Every district in India collects data differently. One might report income in rupees, another literacy in percentages, and a third electricity access as a raw count of villages connected. If you tried to add these numbers directly to rank districts, the results would make no sense. This is exactly the problem the Mean Standardisation Method solves. It is one of the simplest and most widely used normalisation techniques in social science research, and it forms the backbone of many composite indices used to rank districts, states, and even countries on development parameters.
Table of Contents
What is the mean standardisation method
The mean standardisation method is a normalisation technique used to bring variables measured in different units onto a common, comparable scale. It does this by dividing each district’s actual value for a variable by the mean (average) value of that variable across all districts in the dataset.
The logic is straightforward: instead of comparing raw numbers that carry different units, you compare how far each district’s value is from the average. A district performing exactly at the average gets a standardised score of 1. A district performing better than average gets a score above 1, and one performing worse gets a score below 1.
This is part of a broader family of normalisation approaches. The OECD’s handbook on constructing composite indicators lists standardisation, along with min-max scaling and ranking, as one of the core methods analysts use to make indicators comparable before combining them into a single index. Mean standardisation is a simplified variant of this family, centred on 1 rather than on 0, which makes it easier to interpret at a glance.
How to compute standardised values
The formula itself is simple:
Standardised value = Actual value of the variable / Mean value of the variable
Consider a variable like the percentage of workers engaged in non-agricultural labour. Suppose a particular district reports a value of 73, and the average across all districts in the study is 59. The standardised value works out to:
73 ÷ 59 = 1.237
This tells you the district is performing about 23.7 per cent above the average for this indicator. If another district had an actual value of 40, its standardised value would be 40 ÷ 59 = 0.678, meaning it is roughly 32 per cent below average.
This exact process is repeated for every variable and every district in the dataset. So if you are working with, say, ten indicators across thirty districts, you end up with 300 standardised values, each one centred around 1 and each one now comparable to values from a completely different indicator.
Why this eliminates scale effects
The biggest advantage here is that it removes the problem of unit bias. A variable measured in rupees (say, per capita income) and a variable measured in percentages (say, literacy rate) cannot be added or averaged directly, because their scales are wildly different. Once both are converted into standardised values through mean standardisation, they are both expressed as ratios relative to their own mean, so they become directly comparable and combinable.
This step is critical in social science research, where composite indices routinely combine indicators as varied as household income, school enrolment, hospital beds per capita, and road density. Without normalisation, the indicator with the largest raw numbers would dominate the index purely due to scale, not because it is actually more important.
Deriving the composite index
Once every variable has been converted into a standardised value for every district, building the composite index is the easy part. You simply take the arithmetic mean of all the standardised values for a given district. This single number becomes that district’s composite index score.
For example, if a district has standardised values of 1.237, 0.85, 1.10, and 0.95 across four indicators, its composite index would be:
(1.237 + 0.85 + 1.10 + 0.95) ÷ 4 = 1.034
This composite score can then be compared across districts to build a ranking. A score above 1 indicates overall performance better than the average district in the dataset, while a score below 1 indicates the opposite. Because every underlying variable contributes on an equal, unit-free footing, the ranking reflects genuine relative performance rather than an artefact of measurement scale.
This approach mirrors how several real-world development indices work in India. The Indian Economic Service’s documentation on backwardness indices describes how schemes like the Backward Districts Initiative combined multiple parameters with equal weights to rank districts, a structure very similar in spirit to what mean standardisation produces.
Where this method is applied
Composite indices built through mean standardisation, or close variants of it, show up constantly in Indian policy research. The Aspirational Districts Programme run by NITI Aayog ranks districts using a composite score built from dozens of indicators across health, education, and infrastructure, each normalised before being combined. Similarly, government efforts to identify India’s most backward districts have historically relied on quantifying parameters like poverty, health, education, and infrastructure into a single index for the first time using district-level data.
These are not just academic exercises. The resulting rankings determine where government funding gets prioritised, which districts receive special development attention, and how progress is tracked over time. Getting the normalisation step right, therefore, has real consequences for resource allocation.
Application and comparison with other methods
The mean standardisation method is valued for being transparent and easy to compute, but it is rarely used in isolation. Researchers typically cross-check it against alternative approaches, most commonly the Range Equalisation (RE) method, which rescales variables based on the gap between the minimum and maximum observed values rather than the mean.
Why bother with this cross-checking? Because different normalisation choices can, in theory, produce different rankings for the same underlying data. A study on identifying backward districts in India points out that differences in methodology and choice of indicators often lead to noticeable mismatches in how districts get ranked on backwardness scales. This is precisely why validation matters.
In practice, when researchers compute a composite index using both the Mean Standardisation Method and the Range Equalisation method, and then check the correlation between the two sets of rankings, a high correlation coefficient is treated as evidence that the index is robust and not overly sensitive to the specific normalisation technique chosen. This mirrors a broader finding in international composite indicator research, where comparisons of different standardisation methods for performance indices have shown correlation coefficients typically in the 0.98 to 0.99 range, suggesting the underlying story the data tells does not change much regardless of which standardisation technique you apply.
This is an important lesson for anyone building or evaluating a composite index: the choice of normalisation method matters less for the overall conclusion than getting the indicator selection and data quality right in the first place. A well-constructed index built on sound data will tend to tell a similar story whether you standardise by mean, by range, or by z-scores.
Strengths and limitations to keep in mind
The mean standardisation method’s biggest strength is its simplicity. Anyone with a basic understanding of averages can compute it, verify it, and explain it to a non-technical audience, which matters a great deal when the resulting index is meant to inform public policy. It also treats every indicator with equal footing once standardised, avoiding the complexity of weighting schemes that can introduce their own biases.
That said, it is sensitive to outliers. If one district reports an unusually extreme value for a variable, it can pull the mean itself in a way that distorts the standardised values for every other district on that indicator. This is why researchers often examine the underlying data distribution before applying this method, and why cross-validation with a method like Range Equalisation remains good practice.
What do you think? If you were building a development index for your own state, would you trust a method as simple as dividing by the mean, or would you want a more complex statistical technique before relying on the rankings it produces?
References
- https://www.oecd.org/en/publications/handbook-on-constructing-composite-indicators-methodology-and-user-guide_9789264043466-en.html
- https://ies.gov.in/arthapedia/concept/backwardness
- https://www.niti.gov.in/sites/default/files/2018-12/AspirationalDistricts-Book.pdf
- https://www.business-standard.com/article/economy-policy/quiz-do-you-know-which-are-india-s-most-backward-districts-117121901274_1.html
- https://www.sciencedirect.com/science/article/abs/pii/S0038012119302022
- https://www.oecd.org/content/dam/oecd/en/publications/reports/2003/11/composite-indicators-of-country-performance_g17a155e/405566708255.pdf
Leave a Reply