Civil Conflicts and Exchange Rate Misalignment: Evidence from MENA and Arab League Members

1 Introduction

Economic research has shown that macroeconomic stability is an important requisite to achieve sustainable economic growth and build prosperous, inclusive societies. Recently however, several Arab League member States and countries from the Middle-East have experienced episodes of high inflation and large nominal exchange rate movements. While countries from the region have long been affected by civil conflicts, these have been particularly intense in Iraq, Libya, Somalia, Sudan, Syria and Yemen in recent years. The adverse human and macroeconomic consequences of civil conflicts on the real sector have been extensively studied in the literature, especially in sub-Saharan Africa, but relatively less attention has been paid to the implications for the monetary and financial sectors.

In this paper, I combine macroeconomic data with data on civil conflicts occurrence and intensity to assess the effects of civil conflicts on real exchange rate (RER) misalignment in Arab League member countries, Iran and Turkey. A positive RER misalignment, or a real overvaluation, occurs when the RER is higher than its unobserved equilibrium, and does not necessarily correspond to a RER appreciation. Large RER misalignment would create or reinforce macroeconomic imbalances that could threaten the following post-conflict transition.

ElBadawi and Soto (2013b) show that while exchange rate regimes are persistent over time, a significant share of countries tend to adopt a more flexible exchange rate regime in post-conflict transitions. This denotes the need to correct the imbalances created by the conflict: high inflation is one of them and primarily affects the most vulnerable. Because civil conflicts negatively affect aid, capital inflows, agricultural output and exports, and because they lead to a higher demand for military goods and imports, they might lead to a RER depreciation. However, higher public spending and high inflation due to shortages in basic goods, among other reasons, could have an opposite effect and lead to a RER appreciation. Since the response of RER misalignment to civil conflicts is undetermined, I adopt an empirical approach.

Following Collier (1999), among others, the recent literature on the economic causes and consequences of civil conflicts is quite extensive and has identified a great variety of relationships, particularly focussing on sub-Saharan Africa and on country case studies. While the RER is increasingly recognised as an important determinant of growth and structural change (Guzman, Ocampo, and Stiglitz 2018; Rodrik 2008), little attention has been paid to the effects of civil conflicts on this variable.

The contribution of this paper is therefore threefold. First, it assesses the macroeconomic impacts of civil conflicts in the Middle-East and North Africa region, which has received relatively less attention in this literature. Second, it provides country-specific estimations of the RER misalignment over a long period, following the behavioural approach developed in Edwards (1989), ElBadawi (1994) and Clark and MacDonald (1998), among others. Third, while the computation and the macroeconomic effects of RER misalignment have been widely studied, this paper adds to a still scarce literature on the determinants of RER misalignment.

To test the hypothesis that RER misalignment is affected by civil conflicts, I use conflict data from the Center for Systemic Peace, real effective exchange rate data from Darvas (2012) and macroeconomic data mainly from the World Development Indicators, complemented by several other sources. The panel dataset has a yearly frequency, covers the period 1970–2018 and includes 19 Arab League member countries, Iran and Turkey. Libya, Palestine and Somalia are excluded from the sample due to missing data that impede the computation of their RER misalignment. Since the OLS estimates might be biased due to several sources of endogeneity, I adopt an instrumental variable approach and instrument the magnitude of civil conflicts in each country by the magnitude of civil conflicts in the neighbouring countries. The correlation between civil conflict domestically and in neighbouring countries has been established by the literature (see for example Phillips 2015), and I argue that the exclusion restriction is fulfilled since civil conflicts in neighbouring countries only affect international capital flows in the short-term through the increased probability of civil conflicts domestically.

The results show that civil conflicts in the region lead to RER overvaluations. Additionally to the threats posed on the post-conflict transition itself by the resulting dissatisfaction and social unrest, the outcome of the creation of new institutions can be affected and have lasting effects. If left unaddressed, RER overvaluation resulting from civil conflicts can therefore have effects in the long-run, beyond the business cycle, and sum to human and physical capital losses to hinder economic growth and development prospects.

The remainder of the article is organized as follows. Section 2 reviews the literature, Section 3 details the construction of the real exchange rate misalignment index and Section 4 presents the data and stylized facts. Section 5 introduces the empirical strategy and Section 6 presents and discusses the results. Finally, Section 7 concludes.

2 Review of the Literature

Civil conflicts have attracted much attention and are the object of a vast economic literature. Their occurrences have been analysed (Collier and Hoeffler 2002), as well as their social, economic and environmental consequences (Collier 1999; Devadas, ElBadawi, and Loayza 2019; Prem, Saavedra, and Vargas 2020), including their consequences on neighbouring countries (David et al. 2020). A large strand of this literature has focused on civil conflicts’ determinants (Collier and Hoeffler 2004). The occurence, the duration and the intensity of such conflicts are related to climate change and weather shocks (Burke et al. 2009, 2015; Mach et al. 2019), negative growth shocks (Miguel, Satyanath, and Sergenti 2004), income shocks (Dube and Vargas 2013), inequalities (Østby, Nordås, and Rèd 2009), rents (ElBadawi, Ibrahim, and Zaki 2020), or possibly humanitarian interventions in some cases (Nunn and Qian 2014), among other causes. Theoretical and empirical analysis also confirmed that country-specific dynamics in political economy affect civil conflicts’ duration and intensity (Fergusson, Vargas, and Vela 2013, 2016). Another strand of this literature has discussed the economic issues related to the crucial post-conflict transitions (Collier, Hoeffler, and Söderbom 2008; ElBadawi 2008; ElBadawi and Soto 2013b; Fergusson et al. 2021).

The literature on the economics of civil conflicts, especially when using cross-country data, has largely focused on sub-Saharian Africa. Several specific countries, such as Colombia, have also been widely studied. The Middle-East and North Africa is not absent, but it has received relatively little attention compared to the prevalence of such conflicts and their impacts in the region. By using data for Arab League members, Iran and Turkey, this paper contributes to fill this gap in the literature.

The literature that has assessed the economic consequences of civil conflicts has also mainly focused on real variables, such as human and capital loss, growth decline and migrations, for example. With notable exceptions, monetary and financial variables, which do not correspond to the most pressing preoccupations during a civil conflict, have received much less attention. ElBadawi and Soto (2013a), for example, find evidences that managed floating exchange rate regimes have a greater capacity to stabilize inflation in post-conflict transitions.

Several authors have argued that sustained and limited RER undervalations can be growth-enhancing by correcting market failures and fostering exports and structural change (Guzman, Ocampo, and Stiglitz 2018; Rodrik 2008). The RER is unlikely to be the main concern of any group involved in a civil conflict and growth-promotion is at best a distant objective in such a context, but large RER misalignments could weigh on the post-conflict transition and complicate the return to economic and social stability.

Several outcomes of civil conflicts could generate forces, sometimes working in opposite directions, that lead to RER misalignments. Staines (2004) finds that aid usually declines during civil conflicts. The same appears to be true for capital inflows (Li, Murshed, and Tanna 2017) and exports (ElBadawi and Soto 2013b). These shortages in the supply of foreign currency are expected to lead to RER depreciations and are reinforced by the higher demand for legal or clandestine imports of military goods, which constitute an increase in the demand for foreign currency. The U.S. civil war provides a famous historical example, when the then Secretary of Treasury secured three bank loans of fifty million U.S. dollars each to secure access to gold and therefore imports (Mehrling and Sandilands 1999). The decrease in agricultural sector intensity (Eklund et al. 2017) and agricultural output (Jaafar et al. 2015) further depresses exports and increases the demand for imports to satisfy subsistence requirements, making hard currency more valuable and depreciating further the domestic currency.

Civil conflicts also generate appreciating pressures on the domestic currency. Increased remittances and bilateral or multilateral military aid could limit the shortages of foreign currency. Miyamoto, Nguyen, and Sheremirov (2019) find that higher public spendings lead to a RER appreciation in developing countries, and public expenditures increase during civil conflicts, beyond the reallocation of health and education budget to military spending (Gupta et al. 2004). Furthermore, civil conflicts often lead to substantial inflation hikes (ElBadawi 2008). A lower demand for money and higher public financial needs, in an environment of reduced public revenues and access to financing, can account for this fact and explain why government’s tolerance for inflation increases during conflict, as shown in Adam, Collier, and Davies (2008). Together, these consequences of civil conflicts could lead to substantial RER overvaluation and large macroeconomic imbalances that could endanger the post-conflict transition.

Because these mechanisms can work in opposite directions and the response of RER to civil conflicts is uncertain, this paper adopts an empirical approach to assess the effects of civil conflicts on RER misalignment. By focussing on the monetary and financial consequences of civil conflicts, this papers contributes to fill this second gap in the literature.

3 Real Exchange Rate Misalignment

This section details the construction of the real exchange rate misalignment index, following a behavioural approach in the tradition of Edwards (1989), ElBadawi (1994), Clark and MacDonald (1998) and Goldfajn and Valdés (1999), among others. Alternatives approaches, namely the fundamentals and the external sustainability approaches, are possible but imply drawbacks in the case of developing countries. Noureldin (2018) discusses these limitations.

The behavioural approach, based on theoretical foundations, links a theoretical long-term real exchange rate equilibrium to net foreign assets, terms of trade and a Balassa-Samuelson effect, as well as other real exchange rate determinants. The misalignment is then defined as the difference between the observed real exchange rate and the estimated equilibrium real exchange rate. The main advantages of this approach is to incorporate the Balassa-Samuelson effect, instead of the relative prices of traded goods only, and not to rely on normative current account targets. The behavioural approach relies instead on an econometric specification that links the real exchange rate to its determinants, which is verified by construction. Bénassy-Quéré (2015) offers a textbook treatment of this notion and its theoretical foundations.

The first step to estimate the equilibrium RER consists in specifying an empirical model that expresses the RER as a function of a vector of its fudamentals. Equation (1) details the specification:

where t denotes the time and i the country, α _i a country-specific intercept, RER the real exchange rate (in log), Ω a vector of parameters and X′ a vector of fundamental RER determinants comprising the commodity terms of trade (in log), productivity differential with respect to OECD countries, proxied by the ratio of real per capita GDP over that of OECD countries (in log), to capture the Balassa-Samuelson effect, trade openness, government consumption, net foreign assets and the world interest rate, proxied by the Federal Reserve 1-Year Treasury Constant Maturity Rate. ɛ _i,t is the error term.

Several determinants of the RER, and the RER itself, might not be stationary. Furthermore, the RER and its determinants might be cointegrated. If this is the case, the estimator used to estimate equation (1) must take this characteristic of the data into account. Table 1 reports the results of the Im-Pesaran-Shin test (Im, Pesaran, and Shin 2003) and shows that the null hypothesis cannot be rejected for all variables except two, indicating that all panels contain unit roots.

Table 2 reports the statistics of the Kao (1999) and Pedroni panel cointegration tests (Pedroni 1999, 2004) for the RER and its determinants. Both tests share the null hypothesis of no panel cointegration between the variables. All test statistics of the Kao panel cointegration test lead to rejecting the null hypothesis, while the results of Pedroni’s test are more nuanced since two of them indicate that we can reject the null hypothesis with 90 % confidence only. Overall, these results suggest that the RER and its determinants are cointegrated, and the empirical strategy should include an error-correction mechanism in a panel setting.

Equation (1) is estimated following three econometric methods that allow for an error-correction mechanism when using panel data. The dynamic fixed effects (DFE) estimator restricts all short- and long-run parameters to be equal across countries, except for the intercept which is country-specific, and the mean group (MG) estimator allows both the short- and long-run parameters to be country-specific. The pooled mean group (PMG) estimator assumes that all countries share common long-run parameters but allows short-term parameters to differ across countries, and therefore offers the best compromise between consistency and efficiency. ElBadawi, Kaltani, and Soto (2012) offer a detailed description of each estimator and the trade-offs involved.

Table 3 presents the results of the estimation of equation (1) using the DFE, MG and PMG estimators. The signs of coefficients are consistent with the theory, except for net foreign assets which appears to be associated to a RER depreciation. Increases in the commodity terms of trade, productivity, the international interest rate and public consumption are associated to a RER appreciation, while increased trade openness is associated with a RER depreciation.

After estimating equation (1), the equilibrium exchange rate is computed as indicated in equation (2):

where X ═ ′ is a vector of the long-term values of the RER determinants, i.e. the trend components obtained using the Hodrick-Prescott filter, Ω ˆ is a vector of estimated parameters from equation (1), and δ ˆ   i = RER ―   i − β ˆ ′ X ─ ′   i is a country specific intercept that allows to normalize the equilibrium RER so that the long run RER misalignment is equal to zero in each country. RER ― and X ─ ′ correspond to the mean values of the RER and its determinants, respectively.

Equation (2) can therefore be rewritten as:

As evidenced in equation (3), the equilibrium RER corresponds to the mean value of the RER and a term that depends on the weighted difference between the trend components of the RER fundamentals and their mean values. The RER misalignment can then be defined and computed as the difference between the observed RER and the equilibrium RER:

As indicated in equation (4), the RER misalignment depends positively on the difference between the observed RER and its mean value, and depends negatively on the weighted difference between the trend components of the RER fundamentals and their mean values. Noting that RER i , t = α ˆ   i + Ω ˆ X i , t ′ + ε ˆ   i , t , that RER ―   i = α ˆ   i + Ω ˆ X ─ ′   i , that the cyclical component of the RER determinants can be written as X i , t ′ ∼ = X i , t ′ − X i , t ═ ′ and rearranging equation (4), the RER misalignment can be expressed as:

Equation (5) shows that the RER misalignment depends on the weighted cyclical components of the RER determinants and short-term shocks. The next section introduces the data and presence stylized facts on civil conflicts and RER misalignment in Arab League member states, Iran and Turkey.

4 Data and Stylized Facts

The country-level data cover 19 countries from the Arab League as well as Iran and Turkey, between 1970 and 2018. Libya, Somalia and Palestine are not included in the sample because missing data for these countries impede the computation of their RER misalignment index. Appendix Table A.1 details the list of countries included in the sample, and Appendix Table A.2 lists all the data sources used in this paper.

4.1 Civil Conflicts

The data for civil conflicts is retrieved from the Major Episodes of Political Violence (MEPV) dataset elaborated by the Center for Systemic Peace. This dataset differentiates civil violence from civil war based on the degree of militant organization, on tactical and strategic characteristics, and on the expressed level of commitment to the use of violence. The concept of war corresponds to stronger institutional or institutionalized components and more definite objectives. The distinction, however, necessarily implies a subjective judgement, and the same is true for the classification of an intrastate conflict as a civil or ethnic conflict. Therefore, the variable capturing total civil conflict includes both civil and ethnic conflicts, independently from their classification as violence or war.

The magnitude of each of these conflicts is coded on a 0 to 10 scale. A score of 0 denotes the absence of conflict, and scores from 1 to 10 denote increasing intensity of conflicts, from “Sporadic or Expressive Political Violence” to “Extermination and Annihilation”. Magnitude scores are considered consistent and comparable across categories and cases and can vary across year during a single conflict episode, allowing therefore to obtain a country-year indicator for civil violence, civil war and (total) civil conflict.

Figure 1 shows the evolution of the aggregate intensity of civil conflicts in 21 Arab League members, Iran and Turkey between 1970 and 2018 (data for Palestine is absent from the MEPV dataset). After attaining their lowest intensity in 1973, civil conflicts intensified from 1974, and more sharply from 1975, until reaching their highest intensity in 1991, at the end of the Cold War when the Soviet Union disintegration culminated. The intensity of civil conflicts sharply declined in the next decade, reaching a relatively low plateau in year 2000 that was maintained during the first decade of the 21st Century. However, the intensity of civil conflicts during this decade remained much higher than at the beginning of the 1970s. In 2011, massive political upheaval led to a new episode of intensification of civil conflicts in the region. The intensity of these conflicts was maintained during the entire decade, at a level comparable to those observed at the end of the 1970s and throughout the 1980s.

Figure 1:

Civil conflict intensity in Arab League members and MENA countries. Source: Center for Systemic Peace, elaborated by the author. The country-year conflict intensity index takes values between 0 and 10.

Figure 2 shows the aggregate intensity of civil conflicts for each country between 1970 and 2018. Civil conflicts intensity has been the highest in Iraq, Sudan and Somalia during the period and overall, Western Asia appears to have suffered a greater intensity of civil conflicts than North Africa, the Arabian Peninsula being the least affected area. Only six countries did not suffer any civil conflict during this period: Bahrain, Comoros, Kuwait, Qatar, Tunisia and the United Arab Emirates. Appendix Figure B.1 shows the complete time-series of civil conflict for each country.

Figure 2:

Total civil conflict intensity, 1970–2018. Source: Center for Systemic Peace, elaborated by the author. The country-year conflict intensity index take values between 0 and 10.

4.2 Real Exchange Rate Misalignment

Data for real exchange rates are retrieved from Darvas (2012) and correspond to the real effective exchange rates that include 66 trading partners. The determinants of the equilibrium RER are retrieved from several sources. The commodity terms of trade come from Gruss and Kebhaj (2019), the international interest rate is proxied by the U.S. 1-Year Treasury Constant Maturity Rate, retrieved from the Federal Reserve Economic Data, and data on net foreign assets (in % of GDP) and the productivity index, defined as the ratio of per capita nominal GDP to the OECD average per capita nominal GDP (in current USD), comes from the WDI dataset (World Bank – WDI 2020). The share of public consumption in GDP and trade openness, computed as the sum of the shares of merchandise imports and exports in GDP, come from the Penn World Tables version 9.1 (Feenstra, Inklaar, and Timmer 2015).

Figure 3 describes the evolution of RER misalignment in different geographic regions between 1970 and 2019. The RER misalignment shows a downward trend in North Africa, from a substantial overvaluation in the 1970s to a large undervaluation since the 2000s. It has been more volatile in Iran and Turkey, with a large overvaluation until 1992 and around 2000, and a large undervaluation in 1993–1995 and since 2002. In the Levant and Iraq, the RER misalignment appears to be substantially overvalued in 1984–1990, and then steadily increase from a large undervaluation in 1992 to a large overvaluation in 2019. In sub-Saharan African countries that are members of the Arab League, the RER appears to have been relatively closer to its equilibrium over the period. The region shows an overvaluation until 1991, an undervaluation in 1992 and a steady increase towards a modest overvaluation in 2017. The RER has remained close to its equilibrium since. In the Arabian Peninsula, the RER misalignment shows a large overvaluation before 1978, followed by a steady increase from undervaluation to overvaluation between 1980 and 1985. The RER remained close to its equilibrium between 1986 and 2003, before showing a non-negligible undervaluation until 2014. It has been slightly but increasingly overvalued since.

Figure 3:

Real exchange rate misalignment, by region. Source: Darvas (2012), calculations and elaboration by the author. A positive value denotes a RER overvaluation, a negative value a RER undervaluation. The RER misalignment is obtained following the procedure detailed in Section 3.

Figure 4 describes the evolution of RER misalignment in different economic groups between 1970 and 2019, following the IMF classification. Unsurprisingly, oil importers appear to have a relatively less volatile exchange rate. After a modest overvaluation up to 1991, the RER fluctuated around and remained close to its equilibrium since.

Figure 4:

Real exchange rate misalignment, by economic group. Source: Darvas (2012), calculations and elaboration by the author. The classification for this figure follows the IMF classification. A positive value denotes a RER overvaluation, a negative value a RER undervaluation. The RER misalignment is obtained following the procedure detailed in Section 3.

Among members of the Gulf Cooperation Council, the RER shows a large overvaluation before 1978, followed by a steady increase from undervaluation to overvaluation between 1980 and 1985. The RER remained close to its equilibrium between 1986 and 2002, before showing a non-negligible undervaluation until 2014. It has been slightly overvalued but close to its equilibrium since. In other oil-exporting countries, the RER has been more volatile. It went from a large overvaluation until 1990, to a strong undervaluation in 1993. It has since steadily increased, and it became substantially overvalued from 2015.

Appendix B contains additional figures that show the RER fluctuations at the country-level. Appendix Figure B.2 shows the observed and the equilibrium RER for each country, and Appendix Figure B.3 shows the misalignment index for each country.

5 Empirical Framework

To assess the effects of civil conflicts on RER misalignment, I estimate the baseline equation (6) using the OLS method:

where t denotes the year and i the country. RER i , t MIS denotes the RER misalignment index defined in Section 3. The main explanatory variable, CivConf _i,t , denotes one of the three measures of civil conflict intensity in country i during year t. Θ is a vector of parameters and the vector X i , t ′ ∼ denotes the set of fundamental RER determinants’ cyclical components, which are known to affect the RER misalignment and are obtained using both the two-sided and one-sided Hodrick-Prescott filters. α _i denotes country fixed effects and captures country-specific time-invariant factors, such as geographic location or agro-ecological conditions, that may affect RER misalignment. μ _i,t denotes the random error term.

As in equation (1), the set of fundamental RER determinants’ cyclical components comprises commodity terms of trade, productivity differential with respect to OECD countries, proxied by the ratio of real per capita GDP over that of OECD countries, trade openness, government consumption, net foreign assets and the world interest rate, proxied by the Federal Reserve 1-Year Treasury Constant Maturity Rate. Because the world interest rate does not vary accross country, the term φ _t that captures year fixed-effects would induce perfect multicollinearity and is therefore not included in equation (6). Year fixed-effects, however, would allow to control for unobserved factors that affect all countries in a given year, such as global demand shocks. For this reason, a variant of equation (6) includes the term φ _t and removes the proxy for the world interest rate from the vector X i , t ′ ∼ .

The baseline equation might suffer from several shortfalls that impede a causal interpretation of the OLS estimates. First, although the transmission mechanisms might not appear to be evident, the relation between civil conflicts and RER misalignment might suffer from reverse causality (Ambaw and Sim 2021). According to this view, a highly misaligned RER could fuel discontent or lead to severe losses for some agents, and therefore lead to civil conflict. Second, the presence of missing variables in the error term might bias the OLS estimates. A major economic crisis may affect both the RER misalignment and social stability, increasing the probability of occurrence of a civil conflict. It could be possible to control for some of these variables to increase the robustness of the results, but some missing variables are unobservable and time-varying, and cannot be controlled for by a proxy nor by fixed-effects.

To overcome these shortfalls and identify the effects of civil conflict intensity on RER misalignment, I adopt an instrumental variable approach and instrument the intensity of civil conflict in country i at year t by the intensity of civil conflicts in country i’s neighbouring countries at year t. I use a Two-Stage Least Squares (2SLS) procedure, in which equation (7) describes the first stage and equation (8) the second stage:

where n denotes each of country i’s N _i neighbouring countries, α ₀ is a constant, Φ a vector of parameters, λ _i country fixed-effects and υ _i,t a random error term. CivConf ˆ i , t denotes the predicted value of CivConf _i,t obtained from the first stage and depends on the instrument and the control variables. All the other variables remain as previously defined.

For this strategy to be valid and allow for a causal interpretation, the instrument must verify both the relevance condition and the exclusion restrictions. Figure 5 shows that the relevance condition is fulfilled since the intensity of civil conflicts in country i is positively correlated with the aggregate intensity of civil conflicts in neighbouring countries.^[1]

Figure 5:

Correlation between civil conflict and neighbours’ civil conflicts. Source: Center for Systemic Peace, elaborated by the author. The OLS estimate shows the correlation between civil conflict and civil conflicts in neighbouring countries and is significant at the 1 % level. The circles’ size is proportional to the number of observations.

The exclusion restriction cannot be formally tested since the first stage only includes one instrumental variable. This restriction is verified if the aggregate intensity of civil conflicts in country i’s neighbouring countries affect country i’s RER misalignment through its own intensity of civil conflicts but is uncorrelated with the error term.

Civil conflicts in neighbouring countries are unlikely to cause shifts in capital flows such as remittances, foreign direct investment and official development assistance since their determinants depend mainly on structural characteristics of the country. Shifts in international capital flows induced by civil conflicts in neighbouring countries are expected to be observed in bilateral military cooperation, arms purchases, or humanitarian assistance to alleviate the pressures caused by refugee inflows. Such capital flows can affect the RER misalignment, but they do so only because civil conflicts in neighbouring countries increase the risk of civil conflict within the country. International financial flows linked to refugee inflows, for example, only materialize on a large scale when arrivals are numerous and can threaten the economic and social stability of the country. Lebanon is a singular case in this aspect, since its position as a regional financial center and its banking sector regulation, including bank secrecy for non-residents during part of the period covered by this paper, could threaten the validity of the exclusion restriction. However, Appendix Table C.1 shows that excluding the country from the sample does not alter the results, indicating that the empirical strategy is robust to this specific threat.

6 Results

Table 4 reports the OLS estimates obtained from equation (6). Odd columns include the international interest rate as a dependent variable and country fixed-effects, while even columns do not include the international interest rate but add year fixed-effects to the country fixed effects. The results of these two specifications are similar, particularly when focussing on the main explanatory variable.

Table 4 columns (1) and (2) show that civil violence is not associated with RER misalignment when controlling for the cyclical component of the main RER misalignment determinants. Civil warfare, on the contrary, appears to be strongly and positively associated with RER misalignment, as shown in columns (3) and (4): a one-unit increase in civil warfare intensity is associated with a 0.06 unit increase in RER overvaluation. Columns (5) and (6) show no association between total civil conflicts (which include both civil and ethnic violence and war) and RER misalignment. These results suggest that only organised or institutionalised civil conflicts, i.e. those that incorporate the strongest commitment to violence, lead to RER overvaluations.

The OLS estimates presented in Table 4, however, are likely to suffer from endogeneity issues due to reverse causality and missing variables. They are therefore likely to be biased and cannot be interpreted as causal relations. The next step consists in adopting an instrumental variable approach, which is able to address this problem, by predicting civil conflict intensity in each country based on the aggregate intensity in its neighbouring countries.

Table 5 reports the 2SLS estimates obtained from first stage equation (7) and second stage equation (8). Estimates in column (1) show an absence of relation between civil violence and RER misalignment. The second stage coefficient is positive and statistically significant, but the p-value associated with the F-value is greater than 5 percent, indicating that the group of explanatory variable does not show a significant relationship with the dependent variable. Furthermore, the relevance condition for the instrument is not met since the first stage results indicate that civil violence is not correlated with civil violence in neighbouring countries and that civil violence is not reliably predicted by the instrument and the control variables. The relevance condition can be formally tested using Stock and Yogo’s size method, which compares the Wald test F-statistic for the hypothesis that the coefficient of the instrument is equal to 0 with a critical value (Stock and Yogo 2005). The F-statistic of the coefficient of civil violence in neighbouring countries, equal to 2.11, is inferior to the critical value, which impedes the rejection of the weak instrument hypothesis.

Column (2) in Table 5 shows the 2SLS estimates of the effect of civil war intensity on RER misalignment when civil war in neighbouring countries is used as an instrument. While the OLS estimates indicated a positive and significant relationship, the 2SLS estimate show an absence of relationship since the second stage coefficient is not significantly different from 0 and the p-value associated with the F-stat does not allow to reject the null hypothesis that all coefficients are equal to 0. The first stage estimates show a negative correlation between the explanatory variable and the instrument, but the p-value associated with the F-stat indicates again that the instrument and the control variables do not predict reliably civil war intensity. The relevance condition of the instrument is therefore not fulfilled in the case of civil war, as is confirmed by the fact that the F-statistic of the instrumental variable’s coefficient, equal to 7.03, is inferior to the critical value.

Column (3) in Table 5 reports the 2SLS estimates of the effect of total civil conflict intensity on RER misalignment and show that civil conflicts intensity in Arab League members, Turkey and Iran do lead to an increase in RER misalignment, i.e. tend to overvalue the RER. The first stage estimates show a positive relation between the instrument and the explanatory variable: a 1 unit increase in civil conflict intensity in neighbouring countries is associated with a 0.07 unit increase in civil conflict intensity domestically. This correlation is significant at the 1 percent level, and the instrument along with the control variable are jointly significant in predicting civil conflicts intensity since the p-value associated with the F-stat is lower than 5 percent.

The p-value associated with the F-stat of the second stage, reported in the upper part of Table 5 column (2), is also lower than 5 percent, indicating that civil conflicts and the cyclical components of the RER determinants have a statistically significant relationship with RER misalignment. The estimated coefficient of civil conflicts is positive and statistically significant: a 1 unit increase in civil conflict intensity leads to a 0.24 unit increase in the RER misalignment index. Endogeneity issues have been removed from this 2SLS estimate, allowing therefore for a causal interpretation of the relationship, and the weak instrument hypothesis can be rejected since the Wald test’s F-statistic of the instrumental variable’s coefficient is greater than the critical value, at 21.56. The results show that an increase in civil conflicts intensity leads to an increase in the RER misalignment: the RER tends to be overappreciated as a result of civil conflicts. Appendix Table C.2 shows that these results are robust to using the one-sided Hodrick-Prescott filter.

Columns (1) and (2) in Table 5 show that distinguishing between civil conflicts subcategories does not provide useful information to identify the relationship with RER misalignment. The type of civil conflict might not be correlated with the type of civil conflict in neighbouring countries, as suggested by the first stage estimates. However, the reason may also be that the categorisation of civil conflicts itself, which necessarily entails some degree of subjectivity, is not relevant to assess their effects on RER misalignment. Aggregating the different types of civil conflicts allows to overcome these issues and identify the effect of civil conflicts on RER misalignment, as indicated by the results in column (3).

Human losses are the first and most direct consequence of civil conflicts. From an economic standpoint, this loss in human capital combines with losses in physical capital to hamper the future growth path and development prospects of the country. These consequences of civil conflicts have legitimately attracted much attention and have been intensively studied. However, civil conflicts affect long-term prospects in several other ways.

Following the tradition of Acemoglu, Johnson, and Robinson (2005), among others, many consider that institutions are a fundamental determinant of long-term growth. The definition of these institutions can be quite vague, but the fact that they are affected by path dependence and therefore show a high level of time persistence is a common view. Because the end of a civil conflict and the initial stages of the post-conflict transition define a new, or renewed, distribution of power and set the foundations of the institutions that will ensure its stability (see for example Fergusson et al. 2021), the economic conditions and the social context that prevail during these crucial periods may weigh heavily on their outcome and may be determinant to ensure sustainable growth and long-term development.

RER overvaluation resulting from civil conflicts can therefore have enduring economic impacts, way beyond the business cycle. ElBadawi and Refaat (2015) show that RER undervaluations tend to be pro-poor in normal times. Because the disorganised productive sector cannot meet the demand, imports are likely to play a significant role in the early stages of reconstruction, and this is particularly important for food products. The sharp nominal exchange rate depreciation arising from a RER overvaluation or, under a fixed exchange rate regime, the high inflation and the probably high black market premium would represent a heavy burden on the poorest and directly threaten their very subsistence. The resulting discontent and social unrest could threaten the post-conflict transition itself, or indirectly shape the emerging institutions in a way that tends to allocate efforts and capacities to security organs at the cost of reconstruction and growth promotion, leaving an enduring print on the development prospects of the country.

7 Conclusions

This paper estimates RER misalignments in Arab League member countries, Iran and Turkey between 1970 and 2018, and uses an instrumental variable approach to study how RER misalignment is affected by civil conflicts. It contributes to the literature on the macroeconomic effects of civil conflicts by assessing the monetary and financial consequences of such conflicts, in a region that has received relatively little attention. It also contributes to the literature on RER misalignment by identifying a new determinant of its short-term dynamics in developing countries.

The results show that civil conflicts lead to an increase in the RER misalignment index in the region, indicating that the currency tends to be overvalued in such a situation. Since the expected transmission channels have contradictory effects, the results suggest that the consequences of high public spending during the conflict and the inflationary effects of supply shortages, particularly those concerning basic goods, dominate the other mechanisms.

Because RER overvaluations are detrimental to exports and structural change, they impede the creation of the jobs needed to reduce unemployment and increase incomes, threatening both the post-conflict transition and the long-term development prospects of the country. Therefore, economic policy during a post-conflict transition should include a strategy to realign the currency. In the short-term, the informal sector might act as a buffer if shortages of basic goods persist, and subsistence requirements are binding. The existence of such a sector can give time to implement necessary but difficult measures in the short-term. To ease inflationary pressures, the public budget should be tightly controlled while ensuring the supply of basic goods to the population, through imports and the recovery of the productive sectors, and the reconstruction of vital infrastructure that promotes growth. If subsistence is threatened, foreign aid is necessary but might overvalue further the currency. To limit such an effect, which would lead to a lasting deterioration of competitiveness (Rajan and Subramanian 2011), aid in kind could be appropriate, provided that it is temporary so that it does not harm the recovery of the agricultural sector. During the post-conflict transition, a monetary policy that aims at effectively controlling inflation should also help stabilize the real exchange rate and provide foundations for future growth.

This paper finds that civil conflicts lead to RER overvaluation in Arab League members, Iran and Turkey. Additional empirical work could provide deeper insights on the underlying transmission mechanisms. They could also investigate whether this relation is verified in other regions that have suffered civil conflicts, and assess RER dynamics during post-conflict transitions. Finally, further work that seek to identify additional short-run determinants of RER misalignment in developing countries could provide interesting insights and useful policy recommendations.