PROJECT regionalOil

Introduction

California policymakers have adopted the goal of carbon neutrality by 2045 or earlier. Within California, Kern County accounts for over 70 percent of oil produced within the state. To get a sense of how the transition may affect opportunities in Kern—the extent to which Kern’s economy depends on oil—the project estimates a structural vector autoregressive model that jointly explains the global crude-oil market and the evolution of employment within Kern. The estimates are based on data from FRED and the Quarterly Census of Employment and Wages, and the project contains code to retrieve these data (which will make updating the analysis straightforward). In addition, the project contains R code that

• estimates a structural vector autoregressive model,
• computes structural impulse response functions,
• conducts inference for the structural IRFs based on a residual-based block bootstrap procedure proposed by Brüggemann et al. (2016) as discussed by Kilian and Lütkepohl (2017),
• computes historical decompositions and forecast error variance decompositions,
• constructs counterfactual employment series based on the Wold decomposition and moving-average representation of the system.

A summary of the project is available as a working paper at arXiv (Ryan and Michieka 2026). A published version is now available:

Employment, Input–Output Linkages, and the Energy Transition in California’s Top Oil-Producing Region (with Nyakundi Michieka), Networks and Spatial Economics (2026): https://doi.org/10.1007/s11067-026-09762-x

Please note:

• The code relies on the R package kilianr, which was installed by running devtools::install_github(repo = "richryan/kilianr").
• Updating results will require API keys. Replicating results does not.
• The code relies on the targets package for reproducibility.
• Rich Ryan wrote the code and the paper, so please address correspondence to richryan@csub.edu. He’d be delighted to hear from you.

Push-button replication

The project relies on the R packages

• renv
• targets

for replicability. Here is a workflow that will closely reproduce the results in the paper.

1. Create a new project from git
2. To be sure you are updating results, you can delete the contents of out/
3. From within R at the root of the project, source push_button_replication.R. This will install the relevant packages and run the targets pipeline.
   • In a Posit Cloud environment, I ran into several errors. Running update.packages(ask = FALSE, checkBuilt = TRUE) solved the issues.
4. The file paper-contents.tex collects the results featured by Ryan and Michieka (2026)

In just over 6 minutes, I was able to replicate the results in the following environment:

R version 4.6.0 (2026-04-24)
Platform: x86_64-pc-linux-gnu
Running under: Ubuntu 24.04.4 LTS

Matrix products: default
BLAS:   /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3 
LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.26.so;  LAPACK version 3.12.0

locale:
 [1] LC_CTYPE=C.UTF-8       LC_NUMERIC=C           LC_TIME=C.UTF-8        LC_COLLATE=C.UTF-8     LC_MONETARY=C.UTF-8    LC_MESSAGES=C.UTF-8    LC_PAPER=C.UTF-8      
 [8] LC_NAME=C              LC_ADDRESS=C           LC_TELEPHONE=C         LC_MEASUREMENT=C.UTF-8 LC_IDENTIFICATION=C   

time zone: UTC
tzcode source: system (glibc)

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
 [1] vctrs_0.7.3       cli_3.6.6         knitr_1.51        rlang_1.2.0       xfun_0.57         otel_0.2.0        processx_3.9.0    targets_1.12.0    renv_1.2.3       
[10] data.table_1.18.4 glue_1.8.1        prettyunits_1.2.0 backports_1.5.1   ps_1.9.3          evaluate_1.0.5    tibble_3.3.1      base64url_1.4     yaml_2.3.12      
[19] lifecycle_1.0.5   compiler_4.6.0    codetools_0.2-20  igraph_2.3.1      pkgconfig_2.0.3   rstudioapi_0.18.0 R6_2.6.1          tidyselect_1.2.1  pillar_1.11.1    
[28] callr_3.7.6       magrittr_2.0.5    tools_4.6.0       withr_3.0.2       secretbase_1.2.2 

Note: I haven’t tested this workflow much, but I suspect the workflow will break under Box, as the files in renv/library/ will not always be available.

The script _targets.R also contains directions for updating the results; that is, extending the results to include the latest available data.

In an effort of full and forthright disclosure, I don’t think my code is seeding the bootstrap replications correctly, and I haven’t confirmed that the confidence sets are always the same. Nevertheless, the points estimates should be the same.

API key in .Renviron

The project contains code that retrieves data from two sources that both require an API key:

1. FRED
2. US Energy Information Administration

I saved my API key registered with FRED in my .Renviron file: FRED_API_KEY=my-api-key. I added the API key by running usethis::edit_r_environ(). (The command can be useful because it shows you where the file lives on your machine.)

I saved my API key from the US Energy Information Administration in my .Renviron file. Which I accomplished running usethis::edit_r_environ(). (The command can be useful because it shows you where the file lives on your machine.)

You can learn about the API program at https://www.eia.gov/developer/ and https://www.eia.gov/opendata/. You register an API key. The API key allows code to collect data directly from the EIA website. Importantly, the interface means data can be updated automatically.

A good place to read about the .Renviron file is What They Forgot to Teach You About R by Jennifer Bryan, Jim Hester, Shannon Pileggi, and E. David Aja. Another useful reference is Efficient R programming by Colin Gillespie and Robin Lovelace.

Introduction to Computational and Data Sciences by James K. Glasbrenner, Ajay Kulkarni, and Dominic White was also helpfully consulted.

The references on targets also provide information on worflow and project management.

Resources on the renv package

• “Using {renv} for Reproducible R Projects” by Susan B.
• “Introduction to renv”
• The renv::init() creates
  • renv.lock records what packages were used to produce results (including specific version numbers)
  • `renv::restore()`` will match
  • The directory renv/library/ now contains package files (or symbolic links to those files). The command .libPaths() will show this by telling you where packages in the project are being sourced from. With the renv package, the links will not point to the system-wide R package library.
• renv workflow
  • install.package("somepackage")
  • renv::snapshot() will update the renv.lock file
  • When starting from a fresh R environment, renv:restore() will make sure another R environment has the same version of packages
• Version-controlled files (not renv/library/, which are system-specific—for people working on Windows and Mac machines collaborating on a project, eg—this directory is managed by renv [same for renv/staging/])
  • renv.lock
  • .Rprofile (instructs to use renv/activate.R)
  • renv/activate.R (set up how the renv package works)
  • renv/.gitignore
  • renv/settings.json

From the renv package: “projects using renv will normally use a private, per-project R library, in which new packages will be installed. This project library is isolated from other R libraries on your system.” The idea is to work from an R environment within the project directory.

Resources on the targets package

Here are resources on the targets package:

• Building reproducible analytical pipelines with R by Bruno Rodrigues
• Introduction to targets at Carpentries
• Dr Anna Krystalli’s Putting the R into Reproducible Research at 2019-05-14 RSS Sheffield Local Group meeting
• Louisa Smith’s Reproducible Epidemiology in R
• Flight (2022)

Krystalli cites Buckheit and Donoho (1995) who “distill” Jon Claerbout’s insights into the slogan:

An article about computational science in a scientic publication is not the scholarship itself, it is merely advertising of the scholarship. The actual scholarship is the complete software development environment and the complete set of instructions which generated the figures.

A contribution of the project is integrating analysis of policy-relevant question about the clean-energy transition with a software development environment within the R ecosystem.

Wilson et al. (2017) and Noble (2009) discuss managing a project.

Data

The Quarterly Census of Employment and Wages program uses “the North American Industry Classification System (NAICS) as to assign establishments to industries and to report industry data at highly detailed as well as at aggregated levels.” The classification systems are

• For the years 1990 to 2006, QCEW data were coded using the 2002 version of NAICS.
• For periods from 2007 to 2010, QCEW data were coded using the 2007 version of NAICS.
• For periods from 2011 to 2016, QCEW data were coded using the 2012 version of NAICS.
• For periods from 2017 to 2021, QCEW data were coded using the 2017 version of NAICS.
• QCEW data from 2022-forward are and will be coded using the 2022 version of NAICS.

NAICS codes are available from the Census Bureau.

License

Employment, Input–Output Linkages, and the Energy Transition in California’s Top Oil-Producing Region © 2026 by Rich Ryan is licensed under CC BY-NC-SA 4.0

References

Brüggemann, Ralf, Carsten Jentsch, and Carsten Trenkler. 2016. “Inference in VARs with Conditional Heteroskedasticity of Unknown Form.” Journal of Econometrics 191 (1): 69–85. https://doi.org/10.1016/j.jeconom.2015.10.004.

Buckheit, Jonathan B., and David L. Donoho. 1995. “WaveLab and Reproducible Research.” In Wavelets and Statistics, edited by Anestis Antoniadis and Georges Oppenheim. Springer New York. https://doi.org/10.1007/978-1-4612-2544-7_5.

Flight, Robert M. 2022. “Creating an Analysis With a Targets Workflow.” September 27. https://rmflight.github.io/posts/2022-09-27-creating-an-analysis-using-targets.

Kilian, Lutz, and Helmut Lütkepohl. 2017. Structural Vector Autoregressive Analysis. Cambridge University Press. https://doi.org/10.1017/9781108164818.

Noble, William Stafford. 2009. “A Quick Guide to Organizing Computational Biology Projects.” PLoS Computational Biology 5 (7): e1000424. https://doi.org/10.1371/journal.pcbi.1000424.

Ryan, Rich, and Nyakundi Michieka. 2026. Employment, Input-Output Linkages, and the Energy Transition in California’s Top Oil-Producing Region. https://arxiv.org/abs/2602.23462.

Wilson, Greg, Jennifer Bryan, Karen Cranston, Justin Kitzes, Lex Nederbragt, and Tracy K. Teal. 2017. “Good Enough Practices in Scientific Computing.” PLOS Computational Biology 13 (6): e1005510. https://doi.org/10.1371/journal.pcbi.1005510.