. Additionally, ê1 = r1 = 50% and ê2 = r2 = 100%.
DVOSKIN et al.   
   15
|
3.2  |  Medium‐ and long‐term mechanisms
The medium‐term mechanism supposes, for a given level of capital accumulation, a change in the
composition of output in favour of the modern sector T , while the longer‐term channel presumes an
increase in the rate of capital accumulation due to the rise in the general rate of profits.
Let us consider first the medium‐term mechanism. Devaluation raises the demand price of the
tradable good ( pdI); and since the demand price of the non‐tradable good ( pdNT ) does not initially
pd
change, the relative price, pdI , rises too. Therefore, the actual profit rate of sector T (rI ) temporarily
NT

increases relative to that of sector NT (rNT ). This change in relative profitability triggers a virtuous
process of “structural change” (Rapetti, 2016, p. 259) by inducing the reallocation of resources to the
relatively more productive and dynamic tradable industry, until pNT rises sufficiently to restore the
equalization of returns across sectors at a higher level.

pdI yI
↑ e →↑ →↑ rI →↑ →↑ rNT (B)
pdNT yNT

If we now turn our attention to the longer‐run mechanism, NS postulates that capital accumulation
is an increasing function of net profitability. And given that the rise in e has caused an increase in r,
then the rate of capital accumulation, gK , rises:

↑ e →↑ r →↑ gK (C)

This mechanism is known in the literature as the “development channel” (Frenkel & Ros, 2006, pp.
636, 637).35
The difficulty with mechanism (B) is the following. As this channel rightly documents, since in the
model the tradable sector determines the profit rate of the whole system, the price of the non‐tradable
good will rise to accommodate a higher profit rate of industry NT : otherwise, production of this com-
modity will not be convenient. However, under the specific conditions addressed by NS, the rise in
pNT is not expected to be the outcome of the change in the composition of output in favour of sector T
. To see this, it is enough to consider that sectorial effectual demands will not necessarily change in
the direction predicted by the approach. If any, internal demand for good T will decrease, since its
demand price has risen relative to the money wage, while, on the other hand, there is no reason to
expect an increase in foreign demand, since its international price has not changed. Therefore, unless
once again one resorts to the arbitrary assumption of an infinitely elastic demand for exports,36 there
is no reason to expect that the temporarily higher profitability in sector T will permanently increase its
share in total output.37

35
 Rapetti's “tradable‐led growth” channel (Rapetti, 2016, p. 259) includes both the effects on the composition (B) and the
level (C) of investment.
36
 It could be also assumed that the demand for T comes from capitalists' consumption, whose income has increased due to
devaluation. This assumption, however, would not be less arbitrary. And one should not overlook that, while the real income
of sector T‐capitalists has risen, that of sector NT‐capitalists has been reduced by depreciation. Thus, one must further assume
that T‐capitalists' propensity to consume is higher than NT‐capitalists' propensity.
37
 Of course, at the level of the theory of normal prices, it is not possible to exclude the possibility of a temporary rise in the
production of T due to its higher profitability. All that can be determined by the theory are the fully adjusted positions before
and after devaluation. On this point, see Parrinello (1990) and more recently, Fratini and Naccarato (2016).
 |
16       DVOSKIN et al.

It could be finally argued that foreign demand for commodity T rises because devaluation has per-
manently diminished domestic costs in foreign currency, and therefore the domestic economy gains
markets at the expense of its foreign competitors. This claim would be however overlooking that,
under the assumptions of the canonical model, this difference between costs of production and inter-
national prices cannot but be transitory (and in fact will be eliminated due to the rise in profitability
caused by devaluation); otherwise, one would be forced to admit that domestic production conditions
would eventually regulate the international price of the tradable good, hence contradicting the as-
sumption of a small price‐taking country that is at the core of the canonical model.
The main difficulty with channel (C) is that it needs devaluation to be able to persistently alter
income distribution in favour of profits. However, even new‐structuralist scholars document two pos-
sible reasons why this need not be the case, both due to wage resistance. First, since NS generally
assumes that the non‐tradable sector exhibits decreasing returns to scale (see e.g., Rapetti, 2013, p. 11
and Equation (6) of p. 12), devaluation, through a higher non‐tradable output, raises pNT . Therefore,
workers may attempt to re‐establish their initial real wage, which, due to the assumption of homoge-
neous wages across sectors, ultimately erodes profitability in the tradable sector:

↑ yNT →↑ pNT →↑ w →↓ r (WR1)

Second, due to the expansionary effect of devaluation in both sectors, unemployment (U) decreases
and workers' stronger bargaining power should allow them to claim for higher nominal wages:

↑ yT ∧ ↑ yNT →↓ U →↑ w →↓ r (WR2)

While, under the assumption that output increases due to devaluation, channel (WR2) is plausible
enough as it stands, the action of (WR1) demands some qualification. As it is presented by NS, the
argument does not seem sufficiently robust, since it needs the arbitrary assumption that sector NT ,
typically services, operates under decreasing returns to scale. However, if one accepts the plausible
claim that the potential increase in pNT may trigger wage resistance, the risks that the latter may annul
the initial rise in profitability are more severe than what NS seems to acknowledge. The reason is the
following: the moment one recalls that the normal profit rate of the economy is determined by sector
T , the rise in pNT will be the necessary consequence of the tendency to equalization of profits rates.
And, as we have argued above, for this to happen there is no need for an increase neither in the com-
position nor in the scale of output. Since no changes in the quantities produced by NT are actually
needed for pNT to rise, wage resistance can indeed exert its effects much faster than what is believed
by the proponents of the approach and therefore, the alleged virtuous process of growth may even
vanish before gaining momentum.38
Hence, when both WR channels are duly considered, it seems safe to conclude that, over sufficient
time, two counterbalancing forces operate on accumulation: on the one hand, the initial rise in the
profit rate fosters capital accumulation. On the other hand, the faster growth of nominal wages tends
to erode normal profitability. Thus, the final effect on the rate of accumulation is indeterminate, a dif-
ficulty that is even accepted by some new‐structuralist authors (see, for instance, Rapetti, 2013, p. 21).

38
 A further weakness of channel (C) is related to the assumption that aggregate investment positively depends on normal
profitability. This is however hard to justify, since it assumes that the trajectory of expansion of output capacity is indepen-
dent of the behavior of final demand for consumption goods. For reasons of space, but also because it escapes the aims of the
present paper, we cannot deal with the issue in detail here (see, among others, Freitas & Serrano, 2015; Pariboni, 2016).
DVOSKIN et al.   
   17
|
As we have anticipated in the introduction, there is a fourth expansionary mechanism, also of
“structural change”, that works both in the medium and long runs. It is based on the possibility of
diversifying the productive structure by incorporating new tradable sectors that, before devaluation,
were not internationally competitive. “A more depreciated RER [real exchange rate]”, Frenkel and Ros
(2006, p. 635) argue, “encourages tradable activities that were not profitable before”. If we consider
for instance sector C, whose supply price, under the productive structure assumed by NS, is higher
than its demand price, then devaluation raises the latter in a greater proportion than the former, and
thereby its profitability, until both prices are equal. And this allows domestic production of C to start.
Schematically,

( )
↑ e →↑ pdC < psC → pdC = psC → yC > 0 (D)

Unlike the previous three channels, this mechanism does admit an influence of income distribution
in the productive structure. The problem, however, is that it does not seem very easy to accommodate
within the canonical model. Not only because the latter assumes a given pattern of trade (this indeed
may explain why this mechanism is simply stated, but never developed in detail). More problematic,
in any case, is that this mechanism implicitly supposes that the diversification of the structure is not
damaging for the existing sectors of the economy. But this forgets that devaluation, through its effect
on r, also changes relative costs of production, and therefore the relative convenience of investing in a
specific sector. We have seen in fact that, flukes apart, that is, in the intersection of the e − r curves, the
international competitiveness of one sector can only be achieved at the expense of the other. Anyway,
we will return to channel (D) in the following section, when we see that, under conditions of differen-
tial rent, it is indeed possible to diversify the productive structure in the sense adduced by NS.

4  |   T H E WOR K ING OF T H E G ROW TH CHANNEL S IN

LAT IN A M ER ICA N P RO D U C T IV E STRUCTURES

4.1  | The e–r relationship under differential rent

Let us now examine additional problems with the channels discussed in the previous section, once we
consider a pattern of specialization more suitable for Latin American countries. We will focus on the
e − r relation of the previous sections, central to trigger the action of the expansionary mechanisms of
devaluation. Consider, then, a productive structure such that T produces and exports a primary con-
sumption good (the tradable sector is sector C). The industry, on the other hand, cannot compete
abroad at given international prices.39
Before we proceed to formalize the new conditions of production of the tradable sector, it is neces-
sary to make the following observation. Differently from the analytical framework of Section 2, it is
assumed that, due to the existence of a fixed factor in the production of C, typically land, the potential
extra profits accrued by producers operating in the sector when pdC > psC are not eliminated through a
rise in the profit rate; rather, they are eventually appropriated by landowners in the form of differential
rent, 𝜌40:

39
 For a similar characterization of productive structures of developing countries, see Botta (2010).
40
 Note that differential rent can emerge even if there is no marginal land within the domestic economy. On this point see
Birolo (1981).
 |
18       DVOSKIN et al.

𝜌 = Ep∗C − pSC > 0 (22)

This presents, in turn, the following difficulty. In the canonical model of Section 2, for a given e,
the profit rate was univocally determined by the equality between supply and demand prices of the
tradable commodity (given the money wage, the real wage was also endogenously determined once
pNT and pdT were known). When the equality between supply and demand prices no longer holds, a
given e is not enough to determine r, and therefore, the system gains an additional degree of freedom.
Formally, price Equations (1)–(2C) and (22) have the following unknowns: e, r, 𝜌, psC , pNT (implicit in
(22) there is Equation (3C), i.e., pdC = Ep∗C). Therefore, besides e, it is necessary to fix from outside the
price system either (a) the profit rate, or (b) the real wage. In turn, alternative (a) allows for at least
two different possibilities: (a1) for a given normal profit of enterprise, we can assume that r follows
the pace of the money interest rate fixed by the Central Bank (see Panico, 1988; Pivetti, 1991); or (a2)
there is some mechanism that, for a given interest rate, endogenously determines r as a function of
e (for instance, the mechanism that, as we have seen at the end of Section 2, is envisaged by NS see
Equation (14)). On the other hand, closure (b) can be formalized, for instance, by fixing the price of
the wage basket of the average worker as the numéraire.
It is convenient to examine each of the three possible cases in detail. Alternative (a1) formally
supposes that:

r = r̄ (23A)

Here, given r by (23A) and e, Equation (1) determines pNT and Equation (2) determines pSc. The
magnitude of the rent is determined by (22).
Alternative (a2) can be formalized as:

r = g (e) , r� (.) > 0 (23B)

Thus, a given e determines r through (23B), and hence, it is possible to determine, through condi-
tions (1) and (2), respectively, pNT and psC, while the magnitude of the rent is again determined by (22).
Finally, when c = (cNT ;cc ;cI ) represents the unitary wage basket, closure (b) can be formalized by:

∑
P ≡ cNT pNT + E cT p∗T = 1 (23C)
T=C, I

And therefore, a given w̄ determines the quantity of baskets affordable by each worker. In turn,
given e, Equation (23C) determines pNT , while now condition (1) has the role to determine the profit
rate. Equations (2) and (22) again determine psC and 𝜌.41 It should be noticed that, with this closure of
the price system, a rise in e decreases pNT . And hence, the consistency of the system further requires,
for a given money wage, a decrease in r. All this means that r = f (e), with f � (e) < 0.
Figure 3 shows the interaction among distributive variables in the presence of differential rent in
sector C, and hence, when there is an additional degree of freedom left (the fact that the feasible

41
 Recall that under this productive structure, commodity I is still consumed by workers, but it is not domestically produced.
Thus, its supply price is not included among the unknowns.
DVOSKIN et al.   
   19
|

F I G U R E 3   Wages, sectorial profit rates and land rent

distributive configurations are now illustrated by the grey area in the figure rather than by a curve,
shows this additional degree of freedom). The following features can be seen from the figure. For a
given e = ē : (i) ē > ê , since C is assumed to be the most profitable tradable sector; (ii) there is a range
of possible values for the effective profit rate, r̄ , that are compatible with the assumed productive
structure. This rate is necessarily below rC (e) and above rI (e). This means that, while sector C is per-
sistently “supra‐competitive”, and hence yields a rent, commodity I will not be profitability
produced.42
In Figure 3,43 the rent can be observed, so to speak, in two different ways: (i) given ē , through the
difference between the effective profit rate, r̄ , and the maximum profit rate, rC (̄e), that industry C
could afford; (ii) for a given r̄ , as the difference between e, and the highest level (eC ) that the sector
could support.
The possibility of differential rent in sector C has important consequences for the transmission
channels adduced by NS. With the additional degree of freedom, the necessary connection between
the money wage in foreign currency and the profit rate shown in Section 2, is lost. The implication
is that the effect of a rise in e over r becomes indeterminate: if, jointly with e, the profit rate is deter-
mined by condition (23A), it is expected that the profit rate will not change with devaluation. And it
can even decrease if the real wage is exogenously given through condition (23C). To be clear, we are

42
 Actually, r̄ can also be over rI (e). In this case, sector I earns the normal profit rate, and therefore, can coexist with the
rentistic sector C (see Section 4.2 below). On the contrary, if this rate was below rI (e) , the distributive configuration would
not persist, since it would imply that sector I would be earning a higher profit rate than sector C, and therefore, its supply
price would necessarily rise.
43
 The technical coefficients of production that support the shapes of these curves are the same as those of Figure 1, with
ē = r̄ = 2, rC (̄e) = 11∕5 and eC = 9∕5.
 |
20       DVOSKIN et al.

not excluding that devaluation may positively affect the profit rate; this is in fact what happens under
the closure formalized through condition (23B). The important thing, however, is that differently from
the canonical model of Section 2, when the tradable sector works under conditions of differential rent,
this positive relationship loses the necessary character that is needed by NS to establish a sufficiently
general causal relationship between the real exchange rate and economic growth.
To sum up, then, it is expected that when e rises, the level of the rent rises too, while the profit rate
can either rise, fall or remain constant. A first implication is that, even if the transmission channels
work in the “right direction”, devaluation will not generally be able to trigger their action. Some fur-
ther implications will be addressed in the following section.

4.2  |  A reexamination of structural change channel (D)

We have seen at the end of Section 3 that NS envisages a fourth mechanism, of structural change
(channel (D)), which however does not easily fit within their canonical model, since we have shown
that it is not possible to incorporate a tradable sector without negatively affecting the normal profit-
ability of the existing one. However, the coexistence of tradable sectors is possible when one of them
works under conditions of differential rent and therefore, this channel may work in the direction ad-
duced by NS.
Let us see this in further detail. If, for a given e, sector C yields the profit rate r̄ , it is possible to
determine the minimum level of e that would allow the other tradable sector (in this case, industry I )
to yield this same level of the profit rate. This level, eI , is no other than the level that equalizes supply
and demand prices for this sector, and is determined by the following condition:

(1 + r̄ ) lI
eI = (24)
1 − (1 + r̄ )kI

Note that at the level eI both sectors would earn the same profit rate, r̄ , and therefore could coex-
ist (besides the normal profit rate, sector C would also earn a differential rent). But then, this would
further seem to suggest that, for sector I to be internationally competitive, a devaluation of magnitude
ΔeI is needed, with:

ΔeI = eI − ē (25)

Equation (25) provides a sort of sectorial index of competitiveness. In graphical terms (Figure 4).44
The problem here emerges when one attempts to calculate, for practical purposes, the required
magnitude of devaluation for a generic tradable industrial sector T , since this needs very carefully con-
sidering both the direction and magnitude of the possible interactions among distributive variables. In
fact, the attentive reader may have already noticed that the magnitude of ΔeI in (25) is a function of
the normal profit rate, whose behaviour, as we have seen in Section 4.1, must be ascertained case by
case. Therefore, at least the following three scenarios are conceivable:
Scenario 1: consider first the case in which the profit rate is an increasing function of e, g (e), as
in Equation (23B). The initial distributive configuration is represented by point A in Figure 5. In this

44
 The technical coefficients of production that support these shapes of the curves are the same ones than those used in
Figure 1 with ē = r̄ = 2 and eI = 3.
DVOSKIN et al.   
|
   21

F I G U R E 4   “Required” devaluation of sector I

F I G U R E 5   Depreciation and pattern of specialization under scenario 1

 |
22       DVOSKIN et al.

case, Equation (25) underestimates the magnitude of devaluation to include sector I in the productive
structure, since it does not consider the endogenous movement of r when e varies. In the figure, the e
—coordinate of point B, eI , represents the required exchange rate according to (25), while the corre-
sponding of point C, e∗, indicates the higher level of the exchange rate effectively needed, once the
change in normal profitability according to g (e)—whose possible trajectory is depicted by the dotted
black arrow—is considered.
As a result, Figure 545 shows two possible outcomes. If the effective exchange rate is initially raised
up to eI , either devaluation stops there and the final distributive configuration is reflected by D (which,
differently from B, does capture the effective interaction between r and e). This only causes a decrease
in the money wage in foreign currency—and in the real wage— and a rise in r (and eventually in 𝜌),
but it is unable to modify the prevailing productive structure. Alternatively, the policymaker, who is
decided to transform the productive structure, attempts to further devalue the currency to complete the
effective path of g (e), described by trajectory DC. However, in this latter case, since the effective
magnitude of the required devaluation may be considerably higher than the one originally projected,
the possibility that the decrease in the real wage is so drastic that workers are prevented from consum-
ing the minimum quantity of necessary consumption goods, cannot be excluded. As necessary goods,
however, they will inevitably exert an influence on income distribution, borrowing an expression from
Sraffa (1960, p. 10), “in devious ways (e.g., by setting a limit below which wages cannot fall)”. And
this could be exerted, for instance, through wage resistance. To have an idea of this possibility, assume
𝛼 measures the minimum quantity of the necessary consumption good, C. In this case, this threshold,
e𝛼, is simply equal to 𝛼1.
Scenario 2: policymakers could be tempted, when there is more than one possible industrial sector
in the economy, to encourage the generic industry M in Figure 6, either because it is considered “stra-
tegic” for the economy; or simply because, on the basis of (25), it is the sector that exhibits “compar-
ative advantages”; namely the sector that is believed to require the lowest fall in the wage expressed in
foreign currency to be profitably produced, and which is therefore seen as the least costly in distribu-
tive terms. To this end, e is raised from its initial level, ē , up to eM. However, if the effective distributive
interactions are instead reflected by conditions (23C)—and by the corresponding function r = f (e)
(with f � (e) < 0)—it is conceivable that the rise in e ends up promoting, since it negatively affects
normal profitability, the less‐desired sector N . In Figure 6,46 this is shown by point C, whose e—com-
ponent is the inverse of the minimum level of the wage in foreign currency compatible with the given
real wage, eC, that also respects the condition of zero “extra profits” of the operating industrial sectors.
Below this level, in fact, one of these two conditions is necessarily violated, and therefore the attempt
to further devalue the currency is bound to fail.
Scenario 3: Finally, the most problematic scenario seems to emerge in the following case. Suppose
distributive conditions are such that, besides the rentistic sector C, there simultaneously is another
industrial sector N that is already competing abroad. However, due to its higher number of linkages
with the remaining sectors of the economy, policymakers decide to develop a different industrial
sector M . The economy is initially at point A in Figure 7 and therefore, the policymaker uses compet-
itiveness index (25) to determine the level of e needed for sector M to earn the ruling profit rate, r̄ , eM
. Therefore, he fails to capture the actual relationship between e and r, reflected by the function g (e)

45
 The technical coefficients of production that support the shapes of these curves are the same as those of Figure 1, with
A = (2;2) , B = (3;2) , C = (4;11∕5) and D = (3;21∕10).
46
 The technical coefficients of production that support the shapes of these curves are the same as those of Figure 1, where
sector M is the one previously labeled as sector I . Technical coefficients of sector N are lN = 19∕63 and kN = 1∕8. Point
( )
C = (9∕4;9∕5) and function f (⋅) is defined by cNT ,cM , cN = (5∕23;2∕23;2∕23). Finally, r̄ = ē = 2 and eB = 3.
DVOSKIN et al.   
|
   23

F I G U R E 6   Depreciation and pattern of specialization under scenario 2

F I G U R E 7   Depreciation and pattern of specialization under scenario 3

 |
24       DVOSKIN et al.

in the figure. Here exchange rate policy faces two potential problems. First, this rate of devaluation
( )
fails to develop sector M, since at the higher level eM, the actual profit rate rises to g eM , which is
unaffordable by the sector. Second, and perhaps more important, this policy also excludes the existing
( )
sector N from the productive structure, since g eM is higher than the maximum profit rate affordable
by industry N (rN ). The conclusion is that devaluation policy not only does not improve, but also de-
teriorates national competitiveness.

5  |   CO NC LUSION S
Motivated by mixed empirical evidence, we have critically explored the transmission channels from
real exchange rate undervaluation to economic growth adduced by NS. To this end, we have slightly
modified the new‐structuralist canonical model, and endogenized the productive structure of the
economy. Our results can be summarized as follows: a first limitation of the canonical model is that
despite the assumption of free financial capital mobility, the closure of the model is incompatible with
full equalization of profit rates across countries, which not only is undoubtedly restrictive but also,
to avoid an overdetermination of the profit rate, requires the further assumption that differences in
countries' net profit rates are entirely explained by movements in the exchange rate.
We have subsequently shown that the transmission channels work under very restrictive assump-
tions. First, the short‐term mechanism assumes an infinitely elastic demand for exports and takes for
granted that productive methods or sectors can be ordered in terms of their capital/labour ratio inde-
pendently of distribution. Second, for the medium‐term mechanism to work, one is forced to assume
either that there is an infinitely elastic demand for exports that validates the shift in the composition
of output in the direction adduced by NS, or alternatively, to abandon the assumption of a small
price‐taking economy, which is at the heart of the approach. Finally, for the long‐term mechanism to
reach definite results, one must neglect the effects of wage resistance, which, when duly considered,
are even more serious than what NS seems to acknowledge. In sum, the positive results do not resist
minor modifications to the canonical model, for example, that sector NT produces a capital good
employed by the tradable sector, the existence of some limits to foreign demand or the possibility of
wage resistance.
On the other hand, when the tradable sector works under conditions of differential rent, as it is the
case in many Latin American countries, the positive relationship between the real exchange rate and
the profit rate does not necessarily hold, and therefore, even if the above‐mentioned problems with the
expansionary mechanisms are neglected, the latter need not be activated by devaluation. Moreover,
although under these conditions the exchange rate may be used as a tool to boost a particular sector
of the economy, the potential risks of this policy should not be underestimated. In particular, we have
shown that not only does this require identifying a priori, and with high precision, all the possible
interactions—both in their direction and magnitude—among the distributive variables of the system.
We have further shown that the rise of the exchange rate may even have an effect opposed to the one
pursued and erode the profitability of already‐competitive productive sectors.

ACKNOWLEDGEMENTS
We are grateful for the financial support provided by the National Agency for Science and
Technological Promotion (ANPCYT, Argentina). Draft versions of this article were presented at the
2nd New Developmentalism's Workshop, “Theory and Policy for Developing Countries”, São Paulo,
August 4–5, 2017, at the Research Seminar of the Institute of Economics—Federal University of Rio
DVOSKIN et al.   
   25
|
de Janeiro (IE‐UFRJ), August 8, 2017 and at the Economics Research Seminar of the Department of
Economics—Karl‐Franzens‐Universität Graz, March 5, 2019. We would like to thank the participants
of these events for their comments and suggestions. One of these preliminary versions was published
as a working paper in the Centro Sraffa Working Paper Series, Roma Tre University. We would also
like to specially thank Saverio Fratini, Christian Gehrke and two anonymous referees of this journal
for their constructive comments and suggestions. All the remaining errors are ours.

R E F E R E NC E S
Acar, M. (2000). Devaluation in developing countries: Expansionary or contractionary? Journal of Economic and
Social Research, 2(1), 59–83. https​://doi.org/10.1.1.517.3860
Aguirre, A., & Calderón, C. (2005). Real exchange misalignments and economic performance. Central Bank of Chile
Working Papers, 315, 1–56.
Ahmed, S. (2003). Sources of economic fluctuations in Latin America and implications for choice of exchange rate
regimes. Journal of Development Economics, 73, 181–202. https​://doi.org/10.1016/S0304-3878(03)00073-7
Baldone, S. (2001). A comment on Steedman. In T. Cozzi & R. Marchionatti (Eds.), Piero Sraffa's political economy:
A centenary estimate (pp. 359–361). London, UK: Routledge.
Bielschowsky, R. (2009). Sesenta años de la CEPAL: Estructuralismo y neoestructuralismo. Revista CEPAL, 97, 173–
194. https​://doi.org/10.18356/​d8b97​556-en
Birolo, A. (1981). Mezzi di produzione non prodotti ed distribuzione del redito in una economía aperta. Studi Economici,
16, 115–135.
Botta, A. (2010). Economic development, structural change and natural resource booms: A structuralist perspective.
Metroeconomica, 61(3), 510–539. https​://doi.org/10.1111/j.1467-999X.2009.04077​
Braun, O., & Joy, L. (1968). A model of economic stagnation: A case of study of the Argentine economy. The Economic
Journal, 78(312), 868–887. https​://doi.org/10.2307/2229183
Bresser Pereira, L. C. (2008). The Dutch disease and its neutralization: A Ricardian approach. Brazilian Journal of
Political Economy, 28(1), 7–71. https​://doi.org/10.1590/s0101-31572​00800​0100003
Chou, W. L., & Chao, C. C. (2001). Are currency devaluations effective? A panel unit root test. Economics Letters,
2001, 19–25. https​://doi.org/10.1016/s0165-1765(01)00399-8
Couharde, C., & Sallenave, A. (2013). How do currency misalignments' threshold affect economic growth? Journal of
Macroeconomics, 36, 106–120. https​://doi.org/10.1016/j.jmacro.2012.11.002
Damill, M., & Frenkel, R. (2017). Symposium: Dilemmas of exchange rate and monetary policies in Latin America.
Journal of Post Keynesian Economics, 40, 1–8. https​://doi.org/10.1080/01603​477.2017.1343079
Diaz Alejandro, C. (1963). A note on the impact of devaluation and the redistributive effect. Journal of Political
Economy, 71(6), 577–580. https​://doi.org/10.1086/258816
Dvoskin, A., & Fratini, S. (2016). On Samuelson‐Etula master function and the capital controversy. European Journal
of the History of Economic Thought, 23(6), 1032–1058. https​://doi.org/10.1080/09672​567.2016.1186920
Dvoskin, A., & Petri, F. (2017). Again on the relevance of reverse capital deepening and reswitching. Metroeconomica,
68(4), 625–659. https​://doi.org/10.1111/meca.12137​
Economic Commission for Latin America and the Caribbean (ECLAC). (2007). Progreso técnico y cambio estructural
en América Latina. Santiago, Chile: División de Desarrollo Productivo, ECLAC.
Economic Commission for Latin America and the Caribbean (ECLAC). (2018). Perspectivas del comercio internacio-
nal de América Latina y el Caribe, 2018 (LC/PUB.2018/20‐P). Santiago, Chile: Author.​
Fratini, S. M., & Naccarato, A. (2016). The gravitation market process as a stochastic process. Metroeconomica, 67(4),
698–716. https​://doi.org/10.1111/meca.12116​
Freitas, F., & Serrano, F. (2015). Growth rate and level effects, the stability of the adjustment of capacity to demand
and the Sraffian supermultiplier. Review of Political Economy, 27(3), 258–281. https​://doi.org/10.1080/09538​
259.2015.1067360
Frenkel, R., & Ros, J. (2006). Unemployment and the real exchange rate in Latin America. World Development, 34(4),
631–646. https​://doi.org/10.1016/j.world​dev.2005.09.007
Garegnani, P. (1983). The classical theory of wages and the role of demand schedules in the determination of relative
prices. American Economic Review, 73(2), 309–313.
 |
26       DVOSKIN et al.

Goncalves, C. E., & Rodrigues, M. (2017). Exchange rate misalignment and growth: A myth? Department of Economics,
FEA/USP Working Paper Series, 07, 1–14. https​://doi.org/10.5089/97814​84330​135.001
Guzmán, G., Ocampo, J. A., & Stiglitz, J. E. (2018). Real exchange rate policies for economic development. World
Development, 110(2018), 51–62. https​://doi.org/10.1016/j.world​dev.2018.05.017
Kamin, S. B., & Klau, M. (1997). Some multi‐country evidence on the effects of real exchange rates on output. BIS
Working Papers, 48, 1–15. https​://doi.org/10.2139/ssrn.96613​
Kim, G., An, L., & Kim, Y. (2015). Exchange rate, capital controls and output: Developed versus developing countries.
Atlantic Economic Journal, 43(2), 195–207. https​://doi.org/10.1007/s11293-015-9458-2
Krugman, P., & Taylor, L. (1978). Contractionary effects of devaluation. Journal of International Economics, 8, 445–
456. https​://doi.org/10.1016/0022-1996(78)90007-7
La Marca, M. (2010). Real exchange rate, distribution and macro fluctuations in export‐oriented economies.
Metroeconomica, 61(3), 510–539. https​://doi.org/10.1111/j.1467-999X.2009.04077​
Levy‐Yeyati, E., & Sturzenegger, F. (2007). Fear of appreciation. World Bank Policy Research Working Paper Series
No. 4387. The World Bank. https​://doi.org/10.1596/1813-9450-4387
Lewis, A. W. (1954). Economic development with unlimited supply of labour. The Manchester School, 22(2), 139–191.
https​://doi.org/10.1111/j.1467-9957.1954.tb00021
López, J., Sanchez, A., & Spanos, A. (2011). Macroeconomic linkages in Mexico. Metroeconomica, 62(2), 356–385.
https​://doi.org/10.1111/j.1467-999X.2010.04114​
Marques Ribeiro, R. S., McCombie, J. S. L., & Tadeu Lima, G. (2017). Some unpleasant currency‐devaluation arithmetic
in a post Keynesian macromodel. Journal of Post Keynesian Economics, 40, 145–167. https​://doi.org/10.1080/01603​
477.2016.1246949
Mejía‐Reyes, P., Osborn, D. R., & Sensier, M. (2010). Modelling real exchange rate effects on output performance in
Latin America. Applied Economics, 42(19), 2491–2503. https​://doi.org/10.1080/00036​84070​1858117
Missio, F. J., Jayme, F. G., Britto, G., & Luis Oreiro, J. (2015). Real exchange rate and economic growth: New empirical
evidence. Metroeconomica, 66(4), 686–714. https​://doi.org/10.1111/meca.12087​
Mitchell, A. J., & Pentecost, E. J. (2001). The real exchange rate and the output response in four EU accession countries.
Emerging Markets Review, 2, 418–430. https​://doi.org/10.1016/s1566-0141(01)00027-9
Miteza, I. (2006). Devaluation and output in five transition economies: A panel cointegration approach of Poland,
Hungary, Czech Republic, Slovakia and Romania, 19933–2000. Applied Econometrics and International
Development, 6(1), 77–86.
Neto, A. S. M., & Lima, G. T. (2017). Competitive exchange rate and public infrastructure in a macrodynamic of eco-
nomic growth. Metroeconomica, 68(4), 792–815. https​://doi.org/10.1111/meca.12143​
Nouira, R., & Sekkat, K. (2012). Desperately seeking the positive impact of underdevaluation on growth. Journal of
Macroeconomics, 34(2012), 537–552. https​://doi.org/10.1016/j.jmacro.2011.12.002
Oskooee, M. B., & Hajilee, M. (2010). On the relation between currency depreciation and domestic investment. Journal
of Post Keynesian Economics, 32(4), 645–660. https​://doi.org/10.2753/pke01​60-34773​20409​
Oskooee, M. B., & Kandil, M. (2011). Are devaluations contractionary in MENA countries? Applied Economics, 41(2),
139–150. https​://doi.org/10.1080/00036​84060​0994195
Oskooee, M. B., & Miteza, I. (2006). Are devaluations contractionary? Evidence from panel cointegration. Economic
Issues, 11(1), 49–64.
Panico, C. (1988). Interest and profit in the theories of value and distribution. London, UK: Macmillan. https​://doi.
org/10.1007/978-1-349-09199-7
Pariboni, R. (2016). Autonomous demand and the Marglin‐Bhaduri model: A critical note. Review of Keynesian
Economics, 4(4), 409–428. https​://doi.org/10.4337/roke.2016.04.04
Parrinello, S. (1990). Some reflections on classical equilibrium, expectations and random disturbances. Political
Economy‐Studies in the Surplus Approach, 6(1–2), 113–124.
Pivetti, M. (1991). An essay in money and distribution. London, UK: Macmillan. https​://doi.org/10.1007/978-1-349-
21334-4
Rapetti, M. (2013). Macroeconomic policy coordination in a competitive real exchange rate strategy for development.
Journal of Globalization and Development, 3(2), 1–31. https​://doi.org/10.1515/jgd-2012-0009
Rapetti, M. (2016). The real exchange rate and economic growth: some observations on the possible channels. In
M. Damill, M. Rapetti, & G. Rozenwurcel (Eds.), Macroeconomics and development: Roberto Frenkel and the
DVOSKIN et al.   
|
   27

economics of Latin America (pp. 250–268). New York, NY: Columbia University Press. https​://doi.org/10.7312/
dami1​7508-014
Razin, O., & Collins, S. M. (1997). Real exchange rate misalignments and growth. NBER Working Papers Series, 6174,
1–33. https​://doi.org/10.3386/w6174​
Razmi, A. (2007). The contractionary short‐run effects of nominal devaluation in developing countries: Some neglected
nuances. International Review of Applied Economics, 21(5), 577–602. https​://doi.org/10.1080/02692​17070​1474611
Razmi, A. (2012). The exchange rate, diversification and distribution in a modified Ricardian model with a continuum
of goods. The Manchester School, 81(3), 356–385. https​://doi.org/10.1111/j.1467-9957.2012.02283.x
Razmi, A., Rapetti, M., & Skott, P. (2012). The real exchange rate and economic development. Structural Change and
Economic Dynamics, 23(2), 151–169. https​://doi.org/10.1016/j.strue​co.2012.01.002
Rodrik, D. (2008). The real exchange rate and economic growth. Brookings Papers on Economic Activity, 39(2), 365–
439. https​://doi.org/10.1353/eca.0.0020
Ros, J. (2016). The real exchange rate, the real wage, and growth: A formal analysis of the development channel. In
M. Damill, M. Rapetti, & G. Rozenwurcel (Eds.), Macroeconomics and development: Roberto Frenkel and the
economics of Latin America (pp. 229–249). New York, NY: Columbia University Press. https​://doi.org/10.7312/
dami1​7508-013
Ros, J., & Skott, P. (1998). Dynamic effects of trade liberalization and currency overvaluation under conditions of in-
creasing returns. The Manchester School, 66(4), 466–489. https​://doi.org/10.1111/1467-9957.00118​
Serena, J. M., & Sousa, R. (2017). Does exchange rate depreciation have contractionary effects on firm‐level invest-
ment? BIS Working Papers, 624, 1–34. https​://doi.org/10.2139/ssrn.3139024
Siqueira Machado, P. (2017). A Relacao Salario‐Cambio, Distribucao de Renda e Precos Relativos (Unpublished mas-
ter's thesis). Rio de Janeiro, Brazil: UFRJ‐Instituto de Economia.
Smithin, J. (2002). Interest parity, purchasing power parity, “risk premia” and Post Keynesian economic analysis.
Journal of Post Keynesian Economics, 25(2), 219–235.
Sraffa, P. (1926). The laws of returns under competitive conditions. The Economic Journal, 36(144), 535–550. https​://
doi.org/10.2307/2959866
Sraffa, P. (1960). Production of commodities by means of commodities. Cambridge, UK: Cambridge University Press.
Steedman, I. (1999). Production of commodities by means of commodities and the open economy. Metroeconomica,
50(3), 260–276. https​://doi.org/10.1111/1467-999x.00073​
Tadeu Lima, G., & Porcile, G. (2013). Economic growth and income distribution with heterogeneous preferences on the
real exchange rate. Journal of Post Keynesian Economics, 35(4), 651–674. https​://doi.org/10.2753/pke01​60-34773​
50407​
Upadhyaya, K. P. (1999). Currency devaluation, aggregate output and the long run: An empirical study. Economics
Letters, 64(1999), 197–202. https​://doi.org/10.1016/s0165-1765(99)00088-9
Woodford, M. (2008). Comments on the real exchange rate and economic growth. Brookings Papers on Economic
Activity, 2008(2), 420–437.

How to cite this article: Dvoskin A, Feldman GD, Ianni G. New‐structuralist exchange‐rate
policy and the pattern of specialization in Latin American countries. Metroeconomica.
2019;00:1–27. https​://doi.org/10.1111/meca.12265​