measuring how local governments approach resilience using budget statements
This repository is for a paper published in the Journal of Public Administration Research & Theory, [DOI and link to be filled in].
This repository contains all of the data and code materials necessary to replicate the project. For questions, please raise an issue and tag us here on Github.
Local governments consider a wide range of policies to increase resilience in the face of myriad risks, and employ a variety of tactics to communicate about these policies to external actors. An important platform to signal resilience as a policy priority is through the budget process wherein local communities decide ‘who gets what, when, and how’. Using computational text mining techniques, we assess how county governments in California signal efforts toward resilience in their budgets during 2012-2017 fiscal years, as well as whether and how those signals are received by the capital market. To test the relationship between local government resilience signals and capital market outcomes, we focus on county underlying credit ratings issued by counties. Empirical results show that county underlying credit ratings are insensitive to resilience signals in local government budgets. By examining the efficacy of resilience signals and their effects on the capital market, we offer evidence on the link between policy signaling and financial outcomes at the local government level.
- The project requires the R statistical programming language, version 4.1.2 (2021-11-01). Some of the R packages required might require loading underlying dependencies.
The directory has 5 folders with the following files:
-
code/ scripts for replicating the project
- /resilience_concept_coding.R measures resilience emphases in budget documents, makes figures 1 and A1
- /generate_county_demovars.R which aggregates county-year data from the American Community Survey and the CA Employment Development Department.
- /aggregate_inputs.R generates the panel data set with credit ratings, resilience emphases, and county covariates
- /regress_ratings_resilience.R performs the mixed effects orginal regression models
-
input/
- county_year_panel.txt, the final filtered and aggregated dataset used for the regression models
- concept_results.txt, which has estimated resilience emphases by county and year
- county_demographs.txt, which has population, unemployment, political ideology, and median household income by county and year (excepting political ideology, which is based on https://americanideologyproject.com/ and only estimated once per county)
- CAcounty_ratings.csv: county-level ratings
- Combined_DIVER_Data.csv: county financial records
- Macro/ and Micro/, which are the raw budget .txt files
- wiki/ which contains snapshots of the original wikipedia articles used to measure the resilience emphases
- unused/ which is a folder of various data inputs that were not used in the final analysis
-output: log files, calculations, processed text, and figures/tables -scratch: items in this folder are not synced to github. this folder is where the word embeddings file is downloaded if not already stored locally on the machine
- run code/01_resilience_concept_coding.R to measure resilience emphases in budget documents.
- run code/02_generate_count_demovars.R to create dataset of county covariates
- run code/03_aggregate_inputs.R to generate panel dataset for regression analysis
- run code/04_regress_ratings_resilience.R to perform the regression analysis Tables and figures found in the paper are automatically generated by these scripts.
Please feel free to ping us if any links appear broken or you have trouble with the code or data.
Contributors names and contact info
Robert A. Greer, rgreer1@tamu.edu Tima T. Moldogaziev,timatm@psu.edu Ryan P. Scott, ryan.p.scott@colostate.edu Tyler A. Scott, tascott@ucdavis.edu, @tylerscottphd
Shield:
This work is licensed under a Creative Commons Attribution 4.0 International License.
