SNA Guide  /  Overview
Economics — OECD/IMF/Eurostat Handbook

Measuring the Complete Economy

A comprehensive guide to the Non-Observed Economy — covering every component, GDP measurement methods, and why production that nobody directly sees still belongs in official statistics.

12Chapters covered
5NOE categories
3GDP approaches

Jump to a chapter

Chapter 1

The Problem of Exhaustiveness

Why GDP must cover all production inside the boundary — including what surveys never directly observe.
Chapter 2

Conceptual Framework

What counts as production, who produced it, which country gets it, and how all three GDP approaches relate.
Chapter 3

Notions of the NOE

Underground vs informal vs illegal — why these categories differ and how to classify missing production.
Chapter 4

Assessing National Accounts

Detect missing production through data confrontation, upper-bound analysis, and special-purpose surveys.
Chapter 5

Achieving Exhaustiveness

Labour-input, supply-based, demand-based and commodity-flow methods for estimating what surveys miss.
Chapter 6

Improving Data Collection

Business registers, survey design, non-response — fixing the system that creates gaps in the first place.
Chapter 7

Implementation Strategy

Prioritisation, planning, backcasting and the operational management of NOE measurement programmes.
Chapter 8

Underground Production

Legal production deliberately concealed — how it differs from informal and illegal, and how to measure it.
Chapter 9

Illegal Production

Why illegal activity can still belong in GDP, and why theft and extortion definitely don't constitute production.
Chapter 10

Informal Sector

Household unincorporated enterprises — defining, surveying, and including them through mixed surveys.
Chapter 11

Household Own-Use

Own-use goods production, imputed housing rent, and the boundary that excludes most unpaid services.
Chapter 12

Macro-Model Methods

Why currency-demand and electricity models can't replace national-accounting methods — and when models are legitimate.

The SNA handbook opens with a deceptively simple objective: GDP should cover all economic production that falls within the national-accounts production boundary — whether statisticians directly observe it or not. This is called exhaustiveness.

Core Problem
Poor coverage distorts both GDP levels and growth rates. GDP per capita, international comparisons, and even GDP-linked contributions to international organisations can all be wrong if production is systematically missed. Growth estimates become biased when the missing economy grows at a different rate than the observed economy.

Observed vs Measured: A Critical Distinction

Three concepts that are not identical:

Economic activity — what actually happens in the economy
↓
Observed activity — what surveys and administrative sources directly capture
↓
Measured activity — what ultimately enters official GDP (including indirect estimates)
Numerical Example

Suppose true productive activity = ₹100.

Stage 1: Surveys capture ₹85. Non-observed = ₹15.
Stage 2: National accountants estimate +₹10 indirectly.
GDP = ₹85 + ₹10 = ₹95. Non-measured = only ₹5.

Claiming "the shadow economy is 15%, therefore GDP is underestimated by 15%" can be completely wrong — some of that activity may already be incorporated through adjustments.

The Five NOE Categories

Underground
Legal production deliberately hidden from authorities — tax evasion, avoiding regulations.
Informal
Small household enterprises below registration thresholds — legal, but structurally invisible to standard business surveys.
Household Own-Use
Goods/services produced and consumed by the same household — own-account construction, subsistence farming.
Illegal
Activities prohibited by law — but still involving genuine production and exchange.
Statistical Underground
Production missed due to statistical deficiencies — late register updates, non-response, frame undercoverage.

The Production Boundary

⚠ Key Rule
Household produces a GOOD for itself → often included. Farmer grows potatoes for own consumption ✓. Household constructs own dwelling ✓.

Household produces a SERVICE for itself → generally excluded. Cooking ✗. Cleaning ✗. Childcare ✗. Driving family members ✗.

Output ≠ Sales

If a baker produces 100 breads but sells only 80, output = ₹2,000, not ₹1,600. The remaining ₹400 enters inventory. This distinction is fundamental to accurate national accounting.

Who Is the Producer? Institutional Units

An institutional unit can own assets, incur liabilities, and conduct transactions on its own account. When acting as a producer, it is called an enterprise — which is much broader than "registered company."

EntitySNA ClassificationKey Characteristic
Reliance IndustriesNon-financial corporationSeparate legal entity; assets distinct from owners
Ravi's tea stallHousehold unincorporated enterpriseNo legal separation between household and business
Large partnership with full accountsQuasi-corporationEconomically operates like a corporation despite legal form
Government departmentGeneral governmentProduces non-market services

GDP Is Based on Residence, Not Nationality

GDP belongs to the country where the centre of economic interest is located (normally at least one year of presence). A foreign worker contributing to Indian production adds to India's GDP, but their income is part of their home country's national income. This is why GDP ≠ GNI.

Three Valuation Concepts

Basic price → what the producer effectively receives Producer price → basic price + applicable taxes Purchaser price → what the buyer actually pays (includes margins) GDP = GVA at basic prices + Taxes on products − Subsidies on products

The Accrual Principle

If a baker receives flour in December but pays in January, national accounts record the transaction in December — when the change of ownership occurred. Cash movement ≠ time of economic activity.

Key Expenditure Components

The familiar C + I + G + X − M conceals seven categories: household final consumption; government final consumption; NPISH consumption; gross fixed capital formation; changes in inventories; acquisition less disposal of valuables; exports minus imports. "I" in the textbook covers both GFCF and inventory changes.

⚠ Important: Inventory Holding Gains ≠ Production
If a baker's flour inventory rises from ₹100 to ₹120 due to inflation, that ₹20 is a holding gain, not new production. Inventories should be valued at prices prevailing when they enter/leave — specifically to exclude holding gains from GDP.

Chapter 3's central lesson is subtler than it looks. The five familiar NOE categories are only the beginning. The deeper insight is distinguishing:

Nature of Activity
What kind of economic activity is it? Underground, informal, illegal, own-use?
Cause of Non-Observation
Why did the statistical system fail to observe it? Undercoverage, non-response, underreporting?

These are not the same question. An unregistered business might be so because the statistical system failed, because the owner deliberately avoided registration, or because registration isn't legally required — three identical-looking situations demanding different solutions.

Illegal ≠ Excluded from GDP

This conflicts strongly with everyday intuition. The handbook explicitly says illegal production belongs inside the SNA production boundary. The deciding factor is whether production occurred, not whether government approves of it.

⚠ Critical Distinction: Consent
A buyer voluntarily pays ₹1,000 for an illegal product → mutually agreed transaction → potentially production. Someone steals ₹1,000 → no mutual agreement → no new output produced → not production. Theft is redistribution, not creation.

The Istat Analytical Framework: T1–T7

Converts overlapping NOE concepts into seven mutually exclusive statistical types — each corresponding to a different reason why a business might be missing:

TypeCategoryWhy MissedExample
T1Statistical undergroundNon-responseFirm on register, doesn't return questionnaire
T2Statistical undergroundRegister not updatedNew firm not yet in database
T3Statistical undergroundNot registered — statistical reasonFirm not hiding; system just missed it
T4Economic undergroundUnderreportingFirm reports ₹70 when actual is ₹100
T5Economic undergroundDeliberately not registeredOwner avoids taxes/regulation
T6InformalNot registered — not requiredTiny enterprise below threshold
T7IllegalCannot register — illegal activityProhibited goods producer
T3 vs T5 vs T6 — the most important distinction: All three appear as "unregistered business." But T3 = statistical system failure; T5 = deliberate concealment; T6 = not legally required to register. These demand completely different statistical responses. Unregistered ≠ underground.

Five Bakers — One Economy, Five Different Problems

Baker A — Normal
Registered, reports everything. ₹10L observed. Not NOE.
Baker B — Underground T4
Registered, actual ₹10L, declares ₹7L to avoid tax. ₹3L underground underreporting.
Baker C — Underground T5
Operates legally but deliberately refuses registration. ₹5L. Deliberate non-registration.
Baker D — Informal T6
Tiny household bakery, not required to register. ₹3L. Informal, not underground.
Baker E — Statistical T2
Perfectly legitimate, opened last month, not yet on register. ₹4L. Register deficiency.
The point: Calling all ₹15L of missing production "informal economy" would be badly misleading — and would tell the statistician almost nothing about how to recover it. Each type demands a different solution.

Chapter 4 changes the question from "what can be non-observed?" to "given existing GDP estimates, how can we diagnose whether production is still missing, and roughly how much?"

The Forensic Accounting Principle
Hidden activity leaves statistical footprints. Every transaction has relationships to another buyer/seller, labour, income, expenditure, taxes, inputs and outputs. You rarely observe missing production directly — you find it because the rest of the economy stops adding up.

Tool 1: Data Confrontation

National accounts contain identities: GDP_Production = GDP_Expenditure = GDP_Income; and for each product, Supply = Use. These create multiple independent windows into the same economy. If they tell different stories, investigate.

VAT Comparison Example
Theoretical VAT (from supply-use tables) = ₹100
Actual VAT collected = ₹92 → Investigate the ₹8 gap
But if actual VAT = ₹107 → Your measured tax base is probably too small!

The second case is more alarming: government collected more tax than your production estimates should generate — meaning GDP may be too low.

⚠ Discrepancy ≠ NOE
A discrepancy says "Something needs explaining." First remove conceptual differences, legal differences, timing differences and statistical measurement errors. Only then does the remaining unexplained residual potentially signal NOE. A Dutch study found roughly three-quarters of an apparent income discrepancy was explained by definitional differences.

Labour Confrontation

Enterprise Side (Labour Used)
Ask businesses: "How many people do you employ?" → 800,000 reported bakery jobs.
Household Side (Labour Supplied)
Ask people: "Are you working in a bakery?" → 1,000,000 self-reported. Gap = 200,000 unexplained workers.
⚠ The Gap Is Only a Lower Bound
If true employment is 1,200,000 but household surveys see 1,000,000 and enterprise surveys see 800,000, the observed discrepancy is only 200,000 — but actual enterprise undercoverage is 400,000. Both sources can independently miss labour.

Tool 2: Upper-Bound Estimation

Instead of estimating the exact NOE, ask: "Could it even plausibly be 25% of GDP?" For each GDP component, estimate the maximum plausible missing amount under deliberately generous assumptions. Sum them. This ceiling helps test implausibly large claims.

Canada's Upper-Bound Logic

Government expenditure: Upper bound = 0.

Business fixed investment: Upper bound ≈ 0 (businesses rarely hide their own expenditures).

Residential construction and household consumption: Largest vulnerability — cash payments, renovation off-books, tips.

Result: Maximum missing underground GDP ≈ 2.7% of published GDP — far below macro-model estimates of 10–14%.

Skimming Does NOT Always Reduce GDP

If a flour wholesaler hides sales to a bakery, the bakery still pays the full ₹100 and passes it into bread prices. The hidden intermediate transaction may already be embedded in final prices recorded by household expenditure surveys. Hidden turnover ≠ hidden value added.

Tool 3: Special-Purpose Surveys

Ask Buyers, Not Sellers
Household expenditure surveys can reveal what sellers hide. Buyers have less incentive to conceal a transaction they participated in innocently.
Time-Use Surveys
Forcing people to account for 24 hours can reveal secondary activities — including underground or informal work — that job-status questions never capture.
Tax Audits
Strong individual-level evidence, but samples are NOT random — auditors target suspicious firms. Audit results cannot be extrapolated as representative estimates.

Four Types of Adjustment (Eurostat Framework)

Adjustment TypePurposeExample
Data-validationTwo sources contradict each otherCorrect the underlying data
ConceptualSource data don't match SNA definitionsBusiness accounts use historical-cost inventories; SNA needs replacement cost
ExhaustivenessKnown NOE is absentAdd estimated informal-sector GVA
BalancingForce accounting identity to holdRemove residual inconsistency after all other adjustments
⚠ Warning
Balancing adjustment ≠ NOE adjustment. One forces accounting consistency. The other explicitly adds activity believed to be missing. Confusing them misrepresents what GDP actually contains.

Chapter 4 told us where production might be missing. Chapter 5 asks: once we believe production is missing, how do we actually estimate it and put it into GDP? The philosophy is not "apply 10% to GDP" but detailed, specific adjustments based on identifiable economic relationships.

The Four Indirect Production Approaches

Supply-Based
Observe inputs → infer output. 1,000 kg flour × 1.5 kg bread/kg flour = 1,500 kg bread output.
Demand-Based
Observe who used the output. 10,000 vehicles × average repair cost = estimated vehicle-repair output.
Income-Based
1,000 doctors × average receipts per doctor = estimated healthcare output. Requires adjustment since tax records may understate income.
Commodity-Flow
Supply identity: Domestic output + Imports = IC + C + GFCF + Inventories + Exports. Use known sides to infer the unknown.

The Labour-Input Method — Most Important

Core steps: (1) estimate total labour input from household/demographic sources; (2) estimate output and GVA per unit of labour from enterprise data; (3) multiply them.

Output = L × (Output / L) GVA = L × (GVA / L) Where L = standardized labour input (hours or FTE) The gap (L_household − L_enterprise) reveals potential missing production
⚠ Critical: Person ≠ Job ≠ Hours
One person can hold two jobs. Part-time workers differ from full-time workers. Both household and enterprise sources must be converted into hours worked or full-time equivalent employment before comparison. Naive person-count comparison is misleading.

Italy's Labour-Input System (~70% of GDP)

Italy historically estimated around 70% of production using the labour-input method. The system integrates enterprise records + household records + administrative data, distinguishing regular employment, full-time irregular jobs, and multiple jobs. Everything converts to FTE before applying productivity ratios.

Supply and Use Tables: The Master Framework

For each INDUSTRY: Output = Intermediate Consumption + Value Added For each PRODUCT: Supply = Use (Output + Imports = IC + C + GFCF + ΔInventories + Exports)
How the Framework Reveals Missing Production

Baker claims output = ₹200. Households report bread purchases = ₹250.

Supply = ₹200. Use = ₹250. GAP = ₹50.

The table cannot balance. Now investigate: baker output understated? Household expenditure overstated? Imports missing? The balancing process isn't "force the numbers to agree" — it's a structured investigation.

⚠ Don't Estimate GVA Alone
Always estimate Output and Intermediate Consumption separately, then derive GVA = Output − IC. Value added doesn't have an independently observable price × quantity structure, and supply-use tables need them separately.

Industry-Specific Methods

IndustryKey ObservableMethod
AgricultureArea × yield per hectareCrop-cutting surveys, satellite imagery, seed consumption
ConstructionCement, steel, bricks supplyMaterial input × output ratios (adjusting for material mix)
Retail tradeMerchandise turnoverOutput = trading margin, not sales. Commodity flow with margin rates.
TransportVehicle registrations × trips × farePhysical activity indicator
Owner-occupied housingStock of owner-occupied dwellingsComparable-rent method: stock × market rent of similar rented property
Domestic servantsLabour-force survey → workersWorkers × average compensation
The Deeper Lesson: GDP is not "collected" from surveys. It is constructed by combining imperfect observations inside a system of economic identities. The better the statistical office at triangulating those observations, the closer published GDP gets to actual production.

Chapters 4–5 described how to detect and estimate missing production. Chapter 6 asks: why keep correcting bad data forever? Can we improve the statistical system so less becomes non-observed?

Ch. 4: Find the holes (data confrontation, upper bounds)
↓
Ch. 5: Estimate what's inside the holes (indirect methods)
↓
Ch. 6: Fix the system creating the holes (upstream quality control)

Why Institutions Matter for Statistical Quality

A bakery owner who fears that statistical data will be shared with tax authorities will deliberately underreport. Statistical legislation guaranteeing confidentiality isn't just an ethical issue — it directly affects measurement error.

The Three Fundamental Survey Problems

ProblemFirm on Frame?Responds?Main Solution
Undercoverage❌ No—Better register/frame + supplementary survey + NA adjustment
Non-response✅ Yes❌ NoFollow-up, imputation or reweighting — should be handled within the survey programme
Underreporting✅ Yes✅ YesSpecial investigation + national-accounts adjustment. Ordinary editing often cannot detect systematic deliberate underreporting.
⚠ Critical Implication
Non-response should NOT remain an NOE problem — handle it within the basic survey programme through imputation and reweighting. But underreporting — where a firm responds but lies — typically requires national-account adjustments beyond survey-level corrections.

The Business Register: Frame Quality Is Everything

The survey frame has more influence than any other aspect of survey design on survey coverage, and therefore directly affects NOE measurement. A beautiful questionnaire and perfect methodology are useless if the frame is missing 20% of enterprises.

Why Multiple Administrative Sources Are Better

GST database captures firms A, B, C. Payroll database captures B, C, D. Corporate tax captures A, C, E. Customs captures C, F.

Combined → A, B, C, D, E, F. Much better coverage than any single source.

But this requires a common enterprise identifier to match units across databases — otherwise matching becomes prohibitively expensive.

Survey Design Principles That Reduce NOE

Questionnaire Simplicity
150 questions → more non-response. 30 essential questions → more usable data. More questions can produce less information.
Size-Appropriate Design
Large firms: full establishment survey. Medium: list-based. Small/micro: household or area-based survey with shorter questionnaire.
Non-Response Imputation
Never treat non-respondents as inactive/zero. If responding and non-responding firms are similar, expected output ≈ average × total firms in stratum.

The Dangerous Silo Problem

Enterprise survey team, household survey team, and national accounts team each work in isolation → the same statistical problem persists indefinitely. Chapter 6 wants a feedback loop: national accounts detect a discrepancy → enterprise survey investigates → business register updated → questionnaire changes → better survey next year → smaller adjustment.

Strong basic data + Targeted supplementary data + National-account reconciliation

Chapter 7 is the management chapter. All the conceptual and technical tools are in place — now: what do you actually do first, second, third?

The Master Implementation Cycle

1. Define objectives (exhaustive GDP? separate informal-sector statistic?)
↓
2. Choose an analytical framework (adapted to the country)
↓
3. Assess the current system (work backwards from GDP to root cause)
↓
4. Prioritise: importance of problem ÷ cost of improvement
↓
5. Create a specific plan: initiative + deadline + owner + expected output
↓
6. Implement, document everything, evaluate, revise ↺

Short-Term vs Long-Term

Short-Term Fix
Compensate for the hole: better imputations, model-based adjustments, supplementary investigations, better reconciliation. Can often be introduced quickly and cheaply.
Long-Term Fix
Repair the hole: improve business register, redesign surveys, introduce better administrative data, add supplementary collections. Requires substantial resources but reduces the need for adjustments.

The Critical Problem: Methodological Improvements Distort History

Suppose old GDP = ₹110 in 2026, and you discover ₹15 of informal activity previously missed. New GDP = ₹125. A naive user calculates (125 − 110)/110 = 13.6% growth. But actual growth was maybe 5% — the rest is better measurement, not new output.

Solution 1: Backcasting (Preferred)
Recalculate all earlier years using the new methodology. Level is higher throughout, but growth rates remain accurate. Expensive but cleanest.
Solution 2: Delay the Break
Accumulate revisions and introduce them together after several years. Protects short-term comparability, delays improved levels.
Solution 3: Old Levels, New Growth
Publish GDP levels using old methodology but calculate growth rates using the improved method. A compromise preserving continuity.
Whatever approach: Tell users what changed. The handbook says revisions should be documented publicly, and forthcoming revisions should be announced well in advance so their timing isn't suspected of being politically motivated.

Countries with Large Household Production

Four Priority Tools
  • Business register — at minimum, ensure large and medium enterprises are captured. Don't attempt to register every micro-enterprise.
  • Mixed household-enterprise survey — sample households, find which members operate businesses, then survey those enterprises.
  • Labour-input method — use frequent labour-force surveys to extrapolate between expensive comprehensive benchmarks.
  • Time-use surveys — especially important for capturing women's production activities that conventional employment questions miss entirely.
SNA Definition
Underground production = legal productive activity deliberately concealed from public authorities, to avoid income/VAT taxes, social-security contributions, regulations, or administrative procedures.

The Critical Distinctions

Underground (Ch. 8)
Legal activity, deliberately hidden. Baker sells ₹10L, reports ₹7L. Legal product + concealed = underground.
Illegal (Ch. 9)
The activity or product itself is prohibited. Separate category entirely.
Informal (Ch. 10)
Small unregistered household enterprise — but NOT necessarily deliberately hiding anything. Tiny tailor not required to register.

Two Types of Underground Enterprise

Type A — Unregistered
Amit moonlights preparing tax returns evenings/weekends without registering. Tiny individually, but 1,000,000 × ₹3L = ₹30,000 crore economy-wide. Find via mixed household-enterprise surveys or labour confrontation.
Type B — Registered but Underreports
Baker IS registered. Actual sales ₹10L. Tax accounts ₹7L. The same ₹7L probably enters the statistical survey. Ordinary editing cannot detect systematic deliberate underreporting.

Measuring What People Are Deliberately Hiding

Ask Buyers, Not Sellers
"Did you pay cash for plumbing work?" is less threatening than "Did you evade tax?" Buyers may not even know if the seller was evading.
Labour Supply Surveys
Workers are often more willing to report "Yes, I work" than businesses are to report hidden revenue. Gradual questionnaire design helps: start with innocuous questions.
Commodity Balances
If farmers claim milk production = 100M litres but cattle numbers and forage consumption imply ~130M, something doesn't reconcile. Physical identities challenge reported data.

The "Shadow Economy Percentage" Problem

Two Very Different Percentages

Suppose official GDP = ₹900 and underground production = ₹100.

Underground as % of official GDP: 100/900 = 11.1%
Underground as % of total production: 100/1000 = 10.0%

Studies often quote the first (larger) number. And crucially: if ₹70 of the ₹100 is already incorporated through indirect adjustments, the true non-measured underground is only ₹30 — not ₹100.

⚠ The Key Subtraction
Non-measured underground = Total underground production − Underground already included in GDP. Never add a gross underground estimate to published GDP without first deducting what may already be inside it.

Other Concepts in Chapter 8

Shuttle Trade
Entrepreneurs buying goods abroad for resale without full customs declaration. Measure via trips × average imported value, or domestic supply/demand reconciliation.
Cross-Border Shopping
A person buying one cheaper phone abroad for personal use is NOT underground production — purchasing for own consumption is not production.
Capital Flight
Transferring assets abroad is NOT production. It may originate from underground proceeds, but the financial transfer itself doesn't create GDP.

The most counterintuitive chapter in the handbook. GDP measures economic production, not welfare or social desirability. An illegal product whose production involves real mutual exchange can belong in GDP.

Three Reasons to Include Illegal Production

Accounting Consistency
Illegal income is subsequently spent on legal goods, services and assets. Excluding the production while downstream spending remains visible creates internal inconsistencies.
International Comparability
Prostitution is legal in Country A, illegal in Country B. If GDP counts it only when legal, countries appear economically different due purely to legal differences, not real production.
Price Integrity
GDP should value actual market prices — even if illegality makes the price artificially high. National accounts describe the economy that actually exists.

The Most Important Distinction: Consent

Illegal Production (inside GDP)
Buyer voluntarily pays ₹1,000 for an illegal product. Mutual agreement exists. A good or service has been produced and exchanged. → Potentially production.
Theft, Robbery, Extortion (outside GDP)
No mutual agreement. Existing assets are redistributed. No new output is created. → Not production. Treated as "other changes in assets."

Complex Cases

ActivityMutual Agreement?GDP?Reasoning
Illegal drug saleYes✓ IncludeGenuine production and exchange
TheftNo✗ ExcludeRedistribution, not production
Fencing stolen goodsYes (resale)✓ Trade marginThe resale service creates value added even if the underlying good was stolen
Service-linked bribeGenerally yesPotentiallyIf linked to an actual service provided, may be treated as part of its price
Extortion paymentNo✗ ExcludeNo service produced; treated as other asset change
Money laundering feeYes✓ Service marginDifference between illegal cash and "clean" value = payment for laundering service

Measuring the Unmeasurable: Supply-Use-Income Triangulation

Drug Market Estimation Logic
Supply side: Domestic production ₹80Cr + Imports ₹30Cr = Supply ₹110Cr
Use side: Domestic consumption ₹70Cr + Exports ₹40Cr = Use ₹110Cr ✓

Then cross-check against estimated number of users × average consumption × price. The seizure-rate assumption is the weakest link: if you assume police intercept 10% of supply, the estimate is 10× seizures — but if the true rate is 5%, your estimate doubles. Always perform sensitivity analysis.

⚠ Double-Counting Danger
Some illegal production may already be inside official GDP — classified under a legal-sounding industry. A business providing illegal services may register officially as "massage services" or "room rental." Misclassified ≠ missing. Before adding an illegal-production estimate, verify how much is already implicitly captured elsewhere.
Central Lesson
Informal ≠ illegal ≠ underground. The vast majority of informal-sector activities produce goods and services that are perfectly legal. The challenge is not whether their output belongs in GDP — it does. The challenge is how to find and measure it.

What Makes an Enterprise "Informal"?

Informal enterprises are a subset of household unincorporated enterprises — meaning there is no economically distinct corporate entity separating the household from the business. Meena's samosa stall uses her household kitchen, her personal savings, and her family members. Business and household finances are intertwined.

Meena's Samosa Business — The Classic Informal Enterprise

Buys potatoes, flour and oil each morning. Makes 100 samosas. Sells them outside an office. Works herself. Daughter helps occasionally. No formal accounts. No separate company. Uses household kitchen. Earns her family's living.

Sales = ₹1,000. Intermediate inputs = ₹600. GVA = ₹400.

Her lack of corporate structure doesn't make the ₹400 disappear from the economy.

Key Distinctions: What Informal Is NOT

Informal ≠ Illegal
Meena isn't doing anything illegal. Making food and selling it is perfectly legitimate. Most informal activity is lawful.
Informal ≠ Underground
She isn't deliberately hiding from authorities. Her income may simply be too small to trigger registration requirements.
Informal ≠ "Whatever Surveys Missed"
Defining informality by survey exclusion makes the concept change when surveys expand. The economic concept must be defined independently of the measurement instrument.

The Mixed Household-Enterprise Survey — The Key Method

Phase 1: Sample geographical areas → list all households → ask "Does anyone operate a business?"
↓
Identify potential informal entrepreneurs (Meena, Ravi, etc.)
↓
Phase 2: Return to those households → detailed enterprise survey: revenue, inputs, employment, capital, registration status
↓
Calculate: GVA = Output − Intermediate Consumption → Enter into national accounts

Questionnaire Design for Informal Enterprises

Many informal entrepreneurs have no formal education and maintain no usable records. Asking "What was your intermediate consumption last year?" is useless. Instead:

Short Reference Period
"Revenue last week" or "Revenue yesterday" — people remember accurately. Maximum recommended reference period: 1 month.
Spread the Survey
Survey different subsamples across all 12 months to capture seasonal patterns without requiring respondents to recall an entire year.
Ask About High/Low Seasons
"Which months are your best months? By how much do sales vary?" lets you extrapolate from short-period data.

Labour-Force Surveys as Extrapolators

A comprehensive mixed survey in 2025 estimates GVA/worker = ₹2L for informal bakeries. Annual labour-force surveys track workers in 2026, 2027 etc. Multiplying workers × ₹2L extrapolates the benchmark — avoiding the expense of a full enterprise survey every year.

Core Principle
Production does not require a sale. Goods produced by households for their own use can enter GDP. The boundary that matters is the production boundary, not whether cash changed hands.

The Goods vs Services Distinction

Household Produces GOODS for Itself — generally included
✓ Crops, livestock, fishing, firewood
✓ Processed food (butter, preserved meat)
✓ Clothing, pottery, furniture
✓ Own-account construction of a house
Household Produces SERVICES for Itself — generally excluded
✗ Cooking for own household
✗ Cleaning own home
✗ Washing own clothes
✗ Childcare for own children
✗ Driving family members
Two exceptions: owner-occupied housing ✓ and paid domestic servants ✓

Imputed Housing Services

The Twin Apartment Example

Rahul owns Apartment A, rents it to Priya for ₹30,000/month → GDP includes ₹30,000 housing service.

Arjun owns identical Apartment B, lives in it himself → cash flow = ₹0.

If GDP recorded zero for Arjun, countries with more homeowners would mechanically have lower GDP even though identical housing services are consumed. So national accounts impute rent: Arjun is treated as producing ₹30,000 of housing services and consuming them.

Output = ₹3,60,000/year. Household consumption = ₹3,60,000/year. Cash flow = ₹0. GDP contribution = real.

Why Unpaid Cooking Is Excluded but Hired Cooking Is Included

If Person A cooks and cleans herself, GDP = 0 for those services. If Person B hires the same work for ₹50,000/year, GDP rises by ₹50,000 — even though the same real services are enjoyed. The SNA acknowledges the exclusion has historically raised concerns including gender bias — but it's a deliberate accounting compromise, not a claim that household work has no value.

Subsistence Agriculture: The Largest Component

Total production (from agricultural survey) × Proportion retained for own use × Farm-gate/harvest price = Value of own-consumed agricultural output

Valuation Challenges

Comparable Market Price
Use local market price for the same good. If wheat sells for ₹30/kg, own-consumed wheat is valued at ₹30/kg.
Firewood/Water Problem
No local market price may exist. Time-use data may be more reliable than price estimates. More time ≠ more output if the resource is becoming harder to find.
Construction
Own-account house building = Gross Fixed Capital Formation (not consumption). Stock-based estimation: housing stock growth + replacement rate.

Time-Use Surveys as the Key Tool

Asking someone to account for all 24 hours reveals secondary production activities — especially women's contributions — that job-status questions routinely miss. Warning: more hours spent collecting firewood can mean the resource is becoming scarcer, not that more firewood is being produced. Time ≠ output.

Cross-Checking Subsistence Estimates

Plausibility Tests
  • Does implied per-capita food consumption convert to plausible daily calorie and protein intake?
  • Can the cattle herd reproduce fast enough to support the estimated slaughter rate?
  • Does estimated fish output correspond to the number of boats and nets in use?
  • Do hunting estimates match licences, ammunition sales, and trophy trade?
The Handbook's Unusually Strong Position
Macro-model methods are discussed partly because they often produce spectacularly large estimates that attract media and political attention — not because they are considered useful for compiling national accounts. The handbook is unusually blunt: these are interesting research tools, unsuitable for exhaustive GDP measurement.

The Six Fundamental Problems

Why Macro-Models Fail for GDP
  • The thing being measured is poorly defined — different papers call different things "shadow economy."
  • Assumptions are overly simplistic — the proxy variable has many other causes.
  • Results are unstable when assumptions change — reasonable variations can produce 3× different estimates.
  • Different models give wildly different answers for the same country and time period.
  • They produce a single economy-wide number, not the industry/expenditure detail national accounts require.
  • They cannot easily integrate with the detailed data-based estimates already in the accounts.

Monetary Methods

MethodCore IdeaKey Assumption (and why it fails)
Transaction MethodMV = PT = k×GDP; excess transactions reveal hidden outputAssumes PT/GDP is constant — but financial market growth, shifting payment structures, etc. make it unstable
Cash/Deposit RatioRising cash:deposits ratio → more underground cashRatio changes with interest rates, banking habits, ATM spread, savings behaviour — not just underground activity
Cash Demand ModelRegress cash on taxes, income, interest rates; attribute tax-induced excess to underground economyStill needs benchmark + velocity assumption; same observable can arise from multiple causes

The Electricity Method

Why Electricity Growth ≠ Hidden GDP Growth

If official GDP grows 2% but electricity grows 6%, the naive inference: "4% of GDP is hidden."

Problem 1 — Fixed costs: A factory shutting production from 1,000 to 500 units may only cut electricity from 100 to 80 units. Up to ~1/3 of industrial electricity may behave as a fixed cost.

Problem 2 — Energy efficiency: Same output with better machines → same GDP, less electricity. Model wrongly infers decline.

Problem 3 — Structural change: Economy shifts from steel to software. GDP rises while electricity intensity falls systematically.

Russian example: Electricity-based estimates suggested the post-1992 economy was doing far better than official GDP implied — but subsequent analysis found major measurement problems. The method was ultimately not used in the official GDP revision.

The Latent Variable Method

More sophisticated: uses both "causes" (tax burden, regulation, unemployment) and "indicators" (labour participation, working hours, GNP growth) in a structural model to infer the hidden economy. But:

Circularity Problem
To calibrate the model, you need benchmark estimates from two countries. Those benchmarks often come from… monetary methods. Sophisticated maths on uncertain foundations.
Identification Problem
Are weekly working hours a cause or a trace of underground activity? The choice is debatable and affects results significantly.
Instability
In one application, removing one small country (Finland) from the sample made almost all coefficients statistically insignificant — a structural model shouldn't behave that way.

The Canadian Reality Check

Macro-Model Estimates
Various currency/shadow-economy studies estimated Canada's shadow economy at 10–13.5% of GDP during 1990–93.
Statistics Canada Upper Bound
Detailed component-by-component analysis produced a maximum possible underground GDP escaping measurement of 2.7% — potentially 4–5× lower.

A Hierarchy of Evidence

Best: Direct empirical data from surveys and administrative sources
↓
Multiple independent sources cross-checked via supply-use tables
↓
Detailed model for a specific missing item (one industry, one product)
↓
Closely related physical or administrative proxy
↓
Worst: One broad macro-model inferring the entire hidden economy
The Chapter 12 sentence to remember: "Do not infer the whole missing economy from a macroeconomic proxy when you can estimate its specific missing components from empirical data." A residual is what your model failed to explain — not the phenomenon you wanted to measure.