Development Research at High Geographic Resolution:

An Analysis of Night Lights, Firms, and Poverty in India

using the SHRUG Open Data Platform∗
Sam Asher†
Tobias Lunt‡
Ryu Matsuura§
Paul Novosad¶
September 2020

Abstract
The SHRUG is an open data platform describing multidimensional socioeconomic devel-
opment across 600,000 villages and towns in India. We present three illustrative analyses only
possible with high-resolution data. We confirm that nighttime lights are highly significant
proxies for population, employment, per-capita consumption, and electrification at very local
levels. However, elasticities between night lights and these variables are far lower in time series
than in cross section, and vary widely across context and level of aggregation. Next, we show
that the distribution of manufacturing employment across villages follows a power law: the
majority of rural Indians have considerably less access to manufacturing employment than
is suggested by aggregate data. Third, we perform a poverty mapping exercise, identifying
the targeting improvement from allocating programs at village rather than at district level.
The SHRUG can serve as a model for open high-resolution data in developing countries.
JEL Codes: R11/C81/O12

∗
Thanks to Teevrat Garg, Francesca Jensenius, Dan Keniston, and Nishith Prakash for sharing data that
contributed to this dataset. This project and underlying data platform were funded by Emergent Ventures. Early
work on this project was supported by the IGC (project 89414), and a project funded by the UK Department for
International Development (DFID) and the Institute for the Study of Labor (IZA) for the benefit of developing
countries. All errors are our own. The SHRUG dataset can be downloaded at http://devdatalab.org/shrug.
†
Johns Hopkins SAIS, sasher2@jhu.edu
‡
Development Data Lab, lunt@devdatalab.org
§
Northwestern University, ryumatsuura@u.northwestern.edu
¶
Dartmouth College, paul.novosad@dartmouth.edu

0
1 Introduction

Development and economic growth, insofar as they affect the living standards of individuals, are
highly localized phenomena. Employment opportunities, schools, public goods, and welfare schemes
are relevant to individuals primarily if they are accessible within a short distance. However, due
to the nature of data available to researchers, analysis of development has largely occurred at an
aggregate geographic scale. In India, much policy and research has been based on the National
Sample Surveys, which have tracked socioeconomic change at the district level since independence.
But a single Indian district is home to around 2 million people, often with vast variation in living
standards. This variation is demonstrated in Table 1, which shows the share of variation in various
development outcomes at different levels of geographic variation. 55% of variation in mean village
per capita consumption and over 90% of variation in rural non-farm employment occurs below
the district level. The National Sample Surveys often base district statistics on fewer than 100
households per district, missing much of this microgeographic variation.
The computerization of government administration and the proliferation of new data sources such
as satellite imagery has provided researchers with the opportunity to study development at higher
geographic resolution than ever before. However, it has been enormously costly and time-consuming
to discover, obtain, clean, and merge data sources that were not designed for social science research.
When researchers do pay the fixed costs, it is often difficult for other research teams to make use
of that data; despite recent efforts to make data availability a key part of the publication process,
published data is rarely comprehensive or easily linked to other data sources. The high fixed costs
of making use of new data is particularly harmful to researchers lacking the funding and time to
make major investments in data assembly, such as many doctoral students and researchers in the
very countries that development economics seeks to study.
This paper introduces the Socioeconomic High-resolution Rural-Urban Geographic Dataset on
India (SHRUG), a multidimensional socioeconomic data platform with high geographic resolution
data on demography, non-farm employment structure, political outcomes, forest cover, night lights,
and administrative program operation, among other dimensions, on the universe of cities, towns,

1
villages, and legislative constituencies in India from 1990 to 2018 (see Table 2). The SHRUG is an
open dataset but also an open data platform that is structured to facilitate and encourage sharing
of readily linkable data between researchers in the future.
In this paper, we present three analyses that shed light on the distribution of economic opportunity
within highly localized regions, analyses that are only possible with high resolution data like the
SHRUG.
We begin by examining the effectiveness of nighttime lights as a measure of local development.
Nighttime lights, as measured from outer space, have been widely used in recent years to examine
patterns of development in places where local economic data is not available. Nighttime lights are
particularly used in analyses of programs which have geographic variation which is more precise than
what can be observed in aggregate data; the ability to observe an economic outcome in a 1km x 1km
cell is a key comparative advantage of night lights. Given the lack of high-precision economic data in
most developing countries, the assumption that night lights have the same elasticities at both very high
and very low levels of geographic aggregation with a range of outcome variables is largely untested.1
Our results confirm that night lights are a highly statistically significant log-linear proxy for a range
of development outcomes—population, employment, per capita consumption, and electrification—at
a very narrow geographical level, even in specifications with regional fixed effects, in the both the
cross section and in the time series. However, there are two significant caveats. First, because night
lights have independent correlations with each of these outcomes, it is difficult to tell from night
lights alone which of these proxy variables are being measured. Researchers have used night lights
as a proxy for GDP growth, cross-sectional GDP, urban extent, public expenditure, and electricity
supply and demand, among other variables (Henderson et al., 2011; Baum-Snow et al., 2017; Bleakley
and Lin, 2012; Min et al., 2013; Hodler and Raschky, 2014; Baskaran et al., 2015; Burlig and
Preonas, 2016; Harari, 2020; Mahadevan, 2019). Our results suggest that all of these proxies are
valid, but researchers have no guarantee that night lights are proxying the specific variable that
1
Henderson et al. (2011) show that elasticities are similar across countries and across large regions within
countries, but they do not measure elasticities at finer geographic levels. And yet the primary use of night lights
in research is precisely to generate outcome variables at very high geographic resolution.

2
they would like to measure and not a different one.
Second, we find that night light elasticities vary substantially across the level of geographic
aggregation and across contexts. Importantly, we find that night light elasticities are an order of
magnitude smaller in the time series than they are in the cross section. Night light elasticities with
most variables are also considerably higher in electrified villages than in non-electrified villages. This
finding suggests that researchers should use extreme caution in applying regional luminosity-to-GDP
elasticities from the literature to microgeographic contexts without additional information, because
these elasticities may be wildly different. Applying cross-sectional elasticities to time series night
light coefficients is likely to result in a substantial upward bias on GDP estimates.
Moving beyond night lights, we next examine the distribution of manufacturing and services em-
ployment across space. Economic activity is often highly agglomerated and firm size distributions are
known to follow power laws (Ellison and Glaeser, 1997; Gabaix, 2016). The amount of concentration
affects the extent to which regional averages reflect the access to opportunity of the majority of a
region’s residents. We show that the distribution of firms across space also follows a power law, such
that some regions have orders of magnitudes more employment density than others. The extent of
employment concentration is higher for manufacturing than for services firms, and the highest by far
for manufacturing employment in rural areas. The mean Indian village has 15 manufacturing jobs
per 1000 people, but the median has less than 5. Almost 80% of rural Indians live in villages with
fewer manufacturing job opportunities than the rural mean. Regional employment opportunities,
especially for women, may be of limited relevance if they are not geographically proximate. This
analysis demonstrates that analysis of employment at aggregate geographic scales may fail to register
the lack of labor market opportunities for most rural workers.
Finally, we conduct a poverty mapping exercise, considering the problem of a policymaker
targeting a program to poor geographic regions (Ravallion, 1993; Baker and Grosh, 1994; Bigman
et al., 2000; Bigman and Srinivasan, 2002; Elbers et al., 2007; Brown et al., 2019). We measure the
efficiency gain that could arise from the ability to allocate a program reaching 25% of the population
at increasingly finer geography. If the policy-maker can observe poverty only at the district level

3
(n=618), then allocating the program to the poorest districts would cover 37% of the rural poor.
Allocating the same program at the village level would cover 44% of the rural poor. There is likely
a tradeoff between precise geographic targeting and administrative cost; highly localized data makes
it possible to estimate the parameters of this tradeoff.
The SHRUG has a number of strengths and weaknesses as an analytic platform. Its primary
strengths are its geographic resolution and consistent location identifiers over a period of dramatic
economic change. There are few other data sources in developing countries that allow tracking of
public goods, economic activity, politician characteristics, election results, and many other outcomes
at the village and town level for almost 600,000 consistently-bounded geographic units.2 The key
comparative disadvantages of the SHRUG are that (i) most of the underlying data sources are based
on narrow surveys with a few dozen questions rather than hundreds; and (ii) the SHRUG describes
locational aggregates, rather than individual or firm microdata.3 Researchers interested in detailed
characteristics of individual households and firms may thus be better served by traditional data sources.
There are a number of contexts in which the SHRUG may have a comparative advantage:
• Programs with highly local variation are difficult to evaluate with aggregate data. Examples
of analysis of such programs using administrative data (some of which are now in the SHRUG)
include Lehne et al. (2018), Chhibber and Jensenius (2016), Burlig and Preonas (2016), Asher and
Novosad (2017), Adukia et al. (2019), Asher and Novosad (2020), and Muralidharan et al. (2017).
• Analysis at the city/town and legislative constituency level is challenging because neither of these
units are identified in India’s conventional data sources. SHRUG data are aggregated to both
of these levels.
• Researchers running field experiments typically use national population censuses as sampling
frames for new field experiments, but have limited additional data on sample locations before
collecting their own baseline surveys. The SHRUG increases the scope of what is known at a
local level. Field experimentalists can begin to test for divergent trends in field locations even
2
There are approximately 605,000 towns and villages in India. We collapse these to about 590,000 unique
locations that can be tracked over time.
3
Note that it can be rapidly linked to firm-level data in the open Economic Census, which is openly available.

4
 before conducting a baseline survey.
Most of the data underlying both the SHRUG and the papers using administrative data described
above are public. The primary constraint to using data like these is the time input required to
assemble, clean, and match them to other data sources. The returns to this activity are increasing
in the number of other localized data sources that they can be matched to. We therefore hope
that the existence of the SHRUG makes it worthwhile for researchers to assemble and share other
administrative and microgeographic datasets describing additional dimensions of development in India.
We structured the SHRUG to maximize the ease and benefit to researchers of sharing new data.
First, by providing a broad spectrum of data under a consistent set of identifiers, we hope to create
a standard set of location identifiers for India. Such standardization can lower the cost of merging
datasets for all researchers. Second, we have structured downloads from the SHRUG to credit
individual contributors. When researchers download data from the SHRUG, they are asked to cite
the authors who created each component of the SHRUG that they are using.4 This citation structure
is intended to give researchers an incentive to contribute, because clean data integrated into the
SHRUG will receive more users and generate more citations than data on journal web sites that
are suitable for replication only.5
Finally, SHRUG data are released to non-commercial users under a copyleft license (the Open
Database License, or ODbL), which commits researchers who link the SHRUG to non-proprietary
data to also post that non-proprietary data in a complete form with SHRUG identifiers at the time
that their research is published.6 Given the long publication lags in economics, this license gives
researchers ample lead time to work on additional projects with the data that they have collected,
but commits them to sharing data in a reasonable time horizon. Like the limited protection time
offered by a patent, this approach trades off the private incentive to researchers of developing unique
data sources with the much greater public good of making those data sources available to the full
4
For instance, users of the data in the SHRUG on the criminal charges facing Members of Legislative Assemblies
should cite Prakash et al. (2019).
5
The present version of SHRUG already includes contributions from two groups of researchers other than
ourselves. Since making the data public in September 2019, three additional groups of researchers are working on
additional data sources that they plan to integrate with the SHRUG.
6
More information on the ODbL is available at https://opendatacommons.org/licenses/odbl/index.html.

5
network of researchers that can make use of them.7
The current release of the SHRUG (version 1.5) describes: (i) demographic and public goods data
on every town and village in India from 1991 to 2011; (ii) employment and location of every non-farm
firm in India from 1990 to 2013; (iii) legislative election results from 1980 to 2018 (Jensenius, 2017); (iv)
assets, liabilities, and criminal charges of all politicians in office and many additional candidates from
2004 to 2017 (Prakash et al., 2019); (v) remotely sensed night lights from 1994 to 2013; (vi) remotely
sensed forest cover from 2000 to 2014; (vii) the share of labor force in agriculture and small area
estimates of consumption from the Socioeconomic and Caste Census of 2012; and (viii) administrative
data from the implementation of India’s national rural roads program. As new census, remote
sensing, and administrative datasets are released, the breadth of this panel will continue to grow.
This paper first describes the construction of the SHRUG in some detail in Section 2, along
with several validation exercises. Section 3 presents the analyses of night lights, agglomeration,
and poverty. Section 4 discusses the strengths and weaknesses of the SHRUG as an analytical tool.
Section 5 concludes with a discussion of the copyleft license and framework for improving institutions
of data sharing between researchers.

2 Data: A 25-year Multidimensional Panel of Locations

All variables in SHRUG are disaggregated to the level of the village (n=590,648), the town (7528),
and the state legislative constituency (n=2800), covering the entire country. The different components
are summarized briefly in Table 2 and in detail in the SHRUG metadata table.8
All of these data sources were initially generated with diverse geographic identifiers which were not
straightforward to link. The Population and Economic Censuses are published with town and village
codes with incomplete correspondences across different rounds. PMGSY data is published at the
habitation level; most villages consists of between one and three habitations. Remote sensing data
is published in grid cells, while political data is published at the constituency level; constituencies
are approximately the size of subdistricts (a standard unit in the Population Censuses), but with
7
Data under formal proprietary contracts that restrict sharing are excluded from this commitment at this time.
8
The SHRUG metadata table can be found at devdatalab.org/shrug-metadata.

6
different and overlapping boundaries.
Any pair of these datasets can be reconciled and matched to each other, albeit with significant
labor input. Sources of information used to match data include numeric identifiers, names, maps,
data contents, and external data sources. In most cases, none of these information sources provide a
perfect match. Identifiers are rarely consistent across datasets; sometimes they are consistent across
some states but not others. Locations may be known by multiple names, and names may change
over time; they virtually always have different spellings across and even within data sources. Maps
are supplied with different projections and different degrees of error; boundaries which should be
identical in different mapping sources often are not. Locations themselves change, splitting, merging,
and realigning boundaries in complex ways.
The process of matching these datasets consists of developing algorithms to use as many of these
information sources as possible, and then tuning those algorithms based on manual verification of
samples of data. Individual verification of every match is however not feasible when observations
number in the hundreds of thousands. Some units cannot be matched even with manual verification
because there is insufficient information available. Some degree of incorrect matches are inevitable.
The core contribution of the SHRUG is a set of universal location identifers (one at the town and
village level, and one at the constituency level) that span the entire sample period from 1990–2018.
All of the datasets above can be rapidly linked to these identifiers in every period. When using
the SHRUG, linking disparate data sources takes seconds rather than months. By releasing the
data with these universal identifiers, our hope is to make them a standard for economic research
in India, allowing future data sources to be linked to the data sources described here with ease.
This section describes the process of creating the universal identifiers that are at the heart of the
SHRUG. The definitions of individual data fields are described in detail in the SHRUG codebook.9
To keep its size reasonable, the SHRUG 1.5 data package includes only a subset of the fields in
the Economic and Population Censuses. We provide linking keys to each of these source datasets
to make it easy for users to bring in additional fields from these open data sources.
9
The SHRUG codebook can be found at http://www.devdatalab.org/shrug.

7
2.1 Building the Town and Village SHRUG

The key challenge in creating time series administrative data in India is in dealing with changing
unit boundaries. Faced with the challenge of villages being split, merged, and integrated with cities
and towns, the decennial Census has opted to create new location identifiers in every decade since
1991. Further complicating the process of matching locations over time, district boundaries have
changed substantially, with hundreds of new districts created between 1991 and 2011. The Census
provides digital keys to link villages and towns to prior censuses, but they are highly incomplete.
The Census district handbooks contain detailed descriptions of boundary changes in narrative format
only. All of these sources have errors and inconsistencies.
We used both the digital linking keys and the district handbooks to create the best possible
correspondence of villages and towns across the 1991, 2001, and 2011 censuses.10 We supplemented
this with a custom fuzzy string matching program to match village and town names over time.11
We conducted a hierarchical match from the largest to the smallest administrative units. We began
with a match of districts across Population Censuses. A 1991–2001 district correspondence was
shared with us by Kumar and Somanathan (2015). We constructed the 2001–2011 district match
based on the back-referenced village identifiers in the 2011 census, which provided a 2001 census
village identifier for the majority of 2011 villages. Within districts, we then matched subdistricts
on the basis of names where possible, and then we matched villages within subdistricts, again on
the basis of names. Where the district and subdistrict maps indicated substantial changes in district
and subdistrict boundaries, we matched villages and towns within higher-level aggregates of districts
and subdistricts. We validated the data using internal consistency checks and data from multiple
sources, including geospatial village and town data assembled by other research groups. Appendix
Table A1 summarizes the share of population from each Population Census that is matched at the
village and town level to the SHRUG by state. Virtually all towns and villages were matched across
10
The Census District Handbook is a 500+ page book describing all changes to boundaries in a given census
district in each intercensal period. There is one book like this for each of India’s ∼700 districts.
11
The program is called masala-merge and is available at http://github.org/devdatalab/masala-merge. It performs
a Levenshtein string match, customized for common string substitutions used in Indian langages.

8
the census periods: the match rates is 98% or higher for all but two small states in 1991.
To match the Economic Censuses to the Population Censuses, we used the location directories
for 1998 and 2005, which were shared with us by the Ministry of Statistics (MOSPI). For 2013, we
used the fact that the Economic Census location codes corresponded to the Population Census short
codes, which were available with village and town names on the Population Census website. The
final step in all these cases was a match using location names with the algorithm described above.
MOSPI was not able to provide a location directory for the 1990 Economic Census. The EC district
codes were the same as those used in the 1991 Population Census, but the lower level codes were differ-
ent in some states. We worked with MOSPI to identify the set of states that used the same identifiers in
the 1991 Population Census and the 1990 Economic Census. It was straightforward in the data to dis-
tinguish these states from the ones which created new codes, and we matched villages and towns on the
basis of identifiers in these places. For towns that could not be reliably matched on the basis of the town
codes, we obtained a number of additional matches in situations where three conditions all held: (i)
towns could be uniquely matched within districts to the 1991 Population Census based on the number
of wards;12 (ii) their within-district size rank was the same in the Economic and Population Censuses;
and (iii) the number of people per Economic Census job was within an order of magnitude of the dataset
mean, which was approximately 20. Appendix Table A2 summarizes the share of employment in each
Economic Census that is matched to the SHRUG, by state. Because of the absence of the 1990 location
directory, the match rate for the 1990 Economic Census is much lower than for the other censuses.
Additional administrative datasets (such as the PMGSY road data and the Socioeconomic and
Caste Census) were matched using a similar approach. These matches are described in more detail
in Asher and Novosad (2020).
Location splitting and merging over time results in an inordinately complex set of linking keys.
To create a dataset that was unique on the same locations for all of the different underlying data
sources, we aggregated location units until a set of boundaries could be found that was unique across
all years. We gave each of these units a unique SHRUG identifier, or a shrid.
12
For instance, if a district had two towns in the 1991 Population Census, with respectively four and seven wards,
we matched them to the 1990 Economic Census towns with the same number of wards.

9
 A shrid describes a geographical unit that can be mapped consistently across all rounds of
the Indian Population and Economic Censuses from 1990 to 2013. In the majority of cases, a
shrid describes a single village or town. When villages or towns have merged or separated in the
sample period, we have aggregated them in the periods where they appear separately, such that
the aggregation is represented by a single consistent shrid in all of the data. Some shrids are thus
composed of multiple Population Census villages or towns, or a combination of villages and towns.
In some cases, village and town boundaries have changed so dramatically that the aggregated
constant boundary unit is quite large. In the case of New Delhi and Chandigarh, whose internal
boundaries have changed frequently since 1991, the shrid is the entire city-state. In the case of
Mumbai, a shrid is a district, which is the smallest non-changing unit. Creating time-consistent
data within urban boundaries is a challenging and distinct project which we are also working on,
but is beyond the scope of this iteration of the SHRUG.
The use of a single consistent unit for each geographic location is a critical simplification for the
researcher. Without a one-to-one relationship between locations across datasets, each additional
bilateral merge requires an increasingly complex task of handling splits, joins, and unit duplications.
When working with shrids, any number of datasets can be merged without increasing complexity.
The complete keys linking shrids to their original population and Economic Census codes and
location names are posted with the SHRUG, making it easy to bring in additional data that is
already linked to Population Census codes. As described in the codebook, we have designed a
naming convention for shrids that will be forward-compatible with future censuses as well. New
keys will be posted as future censuses are released.

2.2 Creating a Constituency-Level Panel

The boundaries of India’s 543 parliamentary and some 4000 legislative constituencies do not align
with the aggregate administrative boundaries used by India’s data collection agencies. To create a
constituency-level dataset with economic outcomes, we matched villages and towns to constituencies

10
using digital maps and collapsed the data to the legislative boundaries.13 We created a set of
time-invariant constituency identifiers for the 3rd and 4th delimitations.14
Creating a constituency-level panel from town and village microdata poses a number of challenges.
First, because of the fuzzy matching process, there are some villages which were matched to some
Censuses and not to others. Simply aggregating employment in matched villages to the constituency
level would thus overstate employment gain in constituencies that have improving match rates over
time. We corrected these errors with an imputation process which we describe here.
In the 2011 Population Census, we have matched 100% of towns and villages to constituencies,
while the match rates in 1991 and 2001 are very high but imperfect. For each constituency, we
therefore know the 2011 population in towns and villages that were matched to the other censuses,
and the 2011 population in towns and villages that were not matched. We impute the prior years’
population in unmatched locations by assuming that the within-constituency 2001–2011 population
growth rate is the same in towns and villages that we did not match in 2001 as it is in the towns
and villages that we did match. We can repeat the process to obtain the full set of populations
in 1991. Because the match rates in 1991 and 2001 are so high, any error in this imputation process
is likely to be minimal. This process will cause the aggregated constituency population to be closer
to the truth than if we counted missing locations as having zero population.
We repeated the process to aggregate the employment count in the Economic Censuses, and the
public goods counts in the Town and Village Directories. For each Economic Census, we assumed
that the employment-to-population ratio for missing locations is the same as it is for non-missing
locations within the same constituency. For location amenities that are aggregated with means
rather than sums (such as the mean number of hours of electricity), we generated an aggregate
based on the population-weighted mean in non-missing locations. To avoid excess dependence on
imputed values, we set fields to missing in constituencies where locations covering more than 25% of
13
SHRUG 1.5 does not include electoral data for parliamentary constituencies or panchayats, but these will be
included in the future. Boundary data was obtained from ML Infomap.
14
This was necessary because the Election Commission of India (ECI) does not always use consistent numeric
identifiers over time. The keys provided with the constituency-level data make it possible to link to the ECI data
and thus to any other dataset that uses those identifiers.

11
the population would be imputed. This means that different constituencies may be missing different
fields depending on the underlying structure of the data.
Appendix Figure A1 runs a simulation to calculate error rates associated with the imputation pro-
cess. We randomly drop additional locations from the 2001 Population Census and then compare the
calculated constituency population to what would be observed if no locations were dropped. When 25%
of shrids are dropped (the upper limit of what we include in the data), the average constituency pop-
ulation is within 2% of the true value with a slightly negative bias of 0.015%.15 It is important to note
that this imputation process applies only to the constituency-level data; when Economic or Population
Census data are missing in the town and village data, they are reported as missing in the SHRUG.
Another challenge that arises is that the available polygon shapefiles for constituencies and
towns/villages are not perfectly aligned, even though they all use the same WGS84 projection and
were obtained from the same firm. The misalignment is small—on the order of several hundred meters
in the worst cases—but it is enough that some villages and towns cannot be unambiguously assigned
to a single constituency. We dropped constituencies in which more than 25% of 2011 population
is in villages or towns that cannot be decisively assigned. We have explored several alternate sources
of data and spoken with several other experts on Indian spatial data, and to our knowledge there
are currently no higher accuracy shapefiles than these, so this amount of error is unavoidable. There
are several ongoing projects to assign villages to constituencies by digitizing electoral rolls; as these
data become available, we aim to integrate them into future versions of the SHRUG.
A third challenge is that some towns contain multiple constituencies. Because the Population
Censuses do not report consistent identifiers at the ward level or below, it is difficult to identify the
population or other characteristics of these constituencies — we know only the aggregate population
of the combined constituencies.16 We therefore exclude constituencies that include any part of
15
Imputed values for constituencies with high imputation rates are available from the authors, as is the share
of imputed data in each constituency-field. These are not included in the online SHRUG package because the
files are extremely large and have relatively narrow usefulness. While it would be desirable to use these values to
econometrically adjust for the partial imputation that took place in calculating these values to adjust for the partial
imputation of some constituency values, we are not aware of an out-of-the-box method that is well-matched to
this partial imputation structure. The simulated error rates suggest that this source of error is small relative to
other sources of error in the collection of administrative data in India, and thus not a major concern.
16
The Population Census reports data at the ward level, but the wards change across rounds and do not necessarily

12
towns that cross constituency boundaries. Constituencies in large urban areas are therefore missing
population and economic data in the SHRUG. However, the election and remote sensing measures
are included for all villages because they are not aggregated from village and town data.
The constituency SHRUG is therefore not representative — in particular, it excludes large cities.
However, we are not aware of other research that measures or exploits socioeconomic data at the
constituency level for large urban constituencies (other than the remote sensing measures described
above), presumably due to the same boundary misalignment issues that we face here. Constructing
such a dataset using ward maps for India’s largest cities would be a valuable contribution that would
enable better study of politics in India’s growing cities.
Finally, India periodically redraws constituencies to account for population changes. The third
delimitation came into effect in 1976 and the fourth in 2008, in the middle of the period covered by
the SHRUG (Iyer and Reddy, 2013). This is not a problem for data construction, since constituencies
are simply defined as polygons. We therefore match both sets of polygons to villages/towns and create
separate complete constituency-level panels from 1990–2018 for the old and the new constituency
delimitations. Researchers can thus make their own decisions regarding which polygons to use for
which periods. We provide separate identifiers for the third and fourth delimitations; there is no
correspondence between these as nearly all of the constituency boundaries were changed.

2.3 Remote Sensing Data

At present, SHRUG includes data from two remote sensing sources: nighttime luminosity (Henderson
et al., 2011) and forest cover (Asher et al., 2019; Townshend et al., 2011). Nighttime luminosity data are
from the National Oceanic and Atmospheric Administration (NOAA), and provide a luminosity value
from 0–63 for each 1 km x 1 km grid cell. Forest cover data comes from Vegetation Continuous Fields
(VCF), a MODIS product that measures tree cover at a 250m resolution from 2000 to 2014. VCF is
predicted from a machine learning algorithm based on broad spectrum satellite images and trained with
human-categorized data, which can distinguish between crops, plantations and primary forest cover.
Gridded data from both of these sources were matched and aggregated to village, town, and
share boundaries with constituencies.

13
constituency boundaries. About 90% of towns and villages were georeferenced with polygons,
permitting accurate measurement of night lights and forest cover. About 10% of locations, especially
in the Northeast, were georeferenced in the mapping data only by points; we constructed Thiessen
polygons to match these to the forest cover and night light rasters. Locations mapped with Thiessen
polygons are flagged in the data.

2.4 Imputation of Consumption Expenditure

We created a measure of village- and town-level consumption expenditure using data from the
Socioeconomic and Caste Census (SECC), a universal enumeration of household assets conducted
in 2012. Asset censuses like this are commissioned by the federal government approximately every
decade to determine individual eligibility for means-tested poverty relief programs. The SECC
enumerates a list of household assets that can be rapidly assessed (including roof and wall material,
number of rooms, and assets such as agricultural equipment, vehicles and mobile phones).
We used this asset data to generate small area estimates of consumption expenditure, following
the method of Elbers et al. (2003). Using the 2011–2012 IHDS-II (Indian Human Development
Survey, 2011–12, available at https://ihds.umd.edu), we regressed total household consumption
on a set of continuous and dummy variables that are equivalent to all asset and earnings information
contained in the SECC. We then used the coefficients (shown in Appendix Tables A3 and A4) to
predict household-level consumption in the SECC microdata. We aggregate this to create a village-
and town-level measure of annual per capita consumption expenditure. Because the urban and rural
asset lists are different, we use separate estimations depending on whether households were identified
as rural or urban in the SECC.17 Village- and town-level asset ownership shares are also available in
the house listing for the 2011 Population Census and match the asset shares in the SHRUG closely.
In addition to average per capita consumption, we included the share of individuals in each town
17
The coefficients from the model that generate these consumption numbers are estimated with statistical error.
To allow researchers to account for these errors, we used a bootstrap approach, drawing households from IHDS
with replacement and re-estimating village-level per capita consumption 1000 times. These 1000 draws are available as
a SHRUG package for download and reflect the distribution of per capita consumption that arises from the first-stage
estimation process. Researchers can use these 1000 draws in a second bootstrap process to account for the consumption
estimation error. The Data Appendix in Asher and Novosad (2020) describes this process in additional detail.

14
and village who live in households with per capita consumption below the 2012 national poverty rate,
defined in two ways: below $2, or below Rs. 27.2 per day in villages and Rs. 33.33 in cities and towns.18
We have validated these consumption estimates in three ways. First, we affirmed that the
consumption distribution broadly matches the 2011–12 IHDS and 2012 NSS. The consumption
distribution in each dataset is shown in Figure 1. For the sake of comparability, the locations used
are Population Census villages in the SHRUG and the closest analog in the other two datasets (first
stage unit in the NSS and primary sampling unit in the IHDS).19 The figure shows that the SHRUG
has lower variance than the IHDS and NSS but otherwise a similar distribution. These differences are
mechanically related to the construction process of small area estimates through several channels. First,
the SHRUG uses predicted consumption, and is thus lacking the error term found in the other datasets.
Second, the observation counts per location in the NSS and IHDS are far lower than in the SHRUG,
so outlier households have a smaller effect on the variance in village-level consumption in the SHRUG.
Third, the SECC asset measures aim to identify poor households and do not include luxury goods.
Households in the top percentiles of the consumption distribution are thus effectively topcoded in the
SHRUG. Similarly, the structure of the asset regression is such that there is a minimum estimated
household consumption quantity, which is the same for all households with none of the assets on the
asset list. But in the NSS and IHDS there is no mechanical lower limit for household consumption.
The SHRUG therefore has no households with either extremely high or extremely low consumption.
Second, we compared average consumption in SHRUG districts and IHDS districts. Figure 2
presents a binscatter showing that average consumption covaries very strongly across these two datasets
in both rural and urban areas. Finally, we decomposed the difference in average consumption between
the SHRUG and the IHDS. Appendix Tables A3 and A4 compare the mean of each component of the
consumption index in the SHRUG and the IHDS. Columns 1 and 2 show the share of households with
18
$2 threshold converted in PPP to Indian rupees in the following way. It should be noted that GDP per capita
(current US$) of India in 2012 is 1,446.985, GDP per capita PPP of India in 2012 is 4,916.486, and exchange rate
between Indian rupee and US$ in 2012 is 53.427 according to the World Bank. Thus, $2 PPP is equal to 31 rupees
(2∗1,446.985/4,916.486∗53.427) or 31 rupees per day. 27.2/33.33 rural/urban classification is from the 2011 Suresh
Tendulkar Committee report (Planning Commission of India, 2009).
19
Each observation in this density plot is a village. For the sample surveys, we apply sampling weights when
calculating village means (to take into account stratification on affluence and occupation), and when plotting densities
(taking into account cross-village stratification on village size and oversampling in some states).

15
each characteristic in both datasets; Column 3 shows the difference. Column 4 shows the coefficient
on consumption from the prediction regression in the IHDS. Column 5 shows the average difference in
Indian Rupees between consumption estimates in the SHRUG and consumption as measured by the
IHDS, that is driven by this component of the regression.20 There are some large differences in the roof
and wall material variables, suggesting that the two datasets classified materials slightly differently;
but these differences cancel each other out and thus create little spread between SHRUG and IHDS.
The transformation of household assets into consumption values assumes a similar relationship
between consumption and asset ownership in the IHDS and the SHRUG. But there is no mechan-
ical link between the asset ownership shares in the two datasets; the similarity in the estimated
consumption distributions is therefore informative and gives us confidence that the consumption
measures in the SHRUG are good proxies for these direct survey measures.

3 Analysis: The High-Resolution Geography of Development

3.1 What do Night Lights Proxy?

The use of satellite imagery as a data source on local development has been a major innovation
in the last 20 years. The most popular form of satellite imagery in economics has been the use of
night lights as a proxy for cross-national GDP (Henderson et al., 2011). Night lights have been
particularly valuable for the study of low-income countries because they offer a high geographic
resolution proxy of development where no other data may be available.
The research has not been entirely consistent on what exactly night lights are proxying. Night
lights have been assumed in various studies to represent GDP growth (Henderson et al., 2011),
cross-sectional GDP (Bleakley and Lin, 2012), urban extent (Harari, 2020; Baum-Snow et al., 2017),
public expenditure (Hodler and Raschky, 2014), rural electrification (either supply or demand) (Min
et al., 2013; Baskaran et al., 2015; Burlig and Preonas, 2016; Mahadevan, 2019). Many of these studies
work from an underlying assumption of a constant elasticity between night lights and the outcome of
interest in the cross-section and in the time series, as well as across different geographic scales of analysis.
A common approach is to show that night lights are correlated with outcomes at an aggregate scale (e.g.
20
The average exchange rate in 2012 was 53 INR to USD.

16
across countries), and then to use night lights as a proxy for the same variable at a microgeographic
scale, where data to test the relationship between night lights and the outcome do not exist.
In this section, we show that: (i) night lights are indeed strongly and significantly correlated with
many of their presumed proxies even at the highest geographic resolution; (ii) because night lights
are independently correlated with a wide range of correlates of development, it is difficult from light
data alone to determine exactly which proxy is being measured; (iii) night light elasticities vary
widely across outcome and across context; and (iv) high geographic resolution night light elasticities
are an order of magnitude smaller in the time series than in the cross section.
We combine data on village-level luminosity with population, employment in manufacturing and
services (or combined), electricity supply, and per capita consumption. The night lights variable
in each case is the log of average luminosity in a location polygon, where luminosity is a value from
0–63 observed in each 1km x 1km grid cell. We focus on a strictly rural sample because our measure
of electricity is not comparable across villages and towns in the cross-section.21

3.1.1 Night Light and Development in Cross Section

Table 3 describes the relationship between each outcome variable and luminosity, controlling only
for the area of the geographic unit. The estimating equation, based on Henderson et al. (2011), is:

ln(Yi)=β0 +β1ln(lighti)+β2ln(numcellsi)+i, (1)

where Yi is the outcome variable, lighti is the average pixel luminosity in the geographic polygon,
numcellsi is the village area, and i is an error term.22 The unit of analysis i is a district (Column
1), subdistrict (Column 2), or village (Columns 3 and 4). Columns 3 and 4 respectively include
21
In villages in 1991 and 2001 we use an indicator variable for whether the village had any power supply. In
2011 we use the number of hours of electricity available per day. The village directories report hours available in
winter and in summer; we take the mean of these measures. All of these power variables are normalized in this
analysis to make their spatial variation maximally comparable; however, it is difficult to determine whether power
has increased between 2001 and 2011 because of the change in the measure used. The electrification measures used
in the town directories (i.e. for urban locations) are the number of electricity connections for different types of users.
With these measures, it is not possible to compare the difference in electrification between a village and a town.
All of these variables are from the census village directories.
22
The area control is the number of 1km x 1km cells in the geographic unit. We include this to omit mechanical
effects of geographic size of administrative units. Columns 2–4 are clustered at the district level.

17
district and subdistrict fixed effects to limit variation in the analysis to microgeographic variation
within these larger units. All variables are measured in logs, so the coefficients can be interpreted
as elasticities.23 Each entry in the table shows a coefficient β1 from a separate regression.
Night lights are extremely highly correlated with local population, employment in manufacturing
and services, electrification, and per capita consumption, at all levels of geographic aggregation.
Even when variation is limited to be within each of India’s 5000 subdistricts (approximately 100
hundred villages each, Column 4), night lights are extremely strong predictors of all six variables.24
The validity of night lights at high geographic resolution for all these outcomes has been presumed
to be true by a large number of papers using night lights as local proxies of development, but has
seen little testing in a developing country context. Further, the log-log relationship between the
outcomes and night lights is broadly linear across the distribution for all of the variables, as shown
in Figure 3, even at the lowest levels of development.
However, the elasticities are far from constant across geographic scales, and vary widely across
variables. The cross-district and cross-subdistrict elasticity of population with respect to night lights
is 1.3; the within-district and within-subdistrict elasticity is less than a third of that at 0.4. For
employment and power supply, we also find that elasticities at aggregate levels are more than twice
as high as the local elasticities. For per capita consumption, the elasticites are much less variable
but still vary by a factor of two.
The ordering of the elasticities across variables is consistent across all four specifications. The
outcome with the highest elasticity with respect to night lights is employment, followed by population,
then power supply, and finally per capita consumption. In all specifications, the employment
elasticity is more than eight times higher than the consumption elasticity. At the village level, a 50%
increase in night lights in associated with a 26% increase in non-farm employment, a 12% increase
23
In all log specifications, we add one to the employment and power variables before taking logs (reflecting
respectively, one job and one hour of daily electricity), so that zeroes are not dropped. Results are virtually identical if
we use the inverse hyperbolic sine transformation or add a different quantity (e.g. 0.1 hours of electricity). Population
and consumption are never zero, and none of the quantities are zero at the aggregate levels of subdistrict and district.
24
These results are robust to using both level and standardized measures of population, employment, consumption
and power. Measuring population and employment in levels results in a much worse fit (lower correlations, regression
coefficients, and r-squared) because of the presence of outliers.

18
in population, an 8% increase in power, and a 3% increase in per capita consumption.
At high geographic resolution, all of these elasticities remain statistically significant even after
controlling for the other potential proxy variables, with the partial elasticities ranging between 0.05
(for consumption) and 0.71 (for non-farm employment) (Appendix Table A5). A key result is that
night lights cannot be interpreted as a specific proxy for only one of these outcomes in isolation;
they reflect a combination of development outcomes. In any given context where night lights are
used as a proxy for development, it is difficult to determine whether they are describing population,
employment, power, or expenditure. They could be performing different roles in different places,
making interpretation difficult.
Night light elasticities are not constant across place. Figure 4 shows that for all outcomes other
than consumption, night light elasticities are 50–100% higher in electrified villages; for consumption,
the elasticity is similar in electrified and non-electrified villages.25

3.1.2 Night Light and Development in the Time Series

Many of the papers that use night lights as an outcome variable focus on changes in light rather than
cross-sectional differences in light across space.26 It is therefore important to understand whether
elasticities in the time series are similar to elasticities in the cross section. This relationship might not
hold if the elasticities above are driven in part by correlations with additional unobserved variables. It
also might not hold if the covariates have different variance in cross-section and time series; for instance,
in a period where there are very few changes in electricity supply, time series night light elasticities
with respect to electricity supply will be relatively smaller compared with the other elasticities. In
this section, we estimate night light elasticities in a time series setup with village fixed effects.
To align the years across data sources, we extrapolate the values from the 1991, 2001, and 2011
Population Censuses (population and electricity) to the Economic Census years of 1990, 1998, 2005,
and 2013.27 Our only measure of per capita consumption is in 2012, so it is not possible to include
25
Specifically, the sample is split according to whether a village reports non-zero hours of electricity.
26
For instance, many programs are evaluated with models with location fixed effects, which effectively compare
night light changes in treated places to non-treated places.
27
We assume a constant rate of population growth between census years and extrapolate the 1991–2001 growth
rate for 1990 and the 2001–2011 growth rate for 2013. We use a linear extrapolation/interpolation for the standardized

19
changes in consumption in this analysis.
Table 4 shows the partial correlation between each outcome variable and luminosity, at various
levels of geographic aggregation. Each coefficient in the table is the coefficient β1 from a single
regression that takes the form:

ln(Yi,t)=β0 +β1ln(lighti,t)+νi +ψt +i,t. (2)

Yi,t and lighti,t are defined as above. This specification includes fixed effects for year (ψt) and location
(νi), such that all variation picked up by β1 is within the analysis unit over time. Columns 1 and 2
show the aggregate district and subdistrict level regressions, while Columns 3 and 4 show village-level
regressions with subdistrict*year and district*year fixed effects respectively. These fixed effects
isolate the relationship between changes in light and development outcomes from regional changes.
Elasticities vary substantially at different geographic scales. Non-farm employment elasticities
are substantially higher at higher levels of aggregation than at more geographically fine units. At
the village level, once district-year fixed effects are taken out and variation is restricted to changes
across villages within districts, the employment elasticities fall to 0.03 and lower. The population
elasticities are very small in all specifications and are less than 0.01 at the village level. Changes
in power supply are in fact negatively correlated with changes in night lights at aggregate geographic
scales, but have an elasticity of 0.03–0.04 at the village level.
In spite of substantially smaller elasticities, the relationship between night lights and all covariates
remains highly significant in the expected direction at the village level, with the exception of
manufacturing employment in the within-district regression. Across all specifications, night lights
have higher elasticities with respect to employment in the services sector than in the manufacturing
sector, reversing the pattern observed in the cross-section. Services employment in rural India is
concentrated in education, health and retail; the higher elasticity of lights with respect to services
may suggest that night lights are picking up changes in consumption and expenditure patterns
electrification measure.

20
(because higher demand results in a larger retail sector), rather than concentrations of production.28

3.1.3 Discussion: What Can We Learn from Night Lights?

The results in this section provide clarification and nuance for analyses that use night lights as a
proxy for development. Night lights are predominantly used to estimate measures of development
at high geographic resolution where other development proxies are unavailable. As such, they are
often used exactly in contexts where the assumption that lights will be associated with various
development measures at a fine geographic scale is difficult to test.
The results shown here support the idea that night lights do indeed describe highly local variation
in development in contexts where little data is available. We have shown that night lights have
strong cross-sectional associations with all of the variables that they are presumed to proxy in the
literature, even at a highly local scale and in a very rural environment. These correlations are to some
extent independent: population, employment, electricity, and per capita consumption are all strongly
correlated with lights, even after controlling for each other. The relationships are strongly log-linear in
the cross section, holding equally for both rich and poor places, as well as well-lit and poorly-lit places.
This finding is therefore supportive of the dominant methodology in the research of interpreting
night lights as a valid high geographic resolution proxy for a range of development outcomes.
However, transforming the coefficients from night lights into specific development measures is
a fraught exercise. Night lights are independently correlated with all of the covariates described
here. If a policy or program is found to be correlated with increased luminosity, then it will be
difficult for researchers to determine whether that correlation is proxying for population, employment,
electrification, per capita consumption, or something else.
Importantly, our results have implications for the application of luminosity-to-GDP elasticities
from the literature to new contexts. Many studies of night lights estimate time series regressions with
place and year fixed effects. Our results suggest that time series elasticities in these contexts may
be an order of magnitude smaller than the widely reported cross-sectional elasticities. This would
28
In Appendix Table A6, we show that these results are robust to limiting the analysis to the census years with
the highest quality data (2001–2013) and to weighting regressions by population.

21
imply that program effects measured with night lights may in fact be much smaller than presumed.
Further, night light elasticities with respect to various development measures are highly variable
across different levels of aggregation and in different places. For instance, we found that elasticities
can be twice as large in places that are electrified. Our analysis suggests that the application of
rule of thumb conversions between night lights and GDP at highly disaggregated geographic levels
in developing countries should be used with extreme caution.

3.2 What is the Distribution of Jobs across Villages and Towns?

Economic opportunities are strongly contingent on local labor and goods markets. Given high
transportation costs in developing countries, the relevant labor market for an individual may be
geographically very small. Analysis of regional opportunities based on large geographic aggregates
may therefore not reflect the opportunities available to most individuals in a region, especially if
economic opportunities are not evenly distributed across space.
In Figure 5, we rank each town and village according to the number of nonfarm jobs per 1000
people (henceforth job density), and plot this measure in logs against the town and village rank.29
Rank percentiles are adjusted for population, such that the 20th percentile village has higher job
density than villages representing the most job-dense 20% of the rural population. If jobs were
equally distributed across space, the job density would be constant and these lines would have
a slope of zero. Instead, the distribution of job density follows a power law as evidenced by the
linearity of the function across most of its support. In other words, highly ranked places have orders
of magnitude higher job density than lower ranked places.
One way to characterize this relationship is with the coefficient from a regression of location job
density on location density rank; this is the slope of the curve in Figure 5. We estimate the slope
of this function within percentiles 10 and 90 to mitigate the effect of outliers, though we find similar
estimates if we include all data points.
For urban manufacturing jobs, the coefficient is -0.019. Moving from the 20th percentile town
to the 80th percentile town changes the manufacturing job density from 48 jobs per 1000 people
29
On average there are 22 manufacturing jobs and 61 service jobs per 1000 people in India.

22
to 16 jobs per 1000 people. Urban service sector jobs are more evenly distributed, with a regression
coefficient of -0.012. Jobs in rural areas are much more concentrated, especially in manufacturing;
the coefficients for rural manufacturing and services jobs are respectively -0.04 and -0.027. The
20th percentile village has 17 manufacturing jobs per 1000 people (just above the national rural
average), while the 80th percentile village has less than 1. Table 5 summarizes the concentration
of job density, showing the difference between mean and median job density, as well as the slope
of the regression function described above with and without district fixed effects.
The uneven distribution of jobs, especially rural manufacturing, is consequential for local labor
markets. The mean village has 15 manufacturing jobs per 1000 people, but the median has only a
third of that. 79% of rural Indians live in villages with fewer manufacturing sector opportunities than
what is implied by mean values. These results are not driven by variation across districts; as shown
in Table 5, the concentration measures are very similar when calculated with district fixed effects.
The power law distribution of rural manufacturing is important in many domains and implies
that analysis at higher geographic aggregations may be misleading. District means may not reflect
the experiences of the overwhelming majority of individuals.

3.3 How Localized is Poverty?

Development programs are often geographic in scope, targeting regions which are under-serviced,
poor, or populated by marginalized groups. In India, policies are often targeted to the district
and subdistrict level; for instance, India’s District Primary Education Program (DPEP) built
schools, hired teachers, and improved infrastructure in districts with below median literacy rates
(Khanna, 2017). India’s national and state governments have long maintained lists of so-called
“backward” districts and blocks which affect eligibility for a range of programs (Kumar, 2020).30
Targeting programs at high levels of geographic aggregation may be desirable for implementation
reasons or simply because higher resolution data is not available to policymakers. The cost of allocating
programs at aggregate levels is that needy individuals will be left out if they live in better off regions.
30
There are on average 10 blocks per district (compared with 1000 villages per district), making them comparable
to subdistricts, the primary intermediate unit in the Indian Census and used here.

23
 Both policy planning and evaluation are difficult in the absence of high geographic resolution
data, such that the extent of mistargeting driven by geographic aggregation is often not even known.
If poverty is primarily a phenomenon of broad regions, then district-level targeting may be almost as
efficient as village-level targeting. If variation in poverty is mostly local, then district-level targeting
could be much worse. India’s primary sample survey with broad geographic coverage, the National
Sample Survey, is minimally helpful in this regard, because it is representative at the district level
and does not sample enough individuals within primary sampling units to describe the geographic
distribution of development at geographically finer levels.
In this subsection, we examine the effects of using different geographic aggregations on the efficiency
of a hypothetical program targeting individuals who live below the poverty line. Our analysis
follows an established literature examining the social benefit of targeting anti-poverty programs
at various geographic scales (Ravallion, 1993; Baker and Grosh, 1994; Bigman et al., 2000; Bigman
and Srinivasan, 2002; Elbers et al., 2007; Brown et al., 2019).
For each level of aggregation (district, subdistrict, and village), we rank the locations by poverty rate.
Ranks are adjusted for population such that a rank X implies that X% of the national population
lives in a location with a higher poverty rate. Figure 6 shows the average poverty rate at each rank,
based on the geographical granularity of the ranking unit. The poverty rate in the 5th percentile
district is 43%; it is 47% in the 5th percentile subdistrict, and 57% in the 5th percentile village.
If the hypothetical program was targeted to the bottom 25% of the population as ranked by
the regional poverty rate, then it would reach 34.6% of poor people if targeted by state, 40.4% if
targeted by district, 42.4% if targeted by subdistrict, and 49.1% if targeted by village. Relative
to district targeting, we find that targeting at the village level could allow an additional 21% more
individuals below the poverty line to receive program benefits, a greater improvement than that
of moving from state to district targeting.
Even though subdistricts are 10 times smaller than districts, the targeting gain in moving from
district to subdistrict is small, but the gain in moving to village-level targeting is much higher. This
effect size is not something that could be measured from district-level data because the geographic

24
aggregate at which poverty is most concentrated is not known a priori.
A caveat to all of these measures is that targeting exercises based on small area estimates
necessarily rely on assumptions of a uniform relationship between assets and consumption that may
not hold (Tarozzi and Deaton, 2009). Standard errors on targeting estimates are therefore likely
to be underestimated.

4 Comparative Advantages and Disadvantages of the SHRUG

The SHRUG has two main advantages relative to other data sources. First, it describes a wide set
of socioeconomic outcomes over a long period for a wider set of high geographic resolution locations
than any other Indian panel dataset with broad scope.31 This enables analysis of outcomes at the
level of policy and program variation, which is often at the village, town or constituency level.
Second, because the SHRUG’s geographic coverage is comprehensive, it is particularly valuable
when used in tandem with other datasets. Each new administrative or remote sensing data source
that is linked to the SHRUG can be fully integrated with all the other data sources in the SHRUG,
expanding the scope of potential analysis. In contrast, sample surveys run by different research
teams do not have the same complementarities, because there is rarely enough sample overlap for
the same locations to appear in both datasets. Two research teams integrating new administrative
or remote sensing data into the framework of the SHRUG will immediately be able to benefit from
each others’ work due to the linked identifiers and comprehensive geographic coverage. For this
reason, the utility of the SHRUG will increase over time as we continue to extend its breadth.
Like every data source, the SHRUG has several limitations. First, the length of most of the surveys
underlying the SHRUG is smaller than in many other data sources. Because the administrative
censuses are implemented for every household and non-farm firm in India, they are necessarily based on
much shorter surveys than detailed sample surveys like the NSS and ASI. This disadvantage is traded
31
Other specialized data sources have high geographic resolution but cover more narrow topics. For example,
Prowess (maintained by the Center for Monitoring of the Indian Economy) describes the operations of large firms
and contains headquarters addresses. The SHRUG can in principle be linked to Prowess at the village and town
level using georeferences, but we have not yet created a key. There are also several village-level time series datasets
with broad and deep surveys, chiefly ARIS-REDS and ICRISAT. The data in these surveys is much more detailed
than the data found in the SHRUG, but they cover at most a few hundred villages.

25
off against the high geographic precision and wide breadth of data available for towns and villages across
all of the modules of the SHRUG. An NSS consumption survey is much more detailed than the SECC;
but the short asset survey in SHRUG can be analyzed in conjunction with data on night lights, forest
cover, administrative programs, public goods and local firms, across several hundred thousand villages.
Second, the SHRUG consists of locational aggregates, rather than individual or firm microdata.
It is thus not suitable for studying variation across individuals or firms within locations. Note that
we have provided keys to link SHRUG to the firm-level Economic Censuses, which describe every
non-farm establishment in India in four cross sections from 1990 to 2013.
Third, not all villages and towns are matched in all periods. If a researcher’s goal was to estimate
the number of non-agricultural firms in India, for example, then aggregating from NSS or Annual
Survey of Industries (ASI) samples is arguably a better approach, because there are no missing
locations. Economic Census data in the SHRUG has a slight rural bias because rural boundaries are
more easily tracked over time than urban boundaries, and many towns are missing economic data
for 1990. Similarly, as noted above, we were not able to obtain census-based data for predominantly
urban constituencies.
Fourth, the SHRUG is only as good as the collection process for the administrative data that
underlies it. The NSS enumerators spend far more time with each firm owner than the Economic
Census enumerators, and have more quality checks and cross validations in their survey process. Some
of the outliers in the Economic Census (and thus the SHRUG) are almost surely incorrect. This is
inherent in the nature of the process of collecting data from hundreds of millions of respondents in a
short time period. We offer some suggestions in the codebook on how to deal with these observations.
Finally, most of the data in SHRUG is available only for the years when large-scale administrative
data were collected. Consumption data is available only in 2012, and none of the four Economic
Census years are exact matches for the Population Census years. The remote sensing data, however,
is available annually.
Administrative data by no means obviates the need for high quality sample field surveys. Whether
the strengths or the limitations of the SHRUG dominate will depend on the particular research

26
question. The SHRUG is particularly well-suited for research requiring high-resolution geographic
variation, rich location information, or socioeconomic outcomes in units with political boundaries.

5 Conclusions: A Model for Collaborative Data Sharing

Most researcher-initiated data collection projects have a relatively narrow scope. A local survey is
conducted for the purpose of an experimental or observational study, one or several research papers
are written, and the data is reused only for replication or in rare cases, for long-term followup.
The existence of comprehensive administrative data makes possible a paradigm where investments in
data yield many more positive externalities for other researchers. Because administrative data is often
comprehensive at the state or the national level, one researcher’s efforts at collecting and rationalizing
an administrative dataset may yield dividends to many other researchers. Many researchers are
already making use of administrative data in India and in other developing countries. In the absence of
a common platform to link these datasets to each other, there is considerable duplication of effort and
many potential complementarities and externalities across projects are not being realized.32 Our aim in
building the SHRUG is to create a common geographic frame for all these datasets, standardizing their
location identifiers and lowering the cost to researchers of creating positive externalities for each other.
Researchers often face a trade-off between sharing data, which enables more socially valuable
research, and keeping data restricted, which ensures that they will not be scooped on future projects
with that data. Some balance between these objectives is needed; private returns to developing
new data sources are desirable as motivating factors.
The increasingly stringent data policies of economics journals have been effective in generating
replication data, but that data is often posted without sufficient documentation or identifiers to be
usable for other research projects. Replication files often describe only the researcher’s final sample,
which may be a small subset of the national data on which it is based. This is not the result of
malign intent; rather, it can take weeks or months of effort to make replication data useful for future
32
Some examples of Indian administrative programs that have been studied with administrative data include the
NREGS public works and wage support program, the RGGVY rural electrification program, and the ongoing Total
Sanitation Campaign, all of which are the subject of multiple research papers. And yet none of these programs (or
research projects) have easily accessible or linkable data frames, causing each new researcher to have to reinvent the
wheel, and limiting the scope of each research project to the amount of data that its research team is willing to clean.

27
projects, and researchers may not see rewards for doing so.33
In creating the SHRUG, we aim to both lower the transaction costs of sharing data and to change
the institutional incentives around data sharing. Researchers who create data sources that are
compatible with the SHRUG will have their work cited, because the data will be more valuable
when linked with the range of fields in the SHRUG. To add weight to this incentive, the SHRUG is
released under a copyleft license that commits users to share any data that they link to the SHRUG
when their papers are accepted for publication. The time lag to publication in economics all but
ensures that researchers who develop new non-proprietary SHRUG-linked data sources will have
a large lead on any other projects working with their data.34 Our project aims to raise the returns
of data assembly, resulting in new public goods for the research and policymaking communities.
The open source software universe has demonstrated that an equilibrium can exist where highly-
skilled individuals freely share the fruits of their labor, creating valuable externalities but also
receiving enough private benefits to make these contributions worth their while. The work underlying
the SHRUG aims to model a set of norms and protocols to facilitate a similar equilibrium around
the sharing of data for social science research.
We conclude by reiterating our request to users of the SHRUG to cite all of the research papers
that underlie the different components of the SHRUG. When data is downloaded from the SHRUG
platform, a list of citations underlying the download is automatically generated. Citing these
appropriately will further increase the returns to other researchers to investing in developing new
data sources and making them easily usable by others, creating positive externalities for all.

5.1 Addendum: Updates and Contributions

The SHRUG will be regularly updated as new data is brought online and inevitable errors are found and
corrected. The latest version will be maintained at the SHRUG web site, and all prior versions will be
33
The low average quality of public replication data further creates a lemons problem: researchers who become accus-
tomed to finding minimally useful data on journal web sites may be discouraged from searching for the gems that do exist.
34
Even in the case of proprietary microdata, it is often possible to share village/town aggregates that strike
a balance between respecting restrictions on data sharing and benefiting the wider research community.

28
archived on the Harvard Dataverse, so that researchers can always replicate an exact prior download.35
Researchers wishing to make their data easily linkable to SHRUG need only to post their data with
unique SHRUG identifiers. We aim to maintain a listing of external datasets that can be linked directly
to the SHRUG in this way. Following the data organization standards and protocols described on the
SHRUG web site will help to make these contributions as accessible as possible to outside researchers.36
Research teams that assemble national datasets at the shrid or constituency level that are of wide
general interest and are extremely clean and consistent with the SHRUG format can request to have
their data maintained and released directly via our web site, which will maximize that data’s availability
to others. We have made every effort to ensure that users downloading data from our site will appropri-
ately reference all contributors to the specific datasets being downloaded, ensuring proper attribution.

35
The SHRUG web site is http://devdatalab.org/shrug. The URL for the archived versions at the Harvard
Dataverse is https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/DPESAK.
36
Contribution protocols are described in more detail at http://devdatalab.org/shrug contribute.

29
References
Adukia, Anjali, Sam Asher, and Paul Novosad, “Educational Investment Responses to Economic
Opportunity: Evidence from Indian Road Construction,” American Economic Journal: Applied
Economics (forthcoming), 2019.
Asher, Sam and Paul Novosad, “Politics and Local Economic Growth: Evidence from India,”
American Economic Journal: Applied Economics, 2017, 9 (1), 229–273.
and , “Rural Roads and Local Economic Development,” American Economic Review, 2020,
110 (3).
, Teevrat Garg, and Paul Novosad, “The Ecological Footprint of Transportation Infrastructure,”
The Economic Journal (forthcoming), 2019.
Baker, Judy L and Margaret E Grosh, “Poverty Reduction through Geographic Targeting: How
well does it work?,” World Development, 1994, 22 (7).
Baskaran, Thushyanthan, Brian Min, and Yogesh Uppal, “Election Cycles and Electricity
Provision: Evidence from a Quasi-experiment with Indian Special Elections,” Journal of Public
Economics, 2015, 126, 64–73.
Baum-Snow, Nathaniel, Loren Brandt, J. Vernon Henderson, Matthew A. Turner, and
Qinghua Zhang, “Roads, Railways and Decentralization of Chinese Cities,” Review of Economics
and Statistics, 2017, 99 (3).
Bigman, David and PV Srinivasan, “Geographical Targeting of Poverty Alleviation Programs:
Methodology and Applications in Rural India,” Journal of Policy Modeling, 2002, 24 (3).
, Stefan Dercon, Dominique Guillaume, and Michel Lambotte, “Community Targeting for
Poverty Reduction in Burkina Faso,” The World Bank Economic Review, 2000, 14 (1).
Bleakley, Hoyt and Jeffrey Lin, “Portage and Path Dependence,” The Quarterly Journal of Economics,
2012, 127 (2).
Brown, Caitlin, Martin Ravallion, and Dominique van de Walle, “Most of Africa’s Nutritionally
Deprived Women and Children are Not Found in Poor Households,” Review of Economics and
Statistics, 2019, 101 (4).
Burlig, Fiona and Louis Preonas, “Out of the Darkness and Into the Light? Development Effects
of Rural Electrification,” 2016. Working Paper.
Chhibber, Pradeep and Francesca R Jensenius, “Privileging One’s Own? Voting Patterns and
Politicized Spending in India,” 2016. Working Paper.
Elbers, Chris, Jean Lanjouw, and Peter Lanjouw, “Micro-level Estimation of Poverty and
Inequality,” Econometrica, 2003, 71 (1), 355–364.
, Tomoki Fujii, Peter Lanjouw, Berk Ozler, and Wesley Yin, “Poverty Alleviation through
Geographic Targeting,” Journal of Development Economics, 2007, 83 (1).
Ellison, Glenn and Edward L Glaeser, “Geographic Concentration in U.S. Manufacturing Industries:
A Dartboard Approach,” Journal of Political Economy, 1997, 105 (5), 889–927.
Gabaix, Xavier, “Power Laws in Economics: An Introduction,” Journal of Economic Perspectives, 2016,
30 (1).
Harari, Nina, “Cities in Bad Shape: Urban Geometry in India,” American Economic Review, 2020, 110 (8).
Henderson, J. Vernon, Adam Storeygard, and David N. Weil, “A Bright Idea for Measuring
Economic Growth,” American Economic Review, 2011, 101 (3), 194–199.
Hodler, Roland and Paul A Raschky, “Regional Favoritism,” The Quarterly Journal of Economics,
2014, 129 (2).
Iyer, Lakshmi and Maya Reddy, “Redrawing the Lines: Did Political Incumbents Influence Electoral Re-
districting in the World’s Largest Democracy?,” 2013. Harvard Business School Working Paper 14-051.

30
Jensenius, Francesca, Social Justice through Inclusion 2017.
Khanna, Gaurav, “Large-scale Education Reform in General Equilibrium: Regression Discontinuity
Evidence from India,” 2017.
Kumar, Hemanshu and Rohini Somanathan, “State and District Boundary Changes in India:
1961-2001,” 2015. Working Paper.
Kumar, Naveen, “Public School Quality and Student Outcomes: Evidence from Model Schools in India,”
2020. Working paper.
Lehne, Jonathan, Jacob Shapiro, and Oliver Vanden Eynde, “Building Connections: Political
Corruption and Road Construction in India,” Journal of Development Economics, 2018, 131, 62–78.
Mahadevan, Meera, “The Price of Power: Costs of Political Corruption in Indian Electricity,” 2019.
Working paper.
Min, Brian, Kwawu Mensan Gaba, Ousmane Fall Sarr, and Alassane Agalassou, “Detection
of Rural Electrification in Africa using DMSP-OLS Night Lights Imagery,” International Journal
of Remote Sensing, 2013, 34 (22).
Muralidharan, Karthik, Paul Niehaus, and Sandip Sukhtankar, “General equilibrium effects of
(improving) public employment programs: Experimental evidence from india,” 2017.
Planning Commission of India, “Report of the Expert Group on Methodology for Estimation of
Poverty (Tendulkar Committee Report),” Technical Report, New Delhi, India 2009.
Prakash, Nishith, Marc Rockmore, and Yogesh Uppal, “Do Criminally Accused Politicians Affect
Economic Outcomes? Evidence from India,” Journal of Development Economics, 2019.
Ravallion, Martin, “Poverty alleviation through regional targeting: a case study of Indonesia,” in
A Braverman, K Hoff, and J Stiglitz, eds., The Economics of Rural Organization: Theory, Practice
and Policy, World Bank and Oxford University Press, 1993.
Tarozzi, Alessandro and Angus Deaton, “Using Census and Survey Data to Estimate Poverty and
Inequality for Small Areas,” The review of economics and statistics, 2009, 91 (4).
Townshend, J., M. Hansen, M. Carroll, C. DiMiceli, R Sohlberg, and C. Huang, “User
Guide for thet MODIS Vegetation Continuous Fields product, Collection 5 Version 1,” Collection
5, University of Maryland, College Park, Maryland, 2011.

31
 Table 1
Geographic Variance Decomposition of SHRUG Variables

Panel A. Share of Variance in Town Outcomes Explained by State and District

State District
Consumption per capita 0.392 0.640
Nonfarm employment per Capita 0.102 0.254
Night lights per capita 0.056 0.203
Population density 0.162 0.304
Average forest cover 0.465 0.736

Panel B. Share of Variance in Village Outcomes Explained by State, District, Subdistrict

State District Subdistrict
Consumption per capita 0.321 0.450 0.526
Nonfarm employment per Capita 0.021 0.044 0.075
Night lights per capita 0.072 0.141 0.181
Population density 0.202 0.290 0.338
Average forest cover 0.472 0.640 0.729

Table 1 presents the spatial decomposition of the variance of a selection of variables in the SHRUG. Stated values
are the R2 from regressions of each variable on a set of, respectively, state, district, and subdistrict level fixed effects.
Employment per capita is calculated by dividing 2013 Economic Census (EC13) nonfarm employment by the total
population from the 2011 Population Census. Night lights per capita is calculated by dividing calibrated total light
from 2013 by the 2011 Census population. Average forest cover is a measure derived from MODIS VCF. Population
density is the 2001 total population divided by 2001 Census land area.

32
 Table 2
SHRUG Summary
Panel A. Data in the SHRUG
Dataset Years Description Units of observation
Population Census 1991, 2001, 2011 Demographic data, social groups, village & town amenities, Village, Town, Constituency, District
including electrification, roads, post offices, markets,
health centers, schools, land use, water source, etc.
Economic Census 1990, 1998, 2005, 2013 Employment and sector of all non-ag firms, including Village, Town, Constituency
manufacturing and services, formal and informal,
government and private firms
SECC 2012 Consumption estimates, poverty, agricultural labor share Village, Town
Election Results 1980–2018 Candidate name / party / votes Constituency/Candidate
Politician Assets/Crime 2003–2018 Criminal charges, assets, liabilities, education Constituency/Candidate
for all politicians contesting MLA seats
Night Lights 1994–2013 Proxy for electrification and economic activity Village, Town, Constituency
Forest Cover 2000–2014 % Tree cover from Vegetation Continuous Fields Village, Town, Constituency
Rural Road Construction 2000–2013 Administrative data from PMGSY, including Village
33

length, material, cost, milestones, etc.

 Table 3
Cross-Sectional Correlates of Night Lights

District Subdistrict Village Village

(1) (2) (3) (4)
Log Population 1.377*** 1.257*** 0.433*** 0.407***
(0.124) (0.034) (0.003) (0.003)
Log Non-Farm Employment 1.755*** 1.740*** 0.711*** 0.697***
(0.156) (0.049) (0.005) (0.005)
Log Manufacturing Employment 2.291*** 2.090*** 0.714*** 0.687***
(0.174) (0.055) (0.005) (0.006)
Log Services Employment 1.625*** 1.669*** 0.683*** 0.660***
(0.152) (0.047) (0.004) (0.005)
Log Hours Electricity 0.604*** 0.553*** 0.264*** 0.215***
(0.092) (0.031) (0.003) (0.004)
Log Consumption 0.167*** 0.114*** 0.105*** 0.086***
(0.027) (0.008) (0.005) (0.004)
N 632 5756 430982 430788
Fixed Effects None None District Subdistrict
∗
p<0.10,∗∗p<0.05,∗∗∗p<0.01

Table 3 shows the cross-sectional relationship between a set of rural development outcomes and night-time luminosity
at various levels of aggregation. Each coefficient in the table is from a separate estimation of Equation 1—a regression
of the log outcome variable on log luminosity and the log polygon size (the number of 1km x 1km cells in the
geographic unit). Column 1 aggregates data to the district level, Column 2 to the subdistrict level, and Columns
3 and 4 to the village level. Columns 2-4 are clustered at the district level. Standard errors on consumption are
calculated using 1000 bootstraps as described in Section 2.4. Source: SHRUG.

34
 Table 4
Time Series Correlates of Night Lights

District Subdistrict Village Village

(1) (2) (3) (4)
Log Population 0.013 0.026*** 0.006*** 0.004***
(0.011) (0.008) (0.001) (0.001)
Log Non-Farm Employment 0.555*** 0.346*** 0.020*** 0.021***
(0.130) (0.082) (0.003) (0.002)
Log Manufacturing Employment 0.398*** 0.202*** 0.006 0.009***
(0.113) (0.075) (0.004) (0.003)
Log Services Employment 0.513*** 0.329*** 0.019*** 0.020***
(0.124) (0.077) (0.003) (0.002)
Electricity -0.147*** -0.106*** 0.038*** 0.031***
(0.047) (0.034) (0.004) (0.003)
N 2358 22060 1479176 1478715
Fixed Effects District, Subdistrict, Village, Village,
Year Year District * Year Subdistrict * Year
∗
p<0.10,∗∗p<0.05,∗∗∗p<0.01

Table 4 shows the time series relationship between a set of rural development outcomes and night-time luminosity
at various levels of aggregation, controlling for location fixed effects. Each coefficient in the table is from a separate
estimation of Equation 2—a regression of the log outcome variable on log luminosity, with location and time fixed
effects. Column 1 aggregates data to the district level, Column 2 to the subdistrict level, and Columns 3 and 4
to the village level. All regressions are clustered at the district level. Source: SHRUG.

Table 5
Concentration of Non-Farm Employment in Villages and Towns
Mean Employment Median Employment Slope Slope
per 1000 ppl per 1000 ppl (National) (Within District)
Towns
Manufacturing 39.1 26.6 -0.019 -0.012
Services 115.8 104.0 -0.012 -0.009
Villages
Manufacturing 15.3 4.8 -0.040 -0.035
Services 40.2 26.0 -0.027 -0.021
Table 5 shows several statistics describing the distribution of non-farm employment in villages and towns, disaggregated
by manufacturing and services. Columns 1 and 2 are mean and median employment per 1000 people, across the
distribution of towns and villages. The mean is population-weighted, and the median refers to the town or village
with the median individual, so that 50% of the national population lives in places with above-median employment
per capita. Column 3 shows the slope from a regression of employment per 1000 people on the town or village
rank, where each location’s rank is a function of employment per 1000 people in the same category. Column 4
shows the slope from the same regression, but with district fixed effects. Source: SHRUG and Economic Census.

35
 Figure 1
SHRUG, IHDS, and NSS Consumption Per Capita

A. Rural
.0001
SHRUG
NSS
IHDS
.00008

.00006
Kernel Density

.00004

.00002

0
0 10000 20000 30000 40000 50000
Per Capita Consumption (Rs)

A. Urban
.00006 SHRUG
NSS
IHDS

.00004
Kernel Density

.00002

0
0 10000 20000 30000 40000 50000
Per Capita Consumption (Rs)

Figure 1 shows the distribution of location-level consumption per capita from the NSS,
the IHDS, and SHRUG. Consumption in each location is a mean across households,
calculated with sampling weights. Location units are PSUs in IHDS, FSUs in NSS, and
towns or villages in SHRUG.

36
 Figure 2
District-level SHRUG Consumption vs. IHDS Consumption

A. Rural
40000

35000
IHDS Consumption (Rs.)

30000

25000

20000

15000

10000 15000 20000 25000 30000

SHRUG Consumption (Rs.)

B. Urban

45000

40000
IHDS Consumption (Rs.)

35000

30000

25000

20000 25000 30000 35000 40000 45000

SHRUG Consumption (Rs.)

Figure 2 is a binned scatterplot of district-level average per-capita consumption in the

IHDS and the SHRUG, weighted by the count of individuals in each survey. The SHRUG
sample has been restricted to the 331 districts where IHDS matches the Indian Population
Census.

37
 Figure 3
Cross-sectional Night Lights vs. Development Proxies

Log Population Log Employment

7.6
4.5

7.4

Log Employment
Log Population

4
7.2

7
3.5
6.8

6.6
3
1 1.5 2 2.5 3 1 1.5 2 2.5 3
Log Luminosity Log Luminosity
Log Daily Hours of Electricity Log Monthly Consumption
9.85
1.7
Log Daily Hours of Electricity

9.8
Log Monthly Consumption

1.6

9.75

1.5
9.7

1.4
9.65

1.3 9.6
1 1.5 2 2.5 3 1 1.5 2 2.5 3
Log Luminosity Log Luminosity

Figure 3 shows the cross-sectional relationship between log night lights and log population
(top left), log employment (top right), log hours of electricity (bottom left), and log
monthly consumption (bottom right). The graphs are generated by binscatter, which
shows the mean Y variable in equally-dense bins across the distribution of log night lights.
All graphs are run at the village level and include district fixed effects. Source: SHRUG.

38
 Figure 4
Cross-sectional Relationship between Night Lights and Development Proxies
as a Function of Electrification

Log Population Log Monthly Consumption

7.4 9.85
Slope: 0.11
Slope: 0.33
9.8

Log Monthly Consumption

7.2
Log Population

9.75 Slope: 0.10

7
Slope: 0.19
9.7

6.8
9.65

6.6
9.6
1 1.5 2 2.5 3 1 1.5 2 2.5 3
Log Luminosity Log Luminosity
Log Manufacturing Employment Log Services Employment
2.5 4
Slope: 0.68
Slope: 0.60
Log Manufacturing Employment

Log Services Employment

2 3.5

Slope: 0.38
Slope: 0.35

1.5 3

1 2.5
1 1.5 2 2.5 3 1 1.5 2 2.5 3
Log Luminosity Log Luminosity
Villages without Power Villages with Power

Figure 4 shows the village-level cross-sectional relationship between four development

proxies (population, consumption, manufacturing and services employment), with separate
series’ plotted for villages that do and do not report access to electricity. The graphs are
generated by binscatter, which shows the mean Y variable in equally-dense bins across
the distribution of log night lights. All graphs are run at the village level and include
district fixed effects. Source: SHRUG.

39
 Figure 5
Distribution of Job Density across Space

Manufacturing Jobs Across Towns Manufacturing Jobs Across Villages

8 8
Manufacturing jobs per 1000 people

Manufacturing jobs per 1000 people

6 6

4 4

2 2

0 0
0 20 40 60 80 100 0 20 40 60 80 100
Percentile rank (7172 towns) Percentile rank (511912 villages)
Services Jobs Across Towns Services Jobs Across Villages
8 8
Services jobs per 1000 people

Services jobs per 1000 people

6 6

4 4

2 2

0 0
0 20 40 60 80 100 0 20 40 60 80 100
Percentile rank (7172 towns) Percentile rank (511912 villages)

Figure 5 describes the distribution of non-farm jobs per 1000 people across villages and
towns. The Y axes shows the log number of manufacturing- or service-sector jobs per
1000 people. The X axis shows the town or village rank on the same measure. The
rank is population-weighted such that a village with rank 50 has more manufacturing or
services jobs per 1000 people than villages representing 50% of the Indian population.
Each point represents the mean of the measure in one percentile bin, or approximately
either 70 towns or 5000 villages. Source: SHRUG and Economic Census (2013).

40
 Figure 6
Concentration of Poverty at Different Levels of Aggregation

.8

.6
Poverty rate at this location percentile

.4

.2

Village Subdistrict
District State
0
0 20 40 60 80 100
Location poverty rank

Figure 6 plots locations’ average poverty rates against their poverty rate ranking. The four
data series’ represent four different level of aggregation: state (n=35), district (n=600),
subdistrict (n=5600), and village/town (n=590,000). The graph shows, for instance, that
the poverty rate in the 5th percentile district is 37%, while the poverty rate in the 5th
percentile village is 50%. Source: SHRUG.

41
A Appendix A: Additional Tables and Figures

42
 Table A1
Population Share Matched to the SHRUG, by State
States PC91 PC01 PC11
India 826117.89 / 833122.68 (99%) 1028120.50 / 1028349.73 (100%) 1209944.68 / 1210741.69 (100%)
Andaman Nicobar Islands 280.66 / 280.66 (100%) 356.15 / 356.15 (100%) 380.55 / 380.58 (100%)
Andhra Pradesh 65140.70 / 66455.27 (98%) 76210.01 / 76210.01 (100%) 84580.78 / 84580.78 (100%)
Arunachal Pradesh 621.18 / 637.04 (98%) 1097.97 / 1097.97 (100%) 1383.17 / 1383.73 (100%)
Assam 22278.90 / 22311.78 (100%) 26640.04 / 26655.53 (100%) 30999.61 / 31205.58 (99%)
Bihar 86119.25 / 86374.47 (100%) 82825.55 / 82825.55 (100%) 104099.45 / 104099.46 (100%)
Chandigarh 642.01 / 642.01 (100%) 900.63 / 900.63 (100%) 1055.45 / 1055.45 (100%)
Chhattisgarh 20827.74 / 20833.80 (100%) 25544.25 / 25545.20 (100%)
Dadra Nagar Haveli 138.48 / 138.48 (100%) 220.49 / 220.49 (100%) 343.71 / 343.71 (100%)
Daman & Diu 101.59 / 101.59 (100%) 158.20 / 158.20 (100%) 243.25 / 243.25 (100%)
Goa 1155.51 / 1169.79 (99%) 1347.67 / 1347.67 (100%) 1458.55 / 1458.55 (100%)
Gujarat 41284.77 / 41309.58 (100%) 50671.02 / 50671.02 (100%) 60439.69 / 60439.69 (100%)
Haryana 16285.72 / 16459.98 (99%) 21139.38 / 21144.56 (100%) 25193.50 / 25351.46 (99%)
Himachal Pradesh 5165.07 / 5170.53 (100%) 6077.90 / 6077.90 (100%) 6864.45 / 6864.60 (100%)
Jammu Kashmir 10142.76 / 10143.70 (100%) 12539.86 / 12541.30 (100%)
Jharkhand 26945.83 / 26945.83 (100%) 32983.76 / 32988.13 (100%)
Karnataka 44663.16 / 44977.20 (99%) 52785.20 / 52850.56 (100%) 61032.42 / 61095.30 (100%)
Kerala 28631.18 / 29098.52 (98%) 31841.37 / 31841.37 (100%) 33406.06 / 33406.06 (100%)
Lakshadweep 51.71 / 51.71 (100%) 60.65 / 60.65 (100%) 64.47 / 64.47 (100%)
43

Madhya Pradesh 62281.73 / 63026.21 (99%) 60345.27 / 60348.03 (100%) 72626.81 / 72626.81 (100%)
Maharashtra 78363.48 / 78936.42 (99%) 96878.63 / 96878.63 (100%) 112323.51 / 112374.34 (100%)
Manipur 1806.38 / 1837.15 (98%) 2166.79 / 2166.79 (100%) 2851.43 / 2855.79 (100%)
Meghalaya 1764.66 / 1774.74 (99%) 2288.95 / 2318.82 (99%) 2961.91 / 2966.89 (100%)
Mizoram 689.54 / 689.76 (100%) 888.57 / 888.57 (100%) 1094.51 / 1097.21 (100%)
Nagaland 1207.14 / 1209.55 (100%) 1989.66 / 1990.04 (100%) 1978.50 / 1978.50 (100%)
NCT of Delhi 9420.64 / 9420.64 (100%) 13850.51 / 13850.51 (100%) 16787.94 / 16787.94 (100%)
Odisha 31515.51 / 31587.64 (100%) 36799.75 / 36804.66 (100%) 41945.54 / 41969.76 (100%)
Puducherry 771.56 / 807.78 (96%) 974.35 / 974.35 (100%) 1247.95 / 1247.95 (100%)
Punjab 19053.16 / 19053.16 (100%) 24334.90 / 24359.00 (100%) 27650.20 / 27743.34 (100%)
Rajasthan 43354.10 / 43879.50 (99%) 56502.28 / 56507.19 (100%) 68548.43 / 68548.44 (100%)
Sikkim 405.02 / 405.02 (100%) 540.85 / 540.85 (100%) 610.57 / 610.58 (100%)
Tamil Nadu 55111.89 / 55834.15 (99%) 62367.39 / 62405.68 (100%) 72117.59 / 72147.03 (100%)
Tripura 2430.67 / 2757.20 (88%) 3198.93 / 3199.20 (100%) 3666.08 / 3673.92 (100%)
Uttarakhand 166186.02 / 166197.92 (100%) 199763.41 / 199812.34 (100%)
Uttar Pradesh 138452.58 / 138837.84 (100%) 8479.34 / 8489.35 (100%) 10071.41 / 10086.29 (100%)
West Bengal 66929.92 / 67887.31 (99%) 80079.76 / 80088.56 (100%) 91085.92 / 91167.27 (100%)

Table A1 presents the state-level population included in the SHRUG panel (numerator), the state-level population in the Population Census datasets (denominator),
and the share of state-level population captured by the SHRUG, for all states and union territories in India. Population numbers are reported in thousands.
Chhattisgarh, Jharkhand, and Uttarakhand were created in 2000 and are thus left blank in earlier years.
 Table A2
Employment Share Matched to the SHRUG, by State
States EC90 EC98 EC05 EC13
India 43266.88 / 62211.08 (70%) 62851.43 / 70891.77 (89%) 79038.38 / 85388.85 (93%) 107639.65 / 110513.80 (97%)
Andaman Nicobar Islands 12.27 / 31.14 (39%) 48.32 / 48.32 (100%) 17.00 / 39.05 (44%) 61.09 / 61.21 (100%)
Andhra Pradesh 4080.46 / 5263.04 (78%) 5742.84 / 6243.11 (92%) 8568.18 / 8991.79 (95%) 10492.67 / 11563.89 (91%)
Arunachal Pradesh 13.00 / 61.86 (21%) 48.80 / 54.68 (89%) 64.96 / 81.30 (80%) 89.80 / 108.38 (83%)
Assam 994.49 / 1265.52 (79%) 1626.39 / 1914.82 (85%) 1731.44 / 2037.68 (85%) 3606.55 / 3665.87 (98%)
Bihar 2467.22 / 2915.64 (85%) 1715.85 / 2028.94 (85%) 2031.13 / 2096.17 (97%) 2929.19 / 3116.34 (94%)
Chandigarh 137.46 / 137.46 (100%) 148.16 / 148.16 (100%) 185.33 / 185.33 (100%) 244.27 / 244.27 (100%)
Chhattisgarh 1003.77 / 1154.32 (87%) 1154.25 / 1377.39 (84%) 1800.44 / 1834.96 (98%)
Dadra Nagar Haveli 13.23 / 13.23 (100%) 27.36 / 31.04 (88%) 64.61 / 64.61 (100%) 94.31 / 94.31 (100%)
Daman & Diu 18.55 / 18.55 (100%) 29.80 / 29.86 (100%) 59.84 / 59.84 (100%) 81.42 / 81.42 (100%)
Goa 87.27 / 169.84 (51%) 153.98 / 191.81 (80%) 187.36 / 208.13 (90%) 284.58 / 284.92 (100%)
Gujarat 2287.73 / 2831.85 (81%) 3676.17 / 3779.33 (97%) 3957.48 / 4412.87 (90%) 6143.60 / 6246.70 (98%)
Haryana 939.56 / 1190.77 (79%) 1052.97 / 1408.53 (75%) 1742.25 / 1950.83 (89%) 2811.10 / 2845.80 (99%)
Himachal Pradesh 324.97 / 357.05 (91%) 446.01 / 461.38 (97%) 543.54 / 552.25 (98%) 894.05 / 938.60 (95%)
Jammu Kashmir 100.83 / 430.17 (23%) 546.40 / 645.96 (85%) 1043.19 / 1065.65 (98%)
Jharkhand 866.09 / 947.85 (91%) 991.34 / 1030.31 (96%) 1377.32 / 1386.44 (99%)
Karnataka 3571.51 / 6339.23 (56%) 4069.62 / 4228.16 (96%) 5035.00 / 5165.28 (97%) 5790.34 / 5829.52 (99%)
Kerala 2223.42 / 2961.80 (75%) 585.07 / 3249.12 (18%) 2931.26 / 4309.21 (68%) 5649.97 / 5701.44 (99%)
Lakshadweep 5.87 / 12.18 (48%) 8.37 / 8.37 (100%) 9.92 / 10.24 (97%)
44

Madhya Pradesh 2867.56 / 3190.24 (90%) 3142.60 / 3325.93 (94%) 3274.40 / 3531.72 (93%) 4086.12 / 4241.05 (96%)
Maharashtra 7187.69 / 7577.37 (95%) 8134.96 / 8381.88 (97%) 9036.32 / 9526.52 (95%) 11797.80 / 11947.80 (99%)
Manipur 9.93 / 133.45 (7%) 109.61 / 167.68 (65%) 147.97 / 204.65 (72%) 353.88 / 385.92 (92%)
Meghalaya 30.52 / 126.71 (24%) 133.20 / 144.36 (92%) 179.10 / 194.70 (92%) 269.67 / 277.45 (97%)
Mizoram 46.78 / 49.23 (95%) 46.98 / 52.25 (90%) 68.40 / 70.18 (97%) 93.97 / 101.05 (93%)
Nagaland 3.67 / 98.66 (4%) 92.67 / 95.23 (97%) 114.70 / 115.90 (99%) 157.44 / 159.77 (99%)
NCT of Delhi 1860.30 / 1860.30 (100%) 3331.36 / 3331.36 (100%) 3387.83 / 3387.83 (100%) 3003.82 / 3003.82 (100%)
Odisha 738.33 / 2205.11 (33%) 1842.30 / 2738.37 (67%) 3312.57 / 3355.95 (99%) 3891.08 / 4051.32 (96%)
Puducherry 84.80 / 104.51 (81%) 143.85 / 155.09 (93%) 101.85 / 165.52 (62%) 211.31 / 213.67 (99%)
Punjab 1210.66 / 1555.16 (78%) 1844.14 / 1914.10 (96%) 2366.73 / 2399.82 (99%) 3125.31 / 3139.81 (100%)
Rajasthan 1745.15 / 2203.52 (79%) 2687.16 / 2885.55 (93%) 3288.03 / 3569.26 (92%) 4897.19 / 5165.42 (95%)
Sikkim 18.00 / 35.24 (51%) 15.69 / 33.56 (47%) 6.39 / 48.67 (13%) 84.61 / 84.65 (100%)
Tamil Nadu 976.67 / 5266.63 (19%) 5842.72 / 6377.40 (92%) 6903.60 / 8052.45 (86%) 8718.60 / 8812.22 (99%)
Tripura 0.00 / 203.84 (0%) 148.40 / 218.62 (68%) 258.37 / 324.29 (80%) 379.29 / 382.24 (99%)
Uttarakhand 354.44 / 448.05 (79%) 7249.33 / 7328.97 (99%) 11377.23 / 11422.24 (100%)
Uttar Pradesh 5406.84 / 7505.02 (72%) 6045.04 / 6283.58 (96%) 564.74 / 619.01 (91%) 800.46 / 980.15 (82%)
West Bengal 3908.86 / 6539.10 (60%) 7588.40 / 7976.98 (95%) 8958.33 / 9277.06 (97%) 10988.06 / 11065.24 (99%)

Table A2 presents the state-level employment included in the SHRUG panel (numerator), the state-level employment in the Economic Census datasets
(denominator), and the share of state-level employment captured by the SHRUG, for all states and union territories in India. Employment numbers are reported
in thousands. Chhattisgarh, Jharkhand, and Uttarakhand were created in 2000 and are thus left blank in earlier years.
 Table A3
Asset Decomposition of Small Area Consumption Estimates:
Rural Households
(1) (2) (3) (4) (5)
IHDS SECC Difference Coefficient Delta
Income 5000-10,000 Rs 0.12 0.18 0.06 10076.33 590.07
Income Above 10,000 Rs 0.06 0.09 0.03 38933.33 1100.99
Home Ownership 0.99 0.95 -0.04 -1334.48 55.42
Kisan Credit Card 0.07 0.04 -0.03 12441.10 -388.38
Land Ownership 0.61 0.44 -0.17 9657.24 -1613.72
Number of Rooms in Home 2.60 2.15 -0.45 3428.70 -1549.51
Both Mobile and Landline 0.03 0.03 -0.00 31479.48 -86.19
Landline Phone 0.01 0.01 0.00 24639.32 15.93
Mobile Phone 0.68 0.69 0.01 23997.18 339.78
Refrigerator 0.11 0.12 0.01 29476.68 363.90
Brick Roof 0.05 0.07 0.02 -9604.72 -235.39
Concrete Roof 0.12 0.22 0.10 1431.76 149.18
GI Roof 0.16 0.14 -0.02 -3359.10 65.99
Grass Roof 0.23 0.16 -0.07 -2919.61 212.58
Plastic Roof 0.00 0.02 0.02 6473.79 110.06
Slate Roof 0.05 0.04 -0.01 2316.43 -34.52
Stone Roof 0.08 0.05 -0.03 11637.33 -341.08
Tile Roof 0.12 0.28 0.16 -6508.29 -1068.59
Four Wheeled Vehicle 0.02 0.03 0.01 85685.73 532.55
Two Wheeled Vehicle 0.17 0.18 0.01 34253.34 501.96
Brick Walls 0.27 0.43 0.16 23029.78 3703.36
Concrete Walls 0.24 0.04 -0.20 22316.29 -4512.30
GI Walls 0.01 0.01 -0.00 14184.44 -27.89
Grass Walls 0.07 0.11 0.04 12808.09 531.28
Mud Walls 0.33 0.27 -0.06 13371.95 -772.25
Plastic Walls 0.00 0.01 0.01 19748.41 128.43
Stone Walls 0.05 0.11 0.06 17065.06 1066.39
Wooden Walls 0.01 0.01 0.00 9216.71 15.00
Table A3 presents a comparison of IHDS and SECC covariates that were used to generate per capita consumption
small area estimates in the rural SHRUG. The IHDS and SECC columns indicate the value for each covariate in
the SECC and IHDS surveys taken at the village level; because the SECC is a census, no weights were required,
while the IHDS required the use of sampling weights. Column 3 presents the difference between the two. Column
4 shows the coefficient for each covariate when regressing per capita consumption on the set of covariates in the IHDS.
Column 5 multiplies column 4 by column 3, representing the expected difference in per capita consumption between
IHDS and SHRUG that is explained by that covariate. The omitted category for roof and wall materials was “other.”

45
 Table A4
Asset Decomposition of Small Area Consumption Estimates:
Urban Households
(1) (2) (3) (4) (5)
IHDS SECC Difference Coefficient Delta
Air Conditioning 0.04 0.08 0.04 17466.61 715.74
Computer 0.13 0.14 0.01 35989.55 475.83
Indoor Toilet 0.67 0.79 0.12 6061.17 725.41
Home Ownership 0.84 0.77 -0.07 . .
Separated Kitchen 0.71 0.72 0.01 -1356.62 -8.42
Number of Rooms in Home 2.76 2.57 -0.19 4714.67 -873.69
Both Mobile and Landline 0.11 0.07 -0.04 35447.19 -1501.55
Landline Phone 0.01 0.01 0.00 15741.89 66.50
Mobile Phone 0.80 0.80 -0.00 30750.63 -25.67
Refrigerator 0.46 0.46 0.00 23585.75 24.26
Brick Roof 0.02 0.07 0.05 -8717.23 -469.53
Concrete Roof 0.30 0.54 0.24 -1861.32 -450.93
GI Roof 0.11 0.14 0.03 -8522.64 -260.14
Grass Roof 0.06 0.04 -0.02 -10871.80 194.61
Plastic Roof 0.00 0.01 0.01 -6085.06 -64.70
Slate Roof 0.05 0.03 -0.02 -12192.98 275.30
Stone Roof 0.06 0.05 -0.01 144.10 -1.03
Tile Roof 0.07 0.10 0.03 -8480.40 -258.71
Four Wheeled Vehicle 0.07 0.07 0.00 67581.44 234.30
Two Wheeled Vehicle 0.37 0.34 -0.03 35366.95 -1212.34
Brick Walls 0.31 0.66 0.35 19329.30 6740.54
Concrete Walls 0.52 0.13 -0.39 23026.67 -8990.93
GI Walls 0.02 0.01 -0.01 26490.49 -208.85
Grass Walls 0.01 0.03 0.02 6927.12 152.49
Mud Walls 0.07 0.08 0.01 15267.08 172.38
Plastic Walls 0.00 0.00 0.00 33435.91 63.89
Stone Walls 0.06 0.07 0.01 13178.63 94.28
Wooden Walls 0.01 0.01 -0.00 6579.85 -24.11
Washing Machine 0.16 0.22 0.06 19691.06 1171.91
Table A4 presents a comparison of IHDS and SECC covariates that were used to generate per capita consumption
small area estimates in the urban SHRUG. The IHDS and SECC columns indicate the value for each covariate in the
SECC and IHDS surveys taken at the village level; because the SECC is a census, no weights were required, while the
IHDS required the use of sampling weights. Column 3 presents the difference between the two. Column 4 shows the
coefficient for each covariate when regressing per capita consumption on the set of covariates in the IHDS. Column
5 multiplies column 4 by column 3, representing the expected difference in per capita consumption between IHDS
and SHRUG that is explained by that covariate. The omitted category for roof and wall materials was “other.”

46
 Table A5
Cross-Sectional Partial Correlations of Night Lights

District Subdistrict Village Village

(1) (2) (3) (4)
Log Population 0.236*** 1.257*** 0.431*** 0.406***
(0.052) (0.034) (0.003) (0.003)
Log Non-Farm Employment 0.104* 1.740*** 0.705*** 0.693***
(0.058) (0.049) (0.005) (0.005)
Log Hours Electricity 0.130 0.553*** 0.262*** 0.213***
(0.083) (0.031) (0.003) (0.004)
Log Consumption 0.053** 0.083*** 0.097*** 0.080***
(0.026) (0.008) (0.001) (0.001)
N 632 5756 425049 424855
Fixed Effects None None District Subdistrict
∗
p<0.10,∗∗p<0.05,∗∗∗p<0.01

Table A5 shows the partial correlation between rural luminosity and population, employment, electricity, and
consumption. Each estimate is from a separate cross-sectional regression of Equation 1, with added controls for
all the other variables in the table. Each entry also controls the number of 1km x 1km cells in the geographic unit.
Column 1 aggregates data to the district level, Column 2 to the subdistrict level, and Columns 3 and 4 to the
village level. Columns 2-4 are clustered at the district level. The table is comparable to Table 3, but has additional
controls for the other outcome variables in the table. Standard errors on consumption are calculated using 1000
bootstraps as described in Section 2.4. Source: SHRUG.

47
 Table A6
Time Series Correlates of Night Lights: Robustness Checks

2001–2013 Pop-weighted
(1) (2) (3) (4)
Log Population 0.006*** 0.003*** -0.004** 0.000
(0.002) (0.001) (0.001) (0.001)
Log Non-Farm Employment 0.026*** 0.025*** 0.047*** 0.049***
(0.009) (0.005) (0.004) (0.003)
Log Manufacturing Employment -0.019* -0.006 0.001 0.008**
(0.011) (0.007) (0.005) (0.004)
Log Services Employment 0.033*** 0.027*** 0.055*** 0.057***
(0.006) (0.005) (0.004) (0.003)
Electricity 0.066*** 0.049*** 0.035*** 0.026***
(0.013) (0.009) (0.004) (0.003)
N 982594 982024 1766561 1765407
Geographic Aggregation Village Village Village Village
Fixed Effects Village, Village, Village, Village,
District * Year Subdistrict * Year District * Year Subdistrict * Year
∗
p<0.10,∗∗p<0.05,∗∗∗p<0.01

Table A6 shows the time series relationship between a set of rural development outcomes and night-time luminosity
at the village level. Columns 1 and 2 restrict the data sample to 2001–2013. Columns 3 and 4 weight regressions
by village population. Otherwise, estimates are similar to Columns 3 and 4 in Table 4. All variables are measured
in logs. All regressions include village fixed effects and are clustered at the district level. Source: SHRUG.

48
 Figure A1
Error Rates from Imputation Simulation

A. Mean Absolute Imputation Error

.025
Mean absolute deviation as share of population

.02

.015

.01

.005

0
0 10 20 30
% of shrids dropped

B. Signed Imputation Error

.001

0
Average error as share of population

−.001

−.002

−.003

−.004

0 10 20 30
% of shrids dropped

Panel A of Figure A1 shows the mean absolute deviation of simulated estimates of

constitutency population as compared with actual constituency population, under scenarios
where we set a different share of the population to missing. We run a simulation where
before calculating constituency population, we drop village and town observations
representing X% of constituency population, and then use our imputation methodology.
The graph shows, for instance, that when we drop 20% of the data before imputation,
our average constituency has a total population error of approximately 1.8%. Panel B
shows the signed error rather than the absolute error, indicating a very small downward
bias in population estimation from our method. Source: SHRUG.
49