Policy Research Working Paper 10599 Alternative Paths for Yemen up to 2030 A CGE-Based Simulation Analysis Hans Lofgren Martín Cicowiez Gianluca Mele Equitable Growth, Finance and Institutions Practice Group November 2023 Policy Research Working Paper 10599 Abstract Over nine years of violence and conflict have profoundly matrix (SAM) for Yemen. The new social accounting matrix altered the Republic of Yemen’s economy. The war has shat- has the virtue of consolidating sparce and often inconsis- tered the country’s already fragile socioeconomic equilibria, tent Yemeni data from multiple sources (the World Bank, affecting nearly every facet of life. Since the onset of the con- the International Monetary Fund, and the United Nations flict, economic diagnostics have focused on descriptions of system) into a coherent framework that reflects the basic the deteriorating macro-fiscal and poverty conditions, lack structure of the economy, both at the macro and sectoral of food security, and loss of capital accumulation. However, levels. The simulation analysis is built around three broad relatively little attention has gone toward the development scenarios spanning 2022 through 2030. The results suggest of a forward-looking vision for the country, rooted in that if the conflict subsides, governance is strengthened, and Yemen’s current economic structure. This paper helps to the donor community provides crucial aid, considerable fill this gap by presenting and analyzing a set of scenarios progress, including reduced poverty rates and improved for Yemen’s economy up to 2030. The analysis is based on living conditions, can be achieved by 2030. Given Yemen’s a new version of the Sustainable Development Goal Sim- low levels of infrastructure and human development, the ulation model, a dynamic computable general equilibrium potential payoffs from investments in these areas are great. (CGE) model, which is applied to a new social accounting This paper is a product of the Equitable Growth, Finance and Institutions Practice Group. 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://www.worldbank.org/prwp. The authors may be contacted at gmele@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Alternative Paths for Yemen up to 2030: A CGE-Based Simulation Analysis Hans Lofgren Martín Cicowiez Gianluca Mele Key words: computable general equilibrium modeling, fragility conflict and violence, economic growth policy, fiscal policy JEL Codes: C68 , D58, E16, E6, E17, O53 1. Introduction Even before the start of its civil war in 2014, the Republic of Yemen had one of the lowest living standards in the world. Since 2014, its population has suffered from one of the world’s severest humanitarian crises with a serious further deterioration of living standards, manifested in increased poverty rates, severe malnutrition, food insecurity, and physical insecurity. Basic public health and education services have suffered, infrastructure (including road networks) has been destroyed, and government employees have been left without pay for extended periods. In the background, like many other low-income countries, Yemen is vulnerable to accelerating climate change. Given these circumstances, the country was ill-prepared to deal with the Covid-19 pandemic and international price shocks for food and energy. Given this, it is more urgent than ever to put an end to the war and identify policies that will help to overcome the humanitarian crisis and put the country on a path toward a better future. With this aim, this paper explores potential scenarios for the economy of the country up to 2030, focusing on alternatives under which the conflict subsides, and the country manages to start addressing the urgent needs of the population and the constraints that stand in the way of resumed sustainable growth and improved living standards. The analysis is based on simulations with the Sustainable Development Goal Simulation model (SDGSIM), a dynamic computable general equilibrium (CGE) model designed for country-level policy analysis.1 For this analysis, the model was adapted to the Yemeni context and calibrated to a new dataset, including a social accounting matrix (SAM) for 2020, the most recent year with sufficient information. The analysis, which covers the period 2022-2030, presents a base scenario without significant improvements in the economy or the state of security and six alternative scenarios which, compared to the base, reflect varying degrees of improvement. The scenarios are underpinned by different trajectories along seven dimensions: health and education; infrastructure; social transfers; remittances from abroad; foreign direct investment (FDI); extraction growth for oil and gas production; and foreign grant aid. The paper is organized as follows: After a brief overview of the model and its database (Section 2a), it presents a short review of literature and earlier applications of CGE modeling to Yemen (Section 2b), and detailed analyses of the base scenario with stagnant performance (Section 3) and the set of alternative scenarios (Section 4). The main findings are summarized in the conclusions (Section 5). The Appendix provides additional information about the model database. 1 SDGSIM is documented in detail in Lofgren and Cicowiez (2019). The starting point for SDGSIM was MAMS, a model designed for analysis of strategies related to the MDG agenda. See Lofgren et al. (2013) for documentation that also is relevant to SDGSIM. 1 2a. Model and database SDGSIM, the model that is used for this analysis, is a CGE model for country-level analysis of medium- and long-run development policies with a focus on the SDG agenda and structural change. (For a review of literature and earlier CGE modeling applications to Yemen, see Section 2b.) The Yemen version of the model was adapted to better capture Yemen’s economic structure and the policy issues that are addressed in this paper. It is a suitable tool for analyzing issues related to a potential social and economic recovery for Yemen given the fact that it, in an integrated manner, captures fiscal issues, household spending and incomes, government services, private sector production, and links with the rest of the world, via trade, transfers, borrowing, and other foreign exchange flows. Moreover, the model relies on a social accounting matrix (SAM) for the bulk of its data – thanks to the ability of SAMs to impose coherence on data from different sources, they make it feasible to capture the structure of the modeled economy also in data-scarce contexts like Yemen’s. Technically, the model consists of a set of simultaneous linear and non-linear equations. It is economy-wide, providing a comprehensive and consistent view of the economy, including linkages between disaggregated production sectors and the incomes they generate, households, the government (its budget and fiscal policies), and the balance of payments. The linkages are summarized in Figure 2.1: the major building blocks in the model are activities (the entities that carry out production), commodities (activity outputs or, exceptionally, imports without domestic production; linked to markets), factors (also linked to markets), and institutions (households, the government, and the rest of the world). In the figure, the private capital account also has its own box given the multiple links that exist between private investment demands and their financing. In any SDGSIM application (and dataset), most blocks are disaggregated – the disaggregation in the Yemen application is shown below in Table 2.1. In each period, the different agents (producers, households, government, and the nation in its dealings with the outside world) act subject to budget constraints: receipts and spending are fully accounted for and by construction equal (as they are in the real world). The decisions of each agent – for producers and households, the objective is to maximize profits and utility, respectively – are made subject to these budget constraints: for example, households set aside parts of their incomes to direct taxes and savings, allocating what is left to consumption with a utility-maximizing composition. For the nation, the real exchange rate adjusts to ensure that the external accounts are in balance, in a setting with exogenous foreign net financing and foreign reserve changes. Wages, rents, and prices play a crucial role by clearing markets for factors and commodities (goods and services). For commodities that are traded internationally (exported and/or imported), domestic prices are influenced by international price developments. It is assumed that Yemen is a small country in the sense that it is able to demand from and supply to international markets at prices that are exogenous in foreign currency. 2 Figure 2.1. Aggregate payment flows in SDGSIM Factor domestic wages and rents private savings Households Markets trnsfr+interest private consumption dir taxes lending factor demand trnsfr-interest indir taxes gov cons and inv Government Private interm input demand Investment trnsfr-interest Financing lending Activities lending Domestic FDI Commodity Rest of Markets imports World domestic demand exports foreign wages and rents private investment Source: Authors’ elaboration. Over time, production growth is determined by growth in total factor productivity (TFP) and factor employment (made possible by better utilization of existing stocks and/or growth in stock supplies). TFP growth is made up of two components, one that responds positively to growth in government infrastructure capital stocks and one that, unless otherwise noted, is exogenous. Growth in private capital stocks is endogenous, depending on investment and depreciation. Public investment and capital stock changes are determined by policy. For other factors, the growth in employable stocks is exogenous. For labor and natural resources (with sector-specific factors for natural-resource-based sectors), the projected supplies in each time period are exogenous. For natural resources, they are closely linked to production projections. For labor, the projections reflect the evolution of the population in labor-force age, labor force participation rates, and educational attainment. In the labor market, unemployment is endogenous – the model includes a wage curve (a supply curve) that is upward sloping until full employment is reached, at which point it becomes vertical -- see Figure 2.2. 3 Figure 2.2. The labor market in SDGSIM 5 4 3 Wage Wage curve Demand 2 1 0 85 90 95 100 - unemployment rate (%) Source: Authors’ elaboration. The basic accounting structure and much of the data needed to implement SDGSIM is derived from a social accounting matrix (SAM). The model also relies on complementary stock, elasticity, and base-year employment data.2 For this study, a new Yemen SAM was built for 2020, primarily drawing on data from the World Bank, the IMF, and the UN. While the quality of data for Yemen is low, the construction of the SAM had the advantage of imposing consistency between data from different sources, as a result producing a coherent database that captures well-known structural features of the Yemeni economy. The disaggregation of the SAM and the model database is presented in Table 2.1.3 It distinguishes 17 production activities and commodities (or products); seven primary production factors including two labor categories (male and female), private capital, and four natural resources (agricultural land, fish stock, and mineral stocks for oil and gas and other mining); and one household. The government is defined broadly to include donor expenses and receipts. In 2020, the country’s GDP was 13,999 billion Yemeni riyals; with an exchange rate of 743 riyals per US$ and a population of 29.8 million, this corresponds to a GDP per capita of US$623. Tables 2.2a and 2.2b show a macro version of the 2020 SAM and its accounts, respectively. The SAM, which is expressed as percent of GDP, is highly aggregate – among other things, it has only one production sector (i.e., one activity producing one commodity) – making it easy to learn about the macro situation in 2020 from the SAM. The government current account deficit was 4.5 percent (of GDP) (cell [cap-gov,gov]) while government consumption and investment amounted 2 For additional details on the SAM, see the appendix. 3 The disaggregation is primarily based on the most recent UN data. Although it would be possible to further disaggregate the sector assuming that the patterns of the 2014 IFPRI SAM still are valid, this does not seem meaningful: it would not add anything significant to the analysis and, furthermore, the extent to which these patterns also were valid in 2020 is unknown. 4 to 19.4 and 0.7 percent, respectively (cells [c-nagr,gov] and [inv-gov,cap-gov]). The resulting government deficit (5.2 percent) was covered by domestic financing, shown in cell [cap-gov,cap- ngov].4 Yemen’s exports and imports amounted to 6.3 and 44.6 percent, respectively (cells [c - agr,row] and [row,c-agr]), yielding a trade deficit of 38.3 percent. Thanks to a large transfer surplus – the sum of net transfers to households and government amounted to 32.6 percent (cells [hhd,row]+[gov,row])– the current account deficit (foreign savings) was much smaller, at 5.9 percent. Supplemented by domestic savings at 2.9 percent (sum of the cells [cap-ngov,hhd] and cap-gov,gov]), it financed total investment at 8.7 percent – the sum of government investment and private investment (8.0 percent; cell [total,inv-prv]). By international standards this is a very low investment level. Table 2.1. Disaggregation of Yemen SAM and SDGSIM application Group Description Group Description Sectors Qat** Factors Labor - Male (activities & Other agriculture Labor - Female commodities)* Fishing Capital - Private Oil and gas** Land Other mining Natural resource - Fishing Manufacturing Natural resource - Other mining Utilities (electricity, gas, and water) Natural resource - Oil and gas Construction Institutions Households Trade Government Hotels and restaurants Rest of the world Transport Taxes and Taxes - Income Telecommunications subsidies Taxes - Imports Financial serv Taxes - Commodities Real estate and business serv Institutions Non-government Public administration (capital Government Government education and health accounts) Rest of the world Other services Foreign reserves*** Trade & Domestic products Investment Private investment (GFCF) transport Imports Government investment (GFCF) margins Exports *Each sector has an activity and a commodity (output). **Qat and Oil and gas are singled out from agriculture and mining, respectively, using CSO (Central Statistical Organization) estimates for 2017. ***Capital account for payments to foreign reserves; not associated with a current account. 4 The government deficit is defined as the difference between government investment and government savings. (Government savings may also be referred to as the government current account surplus.) 5 Table 2.2a. Macro SAM for Yemen 2020 (% of GDP)* cap-ngov tax-com f-capprv cap-row cap-gov tax-imp f-oilgas inv-gov inv-prv tax-dir total f-lab com row hhd gov act drf act 147.6 147.6 com 49.3 110.2 19.4 6.3 8.0 0.7 194.0 f-lab 59.7 59.7 f-capprv 36.2 36.2 f-oilgas 2.4 2.4 tax-imp 0.8 0.8 tax-com 0.9 0.9 tax-dir 1.2 1.2 hhd 59.7 36.1 4.5 19.7 120.0 gov 2.4 0.8 0.9 1.2 1.1 12.9 19.3 row 44.6 0.1 44.7 cap-ngov 7.4 8.4 15.8 cap-gov -4.5 5.3 -0.1 0.7 cap-row 5.9 2.5 8.3 drf 2.5 2.5 inv-prv 8.0 8.0 inv-gov 0.7 0.7 total 147.6 194.0 59.7 36.2 2.4 0.8 0.9 1.2 120.0 19.3 44.7 15.8 0.7 8.3 2.5 8.0 0.7 *For notation, see Table 2.2b. Source: Authors' calculations based on World Bank, IMF, UN, and WIDER. Table 2.2b. Accounts in macro SAM for Yemen 2020 Account Description Account Description act activity gov government com commodity row rest of the world f-lab factor - labor cap-ngov capital acc - domestic non-gov f-capprv factor - private capital cap-gov capital acc - government f-oilgas factor - oil and gas cap-row capital acc - rest of the world tax-imp tax - imports drf change in international reserves tax-com tax - commodities inv-prv investment - private tax-dir tax - income inv-gov investment - government hhd households Drawing on the disaggregated SAM used with the model, Figures 2.3-2.6 provide a summary of the structure of Yemen’s economy in 2020 from an aggregate sector perspective , showing (a) sectoral shares in value added, employment, exports, and imports (Figure 2.3); (b) export and import intensities (i.e., the shares of sector output that is exported and the share of domestic demand met by imports; Figure 2.4); (c) sectoral factor intensities (i.e., factor shares in the value added of each sector; Figure 2.5); and (d) the demand composition for each sector (i.e., the shares 6 of private consumption, government consumption, investment, intermediate consumption, and exports; Figure 2.6). Figure 2.3 shows that agriculture is relatively important as a source of employment, while mineral extraction – particularly oil and gas -- mainly matters as a source of foreign exchange from exports. In turn, manufacturing is important both in terms of value added and foreign trade but less labor-intensive than agriculture and services. Figure 2.4, which displays the shares of exports in output and imports in domestic demand, points to that (a) agriculture and mining production have a substantial export share while imports play a more moderate role; (b) for manufacturing, both exports and imports are important; and (c) as much as some 59.2 percent of domestic manufacturing demand – including food industry and fuel – is met by imports. Figure 2.5 shows factor shares in the value added of each sector. Among other things, it indicates that agriculture, construction, and services are relatively intensive in their use of (male) labor. On the contrary, mining, manufacturing, and utilities are relatively intensive in their use of physical and natural capital. Currently, women are mainly employed in agriculture, education, manufacturing, and trade; their shares in total employment (male and female) are relatively larger in education, health, manufacturing, and agriculture. In CGE applications, the labor/capital ratios of the production sectors have a major impact on the results obtained from policy simulations. Hence, this information is useful when we analyze the results from the simulations. The main features of the composition of domestic demand, covered in Figure 2.6, are that (a) agriculture and manufacturing are primarily sold domestically, either for private or intermediate consumption; and (b) compared to other products, mining products – particularly oil and gas -- are distinguished by significant exports. 7 Figure 2.3. Yemen: sectoral structure in 2020 (%) a. Value added b. Employment Other services Manufacturi Other ng services Agriculture Agriculture Government Mining Government Mining Transport Manufacturi ng Transport Utilities Utilities Constructio Construction Trade n Trade Other Government Transport c. Exports services d. Imports Constructio Government Transport Other n Trade services Trade Utilities Construction Mining Agriculture Agriculture Utilities Manufacturi ng Manufacturi ng Mining Source: Authors’ calculations based on the 2020 Yemen SAM. 8 Figure 2.4. Yemen: export and import intensities in 2020 (%) 80 70 60 50 percent 40 30 20 10 0 EXP-OUTshr* IMP-DEMshr** *Ratio between exports and production; **Ratio between imports and domestic demand. Source: Authors’ calculations based on the 2020 Yemen SAM. Figure 2.5. Yemen: sectoral factor cost composition in 2020 (%) 100 90 80 70 porcentaje 60 50 40 30 20 10 0 Male labor Female labor Capital Natural resources Source: Authors’ calculations based on the 2020 Yemen SAM. 9 Figure 2.6. Yemen: sectoral demand composition in 2020 (%) 100% 90% 80% 70% 60% percent 50% 40% 30% 20% 10% 0% intermediate use* private cons government cons investment exports *Includes trade and transport margins. Source: Authors’ calculations based on the 2020 Yemen SAM. 2b. Literature overview During the last decades, Yemen’s economy has been analyzed in a large number of applications of CGE models, an indication that it is relatively straightforward to apply such models even in a data-scarce environment like Yemen’s. This section provides a brief survey of the 15 Yemen studies that have been identified, with a focus on method and the issues addressed. The studies were published between 2007 and 2022 in academic journals, as book chapters, or as working papers. Most were undertaken by IFPRI as part of a research effort led by Clemens Breisinger; some work was also done at the World Bank. Readers interested in additional details are referred to the individual references for the studies. The methodological framework of these models is rooted in the pioneering work at the World Bank in the late 1970s and early 1980s that is documented in Dervis, de Melo, and Robinson (1982). The approach that was developed may be described as neoclassical structuralist (Robinson 1989): its starting point is Walrasian general equilibrium theory, but it deviates from this theory whenever this is deemed necessary in order to incorporate essential features of the modeled economy. In terms of the specifics of the Yemen applications, the IFPRI standard model (Lofgren et al. 2002), extended to recursive dynamics by Thurlow (2004), provides the mathematical structure and 10 computer code in GAMS that underpins most if not all Yemen CGE applications.5 This model is user-friendly thanks to the inclusion of pre-programmed alternatives for factor and macro closures, separation between computer code and database (facilitating applications of new databases with different disaggregations), and permits regional disaggregation of sufficient data are available. As a complement to dynamic simulations with these CGE models, many of the Yemen applications have included different types of micro or household simulation modules that, drawing on the CGE simulation results, analyze the consequences for more disaggregated indicators, often related to households and nutrition. The issues addressed in these studies were typically selected in the context of discussions between these international institutions and Yemeni authorities and on the basis of the issues that IFPRI and the World Bank identified as being of high priority for Yemen’s development progress. The analyses were most often forward-looking (at the time when the work was done), addressing the future impacts of packages of complementary policies. However, some studies were also designed to decompose the roles of past external and internal shocks on economic developments. Among these studies, forward-looking analyses have addressed prioritization of public spending between health, education, and agriculture (Chemingui 2007); the long-run impact of climate change (Wiebelt et al. 2011); fuel subsidy reform combined with direct income transfers and infrastructure investments (Breisinger et al. 2012), strategies for achieving the millennium development goals (MDGs; Al-Batuly et al. 2013), the potential roles of growth and targeted measures for improving nutrition among poor groups (Breisinger and Ecker 2014), and conflict recovery (Mukashov et al. 2022). The backward-looking analysis of the determinants of recent economic performance have focused on the impacts of natural disasters and conflict on food security (Ecker et al. 2010), impacts of international price shocks and the financial crisis (Breisinger et al. 2011), and the effects of the 2008 tropical storm and flash flood (Breisinger et al. 2016). In terms of topic, the analysis presented in this paper has most in common with Mukashov et al. (2022) who also address conflict recovery. Methodologically, it is most similar to Al-Batuly et al. (2013) who use the Maquette for MDG Simulations (MAMS), a model developed at the World Bank, which is the predecessor of SDGSIM, the model used in this paper. 5 The computer code is written in GAMS, a software for optimization and simulation modeling developed at the World Bank, also in the late 1970s and early 1980s. For more information, visit www.gams.com. 11 3. Base scenario The base scenario is designed to generate a plausible business-as-usual path for Yemen’s economy during the period 2020-2030, i.e., a path that assumes that fighting continues along the lines of the past eight years. As a result, economic growth is slow and living conditions do not improve. The subsequent analysis compares the base and non-base scenarios during the period 2022-2030. The solution for 2020 replicates the information in the SAM while, in 2021 and 2022, GDP grows at rates that are estimated or projected by the IMF (2022).6 The assumptions are kept as simple and transparent as possible. Most importantly, it is assumed that (a) all international (export and import prices) are constant in real terms; and (b) for most institutional payments, including receipt and spending items in the government budget, the GDP shares are constant, kept at shares computed from the 2020 SAM. At the macro level, it is assumed that the government budget is cleared via tax adjustments, that the real exchange rate clears the balance of payments, and that household savings adjust to help maintain domestically financed private investment with an unchanged GDP share. Figures 3.1-3.7 summarize results from the base scenario (covering indicators related to the macroeconomy, sectors, household welfare, and poverty); Tables 3.1-3.6 provide additional information about both base and non-base scenarios, including macro data (growth and GDP shares), the government budget, the balance of payments, and data on economic structure in 2022 and for the base scenario in 2030. As intended, to facilitate comparisons to non-base scenarios, the picture is one of stable growth starting from 2023. GDP growth is simulated at an annual rate at 2.0 percent between 2022 and 2030 (Figure 3.1).7 Absorption (the sum of domestic final demands) is boosted by Yemen’s trade deficit, permitting it to exceed, by a wide margin, GDP at purchasers’ prices and, even more so, GDP at factor cost (Figure 3.2). Among domestic demands, household consumption dominates (on the left-hand axis in Figure 3.3), followed by government consumption (which includes consumption financed by donors and managed by NGOs and others), private investment, and government investment (right-hand axis in Figure 3.3). In terms of GDP shares (Table 3.2), Yemen’s private consumption represents 108.6 percent while the sum of the other domestic final demands is as low as 28.1 percent, values that are extreme by international standards. The macroeconomic aggregates covered in Figures 3.1-3.3 grow at annual rates of 1-2 percent (Figure 3.4). The growth rate for household consumption coincides almost exactly with the population growth rate, leaving per capita consumption unchanged with a headcount poverty rate of 74.4 percent both in 2022 and 2030. Among the sectors, the private service GDP is more 6 While GDP growth is exogenous for the base scenario, it is endogenous for all non-base scenarios, deviating from the base due to the shocks, policy and other, that are imposed. 7 Unless otherwise noted, the growth rates are geometric averages for the period 2022 to 2030. 12 than twice the level of the second sector, agriculture (Figure 3.5). The aggregate sectors grow at annual rates close to 2 percent with mining as the only exception (Figure 3.6), reflecting a balanced growth path. For mining (oil, gas and other), growth is constrained by the assumption that extraction rates are relatively stagnant. The resulting growth in employment does not keep up with the growth in the labor force, leading to increased unemployment rates both for men and women (Table 3.1).8 In the base scenario, aggregate export growth is at 1 percent, kept in check by a negative growth rate for mining, which in 2022 represented around 62 percent of total exports (Figure 3.7; Table 3.6). For the non-mining sectors, the growth rates are around 3 percent. As shares of GDP, the changes in size of the government and its different receipt and spending items are minor (Table 3.3). Foreign aid, which amounts to around 14 percent of GDP, is assumed to be almost entirely in grant form. Government foreign financing (borrowing net of interest) payments are close to zero, generating a virtually unchanged ratio between foreign government debt and GDP (Table 3.2). 8 The unemployment rate is the combined rate of time-related underemployment and unemployment (i.e., according to ILO terminology, the LU2 rate). 13 Figure 3.1. Base: GDP at factor cost annual growth 2021-2030 (%) 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5 -2.0 -2.5 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 Source: Authors' calculations based on IMF (2022). Figure 3.2. Base: Absorption, GDP, and trade 2022-2030 (bn 2020 YER) 2,500 2,000 1,500 1,000 500 0 Absorption Exports Imports GDP at factor cost Source: Authors' calculations based on simulation results. 14 Figure 3.3. Base: Domestic final demands 2022-2030 (bn 2020 YER) 1,800 400 1,600 360 Household consumption 1,400 320 Other indicators 1,200 280 240 1,000 200 800 160 600 120 400 80 200 40 0 0 Hhd cons. Gov't cons. Private inv. Gov't inv. Source: Authors' calculations based on simulation results. Figure 3.4. Base: Real annual macro growth 2023-2030 (%) 0.0 0.5 1.0 1.5 2.0 2.5 Absorption 2.0 Household consumption 2.0 Government consumption 2.0 Private investment 2.2 Government investment 2.0 Exports 1.0 Imports 1.8 GDP at factor cost 2.0 Source: Authors' calculations based on simulation results. 15 Figure 3.5. Base: Sector GDP 2022-2030 (bn 2020 YER) 800 400 700 350 600 300 Private services Other sectors 500 250 400 200 300 150 200 100 100 50 0 0 Private services Agriculture Mining Manufacturing Other industry Gov't services Source: Authors' calculations based on simulation results. Figure 3.6. Base: Real annual sector GDP growth 2023-2030 (%) Figure 3.6. Base: Real annual sector GDP growth 2023-2030 (%) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Agriculture 1.9 Mining 0.1 Manufacturing 2.4 Other industry 2.1 Private services 2.1 Gov't services 2.0 Total 2.0 Source: Authors' calculations based on simulation results. 16 Figure 3.7. Base: Real annual sector export growth 2023-2030 (%) Figure 3.7. Base: Real annual sector export growth 2023-2030 (%) -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 Agriculture 3.2 Mining -0.9 Manufacturing 3.4 Private services 3.0 Total 1.0 Source: Authors' calculations based on simulation results. Table 3.1. Macro indicators in 2022 and by scenario (% annual growth 2023-2030) 2022 base limbo rec-t rec-ta rec+t rec+ta rec+ta+ Absorption 1,848.5 2.0 2.7 3.6 3.6 5.2 5.2 6.0 Consumption - private 1,460.5 2.0 2.6 3.2 3.2 4.2 4.2 5.1 Consumption - government 268.7 2.0 3.1 2.1 2.1 3.8 3.8 3.8 Fixed investment - private 109.4 2.2 3.1 8.6 8.6 13.6 13.6 14.7 Fixed investment - government 9.9 2.0 2.8 22.0 22.0 30.9 30.9 30.9 Exports 95.1 1.0 2.3 4.2 4.2 6.9 6.9 9.0 Imports 561.4 1.8 2.0 2.9 2.9 4.0 4.0 4.4 GDP at factor cost 1,361.5 2.0 3.1 4.0 4.0 6.0 6.0 7.1 Total factor employment (index) Eps 1.2 1.5 2.1 2.1 3.0 3.0 3.3 Total factor productivity (index) Eps 0.8 1.6 1.9 1.9 3.0 3.0 3.8 GNI 1,382.8 2.0 3.0 3.7 3.8 5.5 5.5 6.6 GNDI 1,849.8 2.0 2.8 3.4 3.4 4.8 4.8 5.7 GNI per capita 0.4 0.0 1.1 1.7 1.8 3.4 3.5 4.5 GNDI per capita 0.6 0.0 0.8 1.4 1.4 2.8 2.8 3.6 Real exchange rate (index) Eps 0.61 1.66 1.90 1.92 2.81 2.83 3.76 Unemployment rate (%) 25.4 31.9 29.8 29.1 29.0 25.5 25.3 23.3 Male (%) 24.4 31.0 29.0 28.2 28.1 24.6 24.5 22.4 Female (%) 36.8 42.3 40.0 39.5 39.4 35.8 35.8 33.7 Headcount poverty rate (%) 74.4 74.4 72.0 70.0 69.8 66.0 65.8 62.4 Note: 1. Unless otherwise noted, column for initial year shows data in bn 2020 YER. 2. For the unemployment and poverty rates, the base-year and simulation columns show base-year rates and simulation-specific final-year rates, respectively. 17 Table 3.2. Macro indicators in 2022 and by scenario in 2030 (% of nominal GDP) Indicator 2022 base limbo rec-t rec-ta rec+t rec+ta rec+ta+ Absorption 136.7 138.5 138.9 138.7 138.7 137.7 137.8 137.7 Consumption - private 108.6 111.0 111.4 109.1 109.1 104.9 104.9 106.3 Consumption - government 19.3 18.8 18.8 15.9 15.9 15.4 15.3 14.0 Investment - private 8.0 8.0 8.0 11.2 11.2 13.7 13.7 13.8 Investment - government 0.7 0.7 0.7 2.5 2.5 3.8 3.8 3.6 Exports 7.5 7.2 8.1 8.9 8.9 10.1 10.1 11.9 Imports 44.2 45.7 47.0 47.6 47.6 47.9 47.9 49.6 GDP at market prices 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Net indirect taxes 3.1 2.6 2.5 2.3 2.3 3.0 2.9 1.9 GDP at factor cost 96.9 97.4 97.5 97.7 97.7 97.0 97.1 98.1 GNI 99.9 99.9 99.9 97.6 97.6 95.7 95.8 95.6 GNDI 136.6 138.5 138.9 135.4 135.4 131.6 131.7 131.6 Foreign savings 0.0 0.0 0.0 3.3 3.3 6.1 6.1 6.1 Gross national savings 8.7 8.7 8.7 10.4 10.4 11.4 11.4 11.3 Foreign government debt 43.1 43.7 44.1 41.7 41.7 38.7 38.6 38.7 Domestic government debt 6.5 10.9 10.4 9.7 9.7 8.7 8.7 8.2 Table 3.3. Government receipts and spending in 2022 and by scenario in 2030 (% of nominal GDP) Indicator 2022 base limbo rec-t rec-ta rec+t rec+ta rec+ta+ Receipts Direct taxes 3.0 2.3 2.0 1.7 1.7 2.5 2.4 1.1 Import tariffs 0.8 0.8 0.9 0.9 0.9 0.9 0.9 0.9 Other indirect taxes 2.3 1.8 1.6 1.4 1.4 2.1 2.1 0.9 Private transfers 1.1 1.1 1.1 1.1 1.1 1.1 1.1 1.1 Foreign transfers 14.0 14.8 14.9 14.1 14.1 13.1 13.0 13.1 Domestic financing 0.3 0.3 0.3 0.3 0.3 0.2 0.2 0.2 Foreign financing 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Total 24.5 23.9 23.9 23.0 23.0 23.7 23.7 21.9 Spending Consumption 19.3 18.8 18.8 15.9 15.9 15.4 15.3 14.0 Fixed investment 0.7 0.7 0.7 2.5 2.5 3.8 3.8 3.6 Private transfers 4.4 4.4 4.4 4.5 4.5 4.5 4.5 4.3 Total 24.5 23.9 23.9 23.0 23.0 23.7 23.7 21.9 18 Table 3.4. Balance of payments in 2022 and by scenario in 2030 (% of nominal GDP) Indicator 2022 base limbo rec-t rec-ta rec+t rec+ta rec+ta+ Outflows Imports 44.2 45.7 47.0 47.6 47.6 47.9 47.9 49.6 Factor payments 0.1 0.1 0.1 2.4 2.4 4.3 4.2 4.4 Change in foreign reserves 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 Total 44.4 46.0 47.3 50.2 50.2 52.3 52.3 54.2 Inflows Exports 7.5 7.2 8.1 8.9 8.9 10.1 10.1 11.9 Transfers to households 22.7 23.8 24.1 23.7 23.7 22.9 22.8 22.9 Transfers to government 14.0 14.8 14.9 14.1 14.1 13.1 13.0 13.1 Private borrowing 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 Government borrowing 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 FDI 0.0 0.0 0.0 3.3 3.3 6.1 6.1 6.1 Total 44.4 46.0 47.3 50.2 50.2 52.3 52.3 54.2 Table 3.5. GDP at factor cost by sector -- level in 2022 (bn 2020 YER) and annual growth by scenario 2023-2030 (%) 2022 base limbo rec-t rec-ta rec+t rec+ta rec+ta+ Agriculture 245.4 1.9 2.9 3.1 3.1 4.0 4.0 5.0 Industry 255.8 1.8 2.9 4.7 4.8 7.2 7.3 8.8 Mining 72.1 0.1 0.3 2.1 2.1 3.9 3.9 5.0 Manufacturing 128.5 2.4 4.1 5.3 5.4 7.8 7.8 9.7 Other 55.2 2.1 3.0 6.0 6.1 9.4 9.4 10.4 Services 797.9 2.1 3.2 4.1 4.1 6.2 6.2 7.2 Private 593.9 2.1 3.2 4.9 4.9 7.3 7.3 8.5 Government 204.0 2.0 3.1 1.3 1.3 2.3 2.3 2.3 Total 1,299.1 2.0 3.1 4.0 4.0 6.0 6.0 7.1 19 Table 3.6. Sector structure in 2022 and for base in 2030 (%) Value Produc- Employ- Export/ Import/ added tion ment Exports Imports output demand 2022 Agriculture 18.9 18.8 28.8 11.9 11.1 2.7 14.3 Industry 19.7 31.6 10.2 81.8 79.5 11.2 45.8 Mining 5.6 3.9 0.2 59.0 1.4 68.9 26.8 Manufacturing 9.9 16.4 4.7 22.8 77.8 5.3 57.8 Other 4.2 11.3 5.3 0.0 0.3 0.0 0.8 Services 61.4 49.6 61.0 6.3 9.4 0.6 4.9 Private 45.7 37.7 39.2 5.5 8.9 0.7 6.1 Government 15.7 11.9 21.8 0.9 0.5 0.3 1.0 Total 100.0 100.0 100.0 100.0 100.0 4.3 24.5 2030 Agriculture 21.1 20.1 28.9 14.2 11.6 3.0 14.4 Industry 19.9 31.8 10.6 78.4 79.4 10.2 45.7 Mining 5.3 3.7 0.1 50.9 1.5 60.8 24.8 Manufacturing 10.5 17.1 5.2 27.5 77.6 6.0 57.4 Other 4.1 11.1 5.2 0.0 0.3 0.0 0.8 Services 59.0 48.1 60.5 7.4 9.0 0.7 5.0 Private 44.0 36.6 38.8 6.4 8.5 0.8 6.2 Government 15.0 11.5 21.7 1.0 0.4 0.4 1.0 Total 100.0 100.0 100.0 100.0 100.0 4.2 24.7 Δ (2030-2022) Agriculture 2.2 1.3 0.2 2.3 0.5 0.2 0.1 Industry 0.2 0.2 0.4 -3.4 -0.1 -1.0 -0.1 Mining -0.3 -0.2 0.0 -8.1 0.1 -8.2 -2.0 Manufacturing 0.6 0.7 0.5 4.7 -0.2 0.7 -0.4 Other -0.1 -0.2 -0.1 0.0 0.0 0.0 0.0 Services -2.4 -1.5 -0.5 1.1 -0.4 0.1 0.1 Private -1.7 -1.1 -0.4 0.9 -0.4 0.1 0.1 Government -0.7 -0.4 -0.1 0.1 0.0 0.1 0.0 Total 0.0 0.0 0.0 0.0 0.0 -0.2 0.2 20 4. Non-base scenarios The non-base simulations explore alternative potential paths for the economy of the country up to 2030, all assuming that, compared to the base, the security situation improves and that, in response, economic conditions improve due to more rapid private sector growth. Beyond this general description, there are significant differences between the non-base scenarios. While the paths are implemented in the context of a rigorous model and informed by cross-country evidence, they should merely be seen as pointing to possible gains from a conflict resolution, not as predictions of what will happen. The non-base simulations start to deviate from the base in 2023 – all simulations are identical up to 2022. Given this, the analysis is based on comparisons between the non-base scenarios and the base scenario during the period 2023-2030, taking 2022 as the identical starting point. Table 4.1 provides broad non-quantitative definitions of all non-base scenarios, linking each scenario to a broader context of political and military developments. In sum, the scenario limbo tries to capture the impact of continued and lasting but still relatively marginal security improvement compared to the preceding years whereas the recovery scenarios (rec- and rec+) assume that the political and military conflict ends; the result is more rapid productivity growth as the economy enjoys improved security, leading to some improvements in most economic indicators. The weak recovery scenarios (rec-) are associated with the emergence of a recentralized state in the context of military victory for one side or a political settlement while the strong recovery scenarios (rec+) assume that the result is a more developmental decentralized state, drawing on the decentralization that has emerged on the ground but strengthened by coalition-building among former enemies and rivals. For all recovery scenarios, the government expands spending in critical areas (health, education, social transfers, and infrastructure) at the same time as the outside world, encouraged by more positive domestic events, boost development via increases in private transfer inflows (“remittances”), foreign direct investment (FDI), and investments facilitating more rapid oil and gas extraction. Foreign grant aid does not increase for a subset of the recovery scenarios (rec-t and rec+t) whereas for others (rec-ta, rec+ta, and rec+ta+), it is above base levels 2023-2029 in support of government efforts to foster the recovery. The weak and strong recovery scenarios differ in that these positive shocks are twice the size for those that are stronger ones. In addition, the final strong scenario, rec+ta+, assumes that the government is able to better manage its infrastructure investments, something that is reflected in a doubling of the marginal productivity (MP) of new infrastructure capital. 21 Table 4.1. Broad and contextual definitions of all non-base scenarios Name Description Emerging status quo Context A fragile peace is brokered without any political settlement with intermittent fighting. limbo Simulation assumptions: Same as base except for …. … moderate growth accelerations in private and government sectors thanks to more rapid productivity growth due to the security improvement. Weak recovery scenarios* Context Emergence of a recentralized state as the military conflict is concluded due to victory for one side or a political settlement. rec-t Simulation assumptions: Same as base except for …. ... moderate expansions in (a) government spending on health, education, social transfers, and infrastructure; (b) transfers to households from abroad; (c) foreign direct investment (FDI); and (d) the rates of oil and gas extraction. rec-ta Simulation assumptions: Same as rec-t except for …. … a moderate increase in foreign grant aid 2023-2029 Strong recovery scenarios* Context Comprehensive political settlement enshrines decentralized state with strong powers at the level of the governorates with coalition-building among former rivals and enemies. rec+t Simulation assumptions: Same as rec-t except for …. … expansions in areas a-d that are twice the size. rec+ta Simulation assumptions: Same as rec+t except for …. … an increae in foreign grant aid 2023-2029 that is twice the increase for rec-ta. rec+ta+ Simulation assumptions: Same as rec+ta except for …. … a doubled marginal productivity (MP) of new government infrastructure capital. *The detailed assumption for the recovery scenarios (weak and strong) are presented in Table 4.2. 22 Table 4.2. Detailed definitions of recovery scenarios Scenario rec+t rec-t rec+ta Shock area Changes compared to base* rec-ta** rec+ta+*** Health & Government health & education spending (% of base GDP) education (evenly split between consumption and investment) +2.7 +5.4 Labor productivity in 2030 (%) -- gradual increase 2027-2030 +7.9 +15.8 Infrastructure Government infrastructure spending (% of base GDP) (32.5% consumption, 67.5% investment) +3.3 +6.7 Social transfers Transfers from government to households (% of GDP) +0.6 +1.2 Remittances Transfers to households from abroad (US$) (% annual growth) (2023-2030) +0.5 +1.0 FDI Foreign direct investment (% of base GDP; in US$) +3.45 +6.9 Oil-gas output Change in oil & gas resource extraction levels (% annual growth) +1.3 +2.4 *Unless otherwise noted, changes are introduced gradually 2023-2026 and after that kept in place up to 2030. **The difference between rec-t and rec-ta is that the latter enjoys foreign grant increases that in 2023- 2026 gradually increase from 0.5 to 2 percent of base scenario 2022 GDP and in 2027-2030 gradually decline, in 2030 being back at the 2030 base level. ***The difference between rec+t compared to rec+ta and rec+ta+ is that the latter two enjoy foreign grant increases that are twice as large as for rec-ta, peaking at 4 percent of 2022 GDP in 2026. rec+ta+ differs from all other recovery scenarios in the MP of government capital is 0.6 instead of 0.3. Table 4.2 provides additional technical and quantitative information about the different positive economic shocks that are implemented for the recovery scenarios. For the rec+ scenarios, the basic assumption is that Yemen by 2030 will have managed to take actions and realize improvements that bring the country up to outcomes typical of the 90th percentile among LICs; for the rec- scenarios, the basic assumptions is that, more modestly, the country closes half of the gap between its current government performance and that of the 90th percentile. Both for the rec+ and rec- scenarios, it is assumed that real government consumption not associated with health-education or infrastructure stays at the 2022 level (instead of growing at an annual rate of 2 percent under the base scenario); this may be seen as an efficiency gain since it makes it easier for the government to channel resources to high-priority spending without increases in taxes or foreign aid. 23 More specifically, for the rec+ scenarios, the increases in government health and education spending (split into consumption [current] and investment [capital]) during the period 2023-2026 are defined to permit the country to deliver services typical of the 90th LIC percentile starting from 2026.9 It is assumed to generate a more gradual increase in Yemen’s human-capital index (HCI) to the 90th LCI percentile during the period 2023-2030 as a growing part of the labor force benefits from improved health and education. In its turn, labor productivity grows relative to the 2022 level at a rate that, according to cross-country evidence, is expected given the HCI increase.10 For infrastructure, the rec+ scenarios impose gradual increases in consumption (operations and maintenance) and investment spending that, on average may be sufficient for low- and middle- income countries to achieve the infrastructure-related sustainable development goals (SDGs; Rozenberg and Fay 2019; Foster et al. 2022). Drawing on cross-country estimates (Gupta et al. 2014, p. 171), it is assumed that the infrastructure capital stock has a marginal productivity (MP) of 0.30 for rec+t and rec+ta, and 0.60 for rec+ta+. The gains are realized in the form of increases in total factor productivity (TFP); more specifically, once installed and other things being equal, value-added increases by 0.30 (0.60) YER for every additional YER of infrastructure capital with the gains spread across all sectors in proportion to their initial shares in value added. Finally, the scenarios rec+ta and rec+ta+ deviate from rec+t in that additional foreign grant aid is received: in each year 2023-2026 foreign grants increase by 1 percent of 2022 GDP (i.e., reaching 4 percent in 2026); the increases are reversed during the period 2027-2030 so that, in 2030, foreign grants are back at the level of the base scenario. Similarly, the assumed increase in social transfers from the government to households would permit Yemen to reach the 90th percentile among LICs in terms of spending on social safety nets, assuming that Yemen’s initial level was in the bottom half among LICs (more specifically in the 25th percentile).11 For remittances (i.e., transfers to Yemeni households from abroad expressed in foreign currency), the base scenario assumes that they grow at the same rate as the population, 2 percent, i.e., that they are unchanged in per-capita terms. For the rec+ scenarios, it is assumed that, starting from 2023, the annual growth rate (in foreign currency) increases by 1 percentage point. Such a development may require that the Gulf Cooperation Council (GCC) countries give preferential 9 More specifically, in each year the health and education spending increases for the rec+ and rec- scenarios are computed relative to base GDP and translated into increase in real service provision and real additions to the capital stock. Thus, the real changes are the same for all rec+ (rec-) scenarios, not depending on changes in GDP or prices in the individual rec+ (rec-) scenarios. 10 More specifically, the spending levels by LIC percentile were computed using World Development Indicators data for 2019 (World Bank 2020). The link between the spending percentile, the HCI, and labor productivity is based on data in World Bank (2021). 11 LIC percentile data for spending on social safety nets were computed on the basis of data in the World Bank’s ASPIRE dataset (World Bank 2018). For FDI, discussed in the following paragraph, the calculation was based on World Bank (2021). 24 treatment to worker migrants from Yemen and facilitate their remittance back home. In the model, the households primarily use the increases in social transfers and remittances to boost their consumption. Currently, foreign direct investment (FDI) appears to be close to zero. In the rec+ scenarios, it is assumed that, during the period 2023-2026, FDI gradually reaches and after that stays at the LIC 90th percentile level among LICs. These increases in FDI accelerate the buildup of the private capital stock in Yemen, contributing to more rapid growth in production and incomes. Finally, oil and natural gas dominate Yemen’s export revenues and, at least during the period up to 2030, these commodities are likely to continue to remain major foreign exchange earners. Under the base scenarios, the rate of extraction is assumed to grow at a modest annual rate of 0.1 percent. Under the rec+ scenarios, it is increased by 2.4 percentage points. For sectors other than oil and natural gas, productivity increases due to investment in human capital and infrastructure may contribute to more rapid export growth, especially if the sectors already have managed to export a significant share of its output and if domestic demand growth does not keep up with output growth. As noted in Table 4.2, the rec- scenarios differ from the rec+ scenarios in that Yemen only eliminates half of the gap between its current situation and that of the 90 th LIC percentile in terms of provision of health and education services, infrastructure, and social transfers; accordingly, the productivity gains that are realized thanks to these changes are only half the size. In other areas – remittances, FDI, and oil-gas output – the increases are half as large as for the rec+ scenarios. The difference between the rec-t and rec-ta scenarios is that the latter enjoys a foreign grant increase; however, the increase corresponds to 0.5 percent of base 2022 GDP (as opposed to 1 percent for rec+ta and rec+ta+ compared to rec+t). Aside from the scenario features discussed above, the non-base scenarios have various assumptions in common that are different from those of the base: GDP is endogenous and, instead of letting private investment drive savings (as was done for the base), private savings (with exogenous savings rates) drive domestic private investment. This means that, other things being equal, increases in private incomes net of direct taxes translate into increases in private consumption, savings, investment spending, real investment, and capital stocks, the latter contributing positively to GDP.12 For the government budget and the balance of payments, the base assumptions are retained: unless otherwise noted, foreign grant aid is at the same level (expressed in foreign currency); foreign financing of the government (borrowing net of interest payments) remains close to zero; and taxes and the real exchange rate adjust to clear the government budget and the balance of payments, respectively. The level of tax adjustments is in 12 The only exception is that, for the limbo scenario, it was for simplicity assumed that, just as for the base, GDP growth was assumed to be exogenous, thus endogenizing the TFP improvements. It would have been more cumbersome to implement the alternative of exogenously shocking TFP with endogenous GDP growth. 25 fact an important part of the analysis as it will depend on government spending changes and the level of foreign grant aid and influence the level of household consumption. The results for the different scenarios are summarized in Figures 4.1-4.9; as noted, additional results for both base and non-base scenarios are provided in Tables 3.1-3.6. Figure 4.1 shows the deviations from the base for growth in aggregate government consumption and investment. In the absence of policy changes, the increases for limbo are small, merely reflecting that government spending responds to economic growth. For the recovery scenarios, they reflect policy changes, including larger increases for the rec+ scenarios. They are much larger for investment due to the fact that, under the base, it was at very low levels while, at the same time, government consumption growth is kept in check by a freezing of non-priority spending at 2022 levels throughout the simulation period. As shown in Figure 4.2, across the board, growth increases with a strong correlation between the different scenarios: the gains in productivity, induced by improved security and policy changes, translate into higher GDP and income levels that generate higher household consumption, savings, private investment, exports, and imports, with the investment increases feeding back to higher growth. For the recovery scenarios, private investment growth is boosted by increases in FDI. The import and export gains may be seen as spillovers from increases in domestic supplies and demands that are imperfectly matched. Growth rates are higher for exports than for imports since the initial export value is smaller and the trade deficit is not permitted to expand compared to the base given exogenous values in foreign currency for other balance of payments items.13 Respect for the balance of payments constraint is imposed via real exchange rate depreciation, encouraging exports and discouraging imports (cf. Table 3.1). The growth gains are weak for the limbo scenarios and strongest for the rec+ scenarios, exceeding the base by around 2-3 percent points for household consumption and 4-5 points for GDP (at factor cost), and exceeding the corresponding rec- scenarios by around 1 point for household consumption and 2 for GDP. Among the recovery scenarios, the growth gains are virtually identical for the scenarios with and without the additional aid (rec-ta vs. rec-t; and rec+ta vs. rec+t); however, as will be noted below, higher foreign aid has other more significant positive effects. On the contrary, the assumption of a higher MP of government capital for the final scenario, rec+ta+, makes a substantial difference, raising growth in household consumption and GDP by 0.8 and 1.1 percent points, respectively. Deviations from base growth rates at the sector level are shown in Figure 4.3. For the limbo scenario, the expansion in manufacturing is strongest due to the combined impact of higher income elasticities of demand from households (compared to agriculture) and the absence of sectoral natural resource constraints. Across all recovery scenarios, the increases are strongest for other industry, reflecting strong increases in construction (part of other industry) to meet growing government and private investment demand. For mining, the growth gains reflect the 13 For the full-period growth figures, this also applies to the scenarios with increased foreign aid since these increases end in 2029. 26 assumption that extraction rates are increased, most strongly for the rec+ scenarios. The small growth changes for government services (defined as public administration and public health and education) reflect the net of a policy-driven increase for education and health (which is stronger for the rec+ scenarios) and a decrease to zero growth for public administration. The gains in GDP growth raise employment growth sufficiently to reduce male and female unemployment rates relative to the 2030 base rate; the extent of the rate cuts is linked to differences in GDP growth gains. For the rec+ scenarios, female unemployment also reaches slightly below the 2022 level; for males, this is only the case for the rec+ta+ scenario (Table 3.1). The sectoral changes in export growth, shown in Figure 4.4, are highly correlated with the changes in sectoral GDP growth; the only partial exception is agriculture, for which exports are boosted by lower growth rate for domestic demands. (Government services are omitted from this figure since their exports are negligible.) The changes in headcount poverty – see Figure 4.5 -- are determined by changes in household consumption per capita, assuming an unchanged consumption distribution. As noted, the base rate in 2030 is identical to the 2022 rate, 74.4 percent, reflecting unchanged per-capita consumption. The rate declines slightly for the limbo scenario, to 72 percent. For the different recovery scenarios, the 2030 rates are in the range of 63-70 percent, corresponding to annual declines of 0.5-1.0 percent points. In addition, for the recovery scenarios, household welfare would improve thanks to better services in health and education, better infrastructure, and improved physical security. As noted above, the differences between the scenarios with and without aid increases are negligible for indicators that only consider data for 2022 and/or 2030 as opposed to the evolution of the economy between these two years. By tracing the annual level of household consumption per capita between 2022 and 2030, indexed to 100 for 2022, Figure 4.6 highlights these differences. Compared to the scenarios rec-t and rec+t, the scenarios rec-ta and rec+ta, which only differ in that they enjoy the aid addition, are able to maintain a more even pace of consumption growth with noticeably stronger performance in the initial years when the gains from the policy shifts still are weak and, in the absence of the aid increases, the increases in domestic taxes are larger. Figures 4.7 and 4.8 show the evolution over time for foreign grant aid and its implication for taxation for the rec- and rec+ scenario pairs; the tax increase in 2026 (the peak tax year) compared to 2022 amounts to 3.5 and 2.0 percent of GDP for rec-t and rec-ta, respectively; and to 8.3 and 5.5 percent for rec+t and rec+ta, respectively. If the total addition to fiscal space from aid, taxes or other resources is not sufficient, it would be necessary to opt for less ambitious government programs. However, given the humanitarian crisis in Yemen, it is evident that temporary additions to the aid flows may make an essential difference for living standards, which also may strengthen the prospects for a sustainable resolution of the conflict. Finally, Figure 4.9 provides an alternative perspective on the macroeconomic gains under the different non-base scenarios and how these are distributed across the different types of final 27 demands. Relative to GDP in 2030 for the base scenario, the total gains in final demands range from less than 10 percent for limbo to close to 50 percent for rec+ta+. For the limbo scenario, which does not assume any significant structural change, the bulk of the modest increases accrue to private (or household) and government consumption whereas, for the recovery scenarios, roughly 80-85 percent of the increases are in the form of higher private final demands. In the background, the recovery scenarios assume a specific division of labor between the private sector and the government: the government represents a small but strategically important growth- instigating part of total final demands, helping the private sector to grow, while the private sector accounts for the bulk of final demands and an even larger part of production, given that the bulk of the supplies that meet government investment demands are provided by private sector producers. Figure 4.1. Government demand growth by scenario (% pt. deviation from base) 28.9 28.9 28.9 30 25 20.0 20.0 20 15 10 5 1.8 1.8 1.8 1.1 0.8 0.1 0.1 0 limbo rec-t rec-ta rec+t rec+ta rec+ta+ Gov't cons'on Gov't inv'ment 28 Figure 4.2. Growth in other macro indicators by scenario (% pt. deviation from base) 12.5 14 11.4 11.4 12 10 8.0 6.5 6.5 8 6.0 5.9 5.1 6 4.0 4.0 3.3 3.2 3.1 2.6 4 2.2 2.2 2.2 2.1 2.0 2.0 1.3 1.2 1.2 1.1 1.1 1.1 0.9 0.7 2 0.2 0 limbo rec-t rec-ta rec+t rec+ta rec+ta+ Hhd cons'on Private inv'ment Exports Imports GDP at f.c. Figure 4.3. Sector GDP growth by scenario (% pt. deviation from base) 9 8 7 Agriculture 6 Mining 5 4 Manufacturing 3 Other industry 2 Private services 1 Gov't services 0 -1 limbo rec-t rec-ta rec+t rec+ta rec+ta+ 29 Figure 4.4. Sector export growth by scenario (% pt. deviation from base) 13 12 11 10 9 8 Agriculture 7 6 Mining 5 Manufacturing 4 3 Private services 2 1 0 -1 limbo rec-t rec-ta rec+t rec+ta rec+ta+ Figure 4.5. Headcount poverty rate in 2022 and by scenario in 2030 (%) 80 74.4 74.4 75 72.0 70.0 69.8 70 66.0 65.8 65 62.4 60 55 50 2022 base limbo rec-t rec-ta rec+t rec+ta rec+ta+ 30 Figure 4.6. Household consumption per capita by year (index 2022=100) 130 125 base 120 limbo 115 rec-t 110 rec-ta 105 rec+t rec+ta 100 rec+ta+ 95 Figure 4.7. Grant aid for selected simulations (% of GDP) 16 14 12 10 8 6 4 2 0 2022 2023 2024 2025 2026 2027 2028 2029 2030 rec-t rec-ta rec+t rec+ta Figure 4.8. Taxes for selected simulations (% of GDP) 16 14 12 10 8 6 4 2 0 2022 2023 2024 2025 2026 2027 2028 2029 2030 rec-t rec-ta rec+t rec+ta 31 Figure 4.9. Increases in final demands in 2030 by scenario (% of base GDP in 2030). 50 45 40 35 30 Gov't inv. 25 Gov't cons. 20 Private inv. 15 Private cons. 10 5 0 limbo rec-t rec-ta rec+t rec+ta rec+ta+ 5. Conclusions This paper explores potential scenarios for Yemen’s economy up to 2030, drawing on a new database, including a SAM, put together from scattered data, and SDGSIM, a CGE model that imposes rigor based on first economic principles. The results suggest that, if the conflict subsides, the donor community is supportive, and the economy responds to increased spending on human development and infrastructure in ways that are typical of low-income countries, considerable progress, including reduced poverty rates and improved living conditions, can be achieved by 2030. Given Yemen’s low levels of infrastructure and human development, the potential payoffs from investments in these areas are great. However, high payoffs will require strong governance in general, including efficient government spending and the ability to raise additional taxes without a negative impact on income distribution or private sector incentives. Nevertheless, foreign aid may play a crucial role by reducing the need to raise domestic taxes, making it easier to realize badly needed improvements in living standards during years of expanded public investments in Yemen’s development. 32 References Aguiar, A., Chepeliev, M., Corong, E., McDougall, R. and van der Mensbrugghe, D. 2019. The GTAP Data Base: Version 10. Journal of Global Economic Analysis, 4(1): 1–27. Al-Batuly, Abdulmajeed; Al-Hawri, Mohamed; Cicowiez, Martin; Lofgren, Hans; and Pournik, Mohammad. 2013. Yemen. Marco V. Sánchez and Rob Vos (eds.) Financing Human Development in Africa, Asia and the Middle East. Chapter 10, pp. 247-277. New York: Bloomsbury Publishing. https://doi.org/10.18356/324d850a-en Annabi, Nabil, John Cockburn, and Bernard Decaluwé. 2006. Functional Forms and Parametrization of CGE Models. MPIA Working Paper 2006-04. Poverty and Economic Policy (PEP) Network. Andrews, Kathryn; Avitabile, Ciro; Gatti, Roberta. 2019. Domestic Government Spending on Human Capital: A Cross-Country Analysis of Recent Trends. Policy Research Working Paper No. 9033. World Bank, Washington, DC. Bander, Balkhi; Dhfer, Alshayban; Nawaf M., Alotaibi. 2021. Impact of Healthcare Expenditures on Healthcare Outcomes in the Middle East and North Africa (MENA) Region: A Cross-Country Comparison, 1995–2015. Frontiers in Public Health 8: 1-9. Breisinger, Clemens; Diao, Xinshen; Collion, Marie-Helen; and Rondot, Pierre. 2011. Impacts of the Triple Global Crisis on Growth and Poverty: The Case of Yemen. Development Policy Review, Volume 29, Issue 2, pp. 155-184. https://doi.org/10.1111/j.1467-7679.h.00530.x Breisinger, Clemens; Engelke, Wilfried; and Ecker, Olivier. 2012. Leveraging Fuel Subsidy Reform for Transition in Yemen. Sustainability, 4(11), 2862-2887; https://doi.org/10.3390/su4112862 Breisinger, Clemens; and Ecker, Olivier. 2014. Simulating economic growth effects on food and nutrition security in Yemen: A new macro–micro modeling approach. Economic Modelling, Volume 43, pp. 100-113. https://doi.org/10.1016/j.econmod.2014.07.029 Breisinger, Clemens; Ecker, Olivier; Thiele, Rainer; and Wiebelt, Manfred. 2016. Effects of the 2008 flood on economic performance and food security in Yemen: a simulation analysis. Disasters. Volume 40, Issue2, pp. 304-326. https://doi.org/10.1111/disa.12147 Chemingui, Mohamed Abdelbasset. 2007. Public Spending and Poverty Reduction in an Oil-Based Economy: The Case of Yemen. IFPRI Discussion Paper No. 701. CSO (Central Statistical Organization). 2019. Statistical Year Book for 2017. Dervis, Kemal; de Melo, Jaime; and Robinson, Sherman. 1982. General equilibrium models for development policy. New York: Cambridge University Press. Ecker, Olivier; Breisinger, Clemens; McCool, Christen; Diao, Xinshen; Funes, Jose; You, Liangzhi; and Yu, Bingxin. 2010. Assessing Food Security in Yemen: An Innovative Integrated, Cross-Sector, and Multilevel Approach. Discussion Paper 982. IFPRI. https://www.ifpri.org/cdmref/p15738coll2/id/2539/filename/2540.pdf 33 ILO (International Labour Organization). 2022. ILOSTAT database. https://ilostat.ilo.org/data/ IMF (International Monetary Fund). 2022. IMF database for Yemen. Unpublished. IOM (International Organization for Migration). 2022. Migration Data Portal. Global Migration Data Analysis Centre. https://www.migrationdataportal.org Lofgren, Hans; Harris, Rebecca L.; and Robinson, Sherman. 2002. A standard computable general equilibrium (CGE) model in GAMS. Microcomputers in Policy Research 5. Washington, D.C.: International Food Policy Research Institute. Lofgren, Hans; and Cicowiez, Martín. 2021. “SDGSIM – Model Documentation.” Unpublished. Lofgren, Hans; Cicowiez, Martín; and Diaz-Bonilla, Carolina. 2013. MAMS – A computable general equilibrium model for developing country strategy analysis. In Dixon, P.B. and Jorgenson, D.W. (eds.) Handbook of Computable General Equilibrium Modeling. Volume 1, Chapter 4, pp. 159–276. Amsterdam, Elsevier. Muhammad, A., Seale, J. L., Meade, B., and Regmi, A. (2011). International evidence on food consumption patterns: an update using 2005 international comparison program data. USDA-ERS technical bulletin, (1929). Mukashov, Askar; Breisinger, Clemens; Engelke, Wilfried; and Wiebelt, Manfred. 2022. Modeling conflict impact and postconflict reconstruction: The case of Yemen. Economic Systems, Volume 46, Issue 1. https://doi.org/10.1016/j.ecosys.2022.100940 Robinson, Sherman. 1989. Multisectoral models. In Chenery, Hollis, and Srinivasan, T.N. (eds.) Handbook of Development Economics. Volume 2, Chapter 18, pp. 885-947. Amsterdam, Elsevier. Sadoulet, E. and de Janvry, A. 1995. Quantitative Development Policy Analysis. Baltimore, USA, The Johns Hopkins University Press. Thurlow, James. 2004. A dynamic computable general equilibrium (CGE) model for South Africa: Extending the static IFPRI model. TIPS Working Paper 1-2004. Pretoria, South Africa: Trade and Industrial Policy Strategies. UN (United Nations). 2020. International Migrant Stock 2020. Population Division, Department of Economic and Social Affairs.https://www.un.org/development/desa/pd/content/international-migrant-stock Wiebelt, Manfred; Breisinger, Clemens; Ecker, Olivier; Al-Riffai, Perrihan; Robertson, Richard; and Thiele, Rainer. 2011. Climate Change and Floods in Yemen: Impacts on Food Security and Options for Adaptation. IFPRI Discussion Paper 1139. World Bank. 2018. The State of Social Safety Nets 2018. Washington, DC: World Bank. https://openknowledge.worldbank.org/handle/10986/29115. World Bank. 2020. The Human Capital Index – 2020 Update : Human Capital in the Time of COVID-19. World Bank, Washington, DC. https://openknowledge.worldbank.org/handle/10986/34432 World Bank. 2021. Investing in Human Capital for a Resilient Recovery: The Role of Public Finance. World Bank, Washington, DC. World Bank. 2021. World Development Indicators. Washington, D.C.: The World Bank. http://data.worldbank.org/data-catalog/world-development-indicators 34 Appendix: Additional details on the model database The SDGSIM database for Yemen developed for this analysis has 2020 as its base year. This appendix provides additional information about the database, showing selected data and outlining the key steps in its development, most importantly the work to build the SAM. The other key data are elasticities (in trade, production, and consumption), factor stocks, and projections for GDP and other indicators. A.1. Social Accounting Matrix The procedure followed when building the SAM was top-down and entailed two steps: (i) construction of an aggregate SAM (hereafter referred to as macro SAM); and (ii) disaggregation of the macro SAM into a matrix with a relatively fine sectoral breakdown (in this paper called meso SAM or, for simplicity, SAM). Under step (ii), we first built a SAM that, in addition to the macro SAM, was based on the Yemeni national accounts data as published by the United Nations Statistics Division (UNSD) and the Central Statistical Organization of Yemen, IMF and COMTRADE (data on sectoral exports and imports), ILO (data on sectoral employment and earnings), and a 2014 SAM documented in Raouf et al. (2019). After this, we adapted this SAM to make it suitable for use as data input to SDGSIM (or any other CGE) model. It should be noted that the macro SAM was based on 2020 data while some of the other data was for recent preceding years. In the following, we describe how the information from different sources and for different years was reconciled. A.1.1. Macro SAM In the first step of building the SAM, a schematic representation of the Yemeni economy in 2020 was generated using, almost exclusively, data in IMF’s Table on Selected Economic Indicators (SEI) including macroeconomic aggregates from the national accounts (GDP, exports, and imports), the balance of payments, and data on government receipts and spending. In addition, we used the 2014 IFPRI SAM to split total value added between labor and capital payments, and national accounts data from the UNSD for Gross Fixed Capital Formation (GFCF). The estimated macro SAM is presented in Table 2.2a, where we use the abbreviations shown in Table 2.2b. The procedure we followed for building the 2020 Yemen macro SAM consisted of two main steps. The first step – direct entry of data in raw form and computation of intermediate data – is covered in Table A.1. In the first section of the table (1-17; the parenthesized numbers refer to rows in relevant tables), data from IMF (SEI, fiscal, and balance of payments) was entered into the SAM directly without transformations – this should be done to the maximum extent 35 possible.14 It was here possible to define GDP at purchaser’s prices (GDP, exports, and imports), indirect as well direct total tax payments, and items in the balance of payments related to factor incomes and current transfers. Consistency between receiving and paying accounts was assured thanks to the fact that each SAM entry applies to both; e.g., government consumption in the national accounts (part of commodity demands) and in the government budget (part of current government spending) is a single cell entry. (In practice, such consistency is not automatic given that data for the national accounts, the government budget, and the balance of payments are produced separately.) As shown in the second section of Table A.1 (18-23), additional data needed for calculations were also assembled in this step. For instance, labor and capital payments were computed on the basis of data from the 2014 SAM and the SEI. In the second step, shown in Table A.2, the remaining cells in the macro SAM were defined. To make the calculations shorter and more transparent, initially data for a set of intermediate variables such as household income, total government revenue, (current) outflows of foreign exchange, and government deficit were specified drawing on data covered in Table A.1 (1-4). After this, the logic behind the different calculations was as follows. Total intermediate consumption was calculated by multiplying GDP at basic prices by the ratio of total intermediate consumption to GDP at basic prices obtained from the 2014 SAM (5). The split of GDP at basic prices between labor and capital payments was conducted using the labor and capital income shares from the 2014 SAM (6-7). Household labor incomes were defined as total labor value-added plus net labor incomes from the rest of the world minus labor income to RoW (9). Household capital income was similarly computed as total capital value-added plus capital income from the rest of the world minus capital income to government and RoW (10). Household consumption was computed as a residual using the basic macroeconomic identity (11). Household savings were also computed as a residual, defined as the difference between total household income and expenditure (12). Similarly, to balance the government account (13), transfers to households from the government were defined as the difference between the total of the row of the current government account (total current revenue) and the total of the column, excluding these transfers but including government savings. Like government transfers to households, RoW savings (i.e., the negative of the current account balance) were defined to balance the current account of the rest of the world; specifically, as the difference between the row (receipt) total and the sum of the other items (expenses) in the column, excluding rest-of-world savings (14). Private GFCF was calculated as total GFCF from the UNSD national accounts minus the government GFCF from the IMF fiscal data (15). Domestic non-government (household) 14 A cell payment is referred to via a bracketed listing of its row and column accounts. For example, [f-lab,act] refers to the payment in the intersection between the row of the account f-lab and the column of the account act; i.e., a payment to the labor account from the activity account. 36 financing of private investment) (16) was defined as the difference between (total) private GFCF and FDI. Government domestic financing was obtained as the product of (a) the domestic share in financing of the government from IMF fiscal data, and (b) the government deficit (see cell [cap- gov,cap-ngov]) (17). Remaining financing to the government had to come from the rest of the world (18). Similarly, remaining financing of the household (the difference between total spending on its capital -- i.e., the sum of its payments for private investment, stock change and to finance the government, and its savings – also had to come from abroad (19). Naturally, as a manifestation of Walras’ Law, the capital account of the rest of the world was also in equilibrium once foreign financing to the households had been defined. Rest-of-world savings are viewed from the perspective of the rest of the world; from the perspective of the SAM country, it is the deficit of the current account of the balance of payments. The procedure followed generated a balanced macro SAM similar in structure to the SAM shown in Table 2.2 in the main text. During the process, it is important to verify that the values for the items that are defined as residuals are reasonable and do not deviate strongly from other extraneous data (if available); if this is not the case, then one or more of the other cells must also be off. For example, the value for RoW savings should be close to its values according to other sources. In our case, all residual cells are close to their values reported in alternative sources. 37 Table A.1. Macro SAM Yemen 2020: Data inputs # SAM cell Definition Source Data entered in macro SAM without transformation 1 [f-hydrocarbon,act] Natural resource rent in oil and gas Fiscal 2 [row,com] Imports of goods and services SEI 3 [tax-com,com] Commodity tax Fiscal 4 [tax-imp,com] Customs and other import duties Fiscal 5 [row,f-capprv] Capital income transferred to RoW BoP 6 [gov,f-hydrocarbon] Government income from oil and gas rent Fiscal 7 [tax-dir,hhd] Taxes on income and other direct Fiscal 8 [gov,hhd] Non-tax revenue Fiscal 9 [com,gov] Government consumption expenditure Fiscal 10 [cap-gov,gov] Government savings Fiscal 11 [com,row] Exports of goods and services SEI 12 [hhd,row] Current transfers from RoW to households BoP 13 [gov,row] Current transfers from RoW to government Fiscal 14 [donor,row] Current transfers from RoW to donor BoP 15 [com,donor] Donor consumption expenditure BoP 16 [drf,cap-ngov] Change in foreign reserves BoP 17 [com,inv-gov] Government gross fixed capital formation Fiscal Data inputs to calculations for macro SAM (intermediate data) 18 gfcftot Gross fixed capital formation SEI and NA-UNSD 19 shrlabva Labor share in value added 2014 SAM 20 shrcapva Capital share in value added 2014 SAM 21 intvarat Ratio between intermediate cons and VA 2014 SAM 22 shrgfindom Domestic share in financing of the government Fiscal 23 shrgfinfor Foreign share in financing of the government Fiscal 24 gdppp GDP at purchaser prices SEI 25 gdpbp GDP at basic prices SEI and Fiscal ## shrgfcfprv Private share in GFCF NA ## shrgfcfgov Government share in GFCF NA 14 [row,hhd] Current transfers from households to RoW BoP 15 [com,npish] NPISHs cons expenditure SUT 16 [row,gov] Current transfers from government to RoW BoP 17 [f-lab,row] Labor income received from rest of world BoP 18 [f-cap+mi,row] Capital income received from rest of world BoP 19 20 [inv-prv,cap-row] Foreign direct investment BoP 21 [com,dstk] Changes in stocks SUT Source: Authors’ elaboration. 38 Table A.2. Macro SAM Yemen 2020: Calculations # Item or SAM cell Definition Formula Computed intermediate variables 1 yh Household income = [hhd,f-lab] + [hhd,f-cap+mi] + [hhd,gov] + [hhd,row] 2 yg Government income = [gov,f-capprv] + [gov,f- hydrocarbon] + [gov,hhd] + [gov,row] + [gov,tax-com] + [gov,tax-imp] + [gov,tax-dir] 3 yrow RoW income = [row,com] + [row,f-lab] + [row,f- capprv] + [row,hhd] + [row,gov] 4 gdef Government deficit = [inv-gov,cap-gov] - [cap-gov,gov] Additional cells in macro SAM 5 [com,act] Intermediate consumption gdpbp * intvarat 6 [f-lab,act] Compensation of employees gdpbp * shrlabva 7 [f-capprv,act] Gross operating surplus gdpbp * shrcapva - [f- hydrocarbon,act] 8 [act,com] Sales of domestic output gdpbp 9 [hhd,f-lab] Household labor income = [f-lab,act] + [f‑lab,row] – [row,f- lab] 10 [hhd,f-capprv] Household capital income = [f-capprv,act] + [f-capprv,row] – [gov,f-capprv] – [row,f-capprv] 11 [com,hhd] Private households cons expenditure gdpp - [com,gov] - [com,row] - [com,donor] - [com,inv-prv] - [com,inv-gov] + [row,com] 12 [cap-ngov,hhd] Household savings = yh - [com,hhd] - [gov,hhd] - [row,hhd] - [tax-dir,hhd] 13 [hhd,gov] Transfers from government to households = yg - [com,gov] - [row,gov] - [cap- gov,gov] 14 [cap-row,row] Foreign savings (current account deficit) = yrow - [com,row] - [f-lab,row] - [f-capprv,row] - [hhd,row] - [gov,row] - [donor,row] 15 [com,inv-prv] Private gross fixed capital formation = gfcftot - [com,inv-gov] 16 [inv-prv,cap-ngov] Non-government financing of private invest = gfcftot - [com,inv-gov] - [inv- prv,cap-row] 17 [cap-gov,cap-ngov] Domestic financing of the government = shrgfindom * gdef 18 [cap-gov,cap-row] Foreign financing of the government = shrgfinfor * gdef 19 [cap-ngov,cap-row] Foreign financing of the households = [cap-gov,cap-ngov] + [inv- prv,cap-hhd] + [dstk,cap-hhd] - [cap-ngov,hhd] Source: Authors’ elaboration. 39 A.1.2. Meso SAM At this stage, the aim was to build a SAM with more disaggregated activities and commodities that would be fully consistent with the macro SAM (i.e., that, if it were properly aggregated, it would replicate all cell values in the macro SAM). To that end, we made use of the sectoral GDP and international trade data available from UNSD and COMTRADE, respectively. Not surprisingly, given the use of several data sources, the resulting meso SAM was unbalanced. Therefore, a cross-entropy balancing algorithm was used to balance the SAM. In the balancing process, we imposed that the meso SAM maintain its consistency with the macro SAM. In this step, we also split labor income by gender, using estimates on (a) total labor payments by activity, (b) male and female employment by activity, and (c) the gender wage gap. Table A.3 shows the estimated number of male and female workers at the one digit ISIC rev.4 classification for 2019 obtained from ILOSTAT. In turn, the gender wage gap was estimated by UNDP Yemen at 77 percent (i.e., women earn 77 cents for every dollar that men get for the same work). Specifically, we used the following formulae: ⋅ ⋅ + = , = ⋅ where : labor : set for activities in ILO estimates , : labor payments in SAM obtained from SUT : male workers : female workers : male wage : female wage : gender wage gap Then, substituting the second equation into the first equation, ⋅ + ⋅ ⋅ = , and solving for : 40 , = + ⋅ Finally, once we know and , it is possible to split , into payments to male labor ( ⋅ ) and female labor ( ⋅ ). Table A.3. Number of male and female workers (‘000) Activity Male Female Total Male% Female% Total 5,517.4 399.9 5,917.2 93.2 6.8 Agriculture; forestry and fishing (A) 1,462.6 169.1 1,631.7 89.6 10.4 Mining and quarrying (B) 8.8 0.1 8.8 99.4 0.6 Manufacturing (c) 226.3 51.4 277.7 81.5 18.5 Utilities (D; E) 8.3 0.1 8.5 98.2 1.8 Construction (F) 308.2 0.7 308.9 99.8 0.2 Wholesale and retail trade; repair of motor 1,529.6 24.8 1,554.4 98.4 1.6 vehicles and motorcycles (G) Transport; storage and communication (H; J) 529.4 2.6 532.0 99.5 0.5 Accommodation and food service activities (I) 113.0 1.2 114.2 98.9 1.1 Financial and insurance activities (K) 18.7 0.2 18.9 98.8 1.2 Real estate; business and administrative 27.1 0.8 27.8 97.2 2.8 activities (L; M; N) Public administration and defence; 868.5 25.6 894.1 97.1 2.9 compulsory social security (O) Education (P) 244.6 93.8 338.4 72.3 27.7 Human health and social work activities (Q) 75.2 19.7 94.8 79.3 20.7 Other services (R; S; T; U) 97.2 9.8 107.0 90.8 9.2 Source: Authors’ elaboration based on ILOSTAT. A.1.3. Additional adjustments The (meso) SAM resulting from the previous step was further adapted to give it the structure needed for use as data input to SDGSIM. Specifically, the following two changes were introduced: (i) Capital rents were split into payments to physical (private) capital and natural resources; and (ii) Trade and transport margins were disaggregated into separate margins for the three types of trade (i.e., imports, exports, and domestic sales of domestic products. For (i), we used GTAP data (Aguiar et al. 2019). For (ii), we assumed that the ratio between the margin value and the value of the traded good is the same for the three trade types. The resulting SAM has 17 activities and commodities, 7 factors, 4 institutions, 3 capital accounts, 2 investment accounts (one for each type of capital), 3 accounts for distribution (i.e., trade and transportation) margins, as well as a set of tax accounts that pass on tax payments to 41 the government. Table 2.1 in the main text shows the accounts in the 2020 SAM for Yemen. In the same section, the 2020 SAM data are used to describe Yemen’s economy. A.2. Non-SAM data The elasticities used are displayed in Table A.4. They were defined on the basis of the literature and author assessments, drawing on a combination of econometric evidence and experience from similar country applications; for a survey, see Annabi et al. (2006). The value-added elasticities of substitution are in the range of 0.20-0.95 (Aguiar et al. 2019). For agriculture and mining, the value-added elasticities are at low levels to let production and export growth be driven by exogenous assumptions regarding use of the natural resource factors (land and extractive resources, respectively). For trade, the Armington and CET elasticities are both in the range of 0.9-2.0 (Sadoulet and de Janvry 1995). For household consumption, expenditure elasticities are based on Muhammad et al. (2011). These were then recalibrated to reflect the consumption structure in the Yemen SAM. In the context of the linear expenditure system (LES) demand functions, which are used in SDGSIM, estimates are also needed for the so-called Frisch parameter (technically the elasticity of the marginal utility of income with respect to income) for each household group. Using a relationship in Lluch et al. (1977), which expresses an inverse relationship between the Frisch parameter and real per-capita consumption, this parameter was set at -7.0 for the single representative household in the SAM. Population data are based on UN (2022). Among the factors, base-year stocks are needed for private capital and labor. The private capital stock was estimated based on capital rents in the SAM, typical rates for depreciation (4.5 percent for private capital and 3.5 percent for infrastructure and other government capital), and a relatively modest rate of net profits for private capital (10 percent). For government capital, estimates are not needed for base-year stocks as simulation results only depend on deviations from base scenario levels for the stocks that influence TFP. Apart from simulated investment levels, these deviations only depend on the depreciation rates used for these capital types. The fact that the model is solved over time generates additional data needs. As noted in the main body of the text, the base scenario is calibrated to replicate a path of growth in GDP at factor cost that is based on data and projections from IMF (2022) for the years 2021 and 2022. Besides, projections are also needed for growth in labor and mining natural resource extraction. For the labor factor, these projections are based on projected growth in the population aged 15-64, 2020 data for the extended labor force and the population aged 15-64, assuming an unchanged ratio up to 2030 between the extended labor force and the population aged 15- 42 64.15 Among these data inputs, population 15-64 is from UN (2019) while unemployment, employment and potential labor force is from ILO (2022). For the mining natural resource factor, stock extraction is assumed virtually stagnant. The post calculations that generate poverty results by simulation are based on the 2021 headcount poverty rate estimated by the World Bank, the assumption of a log-normal consumption distribution, and the 2014 Gini coefficient. Household consumption per capita is used as the welfare measure. We considered the 2021 figure as valid also for 2022. Table A.4. Value-added, trade, and consumption elasticities VA CET Armington LES Qat 0.25 2.00 2.00 0.83 Other agriculture 0.25 2.00 2.00 0.58 Fishing 0.20 2.00 2.00 0.58 Oil and gas 0.20 2.00 2.00 0.90 Other mining 0.20 2.00 2.00 1.20 Manufacturing 0.95 1.50 1.50 1.20 Utilities (electricity, gas, and water) 0.95 0.00 0.00 0.80 Construction 0.95 0.00 0.90 0.80 Trade 0.95 0.00 0.00 1.20 Hotels and restaurants 0.95 0.00 0.90 1.20 Transport 0.95 0.90 0.90 0.90 Telecommunications 0.95 0.00 0.00 0.90 Financial services 0.95 0.00 0.90 1.20 Real estate and business services 0.95 0.90 0.90 1.20 Public administration 0.95 0.90 0.90 1.20 Government education and health 0.95 0.00 0.00 0.00 Other services 0.95 0.00 0.00 1.03 Note: VA = CES value-added function Armington = CES aggregation function for domestic demand (elasticities of substitution between imports and domestic output); CET = Constant Elasticity of Transformation function for domestic output (elasticities of transformation between exports and domestic supply) LES = Linear Expenditure system (elasticities of household consumption with respect to total consumption spending) for the household Sources: Annabi et al. (2006), Muhammad et al. (2010), and authors' assessments. 15 Drawing on ILO data and definitions (ILO 2022b), the extended unemployment rate (in ILO documents referred to as “the combined rate of unemployment and potential labor force (LU3)” is defined as 100(unemployed + potential labor force) / (labor force + potential labor force), where the denominator according to ILO terminology is referred to as the extended labor force. The potential labor force is defined as “persons not in employment who express an interest in this form of work but for whom existing conditions limit their active job search and/or their availability”. 43 44