consumptionsurveyindia • consumptionsurveyindia

India’s National Sample Survey (NSS) Household Consumption Expenditure Surveys, 18 rounds from 1981 to 2025. 51.8 million unit-level records harmonized into a single Parquet file. Query food and non-food consumption by state, food group, sector, and round via DuckDB.

Installation

# install.packages("remotes")
remotes::install_github("saketlab/consumptionsurveyindia")

Data setup

Build the consumption database from MOSPI source data:

library(consumptionsurveyindia)

# Download and parse raw NSS data
ics_build(
  raw_dir    = "~/data/consumption_raw",
  output_dir = "~/data/consumption_parquet"
)

Set the default data directory:

options(ics.data_dir = "~/data/consumption_parquet")

Quick start

library(consumptionsurveyindia)
library(dplyr)

con <- ics_connect("~/data/consumption_parquet")

# What rounds are available?
ics_available_data(con)

# State-level consumption expenditure
ics_consumption_by_state(con)

# Break down by food group
ics_consumption_by_state(con, by_food_group = TRUE)

# Food group expenditure shares over time
ics_expenditure_shares(con)

# Query a specific item (by name or code)
ics_query_item(con, "rice", by = "state")

# Search available food items
ics_search_items("milk")

# Consumption by food group, by state or sector
ics_consumption_by_food_group(con, by = "state")

# Diet-disease linkage with healthsurveyindia
ics_correlate_food_disease(con, health_df)

ics_disconnect(con)

Core concepts

Harmonized data: All 18 rounds are unified into a single consumption table with consistent columns (state, sector, item_code, food_group, value, quantity, weight). Item codes are linked to a bundled catalogue with food group classifications.

Survey weights: All query functions compute weighted means using multiplier / 100. Aggregation runs inside DuckDB; only small summaries are collected.

Methodology break: HCES 2022-23 introduced a three-visit design. Pre-2022 and post-2022 rounds are not directly comparable. The era parameter filters by methodology period.

Recall periods: Section 5.x uses 30-day recall; Sections 6.x/7.x use 7-day recall (normalized to monthly).

Capabilities

Function	Purpose
`ics_connect()` / `ics_disconnect()`	Open/close DuckDB connection
`ics_consumption_by_state()`	State-level expenditure, optionally by food group
`ics_consumption_by_food_group()`	Food group shares by round, state, or sector
`ics_expenditure_shares()`	Food group expenditure shares over time
`ics_query_item()`	Query a specific food item by name or code
`ics_search_items()`	Search the item catalogue
`ics_available_data()`	Round-level summary (record counts, coverage)
`ics_available_items()`	Item availability across rounds
`ics_correlate_food_disease()`	Diet-disease linkage with healthsurveyindia