Overview

The CliC:ME project uses diverse data sources to generate a state-of-the-art representation of the world economy in the reference year 2010. All inputs for our computations are publicly available under open access license and are shared by leaders in providing competent socio-demographic, economic and climatic data at national and global levels.

Population

The main source of data in this step is WorldPop.org. We collect the data on population counts (constrained UN adjusted individual countries at 0.1 km resolution) for 2020, from which we impute 2010 using the 2010 and 2020 global population counts (unconstrained global mosaics at 1 km resolution) (Bondarenko et al., 2020; WorldPop, 2019). Note that all these files are available for 5-year groups and by gender.

Gross Domestic Product

Our main data source is the 1x1 km raster of GDP values provided by Kummu et al. (2018) for 2010. This data is then aggregated at an administrative unit level and used for the computation of Total Factor Productivity (TFP).

Urbanization

First, we collect the pixel data on global built-settlement growth provided by WorldPop.org at 0.1 km resolution (Nieves et al., 2020). We then collect shares of sectoral employment from censuses (IPUMS international and Eurostat) for 1,400 administrative units. From the World Bank, we collect the shares of employment and GDP by sector for all the countries in our sample. All these inputs allow us to impute the sectors that are present in each pixel of the world (either agriculture, industry or services).

Education shares

We compile education data from various sources. First, we collect the country-specific distributions of education attainments from the Barro-Lee database (Barro and Lee, 2013). We then complement this dataset with micro data from IPUMS international, Labor Force Surveys (LFS) and national censuses, aggregated at the administrative unit reflecting the distributions of education across sectors. The latter is complemented with the education distribution database from the International Labour Organization (ILO). Lastly, we use the World Bank WDI for GDP per capita, and its subsequent imputation for the purpose of imputing missing data points for some administrative units in Africa, Asia and Oceania. Note that all the data inputs are disaggregated by four age groups (0-20, 20-40, 40-60 and 60-80) and two gender groups.

Damage Functions (Labour Productivity)

We rely on downscaled climate data provided by WorldClim.org. Our reference source is the historical climate data (Fick and Hijmans, 2017) at 2.5 arc minutes (5 km on the Equator), covering the last three decades of observations. Additionally, we access future climate projections that have been compiled for CMIP6. Our scenario is an average of 22 models (ACCESS-CM2, ACCESS-ESM1-5, BCC-CSM2-MR, CanESM5, CanESM5-CanOE, CMCC-ESM2, CNRM-CM6-1, CNRM-CM6-1-HR, CNRM-ESM2-1, EC-Earth3-Veg, EC-Earth3-Veg-LR, GISS-E2-1-G, GISS-E2-1-H, INM-CM4-8, INM-CM5-0, IPSL-CM6A-LR, MIROC-ES2L, MIROC6, MPI-ESM1-2-HR, MPI-ESM1-2-LR, MRI-ESM2-0 and UKESM1-0-LL), four periods: 2021-2040, 2041-2060, 2061-2080 and 2081-2100 (representing 2030, 2050, 2070 and 2090, respectively) and four RCP-SSP scenarios (RCP2.6-SSP1, RCP4.5-SSP2, RCP7.0-SSP3 and RCP8.5-SSP5). All files are downloaded at 5x5 km resolution and include ‘bioclimatic’ variables; that is, Annual Mean Temperature, Temperature Seasonality (standard deviation ×100), Max Temperature of Warmest Month, Min Temperature of Coldest Month, Mean Temperature of Warmest Quarter and Mean Temperature of Coldest Quarter. These factors allow us to construct scenario and pixel specific distributions of annual temperatures. We then refer to the ILO as our source for data on workers’ productivity losses due to heat exposure (Kjellstrom et al., 2018).

Agriculture Productivity

The key data source is the gaez.fao.org web page, published by The Food and Agriculture Organization of the United Nations (FAO) and the International Institute for Applied Systems Analysis (IIASA). We collect three types of data: yields of each crop in 2010 by pixel, percent of land irrigated in 2010 by pixel and production of each crop type in all future climate scenarios by pixel. We consider the 14 most popular crops in the world: bananas, barley, cassava, maize, potatoes, rapeseed, sunflower, soybean, sugar beet, sugarcane, sorghum, rye, wheat and yam. We also collect the data on 2010 crop prices from FAO, UN and we compute average world prices of each crop as a sum of country inputs.

Prices and International Trade

For the price and purchasing power parity data, we use the International Comparison Program (ICP) 2011 — part of the World Development Indicators database from the World Bank. Trade data originates from the WITS database run by the UN, WTO and the World Bank. We also access the data on sectoral consumption and production aggregates from the World Bank.

Utilities by Pixels

For the elasticities of utility levels to wage rates, we use the data on dyadic 20-year flows of international migrants by individual type in 2010, and we run gravity regressions to identify the wanted parameters by socio-demographic groups. We use several gravity variables from Conte et al. (2022) by CEPII complemented with distances determined from our maps. Lastly, we use extensive data on labour informality from the ILO.

Migration

Our main data source for international migration flows is the dyadic numbers of international migrants by education, age and gender in 2010 from the Database on Immigrants in OECD and non-OECD Countries (DIOC). This data is then complemented with migration data from the United Nations Population Division (UNDP) and data compiled by Abel and Cohen (2019). For internal migration, we use censuses published by respective national statistical offices and IPUMS International to generate country-specific flows of across and within admin-unit migrants.

Sea-Level Rise

First, we use the elevation map of the world at a 1x1 km grid downloaded from SEDAC NASA. The second source includes data on sea-level rise and storm surge, which are probabilistic projections for all times and scenarios in the future, compiled by Tebaldi et al. (2021). In particular, we use the data provided by Kirezci et al. (2020).

SPEI data

The main climate data source for this step is WorldClim.org, from which we collect three types of data: (1) Historical monthly weather data, containing information on minimal and maximal temperatures as well as precipitation ranging from 1960-2010. This is the CRU-TS 4.06 (Harris et al., 2020) downscaled with WorldClim 2.1 (Fick and Hijmans, 2017); (2) The 2.1 version of historical climate data for 1970-2000, from which we collect the average temperatures (min and max), and precipitation measurements (Fick and Hijmans, 2017); (3) Projections of temperatures and precipitation for all RCP scenarios and times for CMIP6. All global rasters are downloaded at 2.5 minutes (5 km grid on the Equator).

We then compute the SPEI indicators for all RCP scenarios and time periods, using the R package SPEI developed by Vicente-Serrano et al. (2010). We first compute the evapotranspiration using the Hargreaves equation. We then compute the water balance by using the difference between precipitation and evapotranspiration. Lastly, we use the SPEI function to compute the indexes for all pixels of the world for the year 2010, as well as all future periods and climatic scenarios.

Losses due to Droughts

Drought frequency and intensity is a function of the projected SPEI indicators, using mapping between (negative) SPEI values and expected losses generated by insufficient precipitation. For each scenario, we first compute the average SPEI values over six consecutive months. We then take the minimal value of the six-month SPEI indicators over the annual collection of 12 months. This minimal value is used to compute losses generated by droughts, which equal: 0 if SPEI is above -1; 10% for SPEI between -1 and -1.5; 25% for SPEI between -1.5 and -2; 40% for SPEI between -2 and -2.5; 60% for SPEI between -2.5 and -3; 75% for SPEI between -3 and - 3.5; 90% for SPEI between -3.5 and -4; and 95% for SPEI below -4. The final drought loss is the relative deviation between a particular scenario and the loss computed for 2010.

Losses due to Floods

We first compute the expected losses caused by river flooding in the reference year 2010 using the data from Dottori et al. (2016). To this end, we compute the expected damage caused by flooding using six frequency maps of river floods (1 in 25, 50, 100, 200, 500 and 1000 years) and the flooding tide sizes mapped with losses: 10% if tide is below 50 cm; 25% if tide is between 50 and 100 cm; 50% if tide is between 100 and 200 cm; 80% if tide is between 200 and 300 cm; 90% if tide is between 300 and 500 cm; and 99% if tide is above 500 cm. Then, using data for (positive) SPEI deviations, aggregated over two consecutive months, we take the maximum of the SPEI indicator across annual estimates for all times and scenarios.

Losses due to Heatwaves

Damage due to heatwaves is computed using pixel-specific annual temperature distributions for all times and scenarios. The change in damage is caused by changes in the frequency of extreme temperature, which is then mapped into losses as follows: 0% for temperatures below 30°C; 10% for temperatures between 30 and 35°C; 25% for temperatures between 35 and 40°C; 50% for temperatures between 40 and 45°C; 75% for temperatures between 45 and 50°C; and 99% for temperatures above 50°C. The final value of damage caused by heatwaves is the relative deviation between a particular scenario and the loss computed for 2010.

Losses due to Cyclones

We use the simulated cyclone tracks up to 2100 for all RCP scenarios from Lee et al. (2020), complemented by generated wind speeds per individual cyclone from Meiler et al. (2022). Using expected economic losses from cyclones in more than 20 Caribbean and Pacific countries published by the World Bank, we construct damage functions that relate maximum wind speed to percentage losses in GDP.

Coral Reefs and the Impact on Tourism

First, we collect the data on projected timing of annual severe coral bleaching compatible with CMIP6 models. This data is provided by unepgrid.ch and created in cooperation with NOAA Coral Reef Watch. Second, using the aggregates provided by Spalding et al. (2017), we impute the GDP losses caused by coral bleaching on the total value added generated from reef-related and non-reef-related tourism in the 64 most affected countries.

Fertility, Mortality, Education and TFP Trends Projections

We use projections of population, total fertility rate, education structure and mortality from the Wittgenstein Centre, IIASA. We use their Human Capital Data Explorer through the wcde package in R. For computing future TFP trends, we use population and GDP data from the IIASA IAMC database.