Measuring and Mapping the Welfare Effects of Natural Disasters A Pilot

Transcription

1 Measuring and Mapping the Welfare Effects of Natural Disasters A Pilot Luc Christiaensen,, World Bank, presentation at the Managing Vulnerability in East Asia workshop, Bangkok, June 25-26, 26, 2008

2 Key Policy Questions in Natural Disaster Risk Management What are the economic losses associated with different natural disasters (droughts, floods, typhoons, landslides)? Planning for NDRM investments (ex ante and ex post) Which areas and which people are more likely to suffer from natural disasters? Targeting How long do the effects of natural disasters last? Linking relief and development

3 Damage assessments not designed to address these policy questions Economic impact of natural disasters typically assessed in aggregate terms focused on lives lost & damages to property E.g. EM-DAT from the CRED; ECLAC (for LAC countries); insurance companies like Munich Re and Swiss Re. Limitations: Backward looking, i.e. for compensation, not for planning Indirect (flow of goods and services, unemployment) and second round effects (macro-economic; disease) poorly captured; Analyses do not account for household coping behavior Smaller, more localized events go unnoticed.

4 Piloted tools to strengthen the empirical base for natural disaster risk management and planning National Hazard Map derived from existing meteorological data Extending the international hotspots study to the national level Constructing statistical distributions permits simulating the effects fects of climate change Micro-econometric assessment of the economic impact of natural disasters Linking the hazard map with socio-economic household surveys to estimate the effect of past and current events on household welfare Include both large covariate as well as smaller localized events. Vulnerability map Mapping at a disaggregated level the likelihood that households will become poor Extending small area estimation techniques from poverty mapping into vulnerability mapping

5 Application to Vietnam Highly Relevant Vietnam defined as highly vulnerable Hotspot country (i.e., vulnerable to two or more natural hazards) in recent World Bank study. Increasing frequency of and aggregate damages due to floods and windstorms in compared to previous decades (ADRC 2002). Feasible Meteorological data (about 200 rainfall stations, more than 200 flood gauges, storm tracks, agro-meteorological models) 3 year socio-economic panel at the community level (VHLSS 2002, 2004, 2006) Agricultural census permits disaggregated vulnerability maps Pilot can be extended to other countries (Philippines, Indonesia)

6 Hazard Map - Challenges Focus on three hazards drought, floods and cyclones Methodology - two steps Extrapolate meteorological data from a limited set of measure points to the surface E.g. how to convert water levels in a river to areas flooded Define hazards based on this data what constitutes a shock when is little too little - historic averages vs.. critical thresholds Agro-meteorological models for droughts Validation based on historical records of droughts, floods, cyclones EM-DAT, Disasters reported in VN, DRM spending data Anticipated output

7 Drought and flood hazard map from Sri Lanka

8 Estimating the welfare effects of natural hazards- methodology and challenges Y β ) δ + λ + η + = X 1 + S γ + ( S X 1 ij j e i, j, t = subscripts denoting household, location and time respectively Y = welfare outcome ( (e.g., consumption, income) X = vector of (observable) time-varying household characteristics, S = vector of natural hazards events, S X is a vector of interactions of household characteristics with natural hazard events λ, η = time-invariant invariant (unobservable) household and location effect, e is a stochastic error term. 2-yr panel allows estimation of future welfare given current conditions. 3-yr panel allows control for time invariant community characteristics tics

9 Vulnerability Map Methodology and Challenges V = prob (Y < z) with z a poverty threshold Y = X + S γ + 1β e From the welfare estimations we know the coefficients β, From the hazard maps we know the statistical distribution of each S From the agricultural census we know X-1 We can predict V at a disaggregated level using small area estimation techniques and generate vulnerability maps using actual info on the distribution of shocks λ

10 Concluding Remarks More comprehensive estimates of the economic loss associated with natural disasters Strengthens the empirical base for NDRM (better planning and targeting) Possible extensions Replication in other countries Simulation of the effects of climate change