Index of Industrial Production: What IIP Captures and What It Misses
An industrial production index measures the volume of goods produced across industrial categories, each weighted by how much it contributes to industrial output. Because it counts physical quantities rather than opinions or rupees, a price change does not move it at all. Because industry is a minority of a modern economy, the index is silent on most of what that economy does.
Compute an industrial production index from a release, then choose what to compare it with
With the category index levels a release prints keyed in, the panel rebuilds the headline index: each category's contribution in index points, the sign and size of every movement, the weights proved to add to 100.00 and the contributions proved to add to the move. The two levels this month is compared against are set next. The same headline level yields three different growth rates depending on which comparison is chosen, and the panel prints all three side by side so the intended one is never the assumed one.
Step one: the category levels this release prints
Step two: the two levels this month is compared against
Step three: the weight the index gives a category that did not exist when the weights were set
The panel opens on a complete worked example, and it is worth having in plain text as well as on a screen. The five category levels read 102.00, 104.00, 97.00, 106.00 and 130.00 against a base period of 100.00, and the four published weights of 14.00, 58.00, 12.00 and 16.00 turn the first four into contributions of 0.28, 2.32, minus 0.36 and 0.96 index points. The four contributions add to a move of 3.20 points, so the headline index stands at 103.20. Compared with a month that also sat at 100.00 that is growth of 3.20 per cent, and multiplied by the 22.00 per cent share of total output that industry holds, it is 0.70 per cent added to all of Sankhya's output. Every figure in that paragraph is a multiplication that can be redone, and a number that can be redone is a number worth arguing with.
Two failures are easier to recognise than to describe, so both are worth producing early. Drag the comparison month down 8.00 points and this month's level stays exactly where it was. The year-on-year rate climbs from 3.20 to 12.17 per cent, so the headline gains almost nine percentage points on something that happened a year ago. Dragging it up 8.00 instead makes the same unchanged 103.20 read as minus 4.44 per cent. The panel prints two rates beside each other, one measured against the base period and one against the same month a year earlier. The two rates agree at the opening setting only because that setting puts the comparison month exactly on the base. Once the comparison month sits off the base the two part company, and quoting the wrong one is how a level that has drifted up over many years gets reported as this year's growth.
Everything the panel does rests on one distinction and one fraction, and both are worked out slowly below. The distinction is between counting things and counting money, and it decides what the index can respond to. The fraction is the share of total output that industry accounts for, and it decides how much of an economy the finished number is entitled to describe.
What does an industrial production index actually count?
An industrial production index counts quantities. Tonnes, units, kilowatt hours, metres, litres. The index does not count the value of what was produced, and that single property is the source of almost everything else that is true about it. Think about a vegetable seller outside a bus depot. In one week she brings 200 kg of produce to the stall and sells it at Rs 20/- a kilo, so her takings are Rs 4,000/-. The next week she brings the same 200 kg and the market price has moved to Rs 30/- a kilo, so her takings are Rs 6,000/-. Her takings rose 50.00 per cent. The quantity she carried to the stall did not change by a single kilo, and a volume index reads exactly the same in both weeks.
Now change the other input instead. She brings 240 kg at the original Rs 20/- a kilo, so her takings are Rs 4,800/-, and Rs 4,800/- is 20.00 per cent more than Rs 4,000/-. The quantity moved too this time, by the same 20.00 per cent. The first week was a price event and the second was a production event, so a volume index registers the second and ignores the first. Separating volume from value is exactly this: the nominalA figure measured in the money of the day, before any adjustment for how prices have moved since. The opposite of a real or volume figure. number contains both movements mixed together, and a volume series has had the price movement taken out of it, usually by dividing through a deflatorA price index used as a divisor, so that a series measured in money can be restated as a series measured in quantities. Constructing one is a separate subject..
Does an industrial production index measure the volume of output or the value of output?
A producer sells the identical number of units this period, and the price of each unit rises. What happens to a volume index of industrial production?
How is an industrial production index built from producer returns?
An industrial production index is assembled in three moves, and none of them is mysterious. First, producers send in returns stating quantities: how many tonnes, how many units, how many kilowatt hours. Second, those quantities are grouped into categories, and each category is given a weight equal to its share of industrial output in a chosen base period. Third, the weighted quantities are added together and expressed as a level against that base period, conventionally set at 100.00. Every step is arithmetic that could be redone by hand if the weights were published, and they are.
The weight is where the interesting behaviour lives. A weight is fixed between revisions, so the index measures today's volumes against yesterday's structure. Picture a household that worked out its monthly budget shares years ago: rent 40 paise in the rupee, food 30, travel 15, everything else 15. If it keeps using those shares to describe its spending today, the description drifts further from the household every year the actual pattern moves. The index has the same property, and it is deliberate: holding the weights still is what makes two periods comparable at all. A weight base is revised from time to time, and a revision changes the finished series without anybody having mismeasured anything.
In the Sankhya index above, what does the weight of 58.00 attached to factory made goods represent?
Who compiles an industrial production index in India?
In India, the Index of Industrial Production is compiled and released by the National Statistical Office under the Ministry of Statistics and Programme Implementation. The Index of Industrial Production is a volume index built from producer returns and aggregated with weights fixed against a base period, and the category list, the weight base and the coverage share are all published alongside it. All three are revised from time to time, so the current construction note at the source is worth reading before any weight, any category boundary or any coverage figure is repeated.
What share of an economy does an industrial production index cover?
Here is the fraction that decides how the finished number may be used. In the Republic of Sankhya, industry accounts for 22.00 per cent of total output at the vintage used throughout these notes. An industrial production index therefore reads a fifth of the economy and says nothing whatever about the other 78.00 per cent. Services, agriculture, government, construction of the non-industrial kind: none of them sends a producer return into this index, and none of them appears in its weights. The exclusion is not a defect anybody failed to fix but the scope the index was built to have, in the same way that a thermometer in one room of a house is not a broken instrument for failing to report the temperature in the other rooms.
Notice what the coverage share does to the weights. The weight of 58.00 attached to factory made goods is 58.00 per cent of industry, not 58.00 per cent of Sankhya. Industry is 22.00 per cent of output, and 58.00 per cent of 22.00 per cent is 12.76 per cent, so factory made goods are 12.76 per cent of all output. Running the same multiplication across all four returns shares that add back to 22.00 per cent, and the figure below draws them. The largest single category in the index is an eighth of the economy, and the smallest is under three hundredths of it.
Industry is 22.00 per cent of the Republic of Sankhya's output. What does that make the industrial production index a measure of?
Why does an industrial production index move more than the aggregates?
Two separate things make an industrial volume index jumpy, and separating them is the difference between reading it well and reading it badly. The first is real: industrial output genuinely swings more than services output does. A cement kiln can run flat out one period and stand idle the next. A school, a bank branch and a barber shop all produce at close to the same rate whatever the weather. The second is arithmetic: a narrow index has fewer components to cancel each other out. Add a thousand small movements together and most of them offset; add four and they do not. Volatility in this index is partly the thing being measured and partly the act of measuring it.
Watch the spreadThe distance between the highest and the lowest figure in a set. It describes how far apart the members are, and says nothing about their average. shrink as the figures aggregate. At the first print vintage the four Sankhya categories moved by 2.00, 4.00, minus 3.00 and 6.00 per cent, a spread of 9.00 percentage pointsThe unit for the difference between two percentages. A move from 4.00 per cent to 6.00 per cent is two percentage points, not two per cent. between the fastest and the slowest. Combine them on their weights and the index moves 3.20 per cent. Multiply that by the 22.00 per cent coverage share and industry has added 0.70 per cent to all of Sankhya's output. The components rattle around across nine points; the contribution to the whole economy sits inside one. Some of the remaining wobble is calendar rather than production, and that is why published series are usually put through a seasonal adjustmentA treatment that strips out whatever repeats at the same point in every year, a harvest or an annual shutdown for instance, so two periods can be set side by side. before anyone reads them.
Take a second illustrative period with the same four weights and different volume changes: 8.00, minus 2.00, 1.00 and 3.00 per cent. Now the components are spread across 10.00 points, wider than before, and yet the index moves only 0.56 per cent. The largest weight is the one that fell. The index growth rate swung 2.64 points between the two periods, and the same swing measured against all of Sankhya's output is 0.58 points. Every step of aggregation flattens the movement, so the narrower the index the larger its numbers will look.
Why does an industrial production index move about more than a whole-economy output measure?
What does a count of volumes capture that a survey of opinion cannot?
Size. Size is the whole answer, and it is worth stating flatly. A survey asks a producer whether output was higher, the same or lower, and records a direction. A volume index asks how many tonnes and records a quantity. Magnitude is the thing a count has and an opinion does not. Suppose three Sankhya producers each report that output improved. On a tick sheet that is three improvements out of three, and the three are indistinguishable. On volume returns the same three read 0.50, 6.00 and 30.00 per cent, so the largest is sixty times the smallest. Both instruments recorded something true. Only one of them can say whether the improvement was worth noticing.
Ask ten shopkeepers in a market whether footfall was better than last week and nine say yes. Nine out of ten is genuine information about direction. Now ask them how many customers walked in. If eight of the nine gained four customers each and the tenth, the largest shop in the row, lost three hundred, the market as a whole had a worse week.
Three producers all report that output improved, and their volume changes are 0.50, 6.00 and 30.00 per cent. What does the volume return carry that the tick sheet does not?
What does an industrial production index miss altogether?
Four things, and they are missing for different reasons. Services are missing because they are not industry, the coverage point already made. Informal activityProduction by units that are not registered, do not file returns and are not in the statistical sample. Measuring it needs surveys built for the purpose, which is a separate subject. is missing because a workshop that files no return is not in the sample, so an economy with a large unregistered base has a genuine hole in its industrial count. Quality changeAn improvement in what a unit of output actually delivers, such as a stronger cement or a longer lasting bulb, at the same physical count. Adjusting an index for it is a specialised exercise. is missing because a count of units cannot see that this year's units are better: two hundred longer lasting bulbs count the same as two hundred ordinary ones. And the fourth is the subtle one that most readers walk straight past.
The weights were set before the fast growing category grew, so a fixed-weight index underweights whatever is growing fastest. The claim deserves to be worked rather than asserted, so work it. Suppose Sankhya now has a fifth industrial category, assembled electronics. Assembled electronics did not exist at the weight base vintage and therefore carries a weight of 0.00 in the published index. At the first print vintage it grows 30.00 per cent, and by then it accounts for 9.00 per cent of Sankhya's industrial output. A weight of 0.00 multiplied by a growth of 30.00 per cent contributes nothing at all, so the published index still combines only the four original categories and moves 3.20 per cent, exactly as before.
Now compute what industry actually did. If assembled electronics is 9.00 per cent of industrial output, the four original categories hold 91.00 per cent between them in their old proportions, and on those current shares they contribute 2.91 per cent. Assembled electronics contributes 9.00 per cent of its 30.00 per cent growth, or 2.70 per cent. Industry moved 5.61 per cent. The index said 3.20 per cent. The gap of 2.41 percentage points is not measurement error and nobody made a mistake: it is the weight base being older than the economy it describes. A weight of 0.00 is the extreme case, and any weight below a category's current share produces the same gap in smaller size.
Why does a fixed-weight index understate a sector that is growing quickly?
Which of these does an industrial production index leave out entirely, rather than merely underweight?
Can the Sankhya index be rebuilt from its published weights?
The Sankhya index can be rebuilt, and rebuilding is the test worth applying to any index anybody quotes. A contribution is the weight multiplied by the volume change and divided by 100; the four contributions add to the move, and the move added to the base gives the level. Nothing in the table below is asserted, and the panel above is the same table with every figure left free.
| Category | Weight, weight base vintage | Volume change, per cent | Contribution, index points |
|---|---|---|---|
| Mined and quarried materials | 14.00 | 2.00 | 0.28 |
| Factory made goods | 58.00 | 4.00 | 2.32 |
| Power generation | 12.00 | minus 3.00 | minus 0.36 |
| Processed food and drink | 16.00 | 6.00 | 0.96 |
| All four categories | 100.00 | 3.20 | 3.20 |
The total row carries two different units that happen to share a number, so read it carefully. The 3.20 in the third column is a per cent change in volume. The 3.20 in the fourth column is a count of index points on a base of 100.00. The two figures coincide only because the base is 100.00, and on a base of any other value they would not. The published level is therefore 103.20. The contribution of industry to all of Sankhya's output is 3.20 multiplied by the 22.00 per cent coverage share, or 0.70 per cent.
Two behaviours in that table are worth chasing in the panel above. Once the published weight reaches 9.00 the weight and the current share are the same number, and nothing is left to be stale, so the gap between the index reading and the industry figure closes to exactly nothing. The contribution to all output, meanwhile, is always the index move multiplied by 0.22, so it stays small whatever the five levels are set to. The coverage share is the ceiling on how much this indicator is entitled to say about an economy, and no volume change in any category can lift it.
How does an analyst actually use an industrial volume index?
An analyst treats it as what it is: a volume series covering a fifth of the economy, published earlier and in finer category detail than the aggregates it sits inside. If the question is whether cement, steel and power volumes are moving, this is the number with the category breakdown to answer it, and those categories are where an equity research view on an industrial producer gets built.
The analyst does not treat it as a proxy for the whole. A lender sizing a working capital limit for a cement supplier cares about volumes in that category and can use them directly. The same lender assessing whether a borrower's customers across services and retail can keep paying gets nothing from this index at all. Retail and services customers sit inside the 78.00 per cent it never looked at. A household reading a headline about industrial output should apply the same discipline: the number is real, it is about a fifth of what is going on, and a large move on a fifth is a small move on everything.
The habit that separates a careful reader from a careless one is multiplying an indicator's move by its coverage share before treating it as a statement about the economy. The multiplication takes one step, and one step is the difference between 3.20 per cent and 0.70 per cent.
Where does the reading of an industrial production index go wrong?
Reading a fall in a fifth of the economy as a fall in the economy
The mistake is common enough to be predictable, and it happens in the gap between a headline and a coverage share. The Republic of Sankhya's industrial index falls 5.00 per cent at an illustrative first print vintage. A reader concludes that Sankhya's output is shrinking, and a commentator repeats it. Neither has multiplied by the coverage share, and the multiplication changes the sentence completely.
Industry is 22.00 per cent of Sankhya's output, and 5.00 multiplied by 0.22 is 1.10, so a fall of 5.00 per cent in industrial volumes is a drag of 1.10 per cent on all output. For total output to be unchanged, the other 78.00 per cent has to add that 1.10 per cent back. Adding it back means growing 1.41 per cent, and 1.10 divided by 0.78 is indeed 1.41. Growth of 1.41 per cent in the services, farming and government part of an economy is not a heroic number. So a 5.00 per cent fall in the industrial index is entirely compatible with total output being flat, or higher, or lower. Seventy-eight per cent of the answer was never in the index, so the index cannot settle the question.
The cost is a view held with more confidence than the evidence supports, and confident views get acted on. The fix is one multiplication done before the sentence is written: multiply the indicator's move by its coverage share, and only then say what it implies about the whole. A large move on a fifth is a small move on everything.
The Sankhya industrial index falls 5.00 per cent, and industry is 22.00 per cent of output. A reader says output is shrinking. What is the missing step?
What falls outside an industrial production index?
Which body publishes the real version of any figure above?
Four things decide what an industrial production index means, and all four are published by the bodies in the table: the construction method, the category list, the weight base and the coverage share. All four also move. A weight base is replaced, a category boundary is redrawn, a coverage share is restated on a new set of national accounts. The live wording at the issuing body is therefore what any reading has to be built on.
| Worth looking up | The body that publishes it | Site | What to check first |
|---|---|---|---|
| How an industrial volume index is compiled, and which producer returns feed it | National Statistical Office, India | mospi.gov.in | Whether the method described there still matches the one assumed |
| The category list and the weight base of the Index of Industrial Production | Ministry of Statistics and Programme Implementation | mospi.gov.in | Which weight base is live, given that a change of base changes every level |
| The sector shares of output that give an indicator its coverage share | Reserve Bank of India, statistical handbook tables | rbi.org.in | Whether the share in use belongs to the period being read |
| What a fixed weight base does to a component that grows after it was set | International Monetary Fund, manual on index number construction | imf.org | Which edition is in hand; the guidance has been rewritten more than once |
| The revision notes issued when a weight base is changed | Ministry of Statistics and Programme Implementation, release notes | mospi.gov.in | The vintage stamp, so two figures are not compared across a revision |
The Republic of Sankhya, its five industrial categories and every weight and level attached to them are invented.
Educational material. Not advice on any investment, tax, budget or market position.