Economic Effects of the Syrian War and the Spread of the Islamic State on the Levant

This paper uses a global computable general-equilibrium framework with new detail on six Levant countries -- the Arab Republic of Egypt, Iraq, Jordan, Lebanon, the Syrian Arab Republic, and Turkey -- to quantify the direct and indirect economic effects of the Syrian war and the advance of the Islamic State on the Levant. Syria and Iraq bear the brunt of the direct economic costs, while the other Levant countries lose in per capita but not in aggregate terms. The fact that the Islamic State's spread has undermined regional trade adds to varying degrees to the direct costs in all Levant economies and in the case of Syria and Iraq doubles the welfare losses. All these countries are foregoing opportunities to expand intra-Levant trade and the associated gains in economic efficiency and diversification. The average welfare effects are not indicative of within-country incidence, which varies among workers, landowners, and capitalists.


Policy Research Working Paper 7135
This paper is a product of the Office of the Chief Economist, Middle East and North Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at eianchovichina@worldbank.org. This paper uses a global computable general-equilibrium framework with new detail on six Levant countries-the Arab Republic of Egypt, Iraq, Jordan, Lebanon, the Syrian Arab Republic, and Turkey-to quantify the direct and indirect economic effects of the Syrian war and the advance of the Islamic State on the Levant. Syria and Iraq bear the brunt of the direct economic costs, while the other Levant countries lose in per capita but not in aggregate terms. The fact that the Islamic State's spread has undermined regional trade adds to varying degrees to the direct costs in all Levant economies and in the case of Syria and Iraq doubles the welfare losses. All these countries are foregoing opportunities to expand intra-Levant trade and the associated gains in economic efficiency and diversification. The average welfare effects are not indicative of within-country incidence, which varies among workers, landowners, and capitalists.

Introduction
On the eve of the Arab Spring, six countries in the greater Levant--Turkey, the Syrian Arab Republic, Iraq, Jordan, Lebanon, and the Arab Republic of Egypt--were considering reforms that would have deepened their trade ties and accelerated economic growth, diversification, and job creation. Specific attention was placed on liberalizing agricultural trade with Turkey, reducing the restrictiveness of non-tariff measures, improving transport logistics, and liberalizing intra-Levant trade in services. These reforms were considered essential for stimulating regional trade and were the main components of a reform package that would have been negotiated and implemented as part of a new Levant economic zone (World Bank, 2013). 1 In 2011, many of these countries embarked on political transitions that took priority over other issues. In Syria, initial demonstrations quickly turned into an uprising which grew into a civil war and resulted in widespread devastation with spillovers to neighboring countries. This war and the subsequent advance of the Islamic State of Iraq and Syria (ISIS) -collectively referred to in this paper as the Levant conflict or war -imposed enormous human, social, and economic costs and put a halt to the regional trade integration process, thus undermining development with serious implications for the future of the Levant.
Despite the widespread interest in the Levant war, there are no systematic analyses of its regional and country-specific economic effects. This paper addresses this gap and contributes to the literature a general equilibrium assessment of the Levant conflict, factoring in both the effects of war and the associated disintegration of regional trade. The paper is related to and combines features of two distinct literatures -on trade reform and restrictions (Trela and Whalley, 1990; 1 We refer to the new Levant economic zone as simply the Levant or the Levant area, although the geographic Levant area includes other countries and territories. The six economies would have composed the new Levant economic zone. greatest, and minimal for Turkey and Egypt, where refugees account for a small share of the population.
The direct effects of the Levant war are an understatement of the real economic costs of the Levant conflict. Recall that these countries were embarking on a process of regional trade integration just before the outbreak of war. If the foregone benefits of this integration, especially those associated with failed services liberalization, are included then the total costs of war for Syria and Iraq almost double, reaching 23% and 28%, respectively, and escalate to 10% for Egypt and 9% for Jordan. Furthermore, the average welfare effects are not indicative of the incidence within countries. In Syria, all economic agents are hurt but landowners lose the most as people abandon their homes and farms in search of security. By contrast, in Lebanon and Turkey, land and capital owners benefit while workers lose because the large number of refugees put pressure on demand and augment labor supply.
The remainder of this paper is structured as follows. Section 2 presents the features of the CGE model and the data modifications. Section 3 discusses the simulation design, including the main features of the pre-war plans for trade integration reforms and the war scenario. Section 4 presents the simulation results focusing on welfare, sectoral outputs, factor prices, and the pure GE effects. Finally, we summarize and offer concluding remarks in section 5.

Features of the CGE model and data modifications
The multi-country, multi-sector CGE model, used in this paper and documented in Hertel (1997), is well-suited and widely used for quantitative, ex-ante investigations of the effects of regional trade agreements. In this model, firms are assumed to produce for domestic and export markets, using constant-returns-to-scale technology and a mix of primary and intermediate inputs.
Intermediate products are either produced domestically or imported and substitute imperfectly, following the Armington structure. Product differentiation between imported and domestic goods and among imports from different regions allow for two-way trade in each product category, depending on the ease of substitution between products from different regions. Land, physical capital, skilled, and unskilled labor, and in some sectors a natural resource factor, are used as primary factor inputs into production. The model takes into account the role of overall resource constraints in determining sectoral output supply, has an explicit treatment of international trade and transport margins, a "global" bank mediating between world savings and investment, and a consumer demand system designed to capture differential price and income responsiveness across countries. The accounting relationships and behavioral linkages constrain outcomes in ways not possible with partial equilibrium models. Each country's exports of a particular good equal total imports of this good in other countries, net of shipping costs; global investment equals global savings; aggregate output determines aggregate income in each country; global supply and demand for individual goods balance; demand equals supply for each factor in a country; increases in total factor productivity which raise competitiveness also raise factor prices and help offset the original increase in competitiveness. The results obtained with the model are indicative of medium term outcomes as factor inputs are perfectly mobile across sectors and returns adjust to changes in economic conditions. The procedure used to construct the individual country information employs data from several sources. The UN Statistics Division data for 2007 is the source for the six components of GDPagriculture, hunting, forestry, and fishing (ISIC A-B); mining, manufacturing, and utilities (ISIC C-E); construction (ISIC-F); transport, storage, and communication (ISIC I); wholesale, retail trade, restaurants and hotels (ISIC G-H); and other activities (ISIC J-P).
We sourced bilateral trade value data from WITS and bilateral tariff data from a medley of sources, presented in Appendix Table A1. As part of this procedure, all entries in the two composite regions (rest of Western Asia and rest of Northern Africa) were split and assigned the split values to the newly created economies, while all entries for the two composite regions from the GTAP database were removed from the database. Each entry was split using the most thematically relevant external source. Sectoral GDP shares were used to split consumption and production values, trade data were used to split export and import values, and tariff information was used to assign tariff values. Export shares were used to split further production and consumption information into the final set of industries presented in Table 1. For internal consistency purposes, the required accounting relationships were imposed on the split database 8 using iterative proportional fitting and the procedure was repeated until the database was balanced and consistent with all external targets.
Another important modification was the implementation of the Pan-Arab Free Trade Area (PAFTA), 2 the bilateral preferences associated with the Euromed Association Agreements (AAs), and the bilateral FTAs with Turkey into the tariff rate structure of the GTAP 8 database. 3 We obtained information on bilateral preferences at the most disaggregate product level from a variety of sources, including MFN and non-MFN rates from WTO data, country tariff data, and in the case of the European Union, Eurostat (see Appendix Table A1). Bilateral rates among PAFTA members were set at zero to reflect free trade in agricultural goods and manufactures.
Whenever bilateral country tariff information and non-MFN rates from WTO sources were not available, we assumed reciprocity and applied the rates extended by the partner. In the absence of such rates we applied the MFN WTO rates. Duties on imports from countries outside the MENA region were left unchanged whenever the importing country was part of the GTAP database. In those cases when the country information had to be created from a composite region, we applied WTO MFN rates or used country information. The detailed data on bilateral tariff lines were aggregated into weighted average rates for the 22 sectors in the paper using bilateral import data from WITS for 2007. 4 Whenever such data were not available, imports were inferred from exports for 2007 or from WITS data for 2008. These tariff rate modifications are essential for this analysis as suggested by the substantial differences between the tariff rates available in the GTAP 8 database, especially those implied for Jordan, Iraq, Lebanon, and Syria (Figure 1), and the updated tariff rates, presented by country, product, and source in Appendix Tables B1-B6. Since the GTAP tariffs attributed to Jordan, Iraq, Lebanon, and Syria are composite rates, they do not correspond to the actual trade profile of these countries. Therefore, the new tariff rates differ from the GTAP ones both because of differences in the tariff lines and trade composition. By contrast, the tariff information on Egypt and Turkey in the GTAP 8 database represents relatively accurately existing preferences ( Figure 1).

Simulation design
The pre-war efforts for deeper trade integration in the Levant are reflected in the pre-simulation analysis. Starting from the newly constructed database, the pre-simulation analysis implements the deep trade initiatives discussed by the Levant countries prior to the onset of the Syrian war in 2011. The context for these reforms and the shocks associated with each of these reforms are presented in section 3.1. The updated database from the pre-simulation analysis, which represents an integrated Levant in a peaceful alternative world, is the starting point for the simulation analysis of the Syrian conflict and the spread of ISIS as well as the disintegration of the deep regional trade ties. goods and processed foods from many of the Levant economies were much higher than tariffs on manufactured imports from these same countries (see Appendix Table A1).
This momentum for deep trade reform is reflected in the pre-simulation scenario in the following way. Tariffs on imported food and agricultural products are set to zero in the six Levant economies. Any tariff revenue loss due to this reform is assumed to be compensated by a consumption tax increase so as to keep the tax revenue constant as a share of income.
Improvements in transport logistics is assumed to result in cost reductions associated with a more efficient process of shipping goods within the Levant area. The shocks are proportionate to the reductions needed to bring down the transport cost of a standard container unit to and from these countries to those of a leading country in the region, including MENA and Turkey. Information on transport costs comes from the World Bank's Doing Business database. 9 In the case of exporting a container, the lowest cost country in the developing part of the Mediterranean region is Morocco. In the case of importing a container Egypt is the lowest cost country, while Jordan is the lowest cost country without access to the Mediterranean.
Since the GTAP model does not differentiate across firms based on their ownership structure, as in Tarr and Rutherford (2010), cross-broader services trade liberalization is reflected following Walmsley et al. (2006) and representing the opening of the service sectors to foreign competition as an efficiency improvement. Doytch and Uctum (2011) show that service FDI spurs growth in the sector while a number of papers (Haskel et al., 2007;Markusen et al. 2005)  presence in services provides substantial benefits to domestic firms. The efficiency boost to service companies engaged in cross-border service trade is implemented as a productivity shock which lowers the effective prices of imported services. In order to estimate the size of the productivity shocks, services productivity Π is represented as a function of the trade restrictiveness policies affecting this sector, given by index Ψ, and other factors, represented as Ω. This way productivity Π is given by Π(Ψ, Ω) and the percentage change in productivity is � = � , where ε is the elasticity of the productivity Π to change in the index Ψ. With the elasticity ε equal to 1, changes in the trade restrictiveness index Ψ translate into changes in productivity. 10,11 Using the World Bank's Services Trade Restrictions (STR) database, which contains values of the STR index (STRI) for several service sectors in the Levant countries, and assuming that trade liberalization will reduce the STRI to the minimum of the corresponding indexes in the Euromed area, we computed the implied productivity changes. Sectoral STRIs were available only for financial services and insurance, communications, trade, transportation, and other business services. In the case of construction and tourism, we used the overall STRI, and in the case of Syria, data were not available so we assigned the average regional STRI to each sector. The shocks differ in size and suggest that the liberalization-associated efficiency improvements will be smallest for Turkey (Table 2), as Turkey's services sectors are the most open and productive in the region.
The opening of the services sectors to foreign investment and competition is also expected to boost value-added productivity in some services sectors, resulting in convergence to the highest value added per worker in the region. This process will be gradual and complete convergence is expected to occur only by the end of a 20-year period. 12 Since results are representative of what is likely to happen in a 3 to 5 year timeframe, we first compute the productivity shocks required for complete convergence over a 20-year period, then we annualize them, and finally cumulate them to represent the productivity growth expected only in the span of 3 years. The resulting productivity shocks are shown in Table 2. They suggest that in the Levant, Turkey is expected to be a productivity leader in transport, communication, finance, insurance and real estate, and business services, while Lebanon in construction and retail trade activities. Table 2 Productivity growth associated with services liberalization (%) 12 We exclude from the analysis all government-related services. The macroeconomic closure for this simulation is consistent with the medium-term timeframe and assumes a constant level of employment, with perfect mobility of skilled and unskilled labor between sectors and none between countries or regions. Since the model does not keep track of differences between foreign and domestic assets, we assume that Levant countries' trade balances are fixed as a share of the size of the economy.

Simulation scenario of war and trade disintegration in the Levant
We consider trade disintegration an indirect but essential effect of the Levant conflict. Therefore, the database obtained from the pre-simulation scenario of deep trade liberalization is the initial point for the war and trade disintegration scenarios. We implement shocks that completely reverse the deep trade reforms discussed in 3.1 as well as shocks reflecting: (i) the change in population and labor force size due to loss of life in Syria; (ii) the change in population and labor force size due to refugee movements across countries; (iii) infrastructure destruction in Syria; (iv) increases in trade costs in the Levant; (v) embargo on trade with Syria; and (vi) deterioration in productivity in Iraq. The CGE framework then helps us assess the implications of these shocks on other economic variables in the model. Next, the paper presents details on the elements of the war scenario and the changes to the macroeconomic closure in order to accommodate labor mobility across countries in response to refugee movements.
The war in Syria triggered massive displacement of people and outflows of Syrian refugees into neighboring countries, especially Jordan and Lebanon. In order to implement cross-border movements of people in the model, we relax the assumption of no international labor mobility and adjust the population and labor force of both refugee-receiving and refugee-sending countries, using information from UNHCR Population STATICS and ILOSTAT Database.
Syria's population and labor force were adjusted downward in order to reflect (a) the loss of life since 2011; (b) the number of Syrian refugees who fled the country during the period from 2011 to 2014; 13 and (c) the number of Iraqi refugees who left Syria during the same period. 14 At the same time, the population and labor force of the refugee-receiving countries were adjusted upward in order to reflect the inflow of refugees from Syria (Table 3). Adjustments were also made in order to reflect the return of Iraqi refugees to their home country during the period 2007-13. In the absence of information on the skill mix and participation rates of Syrian refugees, we assume that skilled and unskilled workers are equally affected by the war and that labor force participation rates among refugees are respectively the minimum of the participation rates in Syria and each refugee receiving country. The shocks to the Levant countries' population and labor endowments are shown in Table 3. Although adjustments were made in all countries/regions in the model, we show only those applying to the Levant as the magnitudes for countries outside the Levant are negligible.  13 We make this adjustment in order to assess accurately the medium-term effects of war and reflect the fact that the majority of Syrian refugees plan to return to Syria only upon the fall of the Assad regime. 14  Oil exports from Syria are assumed to decline dramatically (by 90%) due to a combination of factors, including sanctions imposed by the EU and the US and loss of infrastructure. We assume that 20% of Syria's physical capital has been destroyed -a decline as large as the decline in Syria's labor force. In Iraq, we assume the advance of ISIS has led to a 5 % decline in total productivity. The decline in oil exports from Syria and Iraq is offset by a corresponding increase in the production of oil by the GCC countries so that the effect on the world oil price is negligible. This is a realistic assumption because Saudi Arabia has the spare capacity to fully offset a drop in Iraqi and Syrian oil exports. We also assume that Syria's nonoil exports are affected by restrictions on trade between US and Syria and EU and Syria in specific categories, including equipment and vehicles, chemicals, metals, and capital goods.

Welfare effects
The results suggest that Syria and Iraq bear the brunt of the direct war losses as the conflict drags down their per capita welfare by 14% and 16%, respectively. Neighboring Levant economies lose to varying degrees, with per capita welfare declining by almost 11% in Lebanon, less than 2% in Jordan, and only negligibly in Turkey and Egypt (Table 4). The embargo on trade with Syria has been a major factor behind the deterioration in Syria's per capita welfare, reducing it by more than 15%, while capital destruction and loss of workers are responsible for declines of more than 5% and 7%, respectively. In Iraq, the per capita welfare losses are associated with the deteriorating environment and the resulting decline in productivity. In Lebanon, the main effect comes from the massive inflow of Syrian refugees.

Table 4 Welfare effects of war and trade disintegration in the Levant (%)
Syria's direct aggregate welfare decline is much larger than its per capita welfare loss (Table 4).
Syria's economy shrinks by almost a third due to the massive outflow of Syrian refugees and war casualties. By contrast, Iraq's aggregate welfare loss of 11% is smaller than its per capita welfare  Cumulative effects in US$ 2007 -6,510 -10,483 -834 912 -12,280 -3,997 decline because a large number of Iraqi refugees in Syria have returned to Iraq during the period 2010-14. All other Levant economies gain in aggregate terms as the influx of refugees boosts their population numbers, increasing demand for goods and services and labor supply. These effects are most pronounced in Lebanon, where the refugee-to-citizen ratio is greatest, and minimal for Egypt, where refugees account for a small share of the population. The global effects of the crisis are negligible 15 because the conflict has no effect on the main channel of transmission -oil prices.
If in addition to the direct effects of war we include the effects of regional trade disintegration, we find that the Levant conflict hurts significantly all Levant economies (Table 4). Iraq's direct per capita welfare losses from the conflict are as large as its losses from trade disintegration, which are largest among the Levant economies. Stated differently, Iraq's average per capita income could have been nearly a third larger had the country managed to avoid conflict and liberalize its economy. Syria's trade disintegration losses are slightly lower than Iraq's but still sizable at almost 9% in per capita terms. Thus, Syria's per capita income could have been a quarter larger had the country managed to steer away from the civil war and proceeded with its plans to integrate into the regional economy of the Levant. In Lebanon, the inflow of refugees expands the size of the economy but this output expansion is not as large as the increase in population so in per capita terms welfare declines by almost 13% ( Consistent with the results in the literature, most of the disintegration losses stem from foregone services liberalization, whereas those from foregone agricultural liberalization and transport logistics reform are negligible, although the sectoral effects are sizable for some sectors, as discussed in the next section, and for landlocked Iraq which loses 1% from foregone transport logistics reform. The foregone services liberalization generates sizable welfare losses as barrier to entry hurt productivity and put upward pressure on production costs as well as the costs of importing services within the Levant. The results, however, differ across countries reflecting the different extent of productivity loss. As a productivity leader in the Levant, Turkey's losses are small and stem mainly from failed liberalization of construction and business services. 16 By contrast, Iraq's service sectors are among the most inefficient in the Levant so its welfare loss from failed services liberalization of 11% is largest. Syria and Egypt also lose to a substantial degree, reflecting the fact that their service sectors are more protected and less efficient than those of Jordan and Lebanon. 17

Sectoral effects
The direct sectoral effects of the conflict are negative and sizable across the board only in Syria and Iraq (Table 5), where the war has led to a productivity decline, and in the case of Syria considerable capital destruction and loss of labor. In fact, Jordan and Lebanon register sectoral expansions in response to the refugees' effect on the demand for goods and services and supply of labor. The direct sectoral effects of conflict in Turkey and Egypt are negligible. Regional trade 16 According to the STRI data, Lebanon and Jordan have the least restrictive policies in terms of foreign presence in construction and business services in the Levant, respectively. 17 The inflow of refugees has created serious challenges, among which crime, congestion, and a strain on public systems for delivering basic services. This analysis does not factor in these challenges nor provides estimates of the financing needed to address them.

21
disintegration, however, drags down intra-Levant trade (Figure 2), services productivity as well as the cost of producing and importing services within the Levant area with negative effects for output of services in all Levant economies (  Figure 2).  -30 -8 -8 -11 in bulky products with high transport margins, such as agricultural goods, processed foods, natural resources, and metals. In sum, regional trade disintegration as part of the Levant conflict deprives all six economies of opportunities to trade, transform the structure of their economies, and create quality jobs in services and manufacturing. These results are in line with papers which find a negative association between political instability and foreign direct investment flows to MENA (Burger et al. 2013;Meon and Sekkat, 2012). Importantly, Burger et al. (2014) find that political instability does not have an effect on greenfield investments in resources and nontradables but inhibits investments in tradable activities such as commercial services and nonoil manufacturing, thereby slowing structural transformation.

Incidence within countries
The average welfare effects presented in section 3.1 are not indicative of the incidence of the Levant war across economic agents (Table 6). In Lebanon and Turkey, the direct effects of war on real returns to land and capital are positive, although the magnitudes of these effects differ considerably due to the much larger share of refugees in the Lebanese population. Consequently, real land rents rise by close to 40% in Lebanon and by just 2% in Turkey. Landlords and capitalists benefit because the inflow of refugees increases derived demand for land and capital and supply of labor. By contrast, workers in these two countries lose as refugees compete for jobs and put downward pressure on wages. In Syria, although all economic agents lose in real terms, landowners are hardest hit as people flee the country in search of safety. The results suggest that the war in Syria has triggered a nearly 50% decline in real rental rates and about 20% decline in wages and capital rental rates. The conflict is also associated with a rise in resource rents to varying degrees within the Levant, except Syria where the embargo limits returns to resources. Turkey. As barriers to trade in services increase, factor productivity deteriorates leading to declines in real factor returns, especially land and resource rents.

Pure general equilibrium effects
The global CGE model used in the paper allows for interaction of markets which could lead to significant non-linearities and sizable feedback effects in response to equilibrium price changes, even when underlying preferences and production processes are well behaved. In order to assess the importance of these effects, we compute the pure GE effects, defined as the difference between the nonlinear and the linear solutions of the model. 18 The size of the pure GE results reveals to what extent the CGE model enables us to assess more precisely the effects of war and trade disintegration shocks on different aspects of the Levant economies by capturing the feedback effects between markets. Large pure GE effects would signal that the feedback effects due to changes in prices associated with the war shock are substantial and should not be ignored in such assessments. Furthermore, if the sign of the pure GE effects varies, this would indicate that the linear approach misstates the effects, sometimes overestimating and other times underestimating the "true" effects of war. Thus, it would be difficult to determine the direction of bias.   solutions are mostly small (below 2%).

Summary and conclusions
The paper quantifies the economic effects of the