You can download the replication package here (about 4GB).
The main (root) folder contains the following folders:
codes/: contains all Matlab/R codes used as inputs
in the main replication codes.
input/: contains the needed inputs (data and more)
used in the main replication codes.
output/: where all output is stored and
processed.
Besides, it also contains the README.html file and the
main replication codes (explained next).
Replicating the results consist of running a sequence of
R and Matlab codes. In what follows, we
provide an overview of the sequential workflow of the replication and a
brief description of each code. Sections 2 to 4 provide further
details.
The sequence of codes are the following:
gict_initialize.m: initializes the replication package
by exporting some raw data to be further processed in
R.gict_data.R: processes and exports data from previous
step and produces further inputs for the simulations.gict_replicate.m: runs all simulations
sequentially.gict_results.m: processes all the results from the
previous step, reproduces several plots, and exports inputs needed for
producing the maps in R.gict_maps_and_paper.R: reproduces all the maps of the
paper and compiles the gict.tex source (which reproduces
the paper in pdf format).Throughout, several inputs are needed (some from previous papers). We describe them next.
gict.tex: .tex source of the
manuscript.input/H0_areal.mat: 180x360 grid with fraction of land
per cell.input/C.csv: 180x360 grid of the country indices per
cell. From Desmet et al. (2018).input/desert.csv: 180x360 grid of desert dummies. From
Desmet et al. (2018).input/l.csv and input/pop1.csv: 180x360
grids of the population level (number of people per cell) in the data
for years 2000 and 20011, respectively. From Desmet et al. (2018).input/y.csv: 180x360 grid of cell level GDP (in PPP)
for 2000 divided by cell population. From Desmet et al. (2018).input/Y_A.csv: 180x360 gird of cell level agricultural
GDP (in PPP) for 2000 produced by Conte et al. (2021);input/ubar.csv: 180x360 grid of subjective utility
measure for each country, based on the Cantril ladder (Deaton and Stone,
2013). From Desmet et al. (2018).input/trmult_reduced_baseline.mat: matrix of bilateral
trade costs across land cells to the power -theta (trade elasticity).
From Desmet et al. (2018).input/diffmult_reduced.mat: matrix of technology
diffusion across cells with positive mass of land. From Conte et
al. (2021).input/T0.csv and * input/T.mat: 180x360
grid of initial temperatures in 2000 and the evolution of local
temperatures throughout the 21st century. From Conte et al. (2021).input/ksi.csv: 180x360 grid of downscalling parameters
\(\xi(r)\) estimated in Conte et
al. (2021).input/GDPPC_PPP_WBDI.csv: data table with country GDP
(in PPP) in 2020 from the World Bank Development Indicators.input/ne_50m_admin_0_countries.*: files that compose
the world’s shapefile (used to render the maps).input/ne_10m_lakes.rdata: shapefile of lakes in the
world (used to render maps).input/gaoi_with_data.rdata: spatial grid in
Rdata format. From Conte et al. (2021).input/grid_025_clipped_intersected.rdata: spatial grid
at .25 x .25 degree clipped and intersected with the country shapes
(used to render the maps).codes/inversions/gict_functions.R: loads several
functions used in the main R replication codes.output/figures/graph1.pdf: Figure 3.a from Conte et
al. (2021) that is reproduced in the paper.gict_bib.bib: source of the bibliographic references
cited in the paper.The codes gict_initialize.m and gict_data.R
pre-process the some inputs (after setting the working directory
in lines 12 and 39, respectively). In particular, the former
exports the raw inputs from Conte et al. (2021) y.csv,
l.csv, and Y_A.csv in grid format to be loaded
in R (e.g. Y_A_raw.csv).
Then, gict_data.R imports the raw input and corrects the
grid cell-level agricultural output where it exceeds total GDP (i.e. it
corrects some measurement error from Conte et al. (2021)). In
particular, it assigns to those cells the average of the relative
agricultural GDP of the neighboring cells of the same country.
Moreover, gict_data.R exports regional specific
geographical indexes: 180x360 grids of dummies for cells in a
certain region (e.g. input/index_eu26.csv stands for
dummies of the countries at the European Union).
gict_replicate.m sequentially runs all
of the simulations of the paper (after setting the working
directory in line 11). It calls the .m codes for
each specific simulation, stored in the codes/ and
codes/inversion/ folders. Within the latter, the following
codes are called:
inversion.m: does the inversion for sectoral
productivities and amenities as in Desmet et al. (2018) and Conte et
al. (2021).moving_costs.m: does the inversion for mobility costs
and separates amenities from initial utility as in Desmet et al. (2018)
and Conte et al. (2021).theta_inversion.m: does the inversion for sectoral
productivities and amenities as in Desmet et al. (2018) and Conte et
al. (2021) for different values of \(\theta\) (trade elasticity).theta_moving_costs.m: does the inversion for mobility
costs separates amenities from initial utility as in Desmet et
al. (2018) and Conte et al. (2021) for different values of
\(\theta\) (trade elasticity).omega_moving_costs.m: does the inversion for mobility
costs separates amenities from initial utility as in Desmet et
al. (2018) and Conte et al. (2021) for different values of
\(\Omega\) (migration elasticity).alpha_inversion.m: does the inversion for mobility
costs separates amenities from initial utility as in Desmet et
al. (2018) and Conte et al. (2021) for different values of
\(\alpha_M\) (agglomeration forces for
\(M\)).alpha_moving_costs.m: does the inversion for mobility
costs separates amenities from initial utility as in Desmet et
al. (2018) and Conte et al. (2021) for different values of
\(\alpha_M\) (agglomeration forces for
\(M\)).lambda_moving_costs.m: does the inversion for mobility
costs separates amenities from initial utility as in Desmet et
al. (2018) and Conte et al. (2021) for different values of
\(\lambda\) (congestion forces).Besides, the main simulation codes (stored in codes/)
stand for:1
*reb/: simulations for each rebating scheme
(i.e. noreb = simulation without rebating) for 100 periods
and CO2 tax = 40USD/tCO2 starting in 2021. Stores the outputs in the
output/ folder.*reb_HS/: simulations for each rebating scheme and
several CO2 tax values (0 to 65 USD/tCO2) for the hump-shaped
(_HS) curves. It runs for 21 periods. Stores the outputs in
the output/ folder.locreb_HS_theta/: simulations with local rebating of
CO2 taxes of different values (0 to 65 USD/tCO2) for the
hump-shaped (_HS) curves. It runs for 21 periods
for different values of \(\theta\) (the trade elasticity). Stores the
outputs in the output/ folder.locreb_HS_omega/: simulations with local rebating of
CO2 taxes of different values (0 to 65 USD/tCO2) for the
hump-shaped (_HS) curves. It runs for 21 periods
for different values of \(\Omega\) (the migration elasticity). Stores
the outputs in the output/ folder.*reb*_USA/: simulations with local/no rebating of CO2
taxes adopted in the USA
(e.g. noreb_HS_USA = hump-shaped simulations as
above but for taxes set in the USA). Stores the outputs in the
output/ folder.Note that gict_replicate.m runs all the simulations
sequentially, which makes it a time-consuming task (each
individual simulation can take half a day to run on a rather powerfull
CPU).
To address that, each simulation folder in codes/
contains also a .sh shell code that can be run individually
in high-performance computers (e.g. be submitted as a job in a
server-like computer/cluster). By launching each of them individually in
such a computer/cluster allows one to reproduce all results in in
parallel. Importantly, before that one would need to run all of the
codes of the codes/inversion/ folder explained above
(e.g. inversion.m).
The codes gict_results.m and
gict_maps_and_paper.R process all of the simulation results
and report them as documented in the paper (after setting the
working directory in lines 12 and 39, respectively). In
particular, gict_results.m:
R grid format to be used as
inputs for the maps.Then, gict_maps_and_paper.R:
gict_results.m..tex tables to fit the paper’
layout..pdf using the
gict.tex latex source.Conte, B., Desmet, K., Nagy, D.K. and Rossi-Hansberg, E., 2021. Local sectoral specialization in a warming world. Journal of Economic Geography, 21(4), pp.493-530.
Deaton, A. and Stone, A.A., 2013. Two happiness puzzles. American Economic Review, 103(3), pp.591-97.
Desmet, K., Nagy, D.K. and Rossi-Hansberg, E., 2018. The geography of development. Journal of Political Economy, 126(3), pp.903-983.
Note that all simulations whose name does not include
_USA stand for exercises where carbon is taxed in the EU.↩︎