The Difficulties of Measuring National Income in India

5. Depreciation. The consumption of fixed capital is an estimate based on assumed lives for assets, not an observation. Different assumptions produce different net product.

6. The treatment of illegal and unrecorded activity. Income from smuggling, bribery and unaccounted trade is production in the economic sense and cannot be recorded.

7. Changes in quality and new goods. A given rupee buys a very different phone from the one it bought in 2011. The accounts treat both as one unit of the same thing, so real growth is understated where quality improves and overstated where it declines.

The practical difficulties, which are the India specific ones

1. Non monetised production. A large volume of output in rural India never passes through a market: grain kept for the family, fodder, firewood, milk consumed at home, houses built with family labour, services exchanged between neighbours. Only some of it is imputed. The rest is invisible.

2. The size of the unorganised sector. Most Indian workers are in enterprises that keep no formal accounts: small manufacturing units, retail traders, transport operators, construction workers, domestic workers, street vendors. Their contribution has to be estimated by taking a sample and blowing it up by an assumed number of units, and both the sample and the multiplier are uncertain.

3. Illiteracy and the absence of accounts. Even where an enterprise is willing to report, many small producers do not keep records that would answer the question. A farmer asked for the value of his output in a year is being asked something he has never computed.

4. Absence of occupational specialisation. A rural household commonly farms, keeps animals, drives a vehicle in the off season and runs a small shop. The income is one income and cannot be split between industries, which is why the accounts have a separate head for mixed income of the self employed rather than trying to divide it into wages, rent, interest and profit.

5. The statistical machinery, and the delay in it. Estimates rest on large surveys and censuses that are conducted at long intervals: the periodic labour force survey, the household consumption expenditure survey, the economic census, the agricultural census. Between two rounds the earlier structure has to be projected forward, and if the structure has changed the projection carries the error. India's last completed population census was in 2011, and every per capita figure since then has used a projected population.

6. Double counting in practice. Where the same output passes through several hands and each is surveyed separately, the risk is not theoretical.

7. Regional and seasonal variation. Prices for the same commodity differ across States and across the year, so a single valuation is an average that fits nowhere exactly.

8. Revision. The published figure is not one figure. It is a sequence: Advance Estimate, then Provisional, then First, Second and Third Revised Estimates, each based on more complete data. The Statistical Appendix to the Economic Survey labels the vintage of every figure, and a growth rate quoted without its vintage is not a reliable figure. A revision of half a percentage point is common.

9. Base year revision. The current series has base year 2011-12. Changing the base changes the level and sometimes the growth rate of the entire series, because it changes the weights of industries and the prices at which constant price output is valued.

A worked example: what one household hides from the accounts

The household. Ramesh and Sunita live in a village in Vidarbha with two children and Ramesh's mother.

What the accounts capture. Ramesh sells 18 quintals of cotton at the market, which is recorded as agricultural output. Sunita works 60 days under the rural employment guarantee scheme, and those wages are recorded.

What the accounts partly capture. They keep four quintals of jowar for the family. This is production for self consumption and the national accounts do impute it, but the imputation is based on an assumed retention rate for the district rather than on their actual decision.

What the accounts miss entirely.

• Sunita cooks, cleans, fetches water and cares for her mother in law for about seven hours a day. If a paid worker did the same work it would be counted in national income. Because she does it unpaid, it is not.
• Ramesh helps a neighbour build a cattle shed for four days and the neighbour helps him plough. No money changes hands and no output is recorded, though two sheds and two ploughed fields exist that did not before.
• The family collects firewood and fodder from common land.
• Ramesh drives a hired tempo for two months in the off season and is paid in cash without a record.

The size of the problem. Nobody can say precisely, which is itself the point. What can be said is the direction: India's measured national income is below its real production, that the gap is largest in exactly the households the accounts most need to describe, and that the gap has narrowed as the economy has monetised and as digital payments have brought small transactions into the record.

Where a law student meets it. Compensation and maintenance are frequently assessed on proved income, and a person whose real production is invisible to the accounts is usually also a person whose income is hard to prove in court. The measurement problem and the evidentiary problem have the same root.

What has improved

An answer that lists only difficulties is dated. Four genuine improvements should be named.

The base year revision to 2011-12 brought in enterprise level data from the corporate affairs database instead of relying on sample estimates for the corporate sector, and adopted gross value added at basic prices in line with the international standard.

Digital payments have made a large volume of small transactions recordable that previously were not.

The goods and services tax network produces a continuous record of business to business transactions, which improves the estimation of the unorganised sector where it is registered.

Administrative data from tax returns, provident fund accounts and company filings are increasingly used to supplement surveys.

None of these reaches the genuinely non monetised part of production, which remains the hard core of the problem.

What beginners get wrong

"The difficulties mean the figures are useless." They mean the figures are estimates with a margin of error and a stated vintage. That is true of every economic statistic in every country.

"Only poor countries have this problem." Every country imputes owner occupied rent and omits unpaid domestic work. The difference is one of degree, and the degree is large.

"A revision proves the earlier figure was dishonest." A revision is what happens when better data arrive, and a statistical system that never revised would be the suspicious one.

"Black money is simply added on." There is no way to add it on reliably. Estimates of the unaccounted economy vary so widely that quoting one as a fact is a mistake.

Quick revision

1. Conceptual difficulties: identifying final against intermediate goods; valuing government services, which are valued at cost; separating transfers from payments for production; the arbitrary line in imputation; estimating depreciation; illegal and unrecorded activity; and quality change and new goods.
2. Practical difficulties in India: non monetised production; the size of the unorganised sector; illiteracy and absence of accounts; absence of occupational specialisation, which forces the category mixed income of the self employed; inadequate and infrequent statistical machinery; practical double counting; regional and seasonal price variation; repeated revision; and base year change.
3. Unpaid domestic work is production and is not counted, so measured national income is below real production.
4. The last completed census is of 2011, so every per capita figure since is based on a projected population.
5. Estimates run Advance, Provisional and Revised. Always quote the vintage. The current base year is 2011-12.
6. Improvements: the 2011-12 base revision using corporate filings and GVA at basic prices, digital payments, the goods and services tax network, and greater use of administrative data.

Test yourself

1. Distinguish the conceptual from the practical difficulties in estimating national income. Conceptual difficulties concern what ought to be counted and how it should be valued: whether a good is final or intermediate, how to value government services that are not sold, which payments are transfers rather than payments for production, how far non marketed output should be imputed, how depreciation should be estimated, and how quality change should be handled. Practical or statistical difficulties concern the availability and reliability of data: non monetised production, the size of the unorganised sector, illiteracy and the absence of accounts, the absence of occupational specialisation, infrequent surveys, and the delays and revisions that follow.

2. Why does the absence of occupational specialisation create a difficulty, and how do the Indian accounts deal with it? Because a single rural household commonly farms, keeps livestock, runs a small trade and hires out labour or a vehicle, and receives one undivided income from all of it. The income method requires factor incomes to be classified as wages, rent, interest and profit, and no such division can be made where the same person supplies land, labour, capital and enterprise to the same activity. The Indian accounts therefore use a separate category, mixed income of the self employed, which records the combined return rather than attempting a split that the data cannot support.

3. How are government services valued in national income, and what problem does that create? They are valued at the cost of providing them, principally the salaries of the staff, because they are not sold and therefore have no market price. The problem is that cost is an input measure rather than an output measure, so a department that spends more without producing more is recorded as having produced more, and improvements in efficiency that reduce cost appear as a fall in output. No generally accepted alternative exists, which is why the convention has survived.

4. What is meant by non monetised production, and why does it matter in India? It is production that does not pass through a market and for which no money changes hands, such as grain retained for family consumption, fodder and firewood gathered, houses built with family labour, and services exchanged between neighbours. It matters in India because a considerable part of rural production takes this form and only some of it is imputed. The consequence is that measured national income understates real production, and the understatement is largest for the poorest households, which are precisely the ones policy most needs to measure.

5. Why is a national income figure always accompanied by its vintage? Because the same year's income is published several times as data become more complete: first as an advance estimate based on limited indicators, then as a provisional estimate, and then as first, second and third revised estimates. Revisions of half a percentage point in the growth rate are common. Quoting a figure without saying which estimate it is invites comparison between numbers built on different information, which is why the Statistical Appendix to the Economic Survey labels every entry.

6. State four ways in which the measurement of national income in India has improved. The base year revision to 2011-12 brought company level data from the corporate affairs database into the estimation of the corporate sector in place of sample based methods, and adopted gross value added at basic prices in line with international practice. The spread of digital payments has brought many small transactions into the record. The goods and services tax network provides a continuous record of business to business transactions, which improves estimation for the registered part of the unorganised sector. And administrative data from tax returns, provident fund accounts and company filings increasingly supplement periodic surveys. None of these reaches genuinely non monetised production, which remains the central difficulty.