Limitations of the SHRUG

The Economic and Population Census components of the SHRUG are aggregations of data collected by the Indian Ministry of Statistics and Programme Implementation and the Census of India. We have cleaned these data where possible, but errors in data and linking undoubtedly remain. We advise all researchers to run robustness checks with regard to outliers and otherwise unusual units when conducting analyses based on the SHRUG.

If errors are found in the SHRUG which are not in the underlying data, please send a detailed error description to shrug-feedback@devdatalab.org and submit a bug report to the SHRUG-public repository. We will do our best to correct them for future versions. If you have questions about using the SHRUG or ideas for improvement, please visit the Shrug-India subreddit (https://www.reddit.com/r/ShrugIndia/), which is dedicated to discussion of research and analysis using the SHRUG.

The SHRUG is intended to represent the best time series of local socioeconomic data available in India at this time. The strength of the SHRUG is its geographic specificity. The SHRUG is not intended to be used to generate aggregate national statistics. If locations in the Economic Census data could not be matched to the Population Census (due to missing or incomplete location data, for instance), we have excluded their locations in the SHURG.

In our aggregate datasets, we have imputed those missing data where a large enough share of the aggregate's (e.g. constituency's) data was nonmissing. You can view imputation statistics for data tables in the metadata section of this documentation website, and learn more about how the imputation process works here. But national and aggregate estimates of SHRUG data may still misalign with estimates from the survey enumerators.

The SHRUG also offers geographic data in the form of open-source shapefiles for the 2011 Population Census towns and villages, shrids, Census subdistricts and districts, and legislative constituencies. But these geometries, like all open-source locality shapefiles in India, may be subject to small map misalignments and border smoothing. We advise users to validate the polygons if their analysis relies on the exact placement of a village border (e.g. within 10-100 meters). Our own aggregate data was based on 2011 village polygons which we believe are slightly more accurate but are not made available with SHRUG due to licensing constraints.