Analysis code and aggregated results for "The cost of an outage is more unequal than the outage: utility disruption and household borrowing in a monthly panel from Uzbekistan"
Description
This deposit contains the complete analysis code and the aggregated intermediate results underlying the article "The cost of an outage is more unequal than the outage: utility disruption and household borrowing in a monthly panel from Uzbekistan". The study uses 82 monthly rounds of the Listening to the Citizens of Uzbekistan (L2CU) household telephone panel, covering 121,618 household-months on 2,499 households, to separate exposure to utility disruption from its financial cost. Contents: scripts/ — five R scripts. Three produce the estimates (diagnostics, estimation, mechanisms), one produces the figures, and one generates the LaTeX tables directly from the result files so that no number in the article is transcribed by hand. results/ — 51 CSV files holding every aggregated quantity reported in the article, including the coefficients, standard errors and sample sizes behind each table and figure. figures/ — the six figures of the article in PDF. tables/ — the LaTeX emitted by the table-generating script. README.txt — how to reproduce, which script produces which output, a map from each table and figure of the article to its source file, and notes on two deliberate specification choices. The household microdata are NOT included. The analysis uses the public-use file of the L2CU survey, distributed by the World Bank Microdata Library under Licensed Data terms that do not permit redistribution by third parties. Any researcher may obtain the same file directly from that source by accepting the same conditions. In accordance with the licence, no cell derived from fewer than ten households appears in the article or in the files deposited here. Software: R (>= 4.2) with haven, dplyr, tidyr, ggplot2 and patchwork.
Files
Steps to reproduce
Requirements: R (>= 4.2) with haven, dplyr, tidyr, ggplot2 and patchwork. Figures additionally need a Cairo-enabled graphics device. 1. Obtain the public-use file of the Listening to the Citizens of Uzbekistan (L2CU) survey from the World Bank Microdata Library and place the .dta file in the working directory. The scripts locate it themselves; DATA_DIR defaults to ".". 2. Run the scripts in order: Rscript scripts/50_l2cu_energy_diagnostics.R -> out50/ Rscript scripts/51_l2cu_energy_estimation.R -> out51/ Rscript scripts/52_l2cu_energy_mechanisms.R -> out52/ Rscript scripts/53_l2cu_energy_figures.R -> figures/ Scripts 50 to 52 write CSV files and a plain-text log. Script 53 reads those CSVs and writes each figure as EPS, PDF and PNG at 400 dpi. 3. Generate the LaTeX tables from the CSVs, so that no number in the article is transcribed by hand: Rscript scripts/make_tables.R -> tables.tex, tab/*.tex make_tables.R expects the CSVs in a directory named "data"; the files in results/ here are exactly those. The CSV files in results/ are the output of steps 2 and 3 as run by the authors, so the figures and tables can be regenerated from them without access to the microdata. Only the estimation steps require the .dta file. Two specification choices are deliberate and are discussed in the article. Seasonality is estimated with household and YEAR fixed effects, not household and round, because each survey round is a single calendar month and round effects would absorb the season entirely. Time in the panel enters as bins (1, 2-6, 7-12, 13-24, 25-48, 49+ rounds) rather than linearly, because tenure, entry cohort and calendar round satisfy an exact linear identity, so a linear tenure trend is not identified alongside household and round effects. README.txt in the archive maps each table and figure of the article to the script and result file that produce it.
Institutions
- National University of UzbekistanTashkent, Tashkent