220 likes | 435 Vues
Analyzing the poverty and distributional impacts of macro shocks in developing countries A Microsimulation Approach. Poverty Reduction and Equity Group Poverty Reduction and Economic Management Network World Bank April 2011. Outline. Background Approach Model
E N D
Analyzing the poverty and distributional impacts of macro shocks in developing countriesA Microsimulation Approach Poverty Reduction and Equity Group Poverty Reduction and Economic Management Network World Bank April 2011
Outline • Background • Approach • Model • Basic structure and data requirements • Methodology • Standard outputs • Limitations and assumptions • Illustrative results • Conclusions • Applications to date • Moving forward
Background In wake of financial crisis: • Urgency to design policy to mitigate impact of crisis but… • … lack of real time micro data on employment and distributional impacts • High demand by client countries and regional teams for: • Projections of short-term and medium-term impacts • (Ex-ante) Identification of most vulnerable and affected groups • Microsimulation model (FY09-FY10) • Development by PRMPR • Application to Bangladesh, Mexico, Mongolia, Philippines and Poland • (Partial application to Egypt) Following crisis: • Broader demand for microsimulation model to project employment and distributional impact of macro shocks • Collaboration with ADePT to design and develop simulation module
Approach • What is needed? • To account for multiple transmission mechanisms • To capture impacts: • Over the entire income/consumption distribution • At the individual and household levels • Proposed approach • Micro-simulation model • Similar to methods outlined in Bourguignon et al. (2008) • Macroeconomic projections (not CGE) • Microeconomic data from household/LF surveys • Focus on • Labor markets (employment and earnings) • Non-labor income (remittances) • (Prices – food/non-food, other) • Main outputs • Individual level: Information on LF/employment status and labor earnings • Household level: Information on per capita (labor/non labor) income and consumption • Results • Employment impacts • Poverty and distributional impacts
Basic structure Predictions (t+1…t+n) Baseline (Pre-shock, t) Macro projections (t+1…t+n) “Treatment” (Crisis) Impact (1) Micro data (Household/LF survey) Impact (2) Benchmark
Data requirements • Control variables • Macroeconomic projections (country teams) • Total and sectoral GDP growth • International remittances growth • Inflation (general and food) • Exchange rate (local currency/USD) • Macroeconomic parameters (calculations from historical data) • Output elasticity of employment (total and sectoral) • Output elasticity of labor force participation (total and sectoral) • Population growth rate (by age group and gender) • Microeconomic data • Survey with data on: • Individual and household-level labor and non-labor income • Individual-level labor market outcomes • Poverty line(s)
Model overview Baseline (Calibration) Simulation Assessment of impacts Micro data Macro projections Price data Input adjust predict estimate LF status modelEarnings equationMigration/remittances ∆ in LF status (ind) ∆ real earnings (ind) ∆ remittances (HH) Population Income and consumption (individuals and HH) Output Results Population growth Income/consumption distributions Poverty and inequality measures Rule: Replicate macro proportional changes at micro level Rule: Best fit to micro data
Typical outputs • Datasets • Original (baseline) dataset • Simulated dataset(s) • Results • Employment and earnings estimates • Poverty and inequality aggregates • Poverty profiling • Poverty profiling of “new poor” • Poverty profiling for specific groups/areas • Distributional analysis • Growth incidence analysis • Transition matrices
Limitations and assumptions Limitations • Level of disaggregation • Level of sectoral/regional disaggregation dependent on available level of disaggregation for macro projections • Labor market dynamics • Structural relationships remain constant • Validity of MNL and earnings (pre-crisis) estimates • No explicit modeling of changes in composition of employment within sectors • Changes in the proportion of formal/informal employment within each sector are due to individual transitions Assumptions • Immobility of factors of production (including labor) between regions and between urban/rural areas • No (additional) internal or external migration • Labor income and profits grow at same rate within each sector • Constant marginal propensity to consume • Needed only if poverty is consumption-based
Summary of main findings(on impact of financial crisis) In Bangladesh, Mexico, Philippines and Poland, slower or negative GDP growth generally translated into: • Aggregate employment and poverty impacts • Employment rates/levels fell due to the crisis… • … with sectoral impacts varying by country • Incomes were affected through both labor earnings and remittances… • … although the mix varied across countries • Poverty was expected to increase… • … but there was little impact on aggregate measures of inequality • Poverty profiling: Crisis vulnerable • New poor had specific characteristics that distinguished them from chronic poor and non-poor • Implications for design and targeting of safety net and crisis response packages • Distributional impacts • Large impacts in middle of income distribution, associated with employment/earnings shocks • Implications for political economy of crisis response
Changes in total and sectoral employment • 2010 • Decline in employment levels (shown) and employment rates • Significant cross-country variation in sectoral impacts • Manufacturing affected in all countries but to different degrees • Services as fall-back sector in BD, in contrast with Philippines, Mexico, and especially Poland
Changes in household income • 2010 • Decline in household income associated with reductions in both labor and non-labor incomes • Significant cross-country variation in relative importance of shocks to different income sources, reflecting variation in relative importance of different transmission mechanisms • Distributional implications
Changes in poverty and inequality • 2010 • Increase in level and depth of poverty, especially in MX… • … but little impact on aggregate inequality • Caution: Impact in relative (%) versus absolute (PP) terms (3.4) (3.9) (0.7) (0.3) (1.5) (1.2) (0.2) (0.4) (0.012) (0.001) (-0.004) (-0.002)
Characteristics of the crisis-vulnerable (I) • 2010 • HH of new poor households relatively more skilled than those chronic poor households… • … but less skilled than the average HH • Similar pattern across countries
Characteristics of the crisis-vulnerable (II) • 2010 • New poor relatively more likely to reside in urban areas than chronic poor… • … but less likely to reside in urban areas than the average household • Similar pattern across countries, but differences more acute in Philippines and MX (higher level of urbanization? higher rate of poverty reduction in urban areas in recent years?) and no differences in Poland (higher level of penetration of off-farm activities in rural areas)
Distributional impacts:Growth Incidence Curves • On average larger losses in urban areas, where employment impacts are highest • But important distributional impacts within rural and urban areas as well (except for Philippines) • BD: Remittances • MX: Most severe employment losses • PO: Farm versus non-farm employment (rural areas) Philippines Bangladesh Mexico Poland
Distributional impacts:Transition Matrices • Most people remain within decile but • Important movements up and esp. down – particularly in middle of distribution • MX: Largest impact on poor • PO: Largest movements “up” at bottom due to UI Philippines Bangladesh Mexico Poland
Conclusions (I):Main applications to date • Crisis monitoring: • Identification of main transmission mechanisms • Identification of possible “leading indicators” to monitor the likely poverty impact of an economic crisis • E.g. manufacturing employment/job losses, wages, aggregate remittance flows, change in relative price of food • Design of policy response: • Identification of affected/vulnerable individuals/households/groups • SP for the new poor? • Effectiveness of traditional/existing SN programs and automatic stabilizers • Political economy implications of distributional impacts (esp. impacts on urban middle class) • Poverty/Distributional impact of policy response: • Safety nets versus automatic stabilizers (e.g. Mexico and Poland)
Conclusions (II):Main applications to date (publications) • (Stand alone) Country notes • Bangladesh (PRWP 5238) • Philippines (PRWP 5286) • Mongolia (Policy Note, 2010) • Poland (Policy Brief, 2010) • Contributions to country reports • Mexico • Philippines • Overview and summary of findings • Economic Premise (March 2010) • Summary piece in EUVox.org (April 2010) • Gender impacts • Distributional Impact of the Financial Crisis (forthcoming Fall 2011) • Webpage • Poverty Reduction and Equity Group (PRMPR)
Conclusions (III):Moving forward • Possible extensions, depending on the country context • Commodity/food price changes • More disaggregated treatment of sectors • Export/non-export • Formal/informal • More sophisticated treatment of remittances and internal migration • Timeline • Launch of ADePT SIM 1.0 over summer 2011 • Basic features • Manual • Dissemination and training • BBLs • Training at HQ + customized country/regional training events on demand (e.g. Egypt, Mexico, Tunisia) • (Possible customization on demand)