Skip to content

processing_outline.md

Latest commit

 

History

History
171 lines (144 loc) · 6.55 KB

processing_outline.md

File metadata and controls

171 lines (144 loc) · 6.55 KB

Burke, Hsiang, and Miguel 2015 Replication - Processing Steps Outline

Overview

This document outlines the main processing steps for replicating the Burke, Hsiang, and Miguel 2015 paper "Global non-linear effect of temperature on economic production." The analysis examines the relationship between temperature and economic growth using historical data and projects future impacts under climate change scenarios.

Major Processing Steps

1. Data Preparation and Initial Analysis (Stata)

Files: GenerateFigure2Data.do, GenerateBootstrapData.do

1.1 Baseline Regression Analysis

  • Input: GrowthClimateDataset.dta (main dataset with temperature, precipitation, GDP growth data)
  • Process: 1.1.1. Run baseline quadratic temperature response regression 1.1.2. Estimate global response function with temperature and temperature squared 1.1.3. Generate marginal effects and confidence intervals
  • Output:
    • estimatedGlobalResponse.csv (response function data)
    • estimatedCoefficients.csv (regression coefficients)
    • mainDataset.csv (cleaned dataset)

1.2 Heterogeneity Analysis

  • Process: 1.2.1. Analyze rich vs poor country responses (GDP percentile < 50) 1.2.2. Analyze agricultural vs non-agricultural GDP growth 1.2.3. Analyze early vs late period responses (pre/post 1990)
  • Output:
    • EffectHeterogeneity.csv (rich/poor, agricultural responses)
    • EffectHeterogeneityOverTime.csv (temporal heterogeneity)

1.3 Bootstrap Analysis

  • Process: 1.3.1. Bootstrap regression coefficients (1000 replicates) 1.3.2. Sample countries with replacement 1.3.3. Run multiple model specifications:
    • Pooled model (no lags)
    • Rich/poor model (no lags)
    • Pooled model (5 lags)
    • Rich/poor model (5 lags)
  • Output:
    • bootstrap_noLag.csv
    • bootstrap_richpoor.csv
    • bootstrap_5Lag.csv
    • bootstrap_richpoor_5lag.csv

2. Climate Projections (R)

Files: getTemperatureChange.R

2.1 Temperature Change Calculations

  • Input:
    • CMIP5 RCP8.5 ensemble mean temperature data
    • Population data (Gridded Population of the World)
    • Country shapefiles
  • Process: 2.1.1. Calculate population-weighted country-specific temperature changes 2.1.2. Generate conversion factors from global to country-level temperature changes 2.1.3. Project temperature changes for 2080-2100 relative to 1986-2005 baseline
  • Output: CountryTempChange_RCP85.csv

3. Socioeconomic Scenarios (R)

Files: ComputeMainProjections.R (first part)

3.1 Population and Growth Projections

  • Input:
    • SSP (Shared Socioeconomic Pathways) data
    • UN population projections
    • Historical growth rates (1980-2010 baseline)
  • Process: 3.1.1. Interpolate 5-year SSP projections to annual data 3.1.2. Create baseline scenario with historical growth rates 3.1.3. Process SSP scenarios 1-5
  • Output:
    • popProjections.Rdata
    • growthProjections.Rdata

4. Impact Projections (R)

Files: ComputeMainProjections.R (main projection section)

4.1 Future Impact Calculations

  • Input:
    • Bootstrap regression coefficients
    • Population and growth projections
    • Temperature change projections
  • Process: 4.1.1. Project GDP per capita with and without climate change (2010-2099) 4.1.2. Apply four regression models:
    • Pooled model (no lags)
    • Rich/poor model (no lags)
    • Pooled model (5 lags)
    • Rich/poor model (5 lags) 4.1.3. Calculate global averages and totals
  • Output:
    • GDPcapCC_*_*.Rdata (GDP per capita with climate change)
    • GDPcapNoCC_*_*.Rdata (GDP per capita without climate change)
    • GlobalChanges_*_*.Rdata (global summary statistics)

5. Damage Function (R)

Files: ComputeDamageFunction.R

5.1 Damage Function Construction

  • Input:
    • Impact projections from Step 4
    • IAM (Integrated Assessment Model) temperature scenarios
  • Process: 5.1.1. Calculate damages for different global temperature increases (0.8°C to 6°C) 5.1.2. Match to IAM temperature scenarios (DICE, FUND, PAGE) 5.1.3. Generate damage functions for all model specifications
  • Output:
    • DamageFunction_*.Rdata (damage functions by model)

6. Figure Generation (R)

Files: MakeFigure*.R, MakeExtendedDataFigure*.R

6.1 Visualization

  • Input: All output data from previous steps
  • Process: 6.1.1. Generate main figures (2-5) 6.1.2. Generate extended data figures 6.1.3. Create tables and supplementary materials
  • Output: PDF figures and tables

7. Data Dependencies

7.1 Input Data Sources

  • Main Dataset: GrowthClimateDataset.dta - Historical temperature, precipitation, GDP data
  • Climate Projections: CMIP5 RCP8.5 ensemble mean data
  • Population Data: Gridded Population of the World, UN projections
  • Socioeconomic Scenarios: SSP database (population and growth projections)
  • Country Boundaries: ESRI shapefiles
  • IAM Data: Processed Kopp data for damage function comparison

7.2 Key Intermediate Files

  • Bootstrap coefficient files (4 models × 1000 replicates)
  • Temperature change projections
  • Population and growth scenario data
  • Impact projection arrays (country × year × bootstrap)

8. Model Specifications

8.1 Regression Models

  1. Pooled Model (No Lags): Quadratic temperature response, country and year fixed effects
  2. Rich/Poor Model (No Lags): Separate quadratic responses for rich vs poor countries
  3. Pooled Model (5 Lags): Distributed lag model with 5-year temperature lags
  4. Rich/Poor Model (5 Lags): Separate distributed lag responses for rich vs poor countries

8.2 Climate Scenarios

  • RCP8.5: High emissions scenario used for temperature projections
  • Temperature Range: 0.8°C to 6°C above pre-industrial levels

8.3 Socioeconomic Scenarios

  • Baseline: Historical growth rates continued
  • SSP1-5: Shared Socioeconomic Pathways scenarios

9. Implementation Notes

9.1 Key Assumptions

  • Temperature response constrained at 30°C (out-of-sample protection)
  • Countries can transition between rich/poor categories based on future income
  • Population-weighted temperature changes
  • Linear interpolation between 5-year SSP projections

9.2 Computational Requirements

  • Large arrays for bootstrap analysis (1000 replicates)
  • Memory-intensive projections (country × year × bootstrap)
  • Parallel processing potential for bootstrap loops

9.3 Quality Checks

  • Consistency between file reading and writing operations
  • Bootstrap convergence checks
  • Out-of-sample temperature constraints
  • Population weighting validation