Policy Research Working Paper 8967 Growth after War in Syria Sharmila Devadas Ibrahim Elbadawi Norman V. Loayza Development Economics Development Research Group August 2019 Policy Research Working Paper 8967 Abstract This paper addresses three questions: 1) what would have simulation scenarios, the growth drivers are affected by the been the growth and income trajectory of Syria in the assumed levels of reconstruction assistance, repatriation of absence of war; 2) given the war, what explains the reduction refugees, and productivity improvements associated with in economic growth in terms physical capital, labor force, three plausible political settlement outcomes: a baseline human capital, and productivity; and 3) what potential (Sochi-plus) moderate scenario, an optimistic (robust polit- growth scenarios for Syria there could be in the aftermath ical settlement) scenario, and a pessimistic (de facto balance of war. Estimates of the impact of conflict point to nega- of power) scenario. Respectively for these scenarios, GDP tive gross domestic product (GDP) growth of −12 percent per capita average growth in the next two decades is pro- on average over 2011–18, resulting in a GDP contrac- jected to be 6.1, 8.2, or 3.1 percent, assuming that a final tion to about one-third of the 2010 level. In post-conflict and stable resolution of the conflict is achieved. This paper is a product of the Development Research Group, Development Economics. 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 sdevadas1@worldbank.org and nloayza@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 Growth after War in Syria   Sharmila Devadas    Ibrahim Elbadawi  Norman V. Loayza  Development Research  Economic Research Forum  Development Research  Group, World Bank  Group, World Bank  Keywords: War, conflict, reconstruction, growth, factors of production, Syria  JEL Codes: D74, F51, 011, 040, 053   Note: The  authors are  grateful to staff of the United Nations  Economic and Social Commission  for  Western  Asia  including  Ahmad  Shikh  Ebid  for  sharing  information  and  data,  as  well  as  to  Shanta  Devarajan,  Simeon  Djankov,  Young  Eun  Kim,  Aart  Kraay,  Aljaz  Kuncic,  Rabie  Nasser,  and  Nurlina  Shaharuddin for helpful comments.  All remaining errors are the responsibility of the authors. The  views expressed here  are the authors’, and  do not necessarily reflect those  of the World Bank, its  Executive Directors, or the countries they represent.   1. Introduction   The  scale  and  intensity  of  the  violence  and  destruction  associated  with  the  civil  war  that  engulfed  the  Syrian  Arab  Republic  since  2011  have  very  few  parallels  in  recent  history.  The  Syrian  Observatory for Human Rights (SOHR) estimates the total death toll (from 15 March 2011 to 15 March  2019) at a staggering number of 570,000.  The United Nations Economic and Social Commission for  Western Asia (ESCWA 2018) ‐ which conducted an elaborate sectoral analysis of the economic cost of  the Syrian civil war ‐ puts the destruction of the physical capital stock by end of 2017 at almost USD  120 billion.  And, in terms of the cost to the overall economy, the World Bank (2017b) estimates that,  from 2011 until the end of 2016, the cumulative losses in gross domestic product (GDP) reached a  whopping  USD  226  billion,  about  four  times  the  Syrian  GDP  in  2010.    These  assessments  seem  to  broadly cohere with the calculations of the country’s night‐light intensity by Ceylan and Tumen (2018)  and Li et al. (2017), with the latter suggesting that by 2017, Syria had lost about 80 percent of its city  night light.   Moreover,  in  addition  to  the  massive  death  and  destruction,  this  war  has  also  created  an  unprecedented number of refugees and internally displaced persons (IDPs).  According to the United  Nations  High Commissioner for Refugees (UNHCR), there are about 5.6 million registered refugees  from Syria in neighboring countries.  However, accounting for unregistered refugees in just the three  countries of  Egypt, Jordan and Lebanon would raise the aggregate number to more than 7 million  (UNHCR, 2019).  Adding these numbers to the about 6.3 million IDPs in Syria, we have almost two‐ thirds of the 21 million Syrian citizens who have been forced out of their homes.  To appreciate the  global impact of the Syrian refugees and IDPs crisis, it suffices to note that the former accounts for  more than 23 percent of the total number of refugees worldwide, while the percentage of the latter  is estimated at 20 percent of the total number of IDPs in the world.    The losses incurred by Syria are great, but it is not false hope to look toward recovery and  further  strengthening  of  the  country’s  socio‐economic  fundamentals  beyond  its  pre‐war  situation.  Chen, Loayza and Reynal‐Querol (2008) conduct a comprehensive evaluation of the aftermath of civil  war using event‐study analysis across 41 countries over 1960‐2003. They show that recovery to pre‐ conflict levels and further improvements are possible for a country afflicted by war when relatively  lasting peace is achieved. Other studies focusing on World War II (WWII) indicate countries returned  to their pre‐war trends 15 to 20 years post war (Organski and Kugler 1977, 1980); and that countries  suffering large negative output shocks grew systemically faster during the subsequent decades due to  reconstruction dynamics (Milionis and Vonyo 2015). Because of the massive destruction of the factors  of production in Syria at a scale more common in inter‐state wars than civil conflicts, the lessons from  the post‐WWII reconstruction of Europe and insight from modern growth theory could be useful in  assessing  the  post‐conflict  growth  potential  for  Syria.  Janossy  (1969)  postulates  that  fast  growth  during reconstruction is not only the result of higher returns to physical capital accumulation (which  diminish  as  capital  grows  in  relation  to  output)  but  also  depends  on  structural  factors  like  the  reorganization of economic activity and the reallocation of production factors. One of the key lessons  from the experience of post‐WWII growth in the European countries and Japan, for example, was that  the rapid growth impact of the massive re‐building of physical capital was made possible, not only by  the Marshall Plan resources, but also by the relatively limited war time depreciation of the human  capital base and technological potential (Smolny 2000).   The implication of the above for the post‐conflict economic reconstruction agenda for Syria ‐  with almost two‐thirds of its population internally displaced or living as refugees in foreign countries  ‐ is that the restoration of human capital should be accorded the highest priority. And this should be  2    alongside the rebuilding of physical capital which will unavoidably be a key component of the agenda.  Further, attention also needs to be paid to other factors contributing to total factor productivity (TFP),  including institutions and market efficiency.      However, the prospects for mobilizing meaningful multi‐year financing for reconstruction and  development and for achieving a critical mass of voluntary repatriation of refugees would hinge on  the nature of the ultimate political settlement of the conflict.  A lopsided political settlement may  deter refugees, with strong lingering uncertainty about security and economic prospects, to return.  Some of the main impediments hindering repatriation include the dispossession of refugees’ homes  and  mandatory  military  conscription  for  men  of  age.  Therefore,  and  despite  the  “invitation”  for  refugees to return home and the refugee camps being set up within Syria, it is not surprising that only  a  few  thousands  returned  in  2017,  mostly  motivated  by  push  factors  in  the  recipient  countries.   Indeed, this very limited response did not mark the opening of the flood gates for massive repatriation  in the following years (POMEPS 2018).    Moreover, the volume of the funding required for reconstruction has been estimated at USD  250 billion by the United Nations to as high as USD1 trillion (POMEPS 2018), by far more than could  be provided by Syrian allies.  Thus, a genuine reconstruction plan for Syria would best be served by  robust support from the wider international community, who have indicated a preference for a more  robust political settlement (ERF 2019).  The international community can provide some reconstruction  aid that would support and encourage the return of refugees, infrastructure investment, and policy  reform.    This  includes  aid  for  geographically  dispersed  economic  reconstruction  (rebuilding  infrastructure,  access  to  health  and  education,  etc.)  and  institutional  reform  (including  security,  property rights, and access to justice) that benefit various segments of the population fairly (Yahya  2017).  Djankov and Reynal‐Querol (2010) find that both per capita income and civil war are jointly  determined  by  country‐specific  phenomena,  such  as  colonial  history.  Consequently,  policies  are  needed to rectify structural problems that make countries more prone to conflict.   Subscribing  to  the  context  discussed  above,  this  paper  uses  the  World  Bank  Long‐Term  Growth Model ‐ Public Capital Extension (LTGM‐PC) by Devadas and Pennings (2019) ‐ see Appendix  1 for details ‐ to simulate a counterfactual of no conflict scenario (in Section 2), to estimate the impact  of conflict (in Section 3), and to assess the potential post‐conflict growth for Syria (in Section 4). The  after‐war projections are carried out for three political settlement scenarios: a baseline (Sochi‐plus,  mainly  operated  by  Iran,  Russia,  and  Turkey,  with  some  involvement  from  the  United  Nations)  moderate scenario; a high (robust political settlement, brokered by the United Nations) scenario; and  a pessimistic (de facto balance of power) scenario. The LTGM‐PC builds on the Solow‐Swan growth  model, and is evolved from another World Bank tool, the Long‐Term Growth Model (LTGM) (Loayza  and Pennings 2018; Hevia and Loayza 2012). The LTGM‐PC splits the accumulation of aggregate capital  into public and private portions, while retaining other features of the LTGM including other growth  drivers (demographics and the labor force, human capital, and total factor productivity). We run all  our simulations using the LTGM‐PC Excel‐based toolkit.   Our paper complements earlier modeling work by World Bank (2017b) in four ways. One, it  provides a straightforward and transparent analysis of how GDP evolves based on projections for the  growth drivers. World Bank (2017b) uses a dynamic general equilibrium model to simulate the effects  of  the  conflict  through  three  channels  –  physical  capital  destruction,  casualties,  and  economic  disorganization; with the last calculated as a residual based on estimated GDP losses. Two, data‐wise,  we  use  estimates  of  physical  damage  across  all  types  of  capital  whereas  World  Bank  (2017b)  determines  destruction  in  their  simulations  based  on  physical  damage  assessments  in  the  housing  3    sector.  Three,  with  a  greater  certainty  of  the  end  of  conflict,  we  focus  on  growth  scenarios  in  the  aftermath of war, rather than mostly assessing conflict impact based on different end‐time scenarios.  Four, we also attempt to provide a more up‐to‐date assessment of the impact from the conflict, that  is until the end of 2018.   Under the counterfactual simulation (that is, if war had not occurred), our baseline projection  shows average real GDP growth of 5.3 percent per annum over 2011‐2018 which would have led to  real GDP rising from USD 60 billion in 2010 to USD 91 billion and real GDP per capita rising from USD  2,857 to USD 3,774 by 2018. In contrast, our simulations of the impact of conflict point to negative  annual GDP growth of ‐12 percent on average over 2011‐2018, resulting in a GDP level of USD 23.2  billion in 2018. Comparing the conflict versus no conflict simulations suggest a cumulative loss in GDP  potential of about USD 300 billion over 2011‐2018. The depletion of factors of production alone may  account for about 87 percent of the negative GDP growth on average, and further, about 64 percent  of  the  average  negative  growth  is  due  to  physical  capital  destruction.  Physical  capital  destruction  reflects the compounded effects of large outright damages, low new investments, and a falling output  base, that is adversely affected by all growth drivers. Demographics and labor account for about 15  percent, human capital 7 percent, and total factor productivity 13 percent of negative GDP growth on  average over the conflict years.   In our post‐conflict simulations, we assume that the three political settlement scenarios are  associated  with  different  levels  of  reconstruction  assistance  and  different  degrees  of  voluntary  mobility of refugees. These in turn affect key drivers of growth: public and private investment, and  the labor force. We also make different assumptions for human capital growth and TFP growth across  the three scenarios. Depending on the scenarios, our simulation results suggest that  it would take  between 10 and 20 years for Syria to reach its pre‐conflict GDP level and between 10 and 30 years to  reach its GDP per capita level (both at 2010 constant prices). If there were to be an unsanctioned and  misguided “forced” repatriation of refugees, this would result in significantly lower GDP per capita  compared  to  the  voluntary  mobility  case.  Under  voluntary  return,  labor  would  adjust  gradually  to  capital reconstruction, thus keeping labor productivity from falling.     2. Syria Pre‐Conflict Developments and Projections  This section focuses on building a calibration using the LTGM‐PC for a no‐conflict scenario based  on developments  and projections prior  to the conflict. Recent data indicate that  average real GDP  growth  in  Syria  was  around  4.7%  over  2001‐2010  (Graph  1(a)).  World  Bank’s  World  Development  Indicators (WDI)1 data suggest an average growth of 4.9% over 2001‐2010, while IMF World Economic  Outlook (WEO), points to 4.5%, over the same period.2  Just prior to the conflict, projections for near‐ term growth were still robust: an average of 5.6% over 2011‐2012 in the World Bank Global Economic  Prospects (GEP), January 20113; and an average of 5.0% over 2011‐2016 in the IMF WEO, April 2011  edition4  ‐  see Graph 1(b). In the following subsection we look at pre‐conflict trends and projections                                                               1  WDI data for Syria are publicly available only up to 2007. We supplement these data for the last three years up to 2010  with data on GDP growth from the World Bank’s internal macroeconomic model (MFMod), November 2017 vintage, which  is consistent with the IMF WEO October 2018 data.    2 WDI: GDP is at market prices, 2010 USD. IMF WEO: GDP is at market prices, 2000 local currency.   3 Source: http://documents.worldbank.org/curated/en/293851468162862428/Global‐economic‐prospects‐January‐2011‐ navigating‐strong‐currents. GDP growth of 5.0% for 2010 is an estimate. No GEP projections for Syria are available beyond  the January 2011 publication. GDP is at market prices, 2005 USD.   4 Source: https://www.imf.org/external/pubs/ft/weo/2011/01/weodata/index.aspx.   4    (where  available)  for  key  growth  drivers  (physical  investment,  demographics  and  the  labor  force,  human capital and total factor productivity).   Graph 1: Pre‐Conflict Real GDP Growth and Projection  (a) September/October 2018 Vintage (b) April 2011 Vintage %  %  8.0 8.0 6.0 4.9 6.0 4.0 4.0 5.0 4.5 2.0 2.0 0.0 0.0 ‐2.0 ‐2.0 ‐4.0 ‐4.0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 IMF WEO, Oct 2018 WDI/MFMod 2018 IMF WEO April 2011, 6‐year average, 2011‐2016 WDI/MFMod 2018, 10‐year average, 2001‐2010 IMF WEO, April 2011 (dashed line indicates estimate/projection) IMF WEO Oct 2018, 10‐year average, 2001‐2010   2.1 Growth drivers  In  terms  of  physical  investment,  the  investment‐to‐output  ratio,  ⁄   (gross  fixed  capital  formation (GFCF), as a percentage of GDP) averaged about 21.5% over the 7‐year period, 2001‐2007  (Graph 2(a), blue line with marker).5 For gross capital formation (GCF), which includes inventories, as  a percentage of GDP, the IMF WEO October 2018 showed an average of 23.1% over 2001‐2010 (Graph  2(a) red line). Pre‐conflict data (IMF WEO April 2011) meanwhile suggested a lower average of 21.8%  for GCF (% of GDP), albeit with a projected rise of about 3 percentage points over 2011‐2016 (Graph  2(b)). Of total  ⁄ , how much exactly  can be attributed to public investment ( ⁄  is uncertain ‐  40% seems plausible based on reported figures.6  Graph 2: Pre‐Conflict Investment and Projection  (a) September/October 2018 Vintage (b) April 2011 Vintage % of GDP % of GDP 35.0 30.0 30.0 25.0 25.0 25.2 20.0 20.0 23.1 21.5 15.0 15.0 10.0 10.0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Gross fixed capital formation (GFCF) (WDI, 2018) GFCF 7‐year average, 2001‐2007 (WDI, 2018) GCF (IMF WEO, April 2011), dashed line indicates estimate/projection Gross capital formation (GCF) (IMF WEO, Oct 2018) GCF 6‐year average, 2011‐2016 (IMF WEO, April 2011) GCF 10‐year average, 2001‐2010 (IMF WEO, Oct 2018)                                                                5  MFMod numbers suggest a slightly lower average of 20.5% over the same period (and 20.1% over 2001‐2010).   6 World Bank (2017b) reports  ⁄ of 9% and private fixed capital formation‐to‐output,  ⁄ of 12% in 2010 giving a public  investment share of total investment of 43%. Data in IMF (2010) indicate an average share of 38% over 2005‐2008, based on  gross capital formation‐to‐output averaging 21.8%. Calculations using the IMF FAD Investment and Capital Stock Database  2017  suggest  an  average  share  of  47%  over  2001‐2010,  but  the  average  value  for  total  investment‐to‐output  at  6%  is  relatively low compared to the other reported values.     5    Demographics and the labor force. The growth rates of total population and the working‐age  population share averaged 2.5% and 0.8% respectively over 2001‐2010 based on latest United Nations  (UN)  estimates  (United  Nations  2017)  ‐  Graph  3(a). As  a  gauge  of  projections  prior  to  the conflict,  United  Nations  (2011)  indicated  average  growth  rates  for  these  two  variables  of  1.7%  and  1.0%  respectively over 2011‐2020 and 1.5% and 0.3% over 2021‐2030 compared to 2.5% and 0.5% over  2001‐2010 (Graph 3(b), dashed lines with markers).7   The labor force participation rate (LFPR)8 meanwhile, had been on a moderating trend, declining  from 54.5% in 1995 (Male: 84.3%; Female: 24.4%) to 44.9% in 2010 (Male: 75.6%; Female: 14.0%),  with  an  average  growth  over  2001‐2010  of  ‐1.5%  (Male:  ‐0.9%;  Female:  ‐3.9%)  (Graph  4(a)).  This  phenomenon  occurred  despite  relatively  strong  economic  growth,  distinguishing  Syria  from  other  countries – no other Middle East and North Africa (MENA) economy had a similar rate of decline in  the LFPR over the same period, except Yemen. Nasser and Mehchy (2012) note that a sizeable portion  of  the  economically  active  population  that  went  out  of  the  labor  force  in  the  2000s  consisted  of  women  in  rural  areas,  and  workers  becoming  students.  Women  in  the  agriculture  sector  became  jobless with the drought and higher fuel prices in the second half of the 2000s, and were unable to  find other  work, thus returning  to  domestic  services  in their  homes.  In  the case of  the  increase  in  students, this was partly affected by the easing of entry into universities, and the development of the  financial sector (entry of private banks and insurance companies) that required high‐skilled workers.  It is not clear though, if the developments cited in relation to student enrollment were expected to  have a long‐term impact on the LFPR. Early in the conflict, International Labour Organization (ILO)‐ modelled estimates and projections suggested a stabilizing participation rate after 2010 (Graph 4(b)),  though some caution needs to be exercised in taking this at face value given uncertainty surrounding  the underlying data.9    Graph 3: Pre‐Conflict Population Growth Rate and Projection  (a) 2017 Vintage (b) 2011 and 2013 Vintage Average annual  Average annual  growth growth 4.0% 4.0% 3.0% 3.0% 2.0% 2.0% 1.0% 1.0% 0.0% 0.0% 1991‐95 1996‐00 2001‐05 2006‐10 2011‐15 2016‐20 2021‐30 2031‐40 1991‐95 1996‐00 2001‐05 2006‐10 Total population (UN 2011) Working‐age population/total population (UN 2011) Total population (UN 2017) Total population (UN 2013) Working‐age population/total population (UN 2017) Working‐age population/total population (UN 2013) Note: Dashed lines indicate projections for years beyond UN report dates.                                                                           7  This is based on the medium fertility variant – “…considered the most likely future scenario and that is widely used by the  literature” (Barro and Lee 2015). This variant has “normal” fertility, mortality and migration rates.   8  Percentage of total population aged 15‐64.   9  Labor force participation rates for 2013‐2014 are projections. ILO‐modeled estimates and projections are based on  estimates and projections for GDP‐related variables and population structure. The 2013 estimates and projections draw on  the IMF WEO April 2013 and United Nations (2013). However, the IMF stopped publishing projections for Syria effective  2012, and ILO uses the regional median growth to extrapolate GDP growth for Syria.   6    Graph 4: Pre‐Conflict Labor Force Participation Rate and Projection  (a) 2018 Vintage (b) 2013 Vintage %, labor force  %, labor force  participation rate (LFPR) %, growth %, growth participation rate (LFPR) 60.0 60.0 54.5 54.8 4.0 4.0 50.0 45.8 50.0 2.0 44.9 2.0 40.0 40.0 0.0 0.0 0.2 30.0 ‐2.0 ‐1.3 30.0 ‐1.5 ‐2.0 20.0 ‐4.0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 20.0 ‐4.0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 LFPR (WDI Archive, ILO‐modelled estimates, 2013), LHS LFPR (WDI, ILO‐modelled estimates, 2018), LHS Growth in participation rate Growth in participation rate Average growth, 2001‐2010 Average growth, 2001‐2010 Average growth, 2011‐2014   Human capital growth, based on Penn World Tables (PWT) 9, which for Syria, uses Cohen, Soto  and Leker (CSL) data for the average years of schooling of the population aged 25 and above (Cohen  and Soto 2007; Cohen and Leker 2014), averaged 1.0% for the 10 years up to 2010.10, 11 Graph 5(a)  shows average years of schooling based on select age groups under both CSL and Barro and Lee (BL)  measures. Barro and Lee (2015) projections indicate a continued rise in the average years of schooling  absent conflict, for the population aged 15‐64: 12 1.6 years over 2011‐2030. Graph 5(b) shows human  capital  growth,  based  on  the  schooling  years  under  CSL  and  BL  measures.  While  fluctuations  and  differences  are  obvious  decade  to  decade,  there  is  consistency  in  a  long‐term  average  of  approximately 1.5%.  Graph 5: Pre‐Conflict Human Capital and Projection  (a) Average years of schooling (b) Human capital growth Average years  Average human  of schooling capital growth,  % 10.0 4.0 9.0 3.5 8.0 3.0 7.0 2.5 6.0 5.0 2.0 4.0 1.5 3.0 1.0 2.0 0.5 1980 1990 2000 2010 2020 2030 2040 0.0 1981‐1990 1991‐2000 2001‐2010 2011‐2020 2021‐2030 2031‐2040 Cohen, Soto and Leker (25 and above) Cohen, Soto and Leker (25 and above) Cohen, Soto and Leker (25‐64) Cohen, Soto and Leker (25‐64) Barro and Lee (25 and above) Barro and Lee (25 and above) Barro and Lee (25‐64) Barro and Lee (25‐64) Barro and Lee (15 ‐ 64); dashed line indicates projection Barro and Lee (15 ‐ 64); patterned bar indicates projection                                                                10  See the documentation, “Human Capital in PWT 9.0” (https://www.rug.nl/ggdc/productivity/pwt/). Human capital, ℎ , where s is average schooling years.  = 0.134 ∙ s if s ≤ 4,  =0.134 ∙ 4+0.101 ∙ (s − 4) if 4 < s ≤ 8,  = 0.134 ∙  4+0.101 ∙ 4+0.068 ∙ (s − 8) if s>8 (based on Caselli (2005) and Hall and Jones (1999)). For the first four years of education, a  rate of return of 13.4 percent is assumed, corresponding to the average, reported by Psacharopoulos (1994) for Sub‐ Saharan Africa. For the next four years 10.1 percent is assumed, the average for the world.  For education beyond the  eighth year the OECD value, 6.8 percent is assumed.  11  Using PWT 8.1 data points to a human capital growth rate 0.5% over 2001‐2010, but this is based on a previous version  of Barro and Lee data (version 1.3, which has since been updated).   12  Barro and Lee (2015) do not appear to capture the population loss in Syria from the conflict, as they use projections from  United Nations (2013).     7    Total factor productivity (TFP) growth, averaged 1.4% over 2001‐2010 based on calculations by  The  Conference  Board  (2018).  Our  own  estimations  following  the  methodology  in  Kim  and  Loayza  (2019) also suggest an average growth rate of 1.4% for the same period.13 See Graph 6.    Graph 6: Pre‐Conflict Total Factor Productivity (TFP) Growth  % 8.0 6.0 4.0 1.4 2.0 0.0 ‐2.0 ‐4.0 1.4 ‐6.0 ‐8.0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Kim and Loayza methodology (KL) (2019) The Conference Board (TCB) (2018) KL (2019), average 2001‐2010 TCB (2018), average 2001‐2010     2.2 Simulation of what would have happened in the absence of conflict  Table 1 details the baseline calibration of the LTGM‐PC for simulation under the no‐conflict scenario.  The  baseline  assumptions  are  consistent  with  a  long‐term  GDP  growth  average  of  close  to  5.0%.  Average real GDP growth of 5.3% over 2011‐2018 would have led to real GDP rising from USD 60 billion  in 2010 to USD 91 billion, and real GDP per capita rising from USD 2,857 to USD 3,774 by 2018. We  discuss  the  baseline  assumptions  in  sub‐sections  2.2.1  ‐  2.2.2  below.    Sub‐section  2.2.3  considers  upper and lower estimates for the calibration.     2.2.1  Key parameters and initial conditions  Key parameters and initial conditions (panels A and B of Table 1) either take their 2010 values or are  calibrated. In panel A of Table 1, Syria’s share of labor income of 0.520 is proxied by the 2010 average  value for relatively conflict‐free middle‐income MENA economies from PWT 9. This value is close to  the average for lower‐middle income (LMI) countries (0.510), as well as “oil‐based” middle‐income  economies with fuel exports amounting to less than 60% of their merchandise exports (0.540). The  Conference Board (2018)  also uses  a similar value of 0.500. We  calibrate the  depreciation rate  for  aggregate capital to 4.8%, which is the PWT 8.1 data point for 2010 (PWT 9 gives a value of 5.5% for  the same year). Generally, for Syria, PWT data suggests a relatively low share of structures (compared  to other forms of physical capital) and hence a high depreciation rate, one that is above the MENA  and LMI averages. So, we choose the lower of the two PWT depreciation values given the possibility  of measurement error. The public capital depreciation rate of 3.1% follows the PWT 9 depreciation  rate for non‐residential structures. Calculated as a residual, the private capital depreciation rate of                                                               13  TFP is measured using growth accounting based on the standard Solow‐Swan growth model with human capital. Syria data  for output, physical capital, human capital, and employed persons are taken from Penn World Tables (PWT) 9. Labor share  is proxied by the average for four relatively conflict‐free middle‐income Middle East and North Africa (MENA) economies  (Djibouti, Jordan, Morocco and Tunisia).     8    6.2% appears reasonable, consistent with the average of private capital depreciation rates for low‐  and middle‐income countries used in the IMF FAD Investment and Capital Stock Database 2017.14  In panel B of Table 1, we calibrate the initial capital‐to‐output ratio,  /  at 2.560 assuming  steady‐state properties15 and based on 30‐year averages of the investment‐to‐GDP share (22%), GDP  growth  (4.1%),16  and  aggregate  depreciation  rate  (4.5%  in  PWT  8.1).  Taking  this  approach  at  least  provides us with some basis of setting an initial  /  that is in between the PWT values for 2010  (2.384 based on PWT 9 and 2.632 based on PWT 8.1) especially since the PWT 9 value puts Syria on  the border of the 75th percentile of lowest K/Y ratios, and is below the respective averages of LMI  countries, low‐income countries as well as MENA countries.17   Using the IMF FAD Investment and Capital Stock Database 2017 to calculate the shares of public  and private capital shows that the share of public capital  /  is relatively high in Syria, accounting  for 0.610 of total capital. This is higher than the average for LMI countries of 0.399, but closer to the  average for oil‐based economies in the Middle‐East and North Africa (MENA) region, 0.506.  Still, Syria  appears to have the highest ratio among oil‐based economies that have fuel exports accounting for  less than 60% of exports (for which, the average is 0.397).18 Taking all this into account, we use a “mid‐ point” public capital share of 0.450 for our simulations.   2.2.2   Central projections  The central projected paths of key growth drivers over 2011‐2030 (panel C, Table 1) are based  on:   (i) average  /  over 2001‐2007 of 21.5% (from WDI), and public investment share of total  investment of 40%;   (ii) average growth rates over 2001‐2010 for TFP of1.4% (according to The Conference Board  2018  and  our  own  calculations  following  Kim  and  Loayza  2018)  and  human  capital  of  1.0% (following PWT 9);   (iii) United Nations (2011) projections for growth rates of total population of 1.6% and of the  working‐age population share of 0.6% (as discussed in Section 1.1); and   (iv) zero growth in the LFPR (i.e. the 2010 participation rate level is maintained) given some  signs of stabilization (and possible increase) from the 2013 ILO‐modelled estimations and  projections.                                                                       14  4.25% for low‐income countries and 8.30% for middle‐income countries as of 2015.   15  In steady‐state/equilibrium, output grows at the same rate as capital stock, which allows us to write,     /  where  is average output growth and  the depreciation rate.   16  Investment‐to‐GDP and GDP growth are averages of WDI and WEO data (1981‐2007 and 1981‐2010 respectively).   17  Our calibrated value is also closer to what the calculated K/Y would be based on MFMod data (2.498).   18  Based on data from the IMF database for  /  in 2010, and World Bank income and region classifications for the same  year. Oil‐based economies are defined as those with fuel exports at least 30% of total merchandise exports (using WDI  data).   9    Table 1: No‐Conflict Baseline Simulation ‐   Values for Parameters, Initial Conditions and Projected Variables  Input value:  2010/  Average,  Average,  Parameter/Variable*  Calibrated  Pre‐ 2011‐30  Note  Conflict  A. Constant Parameters  Labor share    (1)  0.520    Aggregate capital depreciation rate    (2)  0.048    Public capital depreciation rate    (2)  0.031    Private capital depreciation rate   (3)  0.062  B. Initial Conditions  Initial capital‐to‐output ratio  /   (4)  2.560      Public capital share of total capital  /  (5)  0.450    Initial public capital‐to‐output ratio  /   (6)  1.152    Initial private capital‐to‐output ratio  /   (6)  1.408  C. Projected Variables, Central Path (2010/11— 30)  Investment‐to‐output ratio /  (7)    0.215      Public investment‐to‐output ratio  /  (7)  0.086    Private investment‐to‐output ratio  /  (7)  0.129  Human capital growth    (8)  0.010  TFP growth    (9)  0.014  Population growth rate     (10)    0.016  Working age‐to‐population share, growth      (10)  0.006  Labor force participation rate, growth    (11)    0.000    *Multiply by 100 to obtain parameter/variable values in percent share or growth terms (%).   Note (source and calculation):   (1) Penn World Tables (PWT) 9: share of labor compensation in GDP at current national prices. There is  no data for Syria. As a proxy, we use the 2010 average for four relatively conflict‐free middle‐income  MENA economies (Djibouti, Jordan, Morocco and Tunisia).   (2)  is PWT 8.1 data for Syria.  is the PWT 9 depreciation rate for non‐residential structures.  (3)  is derived as the residual from a weighted average calculation of   based on  and  ⁄ .   (4) Calibrated  value  based  on  long‐term  averages  of  investment‐to‐output,  GDP  growth  and  the  aggregate capital depreciation rate. See the discussion in Section 2.2.1.  (5) Calibrated based on average shares for lower‐middle income countries and oil‐based economies. See  the discussion in Section 2.2.1.   (6) ⁄  and  ⁄  are derived by applying  ⁄ to   / .   (7) Gross fixed capital formation (% of GDP), average for 1998‐2007 from WDI. Public investment share  assumed at 40% based on World Bank (2017b) and IMF (2010).   (8) PWT 9: index of human capital per person based on years of schooling (Cohen and Soto 2007; Cohen  and Leker 2014) and returns to education (Psacharopoulos 1994).  Average growth rate, 2001‐2010.   (9) The Conference Board (2017)/our own estimates using Kim and Loayza (2019). Average growth rate,  2001‐2010.   (10) United Nations (2011).   (11) Zero growth is assumed based on the stabilizing/marginally increasing participation rate observed in  the 2013 ILO‐modelled estimates and projections.     2.2.3   Lower and upper estimates of projections  Lower and upper estimates are set based on adjustments to some of the projected variables  in panel C of Table 1.  Table 2 displays these calibrations – which are those projections that  are different from the central projections in panel C of Table 1.   10     For  the  lower  estimate,  we  introduce  a  decline  in  the  LPFR  based  on  the  latest  ILO‐ modelled estimates and projections that in turn draw on the most recent UN population  estimates  and  regional  GDP  growth.19  Other  possible  alternatives  for  this,  though  probably less satisfactory, are to: (i) assume the same decline in the participation rate as  in the previous decade, but this would have likely been an unsustainable path; or (ii) use  an earlier average growth, namely for 1991‐2000 (‐0.6%), which would be quite arbitrary.       For the upper estimate, we use a higher  / of 23.5% since the IMF data presented in  Section  2.1  indicates roughly  a 2 ppt increase in average gross capital  formation  (from  23.1% over 2001‐2010 to 25.2% over 2011‐2016) and we assume the same for gross fixed  capital formation. We also use a higher long‐term average growth rate of 1.5% for human  capital based on both CSL and BL data.    Table 2: No‐Conflict Simulation –   Upper and Lower Estimates for Projected Variables  Variable  Note  Input value  Lower estimate on projected variables (average, 2011‐2030)  Labor force participation rate, growth    (1)  ‐0.003  Upper estimate on projected variables (average, 2010/11‐2030)  Investment‐to‐output ratio /  (2)  0.235    Public investment‐to‐output ratio  /  (2)  0.094    Private investment‐to‐output ratio  /  (2)  0.141  Human capital growth    (3)  0.015  *Multiply by 100 to obtain parameter/variable values in percent share or growth  terms (%).   Note (see Section 2.1 for further discussion):   (1) Calculated using female and male LFPRs from ILO‐modelled  estimates and  projections, May 2018. We carry out linear interpolation for annual values  based on the ILO‐reported values of series end points (2010 and 2030) and  turning points (2020‐21 for the male participation rate) so that trends are  smoothened.  Female  and  male  shares  of  working‐age  population,  from  United Nations (2017) are kept unchanged at 2010 levels.   (2) / is based on IMF estimations suggesting higher GCF (% of GDP) after 2010.  The  assumption  of  public  investment  share  of  total  investment  of  40%  as  used in the central projection is maintained.   (3) Long‐term average based on calculations using both Cohen, Soto and Leker;  and Barro and Lee average years of schooling data respectively.  See Graph  4(a)‐(b).    Graph 7 shows the trajectory for the level and growth of GDP in Syria based on the central calibrations  in Table 1 and the calibrations for the upper and lower estimates from Table 2. Note that the lower  estimate  shows  a  slight  upward  trajectory  –  this  is  because  of  the  ILO‐modelled  estimates  and  projections, which show the LFPR for men gradually rising from 2021 onwards, after declining in the  previous decade.                                                                      19  The estimates and projections do not fully account for the conflict. UN projections of population decline are captured  but not GDP growth developments. See footnote 10 for further details.   11    Graph 7: No‐Conflict Simulation for GDP in Syria      3. Impact from Conflict  This  section  provides  a  simulation  of  how  the  different  growth  drivers,  introduced  in  the  previous  section, account for the loss in GDP level and growth over 2011‐2018. The purpose of this simulation  is to provide an up‐to‐date, factual and holistic analysis of what happened to GDP potential based on  its underlying drivers. Such an analysis complements other work done, for example, by the United  Nations Economic and Social Commission for Western Asia (UN‐ESCWA) and the World Bank, which  have documented overall GDP loss by sector and expenditure components. Furthermore, the analysis  in this section is a necessary step towards establishing as best as possible, the initial conditions post  conflict, an important precursor to simulations for the reconstruction period.   3.1 How were the drivers of growth affected by the conflict?  3.1.1  Physical capital stock and investment  The  estimates  of  physical  capital  stock  damage  by  the  UN‐ESCWA  appear  to  be  the  most  comprehensive and up‐to‐date data so far, and we use these in our simulations. ESCWA (2016) puts  the destruction of capital across various economic sectors, including public and private capital, at USD  89.9 billion over 2011‐2015. ESCWA (2018) updates this to USD 101.2 billion and USD 119.7 billion by  the end of 2016 and 2017 respectively.  As pockets of fighting continued in Syria in 2018, we build in  12    further damages of USD 7 billion to total USD 126.7 billion in 2018.20  We assume that the conflict  intensity index in the first three months of 2018 is at the 2017 level as estimated by the ESCWA, but  then  improves  to  somewhere  between  the  2011‐2012  level.  Given  the  calculated  average  conflict  intensity index for the year, we then derive an estimate of physical damage that is proportionate to  the 2017 estimate of physical damage of USD 18.5 billion (see Appendix 2).21 Such an assumption of  improvement in conflict intensity is not inconsistent with the dread  factor22  in World Bank (2019),  which reached new heights in 2017, continuing into the first quarter of 2018 before appearing to taper  off to around the 2012 level by May 2018, the last data point. We also consider that the SOHR notes  that a record low 20,000 lives were lost in 2018, and that big data (conflict‐related events) collated  and analyzed by the Global Data on Events, Location and Tone (GDELT) project suggests that conflict  intensity in 2018 was lower than in 2011.   World Bank (2017b, in their Figure 2.9) estimates a decline in private investment as a share of  GDP  ⁄  from 12% in 2010 to 4% in 2015, and for the public investment‐to‐GDP ratio  ⁄  – a  decline from 9% in 2010 to 1% in 2015. On the other hand, ESCWA (2018) sees a smaller decline from  12%  to  9%  for  private  investment  as  a  share  of  GDP  (averages  for  2006‐2010  and  2011‐2016  respectively) and from 10% to 7% for public investment as a share of GDP. In our simulations, we use  the World Bank estimates for our central projections, assuming   ⁄  and  ⁄ remain stable at 4%  and 1% respectively over 2016‐2018. We then consider the ESCWA estimates as an upper estimate to  our projections.   A deterioration in the efficiency of public investment is a likely concern. No Syria‐specific value  of the Infrastructure Efficiency Index (IEI) (Devadas and Pennings 2019) is available. As a proxy for  Syria’s pre‐conflict efficiency of new public capital, we use the LMI average of 0.734, and build in a  decline in the efficiency of public investment to the low‐income (LI) group average of 0.570 by 2017  (at which point the World Bank’s income classification of Syria switches from LMI to LI).    3.1.2 Demographics and the labor force  In terms of Syria’s total population, UN population statistics (United Nations 2017)23 indicate negative  average annual growth of ‐1.7% over 2011‐2018, with the population declining from 21 million in 2010  to 18.3 million in 2018. UN statistics also indicate for the working age‐to‐population share, negative  average annual growth of ‐0.09% over 2011‐2018. The projections capture developments during the  conflict.  From  mid‐2010  to  mid‐2020,  estimated  deaths  are  higher  at  1,036,445  compared  with  657,131 the previous decade24 and reflect the rise in the proportion of male deaths (68% versus 32%  for women). Further, net migration abroad is also higher at 5,397,896 over mid‐2010 to mid‐2020.                                                               20  In terms of other sources of data on physical damage, SCPR (2016) estimates total capital stock at 2010 prices to be  down by 57% at end 2015 compared with 2010, but with no details of the breakdown of damages. World Bank (2017a), in a  detailed but narrower assessment, estimates the damage to housing and infrastructure for six cities in the range of USD9.6  – USD11.1 billion at pre‐conflict prices, over 2011 – February 2017. The housing damages are extrapolated to eight  governorates and used in the model‐based simulations of physical destruction effects on GDP in World Bank (2017b).  21  We use the most recent year’s value of physical damage as opposed to some longer‐term average measure, as the  former would best reflect possible further damages given that much of Syria’s physical capital would have been destroyed  by 2017/2018.   22  Shown graphically, this is an aggregate measure of the number of conflict events based on tank, artillery and air strikes  of light, medium and heavy incidence levels.    23 Medium fertility variant.   24  607,685 on average per decade over mid‐1980 to mid‐2010.    13    This considers refugee numbers based on the UN Refugee Agency (UNHCR) populations of concern  data up to March 2017 and builds in refugee returns over 2020‐2035. 25   There  is  a  possibility  though  that  the  above  numbers  might  still  understate  fatalities  and  migrants. The cumulative number of fatalities from the war is hard to ascertain. A source currently  relied upon, the Syrian Observatory for Human Rights (SOHR) puts the death toll at 560,000 (civilians  and combatants) over March 2011 – 10 December 2018 (with 367,965 documented victims). Netting  off foreign combatants of 65,108, we would get a total of 494,892 Syrian fatalities.26 This is higher than  what the UN population statistics suggest in terms of conflict‐related deaths – 303,451, based on our  calculations  when  we  take  the  difference  of  average  deaths  between  2011‐2020  and  2001‐2010  (multiplied by eight years).   The  UNHCR  populations  of  concern  data  show  that  compared  to  2016  total  registered  Syrian  refugees and asylum‐seekers increased in 2017 by 746,811 (Graph 8(a)). While the equivalent 2018  data are not available, from the UNHCR Regional Refugee and Resilience Plan database specifically on  refugees in selected neighboring countries ‐ Egypt, Iraq, Jordan, Lebanon, Turkey and North Africa,27  we  note  that  when  compared  to  2017,  refugees  had  increased  by  184,398  to  5,663,675  in  2018  (patterned bars in Graph 8(a), with the age and gender profile of these refugees shown in Graph 8(b)).  Based on this, the UN population statistics may be understating refugee movement by about 900,000.  But if we discount the 1 million refugees born in displacement/exile supposedly included in the UNCHR  data, then the under‐representation of population “loss” in the UN population statistics due to the  use of outdated UNHCR data disappears.28 Perhaps of greater concern then is that the UN population  statistics likely do not include non‐registered Syrians in neighboring countries, for which estimates  vary but tend to go up to more than a million (see for instance, Vignal (2018) and World Bank (2019)).  UNDP and UNHCR (2019) put the difference between estimated total Syrians29 and registered Syrian  refugees  at  about  1.6  million  in  December  2018.  The  difference  is  wholly  accounted  for  by  Egypt,  Jordan and Lebanon.   Consequently,  as  a  possible  lower  estimate  to  population  growth  estimates,  we  calculate  an  added decline in the Syrian population of 1.8 million, building in additional conflict deaths of 200,000,  and  unregistered Syrian migrants of 1.6 million. This would reduce the 2018 population from 18.3  million to 16.5 million giving a negative average growth of ‐3.0% over 2011‐2018.30                                                                   25  The UNHCR data available up to 2016 indicate a change of 5,688,897 in the number of refugees, asylum‐seekers  (pending cases) and others of concern based on the 2010 and 2016 stock positions. UN data on the international migrant  stock for 2017, show net migration for Syria of 5,784,696 over 2011‐2017 (2017 stock compared against 2010 stock).   26 Source: http://www.syriahr.com/en/?p=108723. Civilian casualties totaled 111,330 (22.5% of the 494,892 total Syrian  casualties), and were mostly adult males (70.1%), as 20,819 were children under 18 (18.7%) and 13,084 were women over  18 (11.8%).  27  Syria Regional Refugee Response. https://data2.unhcr.org/en/situations/syria  28  https://www.unhcr.org/refugeebrief/the‐refugee‐brief‐11‐december‐2018/  29  Total estimated number of Syrians are government estimates which include registered Syrian refugees, unregistered  Syrian refugees and Syrians residing in host countries under alternative frameworks.   30  UN population statistics suggest that births remained relatively high at 4,339,946 over mid‐2010 to mid‐2020 (about  434,000 births/year) versus 5,275,028 the previous decade (about 528,000 births/year). Some news reports (for example:  http://english.alarabiya.net/en/variety/2015/11/17/Births‐in‐Syria‐down‐more‐than‐50‐since‐war‐media.html) suggest  that births have fallen substantially, but absent more sources of concrete analysis, we do not make any adjustments to  birth rates. Further, World Bank (2019) notes that adolescent fertility rates have only dropped to 40 per 1,000 women aged  15‐19 in 2016 compared to 45 in 2010, amid child marriages as a conflict coping mechanism.   14    Graph 8: UNHCR Registered Syrian Refugees and Asylum‐Seekers  (a) Total Syrian Refugees and Asylum‐Seekers (b) Profile of Syrian Refugees in Selected Neigboring Countries Million  All countries (source: UNHCR persons populations of concern*) 7.00 Total Turkey and MENA (source: UNHCR 6.46 populations of concern*) 5.72 6.00 5.48 60+ yrs 5.66 Selected neigboring countries 5.13 5.48 4.86 5.00 (source: UNHCR Regional Refugee 4.59 18‐59 yrs and Resilience Plan) 3.97 4.00 3.73 12‐17 yrs 3.00 2.52 2.39 5‐11 yrs 2.00 0‐4 yrs 0.76 0.70 1.00 ... ... 0 10 20 30 40 50 60 0.00 2010 2011 2012 2013 2014 2015 2016 2017 2018 % of total refugees in  Egypt, Iraq, Jordan, Lebanon, Turkey,  Female Male and North Africa (5,663,675 as at 31 December 2018) *Comprises refugees, asylum‐seekers (pending  cases) and others of concern.  Source: UNHCR and Government  of Turkey       ILO (2018) projects the overall LFPR at 43.0% in 2018 (female at 12.5% and male at 73.0%) down  from 44.9% in 2010 (female at 14% and male at 75.6%), giving an average negative growth of ‐0.5%  over 2011‐2018 (female LFPR growth at ‐1.4% and male at ‐0.4%). Data in World Bank (2017b) suggest  the LFPR in 2015 stood at 47.2%, somewhat higher than the 2010 level of 44.6%.31 World Bank (2019)  meanwhile shows latest/2017 LFPRs for men and women above 15 years of age at 79.1% and 11.9%  respectively versus 73.0% and 13.0% in 2010. We use the ILO estimates in our central projection, and  for a possible upper estimate, consider an increment in the LFPR by 2018 based on the changes in  participation rates of men and women reported in World Bank (2019).   3.1.3 Human capital  Human  capital  would  have  been  affected  by  (i)  the  interruption  of  schooling  of  the  younger  population, who represent future entrants into the labor force; and (ii) migration and fatalities which  alter the distribution of years of schooling among the remaining population/existing labor force.   Regarding the interruption of schooling, according to SCPR (2016), almost half (45.2%) of all basic  education school‐age children residing within the country were not attending school by 2014‐2015.  They estimate a decline of 1.5 in the average years of schooling for population aged 15 and above,  based on these non‐attendance rates, from 6.8 in 2010 to 5.3 years in 2015.32   We provide alternative calculations to assess the impact from the interruption of schooling as  well  as  migration  and  fatalities  on  the  national  average  of  years  of  schooling.  To  do  this,  we  use  forward extrapolations, with the following assumptions:33    for  age  groups,  3: 25 29 to  10: 60 64,  the  educational  attainment  (average  schooling years), , of gender   (either men,   or women,   in age group,  , at time,  , is  the same as that of the age group five years younger at time,  5, as we assume these groups  have completed their education, i.e.,                                                                  31  Though this LFPR increase contrasts with the ILO estimate of a decline, data in World Bank (2017b) suggests a sharp  increase in the unemployment rate from 8.6% in 2010 to 52.9% in 2015. ILO estimates, on the other hand, point to a more  moderate increase in the unemployment rate to 14.9% in 2018 (14.7% in 2015).   32  The same contraction rate in enrollment for basic education is assumed for secondary and tertiary education.   33 The nomenclature draws on Barro and Lee (2013, 2015).  15    , ,       ∑ ℎ , , ,             (1)    where  ℎ ,  is the proportion of either men or women (  or  ) in age group  ,  that have attained education level  (primary, secondary and tertiary; incomplete and  complete) and  is the corresponding duration system, both of which we assume  to be unchanged;      for  age  groups,  1: 15 19  and  2: 20 24, who  are  still  in  school,  we  use  the  attainment  in  5   for  the  same  group  ,  adjusted  to  account  for  changes  in  enrollment  ratios, Δ , , 34 for age group  in level  during the transition period from  5 to , i.e.    ,    ∑ ℎ , , Δ , , ,    ∑ ℎ , , , Δ , , ,     , ∑ Δ , , ,           (2)    Average total years of schooling for each age group   is then a composite of the respective average  years of schooling of men () and women ( ) in that group:  , , , ,            (3)  where   is population in age group  at time .   Finally, we derive average total years of schooling for the population aged 15‐64:  ∑              (4)  where   is the total population aged 15‐64 at time .   World Bank (2019) indicates that the net enrollment rate for school‐age children (5‐17 years) declined  from 82% in 2010 to 61% in 2018. The sharpest decrease in the enrollment rate for school‐age children  appears to have occurred in the 2012/13 school year, with the enrollment rate rising slightly thereafter  from 55% in 2013 to 60% in 2015 (UNICEF 2016, 2018a). Primary and secondary net enrollment rates  stood at 93% and 67% respectively in 2010. Last available data from UNESCO UIS indicates that these  rates were down to 63% and 46% respectively in 2013, with the secondary enrollment rate stable at  45% in 2018 (UNICEF 2018b). For our projection of average years of schooling for the still‐in‐school  groups using equation (2), we let primary and secondary enrollment rates decline from 93% and 67%  in 2010 to 75% and 45% respectively by 2015 for both girls and boys, keeping them steady thereafter.35  Regarding tertiary education, UNESCO UIS data on the average gross enrollment rate over 2011‐2016                                                               34  We base the enrollment adjustment factor formula on Barro and Lee (2013) (their Table A.2).   35  We use a simple average calculation to obtain the 2015 primary enrollment rate based on a net enrollment rate for  school‐age children of 60% and secondary enrollment rate of 45%. Taking together the data reported by UNICEF (UNICEF  2016, UNICEF 2018a) and the World Bank (World Bank 2019), net enrollment of school‐age children averaged about 84%  over 2010‐2012, 60% over 2013‐2015, and 61.3% over 2016‐2018.     16    rather  surprisingly  suggest  an  increment  of  about  10  percentage  points  during  the  conflict  period  compared with the 2010 rate of 26%.36  We build in this increment over 2010‐2015, applying the same  enrollment rate for men and women, and keep it unchanged thereafter. Primary school duration is  assumed to be 6 years (6‐11 years), secondary school, 6 years (12‐17 years), and university 4 years  (18‐21  years).  In  Appendix 3,  we provide  further  details on how  we  arrive  at  the  average  years of  schooling in 2018.   By using the above methodology, we constrain changes to the national average of schooling  years  to  arise  from  shifts  in  the  distribution  of  the  total  population  by  age  and  gender,  and  in  enrollment ratios. Because we use past composite values of average years of schooling, we do not  consider changes in completion rates (that is, we are assuming completion rates are unchanged). This  approach also does not consider other types of heterogeneity in educational attainment, for instance,  that may depend on the socioeconomic status or geographical origination of migrants and conflict  victims. Verme et al. (2016), focusing on the profile of refugees in Jordan and Lebanon, find that Syrian  refugees in fact tend to have slightly lower levels of educational attainment than pre‐conflict Syrians.37     To obtain human capital growth, we continue to assume the same calculations as in PWT 9  for the returns to education. Our calculations for the central projection using UN population statistics  suggest that average years of schooling would have declined by 1.467 years for the population aged  15‐64 with human capital contracting by an annual average growth of ‐2.59% over 2011‐2015 and ‐ 0.56%  over  2016‐2018  respectively.  When  we  consider  the  additional  decline  of  1.8  million  in  the  Syrian  population  discussed  in  the  previous  section,  average  years  of  schooling  declines  only  marginally more, that is by 1.499 years. If we split this decline over time following the proportions in  the first case, we obtain annual average growth rates in human capital of ‐2.64% over 2011‐2015 and  ‐0.57% over 2016‐2018 respectively.   3.1.4 TFP   The Conference Board estimates  an average  TFP growth of ‐9.5% over 2011‐2017 (from a low of ‐ 22.9% in 2012 rising to 2.8% in 2017). However, while the loss of TFP would be an important driver of  loss in GDP potential, it could be that this high number reflects an underestimation of the destruction  of physical and human capital.  Thus, as an alternative to The Conference Board’s calculations of TFP growth, we utilize Kim and  Loayza’s (2019) model of determining TFP growth. The key element that feeds into this model is an  overall  index  of  TFP  determinants,  the  determinants  being  education,  infrastructure,  innovation,  institutions, and market efficiency. The composite index stood at 30.33 for Syria in 2010 on a scale  between  1  and  100.  We  estimate  the  trajectory  of  this  index  over  2011‐2018  by  calibrating  its  subcomponents. For the education index, we calculate a decline that is proportionate to the fall in  average years of schooling of the working‐age population. Given that enrollment rates fall, and the  exodus  of  Syrians  increases  noticeably  only  from  2013,  we  keep  the  average  schooling  years  unchanged  from  2010  to  2012,  instead  interpolating  a  decline  over  2013‐2018,  based  on  our  calculations of average years of schooling for 2015 and 2018 respectively. For the infrastructure index,                                                               36  See Milton (2019) for a discussion of how Syria’s higher education system survived quantity wise, despite general  expectations that higher education suffers relatively more during conflict, but that quality had been eroded and political  control over campuses increased.   37  Educational attainment for pre‐conflict Syrians five years and older (2007): 83 percent with primary education and  below, 9.5 percent with secondary education and 7.2 percent with tertiary education. Syrian refugees five years and older  (2014): 83.6 percent with primary education and below, 11.4 percent with secondary education and 5.1 percent with  tertiary education (Jordan); 86.7 percent with primary education and below, 8 percent with secondary education and 5.3  percent with tertiary education (Lebanon).   17    we build in a decline that is proportionate to the relative total light in Syria over time estimated by Li  et al. (2017).  For the institutions index, the estimation is based on the source data originally used by  Kim  and  Loayza  (2019)  ‐  the  Worldwide  Governance  Indicators  (WGI)  across  six  dimensions  (Kaufmann, Kraay  and  Mastruzzi  2010)  updated to 2017.  For  the  innovation  and market  efficiency  indices,  we  assume  that  these  evolve  proportionately  to  a  weighted  average  of  the  indices  for  infrastructure and institutions. This gives an overall TFP determinant index of 15.98 in 2018, almost  half  the  2010  level.  The  associated  average  annual  TFP  growth  over  2011‐2018  is  ‐1.6%.38  Further  details of our calculations are provided in Appendix 4.   3.2 Simulation of the impact from the conflict   Table 3 details the baseline calibrations of conditions during the conflict years 2011‐2018, following  the  discussion  in  Section  3.1.  Regarding  the  public  and  private  capital  to  output  ratios,  ⁄   and  ⁄,  the  simulated  ratios  inclusive  of  damage  in  Table  3,  panel  B  uses  calculations  as  described  below to reflect the damage to capital stocks:   Each  period’s  initial  ⁄   and  ⁄  are  reduced  by  lowering  (for  , )  by  the  amount of the monetary value of physical damage (with   held constant). Damage during a  period (year) affects capital and initial capital‐to‐output ratios for the next period.     , , Initial conflict capital‐to‐output ratios,      where  ,  is capital adjusted for damage,  .    , , Then,     , where   is based on   of 1.152 and 1.408 respectively for public and private capital,  and   = USD 60.043 billion;     growth in adjusted capital per worker, 1 ,    , , ,                                                     (5)  where  , the proportion of capital damaged in 2011.   and growth in output per worker,   1 , 1 Γ 1 , 1 , 1 , 1 , 1 ,     with 1 Γ 1 , 1 , 1 ,                      (6)   The process is repeated for periods 2012‐2018. Damages to   are apportioned across the  conflict period based on the 2011‐2017 annual estimates from ESCWA (2018) while we build                                                               38 As an exercise to gauge the room for underestimation of the contraction in TFP growth, we run the Kim and Loayza  (2019) model with the assumption that the overall index of TFP determinants for Syria declines linearly from 30.33 to the  minimum 1 by 2018. This would give us an average annual growth of ‐3 percent over 2011‐2018, only 1.4 percentage  points lower than our current projection of the average growth rate.    18    in  USD  7 billion for  2018 as discussed in  Section 2.1.1. Damages are apportioned between  public and private capital based on their relative cumulative shares of damages as at end 2015,  made available by ESCWA. We assume the same shares for each time  (that is, 40 percent of  damages are attributable to public capital, 60 percent to private capital).   In Table 3, panel C, we present information on the central paths of projected variables based on  the previous discussions. Under the central projection, both  ⁄  and  ⁄ are lower in 2018 (1.029  and  0.708  respectively)  compared  with  2010  (1.152  and  1.408  respectively)  but  more  so  for  the  former, since the damage for public capital was valued at less than for private capital.     Table 3: Simulation of Syria’s Conflict Years (2011‐2018)   Parameter/Variable  2010  Average  2018  Note  2011‐18  A. Constant Parameters  Labor share    (1)    0.520    Aggregate capital depreciation rate    (1)  0.048    Public capital depreciation rate    (1)  0.031    Private capital depreciation rate   (1)  0.062  B. Capital‐to‐Output (K/Y) Ratios  Initial public capital‐to‐output ratio  /   (1)  1.152    Simulated  / (with damage)  (2)    1.141  1.029  Initial private capital‐to‐output ratio  /   (1)  1.408    Simulated  / (with damage)  (2)  1.144    0.708  C. Projected Variables, Central Path (2011‐2018)  Public investment‐to‐output ratio  /  (3)  0.090  0.025  0.010  Private investment‐to‐output ratio  /  (3)  0.120  0.064  0.040  Efficiency of new public investment    (4)  0.734  0.632  0.570  Human capital growth    (5)    ‐0.018  ‐0.006  TFP growth    (6)  ‐0.016  ‐0.022  Population growth rate     (7)  ‐0.017  0.001  Working age‐to‐population share, growth      (7)  ‐0.001  0.011  Labor force participation rate, growth,     (8)  ‐0.005  ‐0.008  *Multiply by 100 to obtain parameter/variable values in percent share or growth terms (%).   Note (source and calculation):   (1) Unchanged from Table 1. See Table 1 for further details on source and computation.   (2) Calculations follow the discussion in Sections 3.1.1 and 3.2.   (3) Based on World Bank (2017b). Their estimates of 1% and 4% for  /  and  / respectively in  2015 are also assumed for 2016‐2018 in our simulations.   (4) Based on the average IEI for LMI countries for 2010 which is assumed to gradually decline to  0.570 (the average IEI for LI countries) by 2018.   (5) See  Section  3.1.3.  Given  that  enrollment  rates  fall,  and  the  exodus  of  Syrians  increases  noticeably only from 2013, we keep the index of human capital unchanged from 2010 to 2012,  such  that  the  contraction  mainly  occurs  over  2013‐2015  with  average  growth  of  ‐4.28%  (average growth over 2016‐2018: 0.56%).   (6) TFP growth is calibrated using the empirical methodology in Kim and Loayza (2019). See Section  3.1.4.  (7) Based on United Nations (2017).   (8) Female and male LFPRs are based on ILO‐modelled estimates, May 2018 (ILOSTAT). Female and  male shares of the working‐age population follow United Nations (2017).     19    In Table 4, we show the lower and upper estimate outcomes for capital‐to‐output ratios when  we consider (i) lower population growth and the associated human capital growth for this population  trend;  and  (ii)  higher  on  average  investment‐to‐output  ratios  and  a  higher  LFPR.  Other  variables  remain unchanged from their central paths in the simulations. For both the lower and upper estimate  scenarios, the / ratios in 2018 are slightly higher than in the central projections but remain lower  than the 2010 levels.     Table 4: Simulation of Syria’s Conflict Years (2011‐2018)   – Lower and Upper Estimates and Impact on Capital‐to‐Output Ratios  Note  Average  2018  Variable  2011‐18  A. Lower Estimate  Population growth    (1)  ‐0.030  0.000  Human capital growth    (2)  ‐0.019  ‐0.006          Capital‐to‐Output (K/Y) Ratios (with Damage)  Simulated  /  1.180  1.091  Simulated  /   1.175  0.737  B. Upper Estimate  Public investment‐to‐output ratio  /  (3)  0.070  0.070  Private investment‐to‐output ratio  /  (3)  0.090  0.090  Labor force participation rate, growth,    (4)  0.007  0.006  Capital‐to‐Output (K/Y) Ratios (with Damage)  Simulated  /   1.133  1.127  Simulated  /   1.139  0.784  *Multiply by 100 to obtain parameter/variable values in percentage share or  growth terms (%).   (1) Based on a further downward adjustment to the UN population statistics to  reflect higher migration and fatalities than the central projection. See Section  2.2. Population changes over 2011‐2018 are calibrated such that they follow  the trend in the UN population statistics but with steeper slopes.    (2) Human capital growth is calculated based on the adjusted population data.  See Section 3.1.3. As in the central projection, we keep the index of human  capital unchanged from 2010 to 2012, such that the contraction mainly  occurs over 2013‐2015 with average growth of ‐4.37% (average growth  over 2016‐2018: 0.57%).   (3) Investment‐to‐output  ratios  are  based  on  ESCWA  (2018)  estimates.  See  Section 3.1.1.  (4) Calculated based on the growth in LFPRs for men and women above 15 years  of  age  over  2010‐2018  reported  in  World  Bank  (2019).  Female  and  male  shares of the working‐age population follow United Nations (2017).       Graphs 9(a) ‐ 9(b) show the outcomes of simulations for GDP level and growth over the conflict  years given the calibrations in Table 3 and Table 4. Our simulations across the three scenarios (central,  lower and upper estimate projections) indicate that the depletion of factors of production alone may  account  for  about  87  percent  of  the  negative  GDP  growth  on  average,  and  further,  that  about  64  percent of the average negative growth is due to physical capital destruction. Demographics and labor  20    account for about 15 percent, human capital 7 percent, and TFP 13 percent of GDP growth on average  over the conflict years.   The decrease in physical capital reflects the compounded effects of large outright damages, low  net investment rate, and a falling output base (which is adversely affected by all growth drivers). The  prominent effective losses due to physical capital destruction are worsened by the lack of sufficient  investment. This echoes the observation by World Bank (2017b) that capital destruction itself might  have  relatively  subdued  effects  in  a  well‐functioning  economy,  as  in  the  aftermath  of  a  natural  disaster; but in the case of conflicts, the fall in investments due to disruptions in economic organization  reinforces the adverse effects from physical capital damages. Having said that, our estimate of physical  capital  decrease  is  greater  than  the  estimate  in  World  Bank  (2017b)  because  of  methodological  reasons: we take into account the monetary value of physical capital destroyed (as reported by ESCWA  (2018), as well as depreciation and gross investment, directly in the calculation of the capital stock;  while World Bank (2017b) assumes that the resulting capital stock keeps the same proportion with  respect to the initial capital stock as the stock of housing does.     Graph 9(a): Conflict Years Calibration for GDP in Syria                     21    Graph 9(b): Average Impact of Different Growth Drivers on GDP during the Conflict  (across Central, Lower and Upper Estimate Projections)  Real GDP Growth  USD bil, 2010  Real GDP Level 5.0% 60 0.0% 50 40 ‐5.0% 30 ‐10.0% 20 ‐15.0% 10 ‐20.0% 0 2010 2011 2012 2013 2014 2015 2016 2017 2018 ‐25.0% 2011 2012 2013 2014 2015 2016 2017 2018 Physical capital TFP Physical capital All factors of production (excluding TFP) Demographics and labor force Human capital All growth drivers (including TFP)     The projections under conflict versus the no‐conflict scenario suggest a cumulative loss in GDP  potential of between USD289 and USD300 billion over 2011‐2018 (see Graph 10(a)). Our estimates  point to a continued loss in 2017‐2018 because of the damage to physical capital and negative TFP  growth. This varies somewhat from ESCWA (2018) and others like Devarajan and Mottaghi (2017),  Gobat  and  Kostial  (2016)  and  World  Bank  (2017b),  all  of  which  point  to  a  trough  in  actual  GDP  contraction around 2012‐2013. ESCWA (2018) estimates average GDP growth of ‐10% over 2011‐2017,  with growth turning positive in 2017. ESCWA also projects a GDP level of USD 27 billion in 2017 against  a no conflict counterfactual of USD 86 billion. Our estimates seem to mimic these results, pointing to  a loss in potential GDP growth of ‐12% on average over 2011‐2018, with an average GDP level of USD  23.2 billion in 2018 (against a no‐conflict scenario of USD 91 billion).      Graph 10: GDP Loss based on the Conflict Simulation Compared Against the Counterfactual of No Conflict            22    4. Growth Scenarios Post Conflict   This section discusses potential growth scenarios for Syria in the aftermath of war, exploring how long  it would take for Syria to reach its pre‐conflict level of development and the number of “lost decades”  under various assumptions for the growth drivers.   Table 5: Key Factors in the Post‐Conflict Scenarios for Growth Drivers  Broad Case  Factor  Baseline (Moderate)  High (Optimistic)  Low (Pessimistic)  Voluntary  Security  Partial political  Robust political  Political settlement  mobility OR  settlement with  settlement.  largely reflects de‐ Forced  strong guarantees for    facto balance of  repatriation  micro‐security and  power, with limited  property rights.   guarantees for micro‐ security and property  rights.  Reconstruction  A substantial  A large  Limited  program  reconstruction  reconstruction  reconstruction  program of USD 140  program amounting  program of USD 30  billion (average of the  to USD 250 billion to  billion, largely relying  high and low  meet UN‐estimated  on China, Iran and  scenarios), spread  reconstruction bill,  Russia, and spread  evenly over a 20‐year  spread evenly over a  evenly over a 10‐year  period.   20‐year period.   period.   Voluntary  Refugee  Of total refugees in  76 percent rate of  Of total refugees in  mobility   returns  neighboring countries,  refugee returns from  neighboring    43 percent return rate  neighboring  countries, 18 percent  based on the ratio of  countries, based on  return rate based on  reconstruction funds  the UNHCR (2018)  the ratio of  in the first 10 years  survey of refugees  reconstruction funds  (moderate versus high  intending to return  in the first 10 years  scenario).   to Syria one day.    (low versus high    scenario).  Forced  Refugee  100 percent return rate, of refugees in neighboring countries.   repatriation  returns     Notes: Under the voluntary mobility case, refugee returns follow the same time pattern across the three scenarios,  rising to peak around 2023‐2024, and gradually moderating thereafter. World Bank (2019) finds that refugee  mobilization tends to be lower, the lower are security and infrastructure services.       Experiences of other countries in the Middle‐East suggest that longer‐lasting conflicts would  entail a longer recovery period. Sab (2014) notes that it took Lebanon 20 years to reach its pre‐war  GDP level (after the Lebanon Civil War from April 13, 1975 – October 13, 1990), Kuwait, seven years  (Gulf War from August 2, 1990 – February 28, 1991), and Iraq, one year only (2003 Invasion of Iraq  from March 19‐May 1, 2003).39 Lebanon lost 70 percent of its GDP level while Kuwait 55 percent and  Iraq  35  percent  during  their  respective  wars.  Gobat  and  Kostial  (2016)  note  that  under  the  hypothetical assumption of reconstruction starting in 2018 and the Syrian economy growing at about  4.5 percent, it would take the country about 20 years to reach its pre‐conflict real GDP level. ESCWA                                                               39  Sab (2014) highlights that while the 2002 invasion of Iraq did not last long, Iraq had experienced low‐grade civil war for  much of the 2000s until the summer/fall of 2009.   23    (2018) puts Syria’s real GDP at about USD 27 billion in 2017 – some 55 percent below the 2010 level  and close to the level in the early 1990s. Our central projection of Syria’s potential GDP level in Section  3, at USD 21.2 billion in 2018, is 65 percent below the GDP level in 2010, close to the loss experienced  by Lebanon.   To analyze the outlook for growth in Syria, we first consider a voluntary mobility case within  which our projections for the growth drivers are guided by three plausible political settlement/security  outcomes (baseline/moderate, optimistic, and pessimistic). These settings are associated with varying  levels of reconstruction assistance, which in turn influences the voluntary mobility of refugees residing  in  neighboring  countries.40  With  regard  to  linkages  to  the  drivers  of  growth,  the  amount  of  reconstruction funds directly affects public and private investment.     Refugee returns, meanwhile,  affect the size of the labor  force. We also build in variation in  human capital growth based on different assumptions for enrollment rates and vary the projections  for TFP growth across the three scenarios. We then look at a second broad case of forced repatriation  of refugees. Forced repatriation would contravene UN principles that care for the safety and welfare  of refugees,41 but it may be instigated by local and international voices eager for a quick resolution of  the refugee issue.42 Under forced repatriation, refugee returns are disconnected from the size of the  reconstruction  program,  and  instead,  a  “full”  rate  of  refugee  returns  is  assumed  across  the  three  scenarios.  We  thus  have  six  scenarios  in  total  –  three  for  each  of  the  two  different  broad  cases  (voluntary mobility and forced repatriation respectively). These are summarized in Table 5. We discuss  the projections for the growth drivers in detail in Section 4.1 below. Then, we present the resulting  simulations for Syria’s growth over the next 30 years in Section 4.2.  4.1 Prospective developments in growth drivers  4.1.1 Physical investment    The reconstruction and expansion of Syria’s physical capital will largely depend on the extent of  foreign funds made available since its self‐financing capacity is likely to be limited, especially in the  near  term.  We  use  Equation  (7),  a  slight  variant  of  the  saving/investment  ‐  balance  of  payments  accounting identity, to link    to the inflow of foreign funds.43              where ∆ (7)  (Note:  =  saving  excluding  net  current  transfers,  =  current  account  balance  excluding  net  current  transfers,  =  trade  balance,  =  income  balance, =  net  current  transfers, ∆   =  change  in  net  foreign liabilities.)                                                               40  World Bank (2019) finds that pull factors, of which two prominent ones are security and infrastructure, have  unambiguous effects on return behavior, in contrast to push factors in countries of asylum which have mixed implications.   41  The UN principle of non‐refoulement, codified in Article 33 of the 1951 UN Refugee Convention, requires that “no  contracting state shall expel or return a refugee in any manner whatsoever to the frontiers of territories where his life or  freedom would be threatened.”    42  The limitations that the UN non‐refoulement principle places on repatriation is frequently resented by states. Host  countries are often impatient to see uninvited refugees leave. Countries of origin are sometimes impatient to see them  return and signal the end of conflict. Moreover, donor states are eager to bring an end to the long‐term refugee assistance  programs that they fund.  43  The standard identity has net current transfers as a component of the current account balance.     24      In Equation (7), external financing may take the form of (non‐debt creating) aid and grants (which  enter as higher  )  or direct investment and loans (which result in an increase in foreign liabilities,  thus increasing  ∆ ).44 If the foreign funds lead to an equivalent amount being spent on tradables  (for example, the imports of capital goods), the current account  will be in deficit, ceteris paribus.45 If  the foreign funds do not lead to the purchase of tradables, the current account will be in balance,  ceteris paribus (see for instance Elbadawi, Kaltani and Schmidt‐Hebbel (2008) for related discussion  on how the utilization of aid monies affects  current account balances and exchange rates). In our  ∆ simulations,   where ∆   ∆ , the inflow of foreign funds, which leads to a  corresponding amount being spent on tradables. This gives us Equation (8).  ∆  varies across the  three post‐conflict scenarios as described in Table 5: beginning 2019, USD 12.5 billion per year over a  20‐year period under the optimistic scenario; USD 7 billion per year over a 20‐year period under the  baseline scenario; and USD 3 billion per year over a 10‐year period under the pessimistic scenario.   ∆                              (8)    Regarding    ,  if  we  assume  of  5  percent  (as  per  the  central  projection  of  our  conflict  simulation),  and    of  around  ‐30  percent  at  the  end  of  the  conflict,  this  would  give  us      of  approximately ‐25 percent.46  This is about 50 percentage points below Syria’s pre‐conflict long‐term  average:  23  percent  (20‐year  average  over  1991‐2010  based  on  WEO  data).  For  the  post‐conflict  pessimistic scenario, we calibrate the transition for Syria’s    by 50 percentage points to 25 percent,  in  eight  years,  based  on  the  experience  of  Lebanon.    Lebanon  was  subject  to  persistent  political  instability  during  its  recovery,  and  external  assistance  specific  to  its  reconstruction  program  was  limited, though it did receive large capital inflows attracted by high interest rates that enabled it to  run current account deficits.  For the optimistic scenario, we assume that Syria’s saving ratio increases  by 60 percentage points to 35 percent in five years, following the timeline and change experienced by  Kuwait as it recovered to its pre‐conflict saving‐to‐GDP level. Resource‐rich, high‐income Kuwait made  a strong recovery after the sharp decline as its oil production capacity was quickly restored amid a  comprehensive economic recovery and reconstruction program.47 For the moderate scenario, we take  an average of the projections for    under the other two scenarios.  See Appendix 5 for further details.     Once we have the projection for   for each scenario, we then split it between public and private  investment  by  continuing  to  assume  that  40  percent  of  total  investment  is  public,  following  pre‐ conflict trends. This is consistent with the estimated relative shares of destruction between public and  private  capital  during  the  conflict.  In  terms  of  the  efficiency  of  new  public  investment,  we  keep  efficiency unchanged at 0.570 under the pessimistic scenario and assume a rise from 0.570 to 0.734  by 2038 under the baseline. For the optimistic scenario, we assume a rise to the average IEI for the  upper‐middle‐income group of 0.769 by 2038.                                                                  44  This would be the “first leg” before the money is spent and both will have counter entries reflected as an increase in  foreign assets.   45  In either instance of financing spent on imports – transfers or loans ‐ the trade deficit is the same, though the current  account deficit including transfers, will differ.   46  This is based on the 2017 estimate of the trade balance share of GDP by ESCWA (2018).   47  See Sab (2014) for discussion on the conflict and recovery stories of Kuwait and Lebanon.   25    4.1.2 Demographics and the labor force  As shown in Graph 8, registered Syrian refugees in neighboring countries numbered 5,663,675 at the  end of 2018. A survey by UNCHR (2018) of a sample of 4,283 Syrian refugees in neighboring countries  indicated that 76 percent intend to return to Syria one day. In the optimistic scenario of the voluntary  mobility case, if we assume a 76 percent return rate, this would amount to 4.3 million returnees. We  use the UN population statistics (United Nations 2017), which projects an average population growth  of 2.5% over 2019‐2038 and builds in net migration into Syria of 4.21 million over 2020‐2035. Of this  total, 66 percent return over 2020‐2025 (about 556,000 on average per year), 29 percent over 2025‐ 2030 (about 240,000 per year) and 5 percent over 2030‐2035 (about 46,000 per year).48,49   For the moderate and pessimistic scenarios in the voluntary mobility case, we assume refugee  return rates of 43 percent and 18 percent respectively based on the ratio of reconstruction‐related  average funding per year in the first 10 years post conflict. For both these moderate and pessimistic  scenarios, we still assume a similar time path as the optimistic scenario – with refugee returns rising  to peak around 2023‐2024, and gradually moderating thereafter. Consequently, we obtain average  population growth rates of 2.2 percent and 1.9 percent respectively over 2019‐2038 for the moderate  and pessimistic scenarios. For all three scenarios in the forced repatriation case, we adjust the return  rate to 100 percent (but keep the pattern of returns unchanged over time), thus giving us an average  population growth rate of 2.7 percent over 2019‐2038.      We continue to draw on United Nations (2017) for projections of the working‐age population  share, and ILO (2018) for LFPR projections – the latter is available until 2030, and so we keep the values  thereafter stable at the 2030 level. The same working‐age population share and LFPR time series are  used across all scenarios.    4.1.3 Human capital  We  follow  the  same  approach  as  in  Section  3.1.3.  Changes  in  the  average  years  of  schooling  are  estimated based on shifts in the population (including the return of refugees) and improvements in  enrollment rates. Where we reduce (voluntary mobility case – moderate and pessimistic scenarios) or  increase refugee returns (forced repatriation case) compared to the UN statistics (voluntary mobility  case – optimistic scenario), we apportion the adjustment to different age groups based on the UNHCR  profile of the age distribution of refugees.      School  destruction  and/or  non‐functioning  schools  are  the  key  reasons  for  low  enrollment  rates (World Bank 2019). Further, displaced families will likely be hindered in their attempts to access  education services. Since the prospects for reconstruction are relatively weak under the pessimistic  scenario, we assume primary and secondary enrollment rates only return to pre‐conflict levels (93 and  67  percent  respectively)  by  2038,  while  the  tertiary  enrollment  rate  rises  to  50  percent  (from  36  percent) by that time. This timeline from the given initial levels is roughly in line with the trajectory of  estimations/projections of  enrollment  ratios  for developing countries  in  Barro and Lee  (2015) (see  their Chapter 3, Figure 3.5) and is longer than what Syria historically took to reach those rates.50 We                                                               48  In the UN population statistics, the net migration numbers are reported for the five‐year period between the middle of  the respective bracketing years (for example, from mid‐2020 to mid‐2025).   49  From our calculations based on average births and deaths in the UN population statistics, this is in addition to about  112,000 inward migrants from mid‐2019 to mid‐2020, which gives us a total of 4.3 million returnees, or 76 percent of  registered refugees residing in neighboring countries over 2019‐2035.   50  UNESCO UIS data for Syria indicate that the net primary enrollment rate rose from 81.9 percent in 1973 to 94.8 percent  in 1987 (approximately 15 years) while the net secondary enrollment rate increased from 39.3 percent in 2000 to 66.9  percent in 2010 (approximately 10 years).    26    further  assume  that  by  2048  enrollment  rates  reach  100,  80  and  60  percent  respectively  at  the  primary, secondary and tertiary levels. For the optimistic scenario, we assume primary and secondary  enrollment rates reach pre‐conflict levels in half the time, that is, by 2028, and by 2038, 100 percent  and 80 percent respectively. For the tertiary enrollment rate, we assume it rises to 50 percent by 2028  (also  in  half the time compared to the  pessimistic scenario)  and  60 percent  by  2038.  By  2048, we  assume enrollment rates reach 90 and 70 percent respectively at the secondary and tertiary levels.  With these calculations we obtain years of schooling of 8.449 and 6.991 respectively by 2038 in the  optimistic and pessimistic scenarios of the voluntary mobility case. The projected years of schooling  in 2038 under the pessimistic scenario is roughly the same as the pre‐conflict value of 7.080. For the  moderate scenario, we take an average of years of schooling under the other two scenarios, which  gives a value of 7.718 by 2038.     Using the above, we obtain an average annual growth rate in human capital of 1.3 percent over  2019‐2048 under the optimistic scenario (1.3 percent over 2019‐2028, 1.4 percent over 2029‐2038,  and 1.1 percent over 2039‐2048). For the pessimistic scenario, average annual growth amounts to 0.9  percent. That is, 0.4 percent over 2019‐2028, 1.0 percent over 2029‐2038, and 1.2 percent over 2039‐ 2048.51 The human capital growth rate under the moderate scenario is a simple average of the growth  rates under the other two scenarios. The average years of schooling and human capital growth rates  remain similar in the scenarios of the forced repatriation case compared to the voluntary mobility case  as there is little change in the population distribution by age groups.   4.1.4 TFP  Using  the  Kim  and  Loayza  (2019)  empirical  model,  we  assume  a  gradual  rebuilding  of  the  TFP  determinants index which was estimated at 15.98 in 2018. Under the optimistic scenario, we increase  this index to 35.42 by 2028 and 75.76 by 2048 based on the trajectory of the Republic of Korea’s index  over the 30‐year period, 1985‐2014. Korea is the best performer in the sample of countries used in  Kim  and  Loayza  (2019).  This  gives  average  annual  TFP  growth  of  1.4  percent  over  2019‐2048  (0.1  percent over 2019‐2028, 2.1 percent over 2029‐2038, and 2.0 percent over 2039‐2048).      For the pessimistic scenario, we repeat the exercise but based on  index of the United Arab  Emirates  (UAE),  the  best  performer  for  the  sub‐sample  of  MENA  countries.  This  would  imply  an  increase in Syria’s index to only 21.72 by 2028 (still below pre‐conflict level) and 32.74 by 2048.  The  corresponding average annual TFP growth over 2019‐2048 would be 0.3 percent (‐0.9 percent over  2019‐2028, 0.9 percent over 2029‐2048). For the moderate scenario, we assume TFP growth rates  that are the average of the rates under the optimistic and pessimistic scenarios.   4.2 Simulation of the post‐conflict growth outlook  For  the  simulations  in  this  section,  we  keep  the  constant  parameters  (labor  share  and  depreciation  rates)  unchanged  from  the  values  used  in  the  pre‐conflict  and  conflict  projections.  Default initial conditions as of 2018 (GDP level, GDP per capita level, and / ratios) are drawn from  the outcomes of the central projection in Section 3. Table 6 details the calibrations and projections of  growth  drivers  post  conflict  based  on  the  discussion  in  Section  4.1.  Under  the  baseline/moderate  scenario of the voluntary mobility case,  / averages about 43 percent and 39 percent over 2019‐ 2028 and 2029‐2038; and, at 23 percent and 25 percent, respectively, under the pessimistic scenario.  The investment shares are exceptionally high under the optimistic scenario, averaging 63 percent and  46  percent  respectively  in  the  first  and  second  decades;  but  are  closer  to  pre‐conflict  levels.  The                                                               51  The slightly faster human capital growth rate in 2039‐2048 compared with the optimistic scenario is mainly because of  there still being room for improvement in the primary enrollment rate in the pessimistic scenario.   27    investment‐to‐output ratios only differ slightly in the forced repatriation case, at most 1 percentage  point in the first decade.    4.2.1.  Post‐conflict GDP projections across the different scenarios under voluntary mobility  The results of our simulations for Syria’s post‐conflict GDP in the different scenarios under the  voluntary mobility case are shown in Graph 11(a). Under the baseline/moderate scenario, average  GDP growth is 8.4 percent over 2019‐2038. As can be observed from the top right of Graph 11(b), with  the  inflow  or  reconstruction  funds,  the  main  growth  driver  over  the  20‐year  period  is  capital  accumulation. As  /  reverts to something close to pre‐conflict trends especially after the 20‐year  annual inflow of reconstruction funds, the contribution from human capital growth and TFP are just  as relevant as physical capital growth.  In this scenario, Syria reaches its pre‐conflict GDP level by 2031,  and its pre‐conflict GDP per capita level by 2033, thus losing about two decades.       In the optimistic scenario, average GDP growth is 10.9 percent over 2019‐2038 reaching a peak  of 20.9 percent in 2022 with exceptionally high investments, and stronger contributions from other  growth drivers relative to the moderate scenario (see Graph 11(a) and Graph 11(b), bottom left). Even  with this strong performance, it would take Syria 9 years, that is by 2027, to surpass its 2010 GDP level  and  11  years  to  surpass  its  pre‐conflict  GDP  per  capita.  In  the  pessimistic  scenario,  GDP  growth  averages 5.1 percent across the next two decades, only slightly higher than pre‐conflict levels, amid  limited reconstruction funds from external sources (see Graph 11(b), bottom right, for the difference  in growth drivers vis‐à‐vis the moderate scenario). In this case, it would take Syria at least 22 years to  pass its pre‐conflict GDP level and almost 29 years to meet its GDP per capita level. This finding echoes  the simulation in World Bank (2017b) where under the assumption that the conflict ends in its sixth  year (2017), with investment recovering but remaining below its pre‐conflict level, Syria’s GDP remains  below its pre‐conflict level even 20 years after the conflict.      28    Table 6: Simulation for Post‐Conflict Syria – Projected Variables  Parameter/Variable*  Note  2018  Scenario  Optimistic  Moderate  Pessimistic  Average  2019‐ 2029‐ 2039‐ 2019‐ 2029‐ 2039‐ 2019‐ 2029‐ 2039‐ 2028  2038  2048  2028  2038  2048  2028  2038  2048  Public investment‐to‐output ratio  /   (1)       ‐ voluntary mobility case  0.010  0.251  0.185  0.140  0.170  0.157  0.120  0.091  0.100  0.100       ‐ forced repatriation case      0.250  0.183  0.140  0.168  0.155  0.120  0.088  0.100  0.100  Private investment‐to‐output ratio  /  (1)       ‐ voluntary mobility case  0.040  0.376  0.277  0.210  0.255  0.236  0.180  0.136  0.150  0.150       ‐ forced repatriation case      0.374  0.275  0.210  0.252  0.232  0.180  0.133  0.150  0.150  Efficiency of new public investment    (2)  0.570  0.625  0.724  0.769  0.615  0.697  0.734  0.570  0.570  0.570  Human capital growth    (3)    0.013  0.014  0.011  0.009  0.013  0.011  0.004  0.010  0.012  TFP growth    (4)  0.001  0.021  0.020  ‐0.003  0.015  0.014  ‐0.009  0.009  0.009  Population growth rate     (5)      ‐ voluntary mobility case  0.034  0.017  0.011  0.027  0.017  0.011  0.021  0.017  0.012      ‐ forced repatriation case    0.039  0.016  0.010  0.039  0.016  0.010  0.039  0.016  0.010  Working age‐to‐population share, growth      (6)  0.009  0.002  0.000  0.009  0.002  0.000  0.009  0.002  0.000  Labor force participation rate, growth    (7)  ‐0.001  0.000  0.000  ‐0.001  0.000  0.000  ‐0.001  0.000  0.000   *Multiply by 100 to obtain parameter/variable values in percent share or growth terms (%).   Notes (source and calculation):   (1) The 2018 values are from the central projection under conflict. Values under the scenarios are derived using the saving‐investment‐external financing accounting identity, based  on assumptions on reconstruction funds and the saving‐to‐output ratio. See Section 4.1.1.  (2) The 2018 value is from the central projection under conflict. A gradual rise from low‐income efficiency to lower‐middle‐income efficiency and upper‐middle‐middle efficiency are  assumed under the moderate and optimistic scenarios respectively. Efficiency is unchanged at the low‐income level under the pessimistic scenario.   (3) Average years of schooling by age groups are projected based on assumptions on enrollment rates (with the moderate scenario as a simple average of the optimistic and pessimistic  scenario). Overall average years of schooling is based on population projections across the scenarios. See Section 4.1.3.  (4) Based on the Kim and Loayza (2019) model, assuming the TFP overall determinants index follows a similar trajectory as that of Korea (over 1985‐2014) under the optimistic  scenario, and the UAE under the pessimistic scenario. TFP growth under the moderate scenario is a simple average of the values under the other two scenarios. See Section 4.1.4.   (5) The optimistic scenario under the voluntary mobility case is based directly on projections from United Nations (2017). For other scenarios, the UN projections are adjusted to  reflect higher or lower refugee returns. See Section 4.1.2.   (6) United Nations (2017).   (7) Based on ILO projections up to 2030 of the labor force participation rates, held constant thereafter.     29    Graph 11 (a): Post‐Conflict Simulation of GDP in Syria ‐ Scenarios under the Voluntary Mobility Case                                   30    Graph 11 (b): Impact of Different Growth Drivers under the Voluntary Mobility Case    4.2.2.  Comparing between voluntary mobility and forced repatriation  In Graph 12, we present the incremental/decremental growth and level effect for GDP and GDP per  capita of the forced repatriation case against the voluntary mobility case. On one hand, in terms of  GDP, the higher population growth from the forced repatriation contributes to higher GDP growth  rates, particularly over the time the influx of refugees is expected, and a progressively higher level of  GDP given these growth rates. This is a somewhat sanguine perspective, based on the assumption that  there  are  no  changes  to  other  factors  of  production,  particularly  demographic  ratios,  labor  force  participation rates, and human capital characteristics.     On the other hand (and most importantly), regarding GDP per capita, growth rates under the  forced  repatriation  case  are  lower  over  the  refugee  influx  period.  For  instance,  at  the  height  of  repatriation, in the moderate scenario, GDP per capita growth is lower by 1 percentage point. Growth  rates recover thereafter.  However, GDP per capita levels remain lower than the voluntary mobility  case  for  the  entire  period  under  our  review.  In  the  moderate  scenario,  GDP  per  capita  level  is  on  average lower by USD 76 (at 2010 constant prices) over 2019‐2048. This is because of lower physical  capital in per worker terms, which reduces labor productivity and output per capita relative to the  voluntary mobility case. Of all the scenarios, it is the optimistic case where forced repatriation is the  least adverse – as refugees already want to return given relatively good conditions for growth.   31    Graph 12: Post‐Conflict Simulation of GDP in Syria –   Incremental/Decremental Effect of Forced Repatriation versus Voluntary Mobility    4.2.3. How long would it take Syria to reach higher income group thresholds?  Prior to the conflict, Syria’s GNI per capita based on the World Bank Atlas Methodology (USD 1,840 as  of 2007) placed it in the lower‐middle‐income category, and at a level that was about half the then  upper‐middle‐income  threshold.  In  Graph  13,  we  illustrate  possible  trajectories  for  Syria’s  GNI  per  capita with conflict and thereafter across the three different post‐conflict rebuilding scenarios under  the voluntary mobility case. We also show what might have transpired in the absence of conflict.     At the tail‐end of the conflict (using 2018 as a reference point), Syria appears to have fallen  just  below  the  lower‐middle‐income  threshold.  While  once  again  surpassing  this  threshold  is  very  likely in the next few years, it would possibly take 18 and 26 years under the optimistic and moderate  scenarios  respectively  to  breach  the  upper‐middle‐income  threshold,  and  beyond  2050  for  the  pessimistic scenario.   In contrast, in the counterfactual of no conflict, Syria might have passed this  level in about six years, that is by 2024. This means that from 2010, while it could have taken Syria 14  years to become an upper‐middle‐income country, it may now take about double, or even triple that  time.           32    Graph 13: Distance to Higher Income Group Thresholds based on GNI Per Capita    5. Summary and Conclusion  In this paper, we use the Long‐Term Growth Model ‐ Public Capital Extension (LTGM‐PC) to answer  three questions pertaining to Syria’s economic growth in the aftermath of its civil conflict: What might  have been the counterfactual of no conflict? What was the impact of the conflict? And what are the  possible growth paths given different scenarios post‐conflict?    Our simulations of the impact of the conflict suggest an average GDP growth of ‐12% over  2011‐2018, with the level of GDP declining to almost one‐third the pre‐conflict level. Cumulatively,  the loss in GDP amounted to about USD 300 billion when the conflict simulations are compared against  the counterfactual. These results are broadly in line with findings in other studies. An added insight  from  what  we  do  is  that  we  identify  how  the  different  growth  drivers  comprising  physical  capital,  demographics and the labor force, human capital, and TFP might have contributed to the decline in  GDP. Our simulations indicate that the depletion of factors of production alone may account for about  87 percent of the negative GDP growth on average, and further, that about 64 percent of the average  negative growth is due to physical capital destruction. Demographics and labor may have accounted  for about 15 percent, human capital 7 percent, and TFP 13 percent of GDP growth on average over  the conflict years. This breakdown of factors is important as it sets the stage for the analysis of Syria’s  post conflict GDP potential which depends on the projected evolution of these growth drivers.   33    The post‐conflict outlook for the growth drivers in turn, depend on the political settlement  outcome which directly affects the availability of reconstruction funds and the voluntary mobility of  refugees.  Voluntary  mobility  would  not  only  be  preferable  on  humanitarian  grounds  but  also  on  economic terms. The political settlement scenario will also affect significantly human capital growth  and productivity growth.    Focusing  on  the  voluntary  mobility  case,  under  our  moderate  scenario  of  partial  political  settlement with strong guarantees for micro‐security and property rights, the average GDP growth  and GDP per capita growth are 8.4 percent and 6.1 percent respectively over 2019‐2038, assuming  that a final and stable resolution of the conflict is achieved. With the inflow of reconstruction funds  amounting to USD 140 billion spread over 20 years, the main growth driver over the 20‐year period is  physical  capital  accumulation  amid  investment‐to‐output  of  about  41  percent  on  average.  As  investment‐to‐output reverts to a lower level especially after the assumed 20‐year annual inflow of  reconstruction funds, the contributions from human capital and  TFP growth are just as relevant as  physical capital growth. In this scenario, Syria reaches its pre‐conflict/2010 GDP level by 2031, and its  pre‐conflict GDP per capita level by 2033, implying two “lost” decades from conflict.   Under  the  optimistic  scenario  of  robust  political  settlement,  with  exceptionally  high  investment‐to‐output of over 60 percent in the first decade, it would still take Syria about one decade  to  surpass  its  2010  GDP  and  GDP  per  capita  levels.  Under  the  pessimistic  scenario  of  limited  guarantees for micro‐security and property rights, reconstruction funds amounting to only USD 30  billion, and investment‐to‐output close to the pre‐conflict average, it would take Syria at least two  decades to meet its pre‐conflict GDP level and close to three to surpass its pre‐conflict GDP per capita.  Respectively for the optimistic and pessimistic scenarios, the average GDP per capita growth rate over  the next two decades is projected to be 8.2 or 3.1 percent.   While the reconstruction and expansion of physical infrastructure is essential, the importance  of strengthening human capital and the factors underlying TFP growth cannot be overstated. In this  paper,  we  have  only  accounted  for  population  and  enrollment  effects  on  human  capital  growth.  However, the quality of education and health (as illustrated in the World Bank’s human capital index)  will also likely be impeding factors that would have to be addressed in Syria’s quest for growth.         34    References  Barro,  R.J.  and  J.W.  Lee.  2013.  A  new  dataset  of  educational  attainment  in  the  world,  1950‐2010.  Journal of Development Economics 104: 184‐198.  Barro, R. and J.W. Lee. 2015. Education Matters: Global Schooling Gains from the 19th to the 21st  Century. New York, NY: Oxford University Press. Data source:  http://barrolee.com/data/oup_download_c.htm  Caselli,  F.  2005. Accounting  for Cross‐Country  Income Differences.  In Philippe Aghion  and  Stephen  Durlauf (eds.). Handbook of Economic Growth Volume 1A: 679 ‐741. Elsevier.   Ceylan,  E.S.  and  S.  Tumen.  2018.  Measuring  Economic  Destruction  in  Syria  From  Outer  Space:  https://theforum.erf.org.eg/2018/06/19/measuring‐economic‐destruction‐syria‐outer‐space/.    The  FORUM Policy Portal.  Economic Research Forum, Cairo, Egypt.    Chen,  S.,  N.  V.  Loayza,  and  M.  Reynal‐Querol.  2008.  The  Aftermath  of  Civil  War.  The  World  Bank  Economic Review 22 (1): 63–85.  Cohen,  D.  and  L.  Leker.  2014.  Health  and  Education:  Another  Look  with  the  Proper  Data.  CEPR  Discussion Papers No 9940. London: Centre for Economic Policy Research.  Cohen,  D.  and  M.  Soto.  2007.  Growth  and  Human  Capital:  Good  Data,  Good  Results,  Journal  of  Economic Growth 12(1): pp.51‐76.  Devadas, S. and S.M. Pennings. 2019. Assessing the Effect of Public Capital on Growth: An Extension  of the World Bank Long‐Term Growth Model. Journal of Infrastructure, Policy and Development 3(1):  22‐55.  Devarajan, S. and L. Mottaghi. 2017. Middle East and North Africa Economic Monitor, April 2017: The  Economics of Post‐Conflict Reconstruction in MENA. Washington, DC: World Bank.  Djankov, S.  and  M.  Reynal‐Querol.  2010. Poverty  and  Civil  War:  Revisiting  the  Evidence.  Review of  Economics and Statistics 92 (4): 1035‐1041.  Economic  Research  Forum  (ERF).  2019.  Repatriation  of  the  Arab  Conflict  Refugees:  Prospects  and  Potential Political and Economic Implications. FEMISE EUROMED Report (forthcoming): Cairo, Egypt.   Elbadawi  I.A.,  L.  Kaltani,  and  K.  Schmidt‐Hebbel.  2008.  Foreign  Aid,  the  Real  Exchange  Rate,  and  Economic Growth in the Aftermath of Civil Wars. The World Bank Economic Review 22(1):113‐140.  Gobat, J. and K. Kostial. 2016. Syria’s Conflict Economy. IMF Working Paper No. 16/13. Washington  DC: International Monetary Fund.   Guha‐Sapir, D., B. Schluter, J.M. Rodriguez‐Llanes, L. Lillywhite, and M. Hsiao‐Rei Hicks. 2018. Patterns  of Civilian and Child Deaths Due to War‐Related Violence in Syria: A Comparative Analysis from the  Violation Documentation Center Dataset, 2011‐16.  Lancet Global Health 6: 103‐110.  Hall, R. and C. Jones. 1999. Why Do Some Countries Produce So Much More Output Per Worker Than  Others? Quarterly Journal of Economics 114: pp 83 ‐116.   Hevia, C. and N. Loayza. 2012. Savings and Growth in Egypt. Middle East Development Journal. 4(1).  International Monetary Fund. 2010. Syrian Arab Republic: 2009 Article IV Consultation. IMF Country  Report No. 10/86. Washington DC: International Monetary Fund.  35    Jánossy, F. 1969. The End of the Economic Miracle: Appearance and Reality in Economic Development.  ISAP, New York.  Kim,  Y.E.  and  N.  Loayza.  2019.  Productivity  Growth:  Patterns  and  Determinants  Across  the  World.  Policy Research Working Paper No.WPS8852. Washington, D.C.: World Bank Group.   Lee, J.W. and H. Lee. 2016. Human Capital in the Long Run.   Journal of Development Economics 122:  147‐169. Data source: http://barrolee.com/Lee_Lee_LRdata.htm  Li, X., D. Li, H. Xu and C. Wu. 2017. Intercalibration between DMSP/OLS and VIIRS Night‐Time Light  Images to Evaluate City Light Dynamics of Syria’s Major Human Settlement during Syrian Civil War.  International Journal of Remote Sensing 38(21).   Milionis P. and T. Vonyo. 2015. Reconstruction Dynamics: The Impact of World War II on Post‐War  Economic Growth. American Economic Association: 2016 Annual Conference.   Milton, S. 2019. Syrian Higher Education during Conflict: Survival, Protection, and Regime Security.  International Journal of Educational Development 64: 38‐47.   Montenegro, C. E. and H. A. Patrinos. 2014. Comparable Estimates of Returns to Schooling around the  World. Policy Research Working Paper No. 7020. Washington, DC: World Bank.  Nasser,  R.  and  Z.  Mehchy.  2012.  Determinants  of  Labor  Force  Participation  in  Syria  (2001–2010),  Economic Research Forum Working Paper No. 698. July.      Organski, A.F.K., and J. Kugler. 1977. The Costs of Major Wars: The Phoenix Factor. American Political  Science Review 71(14):1347 –66.  ———. 1980. The War Ledger. Chicago, IL: University of Chicago Press.  Project on Middle East political Science (POMEPS). 2018. The Politics of Post‐Conflict Reconstruction.  POMEPS  Studies  30:  https://pomeps.org/wp‐content/uploads/2018/09/POMEPS_Studies_30_Post‐ Conflict_Web.pdf.  Psacharopoulos, G. 1994. Returns to Investment in Education: A Global Update.  World Development  22(9): 1325‐1343. Elsevier.  Sab, R. 2014. Economic Impact of Selected Conflicts in the Middle East: What Can We Learn from the  Past? IMF Working Paper No. 14/100. Washington: International Monetary Fund.  Smolny,  Werner.  2000.  Post‐war  Growth,  Productivity  Convergence  and  Reconstruction.    Oxford  Bulletin of Economics and Statistics, 62, 5: p. 589‐606.  Syrian  Center  for  Policy  Research  (SCPR).  2016.  Syria  Confronting  Fragmentation:  Impact  of  Syrian  Crisis Report 2015. February.  The  Conference  Board.  2017.  Total  Economy  Database™  ‐  Growth  Accounting  and  Total  Factor  Productivity, 1950‐2016. November. https://www.conference‐board.org/data/economydatabase/  United  Nations,  Department  of  Economic  and  Social  Affairs,  Population  Division.  2011.  World  Population Prospects: The 2010, Volume 1: Comprehensive Tables.  ———. 2013. World Population Prospects: The 2012, Volume 1: Comprehensive Tables.  36    ———.  2017.  World  Population  Prospects:  The  2017  Revision.  https://population.un.org/wpp/DataQuery/  United  Nations  Development  Programme  (UNDP)  and  United  Nations  High  Commissioner  for  Refugees  (UNHCR).  2019.  Regional  Strategic  Overview  –  Regional  Refugee  and  Resilience  Plan  in  Response to the Syria Crisis.   United Nations Economic and Social Commission for Western Asia (ESCWA). 2016. Syria at War – Five  Years On. United Nations.   ———.  2018.  Syria  at  War  –  Seven  Years  On  (The  Economic  Implications  of  the  Conflict  in  Syria).  Presentation  at  Consultation  Meeting  on  Policy  Alternatives  for  Post‐Conflict  Reconstruction  and  Economic Recovery. National Agenda for the Future of Syria. August.   Yahya, M. and J. Kassir. 2017. Coming Home? A Political Settlement in Syria Must Focus on Refugees.    Policy Outlook: Carnegie Middle East Center, March.  United  Nations  High Commissioner  for  Refugees  (UNHCR).  2018. Fourth  Regional  Survey  on Syrian  Refugees’ Perceptions and Intentions on Return to Syria (RPIS). July 2018.   Verme, P., C. Gigliarano, C. Wieser, K. Hedlund, M. Petzoldt, and M. Santacroce. 2016. The Welfare of  Syrian Refugees: Evidence from Jordan and Lebanon. Washington, DC: World Bank.  Vignal,  L.  2018.  Perspectives  on  the  Return  of  Syrian  Refugees.  In  Syrians  in  Displacement.  Forced  Migration Review Issue 57: February.   World Bank. 2017a. Syria Damage Assessment for Selected Cities Aleppo, Hama, Idlib. Washington DC:  World Bank.  ———.  2017b.  The  Toll  of  War  –  The  Economic  and  Social  Consequences  of  the  Conflict  in  Syria.  Washington DC: World Bank.  ———. 2019. The Mobility of Displaced Syrians – An Economic and Social Analysis. Washington DC:  World Bank.       37    APPENDIX 1‐ A MODEL OF LONG‐TERM GROWTH WITH PUBLIC CAPITAL  Underlying  the  simulations  in  this  paper  is  the  following  model,  reproduced  here  in  an  abridged  manner from Devadas and Pennings (2019). All the simulations are run using the Excel‐based toolkit  constructed based on this model.   4.1   The production function  We assume  a Cobb‐Douglas specification, where the public and private capital  stocks have unitary  elasticity of substitution. The following production function at time, :  ℎ                 (1)  Each firm takes technology (TFP),   and public services   as given, that is, these are externalities to  the firm.   is the private capital stock, ℎ  is effective labor, which can be further decomposed into  ℎ , human capital per worker and , the number of workers.  1  and   are private capital and  labor income shares. Next, we consider the following specification for public services  :                                 (2A)  is the efficient physical public capital stock – the public capital that is actually used in production.    captures whether public capital is subject to congestion (or not).   is the usefulness of public capital  (more technically the elasticity of output to efficient public capital).                                 (2B)  Due  to  corruption,  mismanagement  or  pork‐barreling,  only  a  fraction  1 of  measured  public  capital is useful for production. The measured capital stock  is what is recorded in international  statistical  databases,  constructed  using  the  perpetual  inventory  method.    is  the  average  efficiency/quality of the public capital  stock. Equations (1), (2A) and (2B) can  be written in a  more  conventional production function as:  ℎ                           (3)    Population and labor force growth    Equation (3) can be translated into per worker terms by dividing both sides by   :     ≡ ℎ                        (4)  where  is output per worker and   is private capital per worker and   is measured public capital  per worker (note the lower case).  , where   is total population,   is the working age‐ population ratio and   is the labor participation rate (labor force‐to‐working age population ratio).  The above equation can then be used to calculate growth rates of output per worker from  to  1:  =                 (5)  Equation (5) can be rewritten in terms of growth rates from  to  1:     38    1 , 1 Γ 1 , 1 , 1 , 1 , 1 ,                                                                                                                                                                                                 (6)  where the growth rate of a variable   from  to  1 is denoted by  , , and Γ is the growth rate of  the number of workers:  1 Γ 1 , 1 , 1 ,                  (7)  1 Γ  drops out from equation (6) in the congestion default  1 .  To  obtain  output  per  capita,    from  equation  (4),    ≡   .  Rewriting  this  equation in terms of growth rates:   1 , 1 , 1 , 1 ,                             (8)  To obtain output growth, we multiply (8) with population growth:   1 , 1 , 1 ,                                       (9)  4.2   Public and private capital accumulation, and changes in the efficiency/quality of public capital  The measured quantity of public capital (as in international statistical databases) accumulates  according to a standard capital accumulation identity, with the next period’s stock coming from the  previous  period’s  undepreciated  stock,  1   (where  is  the  public  capital  depreciation  rate) and new public investment,  .  1             (10)  The gross growth rate of measured public capital (not per worker) is:   / 1           (11)  The growth rate of measured public capital per worker, which enters equation (6), is:                              1 , ≡ /                      (12)    , , , The stock of efficiency‐adjusted public capital (which is actually used in production) evolves based on  the previous period’s efficiency‐adjusted undepreciated stock and efficiency‐adjusted new investment  .                      1 +                        (13A)    is  the  average  efficiency  of  existing  public  capital  (rather  than  the  efficiency  of  new  investment). Substituting  into Equation 13A and rearranging as 13B, one can see the    evolves  as  a  weighted  average  of  the  quality  of  existing  public  capital  ,  and  the  quality  of  new  investment  .  +                        (13B)  As such, the quality/efficiency of the stock of public capital  only changes when the quality of new  investment projects is different from that of the existing public capital stock:  . Using equation  (13B), the growth in quality which enters equation (6) can be written as follows:   39    1 , ≡ 1 /( / )                     (14)    The quantity of private capital follows the same accumulation process as public capital. But  with  as  the  private  capital  depreciation  rate,  and  as  private  investment.  The  growth  rate  of  private capital per worker is as follows:    1 ,                                      (15)    , , , 4.3  Analysis of the drivers of growth    To better understand and simplify the analysis of the drivers of growth, we take a log‐linear  approximation of equation (6). Specifically, equations (12), (14) and (15) are substituted into equation  (6).  Then,  taking  logs  and  using  the  approximation  ln 1   (for  small  g)  we  arrive  at  the  following:   ⁄ , , , , , 1 ,   ⁄ 1                                                            (16)  4.4 Implementation  The  future growth rates of the  labor participation rate  (  , ), the working age‐population  ratio ( , ), population ( , ) and pure TFP ( , ), are exogenously determined. The growth rate  of measured public capital per worker ( , )  which is given by equation (12), using the growth  rate  of  the  public  capital  stock  (equation  (11))  as  an  intermediate  step.  Private  capital  per  worker  growth ( , ) is as given by equation (15). The growth rate of the efficiency of public capital ( , )  as  given  by  equation  (14)  using  the  growth  rate  of  the  public  capital  stock  (equation  (11))  as  an  intermediate step.  Finally,  the  model  is  closed by  updating public  capital‐to‐output  using equation (17)  and  the  private capital‐to‐output ratio using equation (18) (with the growth rates in per‐worker terms):  ,                                         (17)  , ,                                         (18)  ,               40    APPENDIX 2 ‐ ESTIMATED PHYSICAL CAPITAL DAMAGE DURING CONFLICT  USD billion,  % 2010 35 35 30 30 25 25 20 20 15 15 10 10 5 5 0 0 2011 2012 2013 2014 2015 2016 2017 2018* Physical capital damage (USD billion, 2010) Conflict intensity index (%)   Notes:     Data on physical damage and conflict intensity from 2011 to 2017 is sourced from ESCWA (2018). The  conflict  intensity  index  is  based  on  damage  reports;  estimated  number  of  casualties;  and  the  geographical size and spread of military operations as well as the volume of assets deployed, and the  intensity of weapons used.   *Authors’ calculations where the conflict intensity index is assumed to take the 2017 value of 25 in  the first quarter and an average of the 2011 and 2012 index values (2 and 6 respectively) in the other  three quarters. Given the calculated average conflict intensity index for 2018 of 9.25, we then derive  an estimate of physical damage that is proportionate to the 2017 estimate of physical damage of USD  18.5 billion.                           41    APPENDIX 3 ‐ PROJECTION OF AVERAGE YEARS OF SCHOOLING DURING CONFLICT          42    APPENDIX 4 ‐ PROJECTION OF TFP GROWTH DURING CONFLICT  Table A4.1 Variables Used in Projections of  Selected Determinant Sub‐Indices of the TFP Overall Determinant Index  Infrastructure Education Governance  Political  stability and  Voice and  Control of  Government  absence of  Regulatory  Variable/ Relative total  Schooling  accountability corruption effectiveness violence quality Rule of law Year city light (%)1 (average years)2 (composite indicator)3 2010 100 7.08 ‐1.70 ‐1.13 ‐0.61 ‐0.81 ‐0.90 ‐0.59 2011 70 7.08 ‐1.81 ‐1.09 ‐0.50 ‐2.01 ‐0.95 ‐0.78 2012 40 7.08 ‐1.84 ‐1.21 ‐1.15 ‐2.68 ‐1.53 ‐1.13 2013 33 6.65 ‐1.83 ‐1.26 ‐1.40 ‐2.68 ‐1.55 ‐1.41 2014 22 6.21 ‐1.88 ‐1.55 ‐1.40 ‐2.75 ‐1.69 ‐1.36 2015 18 5.78 ‐1.92 ‐1.55 ‐1.59 ‐2.97 ‐1.64 ‐1.42 2016 21 5.71 ‐1.99 ‐1.57 ‐1.83 ‐2.92 ‐1.67 ‐1.99 2017 21 5.65 ‐1.97 ‐1.56 ‐1.79 ‐2.63 ‐1.83 ‐2.09 2018 21 5.58 ‐1.97 ‐1.56 ‐1.79 ‐2.63 ‐1.83 ‐2.09 1  Data from Li et al. (2017) is for March of each of year (2010‐2016) and January 2017 and pertains to Syria as a whole. We assume  that the 2017 level continues to prevail in 2018.  2  Authors' calculations.  3  Worldwide Governance Indicators (WGI) (Kaufmann, Kraay and Mastruzzi (2010), updated to 2017). We assume the 2017 values  continue to prevail in 2018.    Table A4.2 TFP Overall Determinant Index and Its Sub‐Indices        43    APPENDIX 4 ‐ PROJECTION OF TFP GROWTH DURING CONFLICT (CONT’D)  Graph A4.1:  Projected TFP Growth for Syria based on the TFP Overall Determinant Index    Notes:   TFP growth is calculated using the following equation based on Kim and Loayza (2019), where  is the TFP Overall Determinant Index, and  is the level of   in index form, with 2011  normalized to one.   ℎ , 0.180 0.050 ln ln 0.099 ln ln   For the initial year, 2011, a 15‐year average, [(t‐1) ‐ (t‐16)]/15, is used for the change in the index  and TFP level.                                 44    APPENDIX 5 – POST‐CONFLICT SCENARIOS FOR SYRIA’S SAVING‐TO‐OUTPUT  We refer to the experiences of Kuwait and Lebanon to establish the transition path for Syria’s saving‐ to‐output, S/Y in terms of the change and time to recovery. Kuwait is used to guide the optimistic  scenario and Lebanon, the pessimistic scenario. The moderate scenario reflects the average of the  values under the other two scenarios.   Graph A5.1: Saving‐to‐Output in Kuwait and Lebanon  S/Y, % 50.0 40.0 30.0 20.0 10.0 0.0 ‐10.0 ‐20.0 ‐30.0 ‐40.0 ‐50.0 ‐60.0 Post war, t+1 2 years pre‐war Last war year, t 1 year pre‐war t+2 t+3 t+4 t+5 t+6 t+7 t+8 t+9 t+10 Kuwait Lebanon Source: Authors' calculations based on Sab (2014) and World Development  Indicators (WDI).  S/Y= GFCF/Y + (X/Y‐M/Y), where GFCF = gross fixed capital formation, X=exports of  goods and services and M = imports of goods and services. We focus on the domestic  savings ratio rather than the gross national savings ratio to better gauge domestic  production capacity.  Square markers are where we assume each country's S/Y recovers to its pre‐war  level.    Graph A5.2: Saving‐to‐Output Scenarios for Syria    45