You can use this DOI to cite this repository:

This repository contains the data and code to reproduce results from our study titled “Quantifying the cost savings of global solar photovoltaic supply chains”. All code is written using the R programming language.

Authors: John Paul Helveston, Gang He, & Michael R. Davidson.

DOI: 10.1038/s41586-022-05316-6

Abstract: Achieving carbon neutrality requires deploying renewable energy at unprecedented speed and scale, yet countries sometimes implement policies that increase costs by restricting the free flow of capital, talent, and innovation in favor of localizing benefits such as economic growth, employment, and trade surpluses. Here we assess the cost savings from a globalized solar photovoltaic (PV) module supply chain. We develop a two-factor learning model using historical capacity, component, and input material price data of solar PV deployment in the U.S., Germany, and China. We estimate that the globalized PV module market has saved PV installers in the U.S. $24 ($19 - $31) billion, Germany $7 ($5 - $9) billion, and China $36 ($26 - $45) billion from 2008 to 2020 compared to a counterfactual scenario where domestic manufacturers supply an increasing proportion of installed capacities over a 10-year period. Projecting the same scenario forward from 2020 results in estimated solar module prices that are approximately 20% - 30% higher in 2030 compared to a future with globalized supply chains. International climate policy benefits from a globalized low-carbon value chain, and these results point to the need for complementary policies to mitigate welfare distribution effects and potential impacts on technological crowding-out.

File organization

data

Contains all of the “raw” data used in our analyses as well as a single formatted.Rds file that when loaded into R is a list of formatted data frames, which is generated by running the script at code/2format_data.R.

code

file or folder	description
0a_run.R	A single file to reproduce all analyses.
0functions.R	Custom functions used in our analyses.
0setup.R	Loads libraries, creates dir object (a list of paths to folders and objects), and loads 0functions.R.
1format_data.R	Formats and harmonizes all raw data and saves the result as the list object at data/formatted.Rds.
2learning_curves.R	Estimates learning curve models and saves all results as the list object at output/lr_models.Rds.
3scenarios_hist.R	Uses the estimates learning curves models to compute differences in historical costs under different scenarios. Results are saved as the list object at output/historical_scenarios.Rds.
4scenarios_proj.R	Uses the estimates learning curves models to compute differences in projected future costs under different scenarios. Results are saved as the list object at output/projection_scenarios.Rds.
5charts.R	Code to reproduce all charts used in our analyses.
6summary.R	Code to print out a summary of all main results.
7tables.Rmd	Generates all tables in the output/tables.docx file.
8alt_models.R	Code to reproduce several alternative models that we considered as sensitivity checks. These results are presented in Extended Data Tables 2 - 4.

figs

All charts created in the code/5charts.R file are saved here in 3 formats: eps, pdf, and png.

output

All model / scenario analyses outputs are saved here.

Reproducing the analyses

Installation setup

Reproducing the analyses requires the follow setup steps:

1. Install R.
2. Install RStudio (optional, though recommended).
3. If on a Mac, install XQuartz to enable Cairo graphics (for reproducing figures).
4. Download the files in this repository.
5. Open the solar-learning-2021.Rproj file, which sets the working directory to the root of the files in this repository. The repository root must be set as the working directory otherwise the code will error.
6. Install additional required R packages by running the code in the /code/0install.R file (you’ll only need to install these packages once).

Full reproduction

The code/0a_run.R file contains scripts to fully reproduce the entire set of analyses in sequential order. This is purely for convenience to quickly reproduce everything. Each file can be separately run if desired.

Setup

Most R files in the /code folder start with the following line to execute the /code/0setup.R file, which loads all required libraries and sets several global variables such as starting and stopping years to bound the analyses:

source(here::here('code', '0setup.R'))

Formatting the data

All of the raw data are stored in the /data folder. The script at /code/1format_data.R formats all of this data and saves it as a list stored at /data/formatted.Rds.

To load this list of formatted data, run this line after sourcing the /code/0setup.R file:

data <- readRDS(dir$data_formatted)

# Names of the formatted data frames:
names(data)

##  [1] "pvProduction"            "silicon"                
##  [3] "shipments"               "plantsize"              
##  [5] "irenaCumCapacityMw"      "nrelCapacity"           
##  [7] "nrelCost"                "seiaCapacity"           
##  [9] "lbnlCost"                "usNrel"                 
## [11] "us"                      "china"                  
## [13] "germany"                 "world"                  
## [15] "rates"                   "hist_us"                
## [17] "hist_china"              "hist_germany"           
## [19] "proj_nat_trends_us"      "proj_sus_dev_us"        
## [21] "proj_nat_trends_china"   "proj_sus_dev_china"     
## [23] "proj_nat_trends_germany" "proj_sus_dev_germany"   
## [25] "exchangeRatesRMB"        "exchangeRatesEUR"

Learning curve models

All of the learning curve models are estimated by running the /code/2learning_curves.R file. Results are saved in a list of data frames containing formatted model results stored in /output/lr_models.Rds.

To load formatted results, run this line:

lr <- readRDS(dir$lr_models)

# Names of the LR models:
names(lr)

Historical scenarios

All of the historical (2008 - 2018) cost scenario calculations are computed by running the /code/3historical_scenarios.R file. Results are saved in a list of data frame containing formatted results stored in /output/historical_scenarios.Rds.

To load the results of the historical scenarios, run this line:

cost <- readRDS(dir$scenarios_hist)

# Names of the scenarios:
names(cost)

## [1] "cost"    "savings"

Projected scenarios

All of the future projected (2018 - 2030) cost scenario calculations are computed by running the /code/4projection_scenarios.R file. Results are saved in a list of data frame containing formatted results stored in /output/projection_scenarios.Rds.

To load the results of the projected scenarios, run this line:

proj <- readRDS(dir$scenarios_proj)

# Names of the scenarios:
names(proj)

## [1] "nat_trends"         "sus_dev"            "savings_nat_trends"
## [4] "savings_sus_dev"

Charts

All of the charts are generated by running the /code/5charts.R file. Results are saved as pdf, eps, and png files in the /figs folder.

Summary

A full summary of all results can be seen by running the /code/6summary.R file.

Other

The /code/7tables.Rmd file is a simple template that we used to generate a word-formatted summary table of the regression results.

The 8alt_models.R file contains code to reproduce several alternative models that we considered as sensitivity checks. These results are presented in Extended Data Tables 2 - 4.