Manufactured and massaged macro

By now you have learned to run a print through the drill — a macro variable, its channel, the sector it lands on, the line on the statements, and whether it is already priced. Every step of that drill starts from a number: the inflation print, the growth figure, the deficit. This module asks the awkward question underneath all of them. How much should you trust the number in the first place?

Not because someone is lying to you. That is the wrong frame, and it will make you a worse reader, not a sharper one. The real answer is quieter and more useful: every official statistic is a thing that had to be built. Somebody chose what goes in the basket, which base year to measure against, how big a sample to take, how to adjust for the season, and when to revise. Each of those choices is defensible, and each of them bends the number a little. The figure that lands on your screen as a clean, single truth is the end of a long assembly line, and the line has known kinks.

This is not conspiracy. It is construction. A reader who understands how a number is assembled can read around its biases and stay useful. A reader who decides the whole thing is fake becomes useless — because "it's all fake" explains every number equally and therefore tells you nothing. The skill in this module sits precisely between naïve trust and lazy conspiracy.

Think about what a single inflation number is trying to do. It wants to compress the price of everything a household buys — food, fuel, rent, phone bills, school fees, a haircut — into one figure. That is impossible without choices, and the choices are where the bending happens.

The first choice is the basket: which goods and services count, and how much each one weighs. In the Indian consumer basket, food and beverages carry a very large share, far larger than in a rich country, because that is how a typical household here actually spends. So when vegetables have a good season, the whole headline can soften even if rent and services keep climbing. The number is honest; it is just answering the question "what did the average basket cost?", not "did life get cheaper for a salaried city family?".

The second choice is the base year — the reference period every "up X%" is measured against. (A base year is simply the anchor year a series is compared to; change it and every growth rate is re-measured against a different starting point.) India revised the base year and method for its growth figures in the mid-2010s, and the revision genuinely changed how growth looked — enough to trigger years of debate about a "back-series" that would let old and new numbers be compared. Nobody invented output. The measuring stick changed, and a different stick gives a different reading of the same distance.

The third is coverage. A huge part of the Indian economy is informal — tiny firms, cash trades, the corner shop — and cannot be counted directly every quarter. So it is estimated, often by scaling up from the organised sector that can be measured. That works until the two parts move differently. When a shock hits small informal firms harder than large formal ones — a cash crunch, a lockdown — a method that infers the small from the large can miss the divergence and read too rosy for a while.

The fourth is the deflator — the price adjustment used to strip inflation out of growth. (The deflator is the price gauge that converts a value measured in rupees into a "real", inflation-removed quantity.) If the deflator understates how fast prices rose, it will overstate real growth, because less inflation gets subtracted. This is a technical, unglamorous number that almost nobody watches, and it can quietly flatter or depress the growth headline everyone does watch.

None of these is fraud. Each is a reasonable answer to a genuinely hard measurement problem. But each has a direction — a known way it tends to bend — and knowing the direction is the whole game.

It helps to picture any macro number as coming off an assembly line with five stations. At each station a choice is made, and each choice adds a small, directional bias. The headline is what rolls off the end — clean-looking, and carrying every bend from up the line.

The two most useful defences against a bent number are not clever. They are habits.

The first is to read a family of series, never a single number. No one figure can be gamed if you hold it against its cousins. Growth in gross value added — the sum of what each sector actually produced — sits alongside headline growth, and the gap between them tells you how much taxes and subsidies flattered or dulled the top line. (GVA measures output from the producers' side, before the taxes and subsidies that separate it from the demand-side headline.) The wholesale price series sits alongside the consumer one; when they diverge sharply, something in the basket or the supply chain is doing the talking. A private activity survey sits alongside the official one. You are not looking for the "true" number. You are triangulating — watching whether the cousins agree, and getting suspicious of a headline that has quietly drifted away from all of them.

The second is to watch the definition, not just the value. The number that moves the most is often the one whose definition just changed — a new base year, a re-weighted basket, an item shifted in or out. When you see a series jump, the first question is not "what happened in the economy?" but "did the ruler change?".

Take a concrete, composite case and run the drill through a bent number. illustrative

Suppose the reported fiscal deficit — the gap between what the government spends and what it earns — lands at a reassuring level, comfortably inside the target. The headline says: the government is being disciplined, it will not need to borrow too heavily, so the pressure on interest rates from state borrowing is mild. A reader who stops there feeds a clean number into the drill and gets a clean, wrong answer.

Now look at how the number was built. Suppose a chunk of spending — a subsidy, an infrastructure push — was funded not from the main budget but through a state-owned enterprise that borrowed on the government's behalf, or through a special bond that sits outside the headline deficit. (This is off-budget borrowing: real government spending financed in a way that does not show up in the reported deficit.) The activity is real. The debt is real. It simply did not pass through the line the headline measures.

Run the corrected number through the transmission chain. The government's true call on the nation's savings is larger than the headline suggests. A larger call on savings means more competition for the same pool of lendable money, which tends to keep bond yields higher than the tidy headline implied. Higher yields raise the cost of funds for a lender — say a composite housing-finance company, "an illustrative housing lender" — which either squeezes its spread or forces it to pass the cost on and slow its loan growth. Same starting figure. The reader who traced the off-budget items sees a mild headwind building for the lender; the reader who took the headline at face value sees blue sky.

The lesson is not "the deficit is always understated." Sometimes it is clean. The lesson is that the definition of the deficit is a choice, and the reader's job is to know what the chosen definition leaves out before running it through any chain.

Knowing that numbers are built protects you from a lot. It also cannot do several things, and pretending otherwise swaps naïve trust for a subtler kind of overconfidence.

It cannot hand you the "true" number. There is usually no clean, hidden real figure sitting behind the official one, waiting to be uncovered. There are only several constructions, each with its own biases, that you read against each other. The output of good scepticism is a direction and a doubt — "this headline probably leans a little rosy, and here is why" — not a corrected decimal you can trust to the second place.

It cannot tell you the magnitude of the bias in real time. You can often name which way a series leans without being able to say by how much. That is a real limit, and it is why the honest read stays qualitative: a lean, not a number. Naming the direction of a bias is reading; asserting its exact size is just forecasting with extra steps.

And it cannot license conspiracy. The instant "read around the bias" becomes "they fake everything," you have stopped reading and started believing, and a belief that explains every number equally well is worth nothing. The discipline is to stay specific: this series, this station, this known lean — not a blanket verdict on the honesty of the state.

The same handful of errors catch readers around massaged data, and each has a tidy antidote.

1. Taking the headline as the whole truth. The most common error is simply stopping at the single blended number — one inflation print, one growth figure — and never asking what basket, base year, or coverage produced it. The fix is the family: always read a series against its cousins.

2. Overcorrecting into conspiracy. The opposite and equally lazy error: deciding the number is fabricated and therefore ignoring it. This feels like scepticism but is its own credulity — faith in a hidden plot instead of faith in a printed figure. A bias you can name is still information; a conspiracy is not.

3. Cherry-picking the base year. Any growth rate can be made to look wonderful or terrible by choosing what it is measured against — a collapsed year makes the next year's bounce look heroic. When someone quotes a dramatic "up X% since," the first question is "since when, and why that year?".

4. Confusing GVA with GDP, or WPI with CPI. Different constructions answer different questions. Treating them as interchangeable — or quoting whichever one flatters your view — is not reading, it is shopping. Know which question each series answers before you cite it.

5. Trusting one private survey as the antidote. Swapping the official number for a single private one you have not examined just moves your faith to a different construction. The move is triangulation across several, not the anointing of a new single truth.