Common mistake 1 with India - state map

Comparing per-capita figures across states without noticing that the small UTs sit at the extremes of almost every Indian distribution - Delhi and Goa at the top, and they are not comparable to a state of 200 million.

Common mistake 2 with India - state map

Mapping a level when the question is a change. A map of GSDP is a map of state size; a map of GSDP growth is a map of the thing you probably meant.

Common mistake 3 with India - state map

Treating the map as the finding. Spatial pattern is a prompt for a test, and this app has them - Moran's I and the spatial regressions under Tests and breaks.

Common mistake 4 with India - state map

Comparing two state maps cut at different class breaks. The same colour then means different numbers, and the comparison a reader makes is with their memory of the first map.

EconStatsCraft › India - state map

India - state map

Shades India by state or union territory. Spellings that differ across sources - Orissa and Odisha, Pondicherry and Puducherry, NCT of Delhi, UP, MP, TN - all match. Drawn on an equal-area projection, so a large state is large because it is large.

What it needs

State column - Full names or the usual abbreviations.
What to shade by

The intuition

A state map has every hazard of the world map and one more that bites harder: at a sub-national scale the regions are units of administration rather than units of anything the data is about. State boundaries were not drawn to make statistics comparable, so a state-level average can be an average of two completely different places, and the colour hides that.

When to use it

Any state-level Indian data: GSDP, per-capita income, literacy, unemployment, power consumption, fiscal deficit, poll results. Also the natural first look before a state panel regression, because it shows you whether the variation is north-south, coastal-inland, or neither.

Where the estimator comes from

The estimate shown for state s is an aggregate over its constituent units, \bar{v}_s=\sum_{i\in s}w_iv_i/\sum_{i\in s}w_i. The MODIFIABLE AREAL UNIT PROBLEM is that \bar{v}_s depends on where the boundaries fall, and redrawing them changes the map and can change the sign of any relationship computed from it. Gerrymandering is the deliberate exploitation of exactly this arithmetic. It is not a defect of the chart - the aggregation is real - but the map presents one particular partition as though it were the only one.

What it assumes

That your column names states. Spellings differ across Indian sources and all the common ones match: Orissa or Odisha, Pondicherry or Puducherry, NCT of Delhi or Delhi, Uttaranchal or Uttarakhand, and the usual abbreviations (UP, MP, TN, WB).
One value per state, as with any choropleth. A state panel needs its time column named so the map knows which year it is drawing.
Boundaries are as India defines them, including Jammu and Kashmir and Arunachal Pradesh. The file is the pre-2019 configuration, so Jammu and Kashmir is one unit rather than being split from Ladakh, and Daman and Diu is separate from Dadra and Nagar Haveli.
The state list matches the vintage of the data - Telangana exists from 2014, and Jammu and Kashmir's status changed in 2019. A series spanning those dates has to decide which geography it is on.

Statistical properties

The modifiable areal unit problem is not avoidable, only acknowledged. Any statistic computed on areal aggregates depends on the boundaries used.
The ecological fallacy applies with full force: a state where average income and average literacy are both high does not imply that richer individuals are more literate. Robinson (1950) measured this exact reversal across US states.
Boundaries change. India's districts have been repeatedly split and reorganised, so a map joining 2011 data to current boundaries mismatches silently unless the vintages are aligned.
A state's value can rest on very few observations. Colour carries no sense of sample size, so a state with 20 respondents looks exactly as authoritative as one with 20,000.
Neighbouring states are not independent, so any regression on state-level values needs the spatial dependence handled - which is what the spatial modules are for.
Population-weighted quantities are usually what the reader wants; unweighted state averages give a small state the same say as a large one.

How to read the output

Equal-area matters here more than on a world map, because India spans 8 to 37 degrees of latitude: on Mercator, Jammu and Kashmir would be drawn nearly 40 per cent larger relative to Tamil Nadu than it is. This projection does not do that.
The union territories are tiny and easy to miss - Chandigarh, Delhi, Puducherry, Lakshadweep. When they matter, read them off the table rather than the map.
A state left grey had no matching row. With 36 units, one missing is three per cent of the country and worth chasing.
With 36 units the classification does most of the work. Quantiles put the same number of states in each class regardless of the gaps between them, which can make a small difference look like a large one.

A worked reading

A state map showing literacy against income, both averaged to the state, can produce a strong positive relationship while the individual-level relationship within every state is flat or negative. Robinson found precisely this in 1950: the correlation between illiteracy and being foreign-born was +0.53 across US states and -0.11 across individuals. The state map is not wrong - those really are the state averages - but the conclusion nearly every reader draws from it is.

Symptom → diagnosis → what to do

If you see this	It means	Do this
A striking regional pattern	Check the sample size per state	Sparse states produce extreme values by noise
Conclusions about individuals	Ecological fallacy	Areal averages do not license individual claims
A state missing entirely	Boundary vintage mismatch	Align the geometry to the data year
Results changing when districts are used	The modifiable areal unit problem	Real, not an error. Report both scales
Unweighted state averages	A small state counts as much as a large one	Weight by population
Adjacent states looking alike	Spatial autocorrelation	Test it before modelling

Common mistakes

Comparing per-capita figures across states without noticing that the small UTs sit at the extremes of almost every Indian distribution - Delhi and Goa at the top, and they are not comparable to a state of 200 million.
Mapping a level when the question is a change. A map of GSDP is a map of state size; a map of GSDP growth is a map of the thing you probably meant.
Treating the map as the finding. Spatial pattern is a prompt for a test, and this app has them - Moran's I and the spatial regressions under Tests and breaks.
Comparing two state maps cut at different class breaks. The same colour then means different numbers, and the comparison a reader makes is with their memory of the first map.

Where to read this properly

Robinson (1950), "Ecological Correlations and the Behavior of Individuals", American Sociological Review 15(3).
Openshaw (1984), The Modifiable Areal Unit Problem, CATMOG 38.
Monmonier, How to Lie with Maps (2018).

All analyses

Data & variables

Each example is simulated from a known model, so you can check whether the estimator recovers the truth - and nothing here is anyone's private data.

Filter variables

Drag a variable into a slot on the right - or just click it.

Choose an example dataset above to begin.

Choose a transformation - it opens in the data editor, with a live preview.

Apply to

Multiply by

Columns ctrl-click for several

Group by ctrl-click for several

Within groups

Statistic

Bands

How many

Cut points

Keep rows where

Value

and

Fill with

Value

Also mark which rows were filled

Rules

Base period = 100

Which part

Percent at each tail

Threshold

Survey weight optional

Direction

Names column

Values column

Largest first

Delete them, not just count them

Read this column as

Tidy the text

New name optional

EconStatsCraft rigorous, plain-spoken econometrics

Econometrics that explains itself.

Regression, panel data, time series and the whole battery of tests. Every result is followed by a verdict you can defend.

Load a dataset to see it here.

India - state map

What it needs

The intuition

When to use it

Where the estimator comes from

What it assumes

Statistical properties

How to read the output

A worked reading

Symptom → diagnosis → what to do

Common mistakes

Where to read this properly

See also

Related Maps

All analyses

Charts

Maps

Explore

Regression

Time series - one series

Time series - several series

Panel data

Multivariate statistics

Marketing research

Causal

Finance & risk

Actuarial

Tests & breaks

Machine learning

Econometrics that explains itself.