Household Consumption and Expenditure Survey 2022-23

This repository consists of the clean data dump generated from the latest household consumption and expenditure survey (HCES), by the NSS. The raw microdata is publicly accessible and can be downloaded here.

Repo Structure

The repository consists of the following files:

1. 01_docs: This consists of all the survey documentation, including questionnaire, file structure, caveats, state codes etc
2. 02_code: This has Stata do files which generate clean data files
3. 03_raw: This consists of all the raw files, taken from the NSS site. User is supposed to create this
4. 04_clean: This has all the level-wise cleaned files, in parquet format.

Code

The folder 02_code has the following files

1. 0_master.do: Sets root directory, folder names etc
2. 1_clean.do: Extracts all variables from raw txt files, by each level and preliminary cleaning
3. 2_mpce.do: This has code to calculate and replicate sector and fractile wise MPCE
4. convert_csv.ipynb: Python code to convert csv to parquet and back

Clean files

To generate the files in 04_output, do the following

1. Download and unzip the repo
2. Create a folder called 3_raw and add the raw txt files from the NSS site, available here. You should get 15 .txt files.
3. Open the state project hces.stpr. From within the project, run first 0_master, followed by 1_clean
4. This creates Stata and csv files for each level
5. [OPTIONAL] If you want to get Parquet files, run the convert_csv file (change directory as needed) to do so.
6. [FOR NON-STATA USERS] To convert the Parquet files to csv, open the convert_csv file and run the section which converts parquet to csv.

Replicating Monthly Per Capita Expenditure numbers

To replicate MPCE numbers by sector and fractile from report, follow steps below:

1. Open hces.stpr
2. Run the whole 2_mpce.do file after running 0_master
3. Run till line 160, to get MPCE by rural and urban sectors
4. Run till line 223, to get MPCE by rural/urban for various fractiles.

These match the report numbers upto 2 decimal places.