For this research seminar, I conducted an analysis of the effectiveness of campaign spending in California using new data from the California Civic Data Coalition. The results corroborate findings at the federal level: campaign spending in California Assembly races, regardless of category and accounting for measurement error, is remarkably ineffective. This means that the puzzle of why candidates continue to spend money on campaigning remains. The sole exception is advertising spending for challengers, which lends support to theoretical models of campaign finance that see campaign spending as directly informative.
If you just want to see the final paper, head to paper/paper.pdf.
The directory structure is as follows:
datacontains raw and processed data from CAL-ACCESS (cleaned up by CCDC) and OpenElectionscodecontains Jupyter notebooks and Stata code used to process and analyze the datatexcontains some TeX tables used in the paperchartscontains some charts used in the paperpapercontains the TeX file used to produce the final paper, a bibliography, and the paper itself
Below, a description of the analysis code:
scrape_candidates.py- scrape certified candidates from CAL-ACCESS site
- produces
scraped_candidates.csv
candidates.ipynb- uses
scraped_candidates.csv - produces
candidates.csv, a list of candidates and filer ids
- uses
expenditures.ipynb- uses
candidates.csvandcommittees.csv - produces
expenditures.csv, a list of candidates and a breakdown of their expenditures for each election cycle
- uses
votes.ipynb- uses data from OpenElections to populate list of candidates with vote share, winning status, and incumbency
- produces
votes.csv
merge.ipynb- merges expenditure and vote share data
- produces
all.csv, a list of all candidates with their expenditure breakdown and vote share
races.ipynb- aggregates candidate expenditure at the race level, as required by the regression specification
- produces
races.csv
summary.do- produces
summary.tex, a table of summary statistics andcorr.tex, a correlation matrix for the spending variables
- produces
general.do- produces
general.tex, a regression table for the effectiveness of campaign spending
- produces
breakdown.do- produces
breakdown.tex, a regression table for the effectiveness of campaign spending by category
- produces
This is not a fully robust analysis of campaign spending in California Assembly races; in particular, the subset of data used for the identification strategy I attempted to employ doesn't have quite enough variation to result in precise estimates. Still, the results are suggestive, and provide a direction in which to pursue further empirical and theoretical work. Comments and suggestions are welcome.