A study on the determinants of the real exchange rate misalignments^*

Um estudo sobre os determinantes dos desalinhamentos da taxa de câmbio real

ABSTRACT

Departing from the literature that established a positive relation between exchange rate undervaluation and economic growth, this paper explores the causes of persistent undervaluation (or overvaluation) of the exchange rate. We first discuss the variables that could explain persistent misalignment of the real exchange rate on a theoretical basis. Then we present empirical evidence of the relation between those variables in selected country groups. We proceed by employing econometric models to estimate the overvaluation or undervaluation based on a panel dataset comprising 119 countries from 1980 to 2022. Our findings indicate that countries with higher savings rates, government debt, and openness to trade tend to exhibit a more undervalued currency, the opposite happening to countries with better terms of trade and larger capital inflows.

KEYWORDS:
Real exchange rate; misalignments; undervaluation; growth

RESUMO

Partindo da literatura que estabeleceu uma relação positiva entre a subvalorização cambial e o crescimento econômico, o presente trabalho explora as causas de uma subdesva­lorização (ou sobrevalorização) cambial persistente ao longo do tempo. Primeiramente realizamos uma discussão teórica das variáveis que poderiam determinar tal viés permanente na taxa de câmbio real. Em seguida apresentamos evidências empíricas não econométricas da relação entre subvalorização e um conjunto de determinantes possíveis, para grupos selecionados de economias. O texto prossegue empregando modelos econométricos para estimar a subvalorização ou sobrevalorização, baseado num painel de 119 países, de 1980 a 2022. Os resultados econométricos indicam que países com maiores taxas de poupança, dívida pública e abertura tendem a ter moedas mais subvalorizadas, o inverso ocorrendo com países que contam com melhores termos de troca e maior influxo de capitais.

PALAVRAS-CHAVE:
Taxa de câmbio real; desalinhamento cambial; subvalorização; crescimento

1. INTRODUCTION

The long-term macroeconomic performance of the world economy, over the last four decades, as presented in statistics from the International Monetary Fund (IMF, 2023) reveals remarkable differences between regions and countries, notably the gap among growth rates of different groups of emerging economies. The per capita income of Emerging and Developing Asia, for instance, increased 9.4 fold between 1980 and 2022, while that of Latin America grew by only 34% in the same period.

Many factors can explain such large differences among countries’ growth rates. Growth in the stock of physical and human capital, and increased productivity are certainly the decisive factors. But the conditions for such factors to materialize depend on institutions, cultural habits, and economic policies. In the new-developmentalist theory, one of these factors is the existence of a macroeconomic framework favorable to growth, with an alignment of basic prices, with emphasis on the exchange rate. In a classic text, Rodrik (2008) sought to show, empirically, that not only was an appreciated exchange rate detrimental to growth but also that an undervalued real exchange rate contributed significantly to the growth of low and middle-income economies. A vast empirical literature followed confirming Rodrik’s results, as shown in a survey by Rapetti (2020).

Assuming the validity of this proposition, it is worth investigating which factors contribute to a country’s exchange rate misalignments. This issue is particularly important because, as we will see later, there are countries and groups of countries that tend to have exchange rates systematically depreciated (while others appreciated) in the long term.

Bresser-Pereira (2009) discussed the tendency towards over appreciation of the exchange rate in developing economies, underscoring the so-called Dutch disease as a determining factor, along with the attraction of external capital through high interest rates and the use of the exchange rate as an anchor to fight high inflation. The argument is essentially theoretical. In this paper, we explore this issue from an empirical point of view. Using databases from the IMF (World Economic Outlook, International Financial Statistics and Balance of Payments Statistics), BIS (Data Portal), and WTO (International Trade Statistics), we estimate models for determining the overvaluation (or undervaluation) of the real exchange rate based on a set of structural and economic policy variables. In particular, we investigate whether an above than average savings rate, terms of trade, openness, influx of capital, and relative participation of commodities in exports may help explain real exchange rate misalignments. We investigate the empirical association between real exchange rate misalignments and macroeconomic conditions.

We investigate, in particular, whether an above than average savings rate, terms of trade, debt level, openness, influx of capital, and relative participation of commodities in exports may help explain real exchange rate misalignments.

Our results indicate that the savings rate, the debt level, and the degree of openness have a positive effect on devaluations, while terms of trade, the relative participation of commodities in exports and capital inflow have a negative effect, on average. The partial positive effect of the savings rate on devaluations seems stronger for emerging economies that are not oil producers and have a floating or an intermediary exchange rate regime, an average impact of 0,62% compared to 0,51% for the whole sample. We also find that for this group of countries, interest rate differentials have a greater effect on misalignments when the currency is undervalued vis-à-vis overvalued. That is, the expected appreciation effect of an increase in capital influx is stronger, on average, if the local currency is undervalued.

This article is organized into six sections, including this introduction. Section 2 presents a brief discussion about the concepts of equilibrium exchange rate, over and undervaluation, that will serve as the basis for our empirical analysis. In section 3, we discuss the determinants of exchange rate misalignments. Section 4 presents some empirical (non-econometric) evidence on the relationship between exchange rate undervaluation and its determinants, for three groups of economies: emerging and developing Asian countries, Latin American, and some European countries. In section 5, we first present a general econometric model that describes the relationship among undervaluation, structural fundamentals, and macroeconomic conditions, as well as the data used to estimate the models. Completing the section, we discuss the results. Finally, section 6 concludes the article.

2. EQUILIBRIUM EXCHANGE RATE, EXCHANGE RATE OVERVALUATION AND UNDERVALUATION¹ (THE ELUSIVE CONCEPT OF EQUILIBRIUM EXCHANGE RATE)

When discussing the effect of a depreciated exchange rate on growth, there is a preliminary conceptual issue that consists of defining the reference value against which the level of exchange rate may be seen as appreciated or depreciated. That is, we have to discuss the slippery notion of the equilibrium level for the exchange rate.

Driver and Westaway (2004) argued that the concept of an equilibrium exchange rate depends on what seems relevant to each analyst and his/her question of interest.

A notion of an equilibrium exchange rate that has gained support among economists who seek to measure the degree of misalignment of the real exchange rate is the BEER (Behavioral Equilibrium Exchange Rate) proposed by Clark and MacDonald (1998). The popularity of their model is probably related to its ability to predict the exchange rate for periods outside the sample, with a greater degree of accuracy than a “random walk”. The BEER constitutes a response to the work of Meese and Rogoff (1983) according to which no economic model predicted the exchange rate better than a statistical random walk model.

Although representing an advance in terms of predictive capacity, models based on BEER may not be a good choice when the question of interest consists of measuring exchange rate misalignment - exchange rate under or overvaluation.

In Brazil, some studies have used the BEER model either to show that “exchange rate in Brazil has evolved according to its fundamentals” (BNDES, 2016), or to assess which factors are responsible for departures from the equilibrium exchange rate. In this type of model, the objective is to predict the real exchange even if at the expense of a large econometric specification. In that sense, the model is better as it can track the real exchange rate with a certain degree of accuracy. Ideally, only random (unpredictable) facts would lead to deviations of the real exchange rate from its predicted value. We want to understand misalignments, therefore, BEER type models do not provide an adequate concept of an equilibrium exchange rate for this article. Our question of interest involves factors that may help explain a systematic undervaluation (overvaluation) of the real exchange rate in relation to an equilibrium position, which brings greater (lesser) external competitiveness and better (worse) current account balance and affects economic growth. We are not interested in temporary deviations from the equilibrium exchange rate, whichever level that might be. Given this objective, it does not make sense to estimate the equilibrium rate based on a model that seeks to track, over time, the actual real exchange rate, as is the case with BEER.

At first, three alternative notions of equilibrium exchange rate seemed to us to be the most appropriate for our objectives: purchasing power parity, adjusted by the Balassa Samuelson effect, Williamson’s fundamental equilibrium exchange rate (FEER), and the industrial equilibrium exchange rate taken as a reference by the new-developmentalist theory (Marconi, 2012; Bresser-Pereira, 2012).

Our choice fell on the first alternative for two reasons. First, most studies that tested the relationship between exchange rate undervaluation and growth estimated undervaluation as the difference between the real exchange rate and its PPP value, adjusted by the Balassa-Samuelson effect, so that using this concept facilitates comparison of results. The second reason is related to data availability. The scarcity of data on certain variables and the operational difficulties involved in calculating the equilibrium exchange rate in the latter definitions, for many countries and for a long period, favored the PPP adjusted by the Balassa-Samuelson effect as the equilibrium real exchange rate.

Before proceeding, it is worth noting that, in a world with varying inflation rates, the most appropriate notion of equilibrium exchange rate to assess the competitiveness of an economy, is the real exchange rate. This can be defined as:

$Q=\frac{EP\text{'}}{P}$ (1)

Where Q is the real exchange rate, E is the nominal exchange rate measured as the value of the foreign country’s currency in terms of the domestic country’s currency, P* is the price of a broad basket of goods and services in the reference country, measured in its currency, and P is the price of a broad basket of goods, equivalent to the external one, in the domestic country. The ratio therefore represents the relationship between the prices of external and internal baskets of goods and services, measured in the same currency. The higher Q, the higher the relative price of external goods, and therefore the more competitive domestic goods are, which is equivalent to saying that the domestic currency is depreciated in real terms.

According to the purchasing power parity theory (PPP), in its absolute version, Q=1. In the relative version, adjusted by the Balassa-Samuelson effect, the equilibrium real exchange rate, Q, is different from 1. How much above (or below) 1 will depend, in this model, on the relative productivity of the economy concerning its reference commercial partner (in practice the USA) and, ceteris paribus, on relative per capita income.

In (1), P and P* are weighted averages of the prices of tradable and non-tradable goods. Therefore, assuming that the prices of tradable goods are determined internationally, and prices of non-tradable goods are lower (measured in the same currency) the lower the country’s average income level, in relation to that of the reference country; we can conclude that EP*/P will be smaller the higher the country’s per capita income, that is:

$Q=f\left(Y_r\right);\frac{dQ}{d{Y}_r}<0$ (2)

Where Yr is the ratio of the per capita income of the domestic country to the per capita income of the USA (reference country).

The relationship between real exchange rate and per capita income, represented in (2) is empirically robust and known as the “Penn effect” (Devereux, 2014; Isard, 2007). This relationship can also be taken as a proxy for the Balassa-Samuelson effect², according to which the increase in the relative productivity of the domestic country in the production of tradables leads to an appreciation of the real exchange rate (reduction in Q) of the domestic country.

The literature on the effects of exchange rate misalignments on growth shows that some countries tend to experience under or overvalued real exchange rates (in the concept above) for long periods. Our interest lies in these long-term deviations (or misalignments) so that we will focus on structural variables that might explain long-term misalignments.

3. DETERMINANTS OF EXCHANGE RATE OVERVALUATION/UNDERVALUATION: THEORETICAL ASPECTS

Assuming the existence of an equilibrium real exchange rate as defined above, we can anticipate that a deviation from equilibrium will be influenced by factors that lead to either a surplus or a deficit in the supply or demand for foreign exchange at that exchange rate level. The supply and demand for foreign currency in the foreign exchange market manifest in the balance of payments, encompassing both the current account and the financial account. Thus, we can analyze the drivers of the surplus (or deficit) of foreign currency separately by examining these two components of the balance of payments.

Capital flows, as measured in the financial account of the balance of payments, assumed a significant role following the liberalization of capital movements in the 1980s (in advanced economies) and 1990s (in many emerging economies). Large capital inflows, when not regulated by prudential policies, have often triggered prolonged processes of exchange rate appreciation in developing economies (Ghosh, Ostri, and Qureshi, 2017). Hence, this variable is expected to contribute to exchange rate over (or under) appreciation. However, research by Combes, Kinda, and Plane (2011) demonstrates that the effects of various flow modalities on the exchange rate vary significantly, with the portfolio category exerting an impact nearly seven times greater than other modalities, such as direct investment.

The fixed income securities component of portfolio flows can be analyzed using the risk-adjusted uncovered interest parity equation, which can be written as:

$\begin{array}{c}{Ê}^e=\left(i-i*\right)-\rho \\ or\\ \frac{E_{t+1}^e}{E_t^s}-1=\left(i-i*\right)-\rho \end{array}$ (3)

Where Ê ^e =expected exchange rate change; $E_t^s$=spot exchange rate at t; i=domestic interest rate; i*=external interest rate; ρ=country risk.

Temporary and cyclical fluctuations in the interest rate differential and country risk are unlikely to significantly impact long-term exchange rate expectations; rather, their influence typically manifests in the spot exchange rate. Conversely, persistent and systemic disparities between domestic and foreign interest rates tend to result in long-term exchange rate appreciation. Conversely, consistently higher country risk, stemming perhaps from a rise in the public debt/GDP ratio, tends to lead to a depreciated real exchange rate. Consequently, these factors tend to create a prolonged disparity between the prevailing exchange rate and the equilibrium rate determined by Purchasing Power Parity (PPP) adjusted by the Balassa-Samuelson effect. In essence, as mentioned earlier, our focus lies solely on factors contributing to long-term deviations from the equilibrium exchange rate, disregarding cyclical or short-term influences. Hence, disparities in interest rates (both domestic and foreign) stemming from monetary policy are not considered.

However, persistent disparities in interest rates have been observed (MacDonald, 2000) and should be accounted for either by the anticipated future exchange rate or the risk premium. These factors necessitate an explanation beyond the interest rate parity condition (Driver and Westaway, 2004). Structural factors such as investment opportunities or the slow adjustment of the current account, as well as institutional and political weaknesses leading to consistently high-country risk, must be considered. Examining Equation (3), we can deduce that enduring positive interest rate differentials, linked to the appeal of a country’s economy, typically contribute to an appreciated exchange rate. Conversely, elevated country risk, perhaps associated with high public debt, tends to result in a depreciated exchange rate relative to equilibrium.

Now, let’s redirect our focus to current account flows and their determinants. The outcome of current transactions is primarily associated with macroeconomic patterns of savings and investment:

$S-I=TC=F$ (4)

Where S represents domestic savings, I denotes investment, TC is the current account balance and F indicates the financial account balance or, alternatively, the economy’s external financing capacity.

Equation (4) suggests that a surplus of savings over investment typically results in an excess demand for external assets and thus a rise in the demand for foreign currency (to procure external assets), which, all else being equal, leads to an exchange rate devaluation. Consequently, this contributes to an increase in net exports and thereby to a surplus in current transactions, ultimately leading to a balance of payments equilibrium in the medium term. This approach to exchange rate analysis is commonly referred to as the macroeconomic balance framework (Isard, 2007). If the exchange rate does not depreciate enough to generate a surplus in the current account (equivalent to the potential excess of savings over investment), it will lead to an aggregate demand shortfall in the economy, resulting in a recession. Other mechanisms may then come into play to compel depreciation, such as net capital outflows associated with the recession, or in extreme cases, aggregate price reductions that would foster real exchange rate depreciation.

A depreciated exchange rate is therefore the instrument that induces a surplus in the current account in a country that has a high savings rate (higher than investment) and the corresponding surplus in the financial account.

In summary, all else being equal (including the investment rate), we can anticipate that a rise in the savings rate will correlate with a more devalued exchange rate. However, some authors have identified a reverse effect, whereby an undervalued exchange rate stimulates savings (Gala and Rocha, 2011). The preceding argument was formulated under the implicit assumption of constant prices, including those of the terms of trade. However, if export prices surge relative to import prices (increasing in terms of trade), the current account balance will improve, all else being equal, leading to an excess supply of foreign currency in the foreign exchange market. Consequently, this would trigger an exchange rate appreciation that, over time, would diminish net exports to their previous levels. In essence, for a given current account balance, an enhancement in the terms of trade tends to foster an appreciation of the real exchange rate (Coudert, Couharde, and Mignon, 2008). Furthermore, if the escalation (or decline) in the terms of trade is not merely a cyclical occurrence, it tends to appreciate (or depreciate) the exchange rate in the long term. A factor closely related to the previous one is the phenomenon known as the “Dutch disease” (Corden, 1984). A surge in the production of extractive industries (for instance, due to the discovery of new resources), characterized by highly profitable exploration and significant potential for competitive exports, can yield the same outcome in the current account (compared to pre-boom levels) even with a more appreciated exchange rate. Hence, it is reasonable to assume that a substantial share of natural resource-based products in the export mix tends to reduce the real exchange rate below the equilibrium level as defined earlier.

Moreover, concerning the prices of exported and imported goods, the level of commercial protectionism can be a crucial factor. An increase in import tariffs, by elevating import prices for domestic consumers, diminishes imports and bolsters the trade balance. Assuming all other factors remain constant, this fosters a surplus of dollars in the foreign exchange market, thereby leading to exchange rate appreciation. Conversely, a policy aimed at augmenting trade openness tends to promote exchange rate depreciation and, consequently, all else being equal, an increase in exchange rate undervaluation (Li, 2003; Gantman and Dabós, 2018). Indeed, all the aforementioned factors interconnect balance of payments flows with the exchange rate. Nevertheless, the accumulated flows in each period influence a country’s stocks of external assets and liabilities, representing its net external wealth, or International Investment Position, in the accounting terminology of the IMF (International Monetary Fund, 2009). The analytical treatment of the relationship between net external wealth and the exchange rate (known as the stocks approach) originated from the debate between Keynes and Ohlin concerning the transfer problem (Lane and Milesi-Ferretti, 2000). The question can be framed as follows: a country with a negative International Investment position (i.e., a net external liability) tends to experience a deficit in the primary income account (comprising interest, profits, etc.) of the balance of payments. Consequently, it needs to generate a compensatory surplus in its balance of goods and services, necessitating a more depreciated exchange rate than would be required if its Net Foreign Assets were greater than or equal to zero.

Lane and Milesi-Ferretti (2000) provided empirical evidence of the transfer effect, indicating that debtor countries tended to have more depreciated exchange rates compared to others. The primary transmission channel for this effect was found to be the price of non-tradables rather than the terms of trade. In essence, an outflow of foreign exchange (such as interest payments) would exert upward pressure on the exchange rate and, consequently, on the relative price of tradables, while simultaneously exerting downward pressure on the relative price of non-tradables. However, caution should be exercised when applying Milesi and Ferretti’s findings to economies with fixed exchange rates or a common currency. Comunale (2019) extended the Lane and Milesi-Ferretti model to the case of EU countries, including the specific scenario of Euro area countries, where the adoption of a common currency implies that real exchange rate misalignment solely stems from changes in internal price levels. Her conclusions reveal highly diverse outcomes among countries, with the transfer effect linked to net foreign assets not consistently manifesting. Table 1 summarizes the variables and their effects on exchange rate undervaluation based on theoretical assumptions discussed above.

Table 1
Determinants of Under/Over Valuation of the Exchange Rate

Variable	Expected Effect	Selected References
Capital Inflows (+)	Overvaluation	Combes, Kinda, and Plane (2011); Ghosh, Ostri and Qureshi (2017)
Interest Rate Differential (>0)	Overvaluation	MacDonald (2000); Driver and Westaway, 2004
Country Risk (Public Debt) (+)	Undervaluation	Driver and Westaway, 2004
Excess Saving	Undervaluation	Isard (2007); Gala and Rocha (2011)
Terms of Trade (+)	Overvaluation	Coudert, Couharde and Mignon (2008)
Dutch Disease	Overvaluation	Corden (1984); Bresser-Pereira (2009)
Trade Liberalization	Undervaluation	Li (2003); Gantman and Dabós (2018)
Net External Wealth (LIabilities)	Overvaluation (Undervaluation)	Lane and Milesi-Ferretti (2000); Comunale (2019)

4. DETERMINANTS OF EXCHANGE RATE UNDERVALUATION: STYLIZED FACTS AND NON-ECONOMETRIC STATISTICAL EVIDENCE FOR SELECTED REGIONS

In order to provide a comprehensive overview of the relationship between exchange rate undervaluation and various structural and macroeconomic determinants, the following table presents data for groups of emerging countries from three regions: Asia, Latin America, and Europe. The selection of these regions was primarily based on the significant disparity in experiences among countries in Emerging Asia and those in Latin America regarding the subject matter of our study. Additionally, we included the group of countries from Emerging and Developing Europe as it represents an intermediate case in terms of the variables of interest when compared to the other two regions. The critical variable - exchange rate undervaluation - was estimated here in a straightforward manner. We computed the percentage difference between the average real exchange rate of each group of countries and the average real exchange rate of emerging and developing economies. Positive values indicate an undervalued exchange rate compared to the emerging market average, while negative values indicate an overvalued rate, expressed as a percentage of appreciation. Our comments will primarily concentrate on comparing Emerging Asia and Latin America. However, it’s noteworthy that the results for Emerging Europe consistently represent an intermediate scenario between these two extremes.

The initial observation, in line with the empirical literature indicating a positive relationship between growth and exchange rate undervaluation (Rapetti, 2020), underscores a stark contrast between Emerging Asia and Latin America. Emerging Asia exhibits an average undervaluation of 6.6% and an average growth rate of 7% per annum during the period 1999-2022, whereas Latin America experiences an average real exchange rate appreciation of 21.3% and very modest growth of 2.2%. Importantly, these trends persist over time, as demonstrated in Tables A1 and A2 in the Appendix, which detail the values of these two variables by decade from 1980 to 2022. The exception, as previously noted, is Latin America’s experience in the 1980s³.

The other five variables outlined in Table 2 appear to be correlated with the initial two and are thus potential factors in explaining either undervaluation or the rate of economic growth. Subsequently, we will discuss these relationships for the regions depicted in Table 2, drawing from the theoretical underpinnings of open macroeconomics outlined in the preceding section.

Table 2
Exchange Rate Undervaluation and Macroeconomic Variables: Average 1999-2022

Variable	Emerging and developing Asia	Latin America and the Caribbean	Emerging and developing Europe
Exchange Rate Undervaluation(1)	6,6	21,3	5,2
Rate of Growth (% yearly)	7,0	2,2	3,6
Savings Rate (% of GDP)	39,4	19,0	23,7
Current Account (% of GDP)	2,2	1,5	0,4
Inflation Rate (Average CPI, %)	3,8	6,5	12,8
Share of Commodities in Total Exports (%)	16,7	45,7	43,3
Openness Coefficient (%)	52,8	43,2	64,9

• Note: (1) [(Average Real Exchange Rate for the Country Group/Average of the Real Exchange Rate of Emerging and Developing Economies)-1)x100. Positive values=undervaluation].
• Source: IMF, World Economic Outlook, Database, Oct. 2023. Authors’ elaboration.

Commencing with savings rates, the disparity between Emerging Asia and Latin America is substantial, with the former’s rate simply double that of the latter. This contrast is mirrored in their respective growth rates - a correlation grounded in growth theory, which also aligns with the levels of exchange rate undervaluation observed in each region. We have observed that the “savings effect” on the real exchange rate can manifest through various channels, ultimately resulting in a devaluation linked to a surplus in current transactions. Once more, the data presented in Table 2 underscores this relationship and highlights the contrast between Emerging Asia, which boasts an average current account surplus of 2.2% of GDP, and Latin America, which maintains an average deficit of 1.5% during the period 1999-2022.

Moreover, the value of the current account, when added to the capital account, equals the financial account. Given that the values of the capital account are generally very small compared to those of the other two flows, it’s feasible to roughly identify a positive balance in the current account accompanied by an outflow of capital (and conversely, a deficit accompanied by an inflow). This perspective allows us to perceive the previous relationship between the current account and exchange rate as a relationship between capital flows and exchange rate undervaluation. In Latin America, the substantial inflow of capital seems to have contributed to maintaining an overvalued exchange rate, while in Emerging Asia, the capital outflow appears to be consistent with an undervalued exchange rate. It’s important to note that while the financial account provides an approximate measure, it may not be the most accurate gauge of the pressure exerted by capital inflows (or outflows) on the exchange rate. This is because it encompasses the sum of private (or autonomous) capital flows as well as official flows (such as changes in reserves and transactions with the IMF). For our econometric model, outlined in the following section, we rely on the measure of autonomous capital to evaluate the influence of capital flows on the exchange rate.

The relationship between inflation and the aforementioned variables exhibits some ambiguity. On one hand, countries grappling with chronic high inflation often resort to monetary (and exchange rate) policies to appreciate the exchange rate as a means of curbing inflation. On the other hand, high inflation tends to breed macroeconomic and political instability, heightening economic uncertainty and deterring foreign capital, thereby contributing to keeping the currency depreciated.

In the aggregated data from the groups of emerging countries presented here, the contrast between Asia and Latin America is once again evident, with the former region (where the currency is, on average, undervalued) exhibiting much lower inflation than the latter (where the currency is, on average, overvalued). However, the results for Emerging and Developing Europe do not confirm this inverse relationship between inflation and undervaluation⁴.

The subsequent variable in Table 2 represents the “Dutch disease” effect and reveals a clear relationship: the greater the share of commodities in the export basket of country groups, the more overvalued the currency tends to be. In Latin America, commodities account for an average of 45.7% of total exports over the period 1999-2022, whereas in Asia, the significance of this group of products in total exports is merely 16.7% during the same period.

One might inquire why the variable “terms of trade” is absent from the table. The rationale is that its absolute value would pose challenges in comparison between countries, as it is measured as an index that permits comparability only within each country or region over time. Nonetheless, it is a variable linked with the “Dutch disease.” Extended periods of rising terms of trade for commodity-­exporting countries would typically result in exchange rate appreciation.

Lastly, concerning the external openness coefficient variable, the literature typically suggests that it contributes to augmenting the degree of undervaluation of the currency, as observed earlier. The rationale behind this is that an economy shielded by protectionist barriers, upon opening up to foreign trade, tends to increase imports and incur trade deficits. This induces exchange rate depreciation, which in turn compensates for the absence of protectionist barriers, thereby improving current transactions. The comparison between Latin America and Asia, as depicted in Table 2, is consistent with this proposition: an external openness coefficient of 52.8% for Emerging Asia and 43.2% for Latin America. In the following section, we introduce an econometric model that delineates the relationship among undervaluation, structural fundamentals, and macroeconomic conditions, as well as the data utilized to estimate the models. The findings are subsequently presented in section 6.

5. EXCHANGE RATE MISALIGNMENTS: CALCULATION AND STRUCTURAL DETERMINANTS

The concept of a misalignment relates to deviations of the real exchange rate from a value that would be consistent with macroeconomic balance, current account sustainability and underlying capital flows. Rodrik (2008) defines a misalignment index departing from the exchange rate, adjusted for purchasing power parity factors, and for the Balassa-Samuelson effect. His measure of under or overvaluation follows a three-step procedure. First, the author calculates the real exchange rate (RER) as the ratio between the nominal exchange rate (XRAT) and the purchase price parity factor (PPP):

$\ln RE{R}_{it}=\ln \left(XRA{T}_{it}/PP{P}_{it}\right)$

As both the nominal exchange rate and the PPP factor are expressed in local (national) currency per unit of US dollars, a rise in the real exchange rate (RER) indicates a depreciation of the local currency.

As a second step, Rodrik (2008) adjusts the RER for the Balassa-Samuelson effect by running a regression of the real exchange rate against the GDP per capita:

$\ln RE{R}_{it}=\alpha +\beta \ln GDPp{c}_{it}+f_t+u_{it}$

The estimated Beta coefficient is interpreted as the Balassa-Samuelson effect on real exchange rates⁵. We follow Rodrik, but we adjust the real exchange rate to the Balassa-Samuelson effect by using the ratio, and not only the country’s GDP per capita. That is, we estimate lnRER against the ratio of each country’s per capita income and the American’s income per capita observed for the same period. We believe that by using the ratio, we can better capture the differences in productivity suggested by the Balassa-Samuelson effect.

As a third step, the misalignment index (lnUnderval) is defined as the difference between the observed real exchange rate (lnRER) and its predicted value from the adjusted Balassa-Samuelson effect, as described above. That is, the measure of misalignment is given by:

$\ln Underva{l}_{it}=\ln RE{R}_{it}-\ln {\widehat{RER}}_{it}$

We estimate a misalignment index from a panel of 119 countries observed between 1980 and 2022. The results of the Balassa-Samuelson effects are reported in Table 3. For the complete sample, a 10% increase in GDP per capita is associated with an average decline of approximately 1.50% in the real exchange rate. Exchange rate movements follow a different logic when the country is an oil exporter. In the second column, we exclude countries classified by the IMF as oil producers from our sample. As expected, the Balassa-Samuelson effect is stronger for non-oil countries: a 10% increase in relative income per capita is associated with an average decrease of 2.02% in the real exchange rate⁶. For emerging markets (or developing countries, according to the IMF classification), a 10% increase in relative income is associated with an average decline of 0.054% in the real exchange rate, but this effect was not statistically significant. However, emerging countries include cases of non-floating exchange rate regimes, with countries that do not have control over their own currencies (currency boards or no separate legal tender). When we exclude the cases of currency boards and countries with no separate legal tender, the expected negative effect re-emerges, with a 10% increase in relative income associated with an average decline of 1.71% in the real exchange rate.

Table 3
Balassa-Samuelson Effect

Dependent Variable: ln RER
Sample	All	Non-oil	Emerging	Floating and intermediary regimes	Floating and intermediary regimes, Emerging, Non oil
lnGDPpc it /GDPpcUSA t	0,48037 [0.0174]***	0,338902 [0.0187]***	0,843352 [0.0092]***	0,472974 [0.0197]***	0,621885 [0.0346]***
	-0,128459 [0.0107]***	-0,18425 [0.0110]***	0,017595 [0.0066]***	-0,138817 [0.0118]***	-0,126595 [0.0219]***
Time effects	yes	yes	yes	yes	yes
Fixed effects	no	no	no	no	no
Observations:	6817	6028	3394	6205	2322
Adj. R-squared:	0,1275	0,1836	0,0941	0,1399	0,101
F-statistic:	24,1666	32,5221	9,1924	24,4666	7,0626
Prob(F-stat):	0,0000	0,0000	0,0000	0,0000	0,0000

• Notes: OLS panel, 177 countries, 1980-2022. White-cross section standard errors in brackets: ***p<0.01, **p<0.05, *p<0.1.

Finally, we restrict our sample to developing (or emerging) countries, with a floating currency regime (or an intermediary regime) that are not oil producers, and we obtain the following estimates for the Balassa-Samuelson effect: a 10% increase in relative income is associated with a 1.86% average decline in the real exchange rate. This group of countries will be our focus for the analysis that follows.

Given the importance of the relationship between exchange rates and growth, our second objective is to investigate which structural variables may possibly contribute to exchange rates being systematically misaligned. A factor highlighted in the literature is the so-called “Dutch disease”, as studied, in the Brazilian case, notably by Bresser-Pereira (2009). In this section, we estimate models for real exchange rate misalignments based on a set of structural and economic policy variables.

We estimate the following basic model:

$\ln Underva{l}_{it}=\alpha +{\beta }_{1}Sa{v}_{it-1}+{\beta }_{2}T{T}_{it-1}+{\beta }_{3}Deb{t}_{it-1}+{\beta }_{4}Ope{n}_{it-1}+{\beta }_{5}Comm/X_{it-1}+{\beta }_{6}Influ{x}_{it-1}+f_i+f_t+u_{it}$

lnUnderval is a measure of undervaluation of country i during the period t. We are interested in the relationship between real exchange misalignments and long-term structural variables so that our t period represents a moving average of 10 years. To control for potential endogeneity arising from simultaneity, all models are estimated with lagged regressors. According to our previous discussion, we should expect the following signs for the coefficients. The savings rate should have a positive effect on the undervaluation index, as the higher the savings rate, the more devalued the real exchange rate. We should also expect a positive effect from the government debt/GDP level and degree of openness of the economy; we should expect the terms of trade and the relative participation of the commodities in exports to have a negative effect on lnUnderval. Finally, the influx of capital should have a negative effect on the undervaluation index as a greater influx of capital tends to appreciate the local currency. We used two different variables to capture this effect: the autonomous capital inflow and the interest rate differential^7. The autonomous capital flow is defined as the negative of the current account added to the change in international reserves, all over GDP, and divided by the real bilateral exchange rate:

$F_A=\left(-CA+\Delta Reservers\right)/RER$

All models were estimated by OLS with country and time fixed effects so that non observed factors that are specific to each country can be controlled for, and macroeconomic changes that may have affected all countries but have not been explicitly included in the model may be captured by the time trend.

In all models, the coefficients for the savings rate, the debt/GDP ratio, and the degree openness are positive, as expected, and statistically significant. The coefficients for relative participation of commodities in exports are statistically significant but present the expected negative sign only in model (1). The autonomous capital flow has a negative estimated coefficient, as expected, but it is not statistically significant in our sample of interest (emerging countries, with a floating currency regime or an intermediary regime that are not oil producers); instead, the interest rate differential seems to be a better variable for these group of countries. Comparing the results of models (3) and (4), we see that in model (4), not only the estimated coefficient for the interest rate differential is negative and statistically significant, but also the coefficient for terms of trade becomes statistically significant, a result that is expected. The results for model (4) suggest that a one percentage point increase in the savings rate is associated with an increase of 0.44% in the real exchange rate, on average and given the other variables. For the interest rate differential, a one percentage point increase in the differential is associated with 1.36% reduction in the real exchange rate, on average and given all other variables.

Table 4
Structural Determinants of Misalignments

Dependent Variable:	ln Underval>0	ln Underval<0
Sample	Floating and intermediary regimes, Emerging, Non oil	Floating and intermediary regimes, Emerging, Non oil
C	-0,101978 [0.0364]***	-0,02372 [0.0752]
Savings/GDP (%)	0,001496 [0.0008]*	-0,003089 [0.0015]**
Surplus/GDP (%)	0,002605 [0.0014]*	0,006381 [0.0022]***
Terms of Trade	-0,001917 [0.0003]***	-0,00158 [0.0008]**
Debt/GDP (%)	0,001816 [0.0004]***	-0,000573 [0.0004]
Inflation Rate	0,005943 [0.0008]***	-0,000571 [0.0041]
Openness	0,219546 [0.0410]***	0,02432 [0.0103]**
Commodities/Exports (%)	0,004645 [0.0005]***	0,000698 [0.0007]
Autonomous Capital Flow/GDP	-0,008343 [0.0024]***	-0,002843 [0.0022]
Time effects	yes	yes
Fixed effects	yes	yes
Observations:	615	459
Adj. R-squared:	0,7739	0,8257
F-statistic:	25,7317	29,9372
Prob(F-stat):	0,0000	0,0000

• Notes: OLS panel, 177 countries, 1980-2022. White-cross section standard errors in brackets: ***p<0.01, **p<0.05, *p<0.1.

We are particularly interested in understanding what aspects are most related to an undervalued currency or positive lnUnderval. We then separated the effects on undervaluation (lnUnderval>0) and overvaluation (lnUnderval<0) for our sample of interest. The results are presented in Table 5.

Table 5
Structural Determinants of Misalignments. Floating and intermediary regimes, Emerging, Non-oil Countries.

Dependent Variable:	ln Underval>0
Sample	Floating and intermediary regimes, Emerging, Non oil
C	0,097074 [0.0316]***	-0,08394 [0.0329]**	0,173753 [0.0268]***
Savings/GDP (%)	0,000481 [0.0008]	0,001481 [0.0007]**	0,001397 [0.0007]*
Terms of Trade (TT)	-0,001556 [0.0003]***	-0,00186 [0.0003]***	-0,004421 [0.0004]***
Debt/GDP (%)	0,002596 [0.0005]***	0,001825 [0.0004]***	0,001991 [0.0004]***
Inflation Rate	0,006337 [0.0007]***	0,005926 [0.0008]***	0,00581 [0.0008]***
Openness	0,147793 [0.0324]***	0,222326 [0.0391]***	0,220019 [0.0390]***
Commodities/Exports (%)		0,004087 [0.0004]***
Autonomous Capital Flow/GDP	-0,009279[0.0025]***	-0,00925 [0.0022]***	-0,009274 [0.0023]***
Terms of Trade*Commodities/Exports			0,000039 [0.0000]***
Time effects	yes	yes	yes
Fixed effects	yes	yes	yes
Observations:	682	627	627
Adj. R-squared:	0,739	0,7717	0,7745
F-statistic:	24,2282	26,1858	26,5891
Prob(F-stat):	0,000	0,000	0,000

• Notes: OLS panel, 177 countries, 1980-2022. White-cross section standard errors in brackets: ***p<0.01, **p<0.05, *p<0.1

For the first column, undervaluations, we see that, except for Openness and the relative participation of commodities in exports, all coefficients have the expected sign and are statistically significant. For the savings rate, for example, a one percentage point increase in savings rate is associated with a devaluation of 0.7%, on average and given the other explanatory variables. Asian countries traditionally have a much higher savings rate when compared to Latin American, with a difference over 15 percentage points. If we were to apply a 15 percentage points, this would have an associated (real) devaluation of 10.5%, not a trivial value. For the second column, overvaluations, all but one variable (the relative participation of commodities in exports)⁸ have the expected sign and are statistically significant. Comparing the two columns, we note that except for the interest rate differential and the government debt-ratio, most coefficients are smaller in absolute value in the case of positive misalignments (devaluations).

6. CONCLUDING REMARKS

There is a vast empirical literature on the effects of exchange rate misalignment on economic growth. Rodrik (2008) analyzed specifically the relationship between undervaluation and economic growth for a sample of 188 countries and a span of time from 1950 through 2004 and found that a 50% undervaluation is associated with an increase in annual growth of income per capita of about 1.3 percentage point. A growing body of research confirmed his results since then.

A question arises, however, as to the determinants of persistent undervalued exchange rates in some countries while the opposite tendency is observed in others. As we showed in section 4 of the present paper, in Asian Emerging and Developing Economies, whose growth rates averaged 7% annually in the period 1999-2022, exchange rates were undervalued by an average of 6,6%, while in Latin America, with poor growth performance, the real exchange rate was overvalued by 21,3% in the same period. What factors could explain those persistent real exchange rate undervaluations or overvaluations?

In this article, we investigate the association between real exchange rate misalignments and structural macroeconomic conditions, and we estimate models for determining misalignments of the real exchange rate based on a set of structural and economic policy variables. For that we use the databases from the IMF (World Economic Outlook, International Financial Statistics and Balance of Payments Statistics), BIS (Data Portal) and WTO (International Trade Statistics). In particular, we investigate whether an above than average savings rate, terms of trade, openness, inflow of capital, and relative participation of commodities in exports may help explain real exchange rate misalignments.

Our findings indicate that the savings rate, public debt level, and degree of openness exhibit a positive impact on undervaluations, whereas terms of trade, and capital inflows demonstrate a negative influence, on average. Specifically, the partial positive effect of the savings rate on devaluations appears more pronounced for emerging economies that are non-oil producers and operate under a floating or intermediary exchange rate regime, with an average impact of 0.62% compared to 0.51% for the entire sample. Furthermore, we observe that within this subset of countries, interest rate differentials exert a stronger influence on misalignments when the currency is undervalued compared to when it is overvalued. In other words, the anticipated appreciation resulting from an increase in capital inflows is typically greater, on average, when the local currency is undervalued. These results are averages for large groups of countries. Further insights can be gleaned through additional investigation into the relationship among macroeconomic conditions, utilizing case studies.

REFERENCES

• BNDES (2016) Taxa de Câmbio no Brasil: o que Dizem os Fundamentos? Nota Informativa APE/DAE, VI, 2. Banco Nacional de Desenvolvimento Econômico e Social.
• Bresser-Pereira, L. C. (2012) “A Taxa de Câmbio no Centro da Teoria do Desenvolvimento”, Estudos avançados 26(75).
• Bresser-Pereira, L. C. (2009) A Tendência à Sopreapreciação da Taxa de Câmbio. Textos para Discussão, 183. FGV-EESP.
• Clark, P. B., and MacDonald, R. (1998) Exchange Rates and Economic Fundamentals: a Methodological Comparison of BEERs and FEERs. Working Paper, 98/67. International Monetary Fund.
• Combes, J.-L., Kinda, T., and Plane, P. (2011) Capital Flows, Exchange Rate Flexibility, and the Real Exchange Rate. IMF Working Paper, WP/11/9. International Monetary Fund.
• Comunale, M. (2019) “Long-Run Determinants and Misalignments of the Real Effective Exchange Rate in the EU”, Scottish Journal of Political Economy, 66(5).
• Corden, W. M. (1984) “Booming Sector and Dutch Disease Economics: Survey and Consolidation”, Oxford Economic Papers, 36(3), pp. 359-380.
• Coudert, V., Couharde, C. and Mignon, V. (2008) Do Terms of Trade Drive Real Exchange Rates? Comparing Oil and Commodity Currencies. CEPII Working Paper, No 2008 - 32 December.
• Devereux, M. B. (2014) Real Exchange Rates and the Balassa-Samuelson Effect Revisited. The Reporter, 4. Acessed on January 9, 2024. Available in Available in https://www.nber.org/reporter/2014number4/real-exchange-rates-and-balassa-samuelson-effect-revisited
  » https://www.nber.org/reporter/2014number4/real-exchange-rates-and-balassa-samuelson-effect-revisited
• Driver, R. L., and Westaway, P. F. (2004) Concepts of Equilibrium Exchange Rates. Working Paper, 248. Bank of England.
• Gala, P. and Rocha, M. (2011) Real Exchange Rates, Domestic and Foreign Savings: the Missing Link. Proceedings of the 37th Brazilian Economics Meeting from ANPEC (Brazilian Association of Graduate Programs in Economics)
• Gantman, E. R., and Dabós, M. P. (2018) “Does Trade Openness Influence the Real Effective Exchange Rate? New Evidence From Panel Time-Series”, Journal of the Spanish Economic Association, 9(1), pp. 91-113.
• Ghosh, A.R., Ostry, J. and Qureshi, M.S. (2017) Managing the Tide: How do Emerging Markets Respond to Capital Flows? IMF Working Paper WP/17/69.
• IMF (2023) World Economic Outlook, International Monetary Fund, October.
• IMF (2009) Balance of Payments and International Investment Position Manual. Sixth Edition (BPM6).
• Isard, P. (2007) Equilibrium Exchange Rates: Assessment Methodologies. IMF Working Papers, 07/296. International Monetary Fund.
• Lane, P. R., and Milesi-Ferretti, G. M. (2000) The Transfer Problem Revisited: Net Foreign Assets and Real Exchange Rates. IMF Working Paper, WP/00/123. International Monetary Fund.
• Li, X. (2003) Trade Liberalization and Exchange Rate Movement. Working Papers, 03/124. International Monetary Fund.
• MacDonald, R. (2000) Concepts do Calculate Equilibrium Exchange Rates: an Overview. Deutsche Bundesbank, Discussion Paper 3/00.
• Marconi, Nelson (2012) “The Industrial Equilibrium Exchange Rate in 2000: An Estimation”. Brazilian Journal of Political Economy 32(4): 656-69, 2012. https://centrodeeconomiapolitica.org.br/repojs/index.php/journal/article/view/375
  » https://centrodeeconomiapolitica.org.br/repojs/index.php/journal/article/view/375
• Meese, Richard A. and Kenneth S. Rogoff. (1983), “Empirical Exchange Rate Models of the Seventies: Are any fit to survive?”, Journal of International Economics 14: 3-24.
• Nassif, A., C. Feijó and E. Araújo (2015) “Overvaluation Trend of the Brazilian Currency in the 2000s: Empirical Estimation”. Brazilian Journal of Political Economy, vol. 35, nº 1 (138), pp. 3-27, January-March.
• Rapetti, M. (2020) “The Real Exchange Rate and Economic Growth: a Survey”, Journal of Globalization and Development, 11(2), pp. 1-54.
• Rodrik, D. (2008) The Real Exchange Rate and Economic Growth. Brookings Papers on Economic Activity.

• 1
  In this text, for the sake of simplicity, we will use the word pairs “depreciation/appreciation” and “devaluation/valuation” interchangeably. Although strictly speaking, the first term refers to movements determined by the market (endogenous), and the second one to changes promoted by economic policy (in regimes where the exchange rate is fixed by the government), for our argument, the distinction is not crucial and often results in less fluid language.
• 2
  Per capita income, that is, income (Y) divided by population (POP), can be decomposed into: Y/POP=Y/PO×PO/POP, where PO=Employed population. Therefore, if the employed fraction of the population (PO/POP) remains constant, per capita income increases in line with total productivity (Y/PO). The Balassa-Samuelson model refers to the effects of the increase in productivity in the tradable sector of the domestic country relative to that of the reference country. Hence, the more the aggregate relative productivity reflects the growth in the relative productivity of tradables, the better the proxy (per capita income) will be.
• 3
  Averages for a longer period, 1980-2022, are presented in the Appendix (Table A1). There we see that values are not substantially different from those presented in Table 2 above, except for the average growth of Latin America in the 1980s and European emerging countries in the 1990s. For these groups of countries, these periods were characterized by severe macroeconomic imbalances that involved debt crisis in Latin America and political transition in Europe. In both periods, these groups of countries experienced large devaluations associated with balance of payments and currency crises, along with prolonged recessions. The well-established positive association between real exchange rate and growth is in fact not observed during periods of acute crisis for these groups of countries. We believe, however, that episodes of low economic growth along with macroeconomic imbalances, political turmoil, and economic transitions do not reflect the structural relations we investigate here.
• 4
  In the estimation of our econometric model (outlined in the following section), we excluded this variable as it was not significant - likely due to the aforementioned ambiguity.
• 5
  The -0,24 coefficient estimated by Rodrik, through a panel with time dummies, indicates a strong Balassa-Samuelson effect: a 10% increase in GDP per capita is associated with an average decline of 2,4% in the real exchange rate.
• 6
  For oil countries, the estimated coefficient is 0.0085, positive and not statistically significant at 5% (not reported).
• 7
  We thank a participant of the 6th Workshop on New Developmentalism held at EESP-FGV for suggesting us the use of the interest rate differential.
• 8
  The reasons why we have the wrong sign for this variable demand further investigation. One of the possible causes is the fact that during downward cycles of commodity prices, a high proportion of commodities in exports should cause depreciation of the exchange rate.

• JEL Classification: F31; F41; F43.

APPENDIX

Table A1
Real Undervaluation of the Exchange Rate relative to the Average of Emerging Economies (%)⁽¹⁾

Período	ASEAN-5(2)	Emerging and Developing Asia	Latin America and the Caribbean	Emerging and Developing Europe
1980-89	15,5	-3,2	8,8	-
1990-99	8,1	13,8	-27,4	33,8
2000-09	19,7	12,5	-25,8	-1,0
2010-2019	20,6	2,0	-19,1	6,1
2020-22	20,7	-7,8	-4,1	19,3
Average 1980-2022	15,9	6,2	-16,1	14,0

• Notes: (1) [(Average Real Exchange Rate of the Country Group/Average Real Exchange Rate of Emerging Economies)-1]*100
• (2) Asean-5: Indonesia, Malaysia, the Philippines, Singapore, and Thailand.
• Source: IMF, Word Economic Outlook, Database, Oct. 2023. Authors’ elaboration.

Table A2
Annual Average Growth Rates of Real GDP by Country Groups (%)

Período	ASEAN-5	Emerging market and developing economies	Emerging and developing Asia	Latin America and the Caribbean	Emerging and Developing Europe
1980-89	5,8	3,2	6,7	2,1	2,2
1990-99	5,0	3,7	7,2	3,0	2,4
2000-09	4,9	6,0	8,1	2,9	4,8
2010-2019	5,2	5,0	7,0	1,9	3,1
2020-22	1,6	3,0	3,8	1,3	2,1
Average 1980-2022	5,0	4,4	7,0	2,4	1,9

• Note: (1) Asean-5: Indonesia, Malaysia, the Philippines, Singapore, and Thailand.
• Source: IMF, Word Economic Outlook, Database, Oct. 2023. Authors’ elaboration.

Table A3

Publication Dates

• Publication in this collection
  07 Oct 2024
• Date of issue
  2024

History

• Received
  24 Dec 2023
• Accepted
  25 Apr 2024