Official Statistics: Measuring Society

Every month, governments publish numbers that move markets, set pay, and shape elections: the unemployment rate, the inflation rate, the size of the economy. It is tempting to treat these as facts read straight off reality, like a thermometer reading. They are not. Each one is the output of careful, documented choices about what to count and how to count it.

That does not make them fake or untrustworthy. National statistics offices are among the most rigorous data producers in the world. But once you see that a headline number rests on a definition, you can read it properly: you know what it includes, what it leaves out, and why two honest analysts might report different figures for the same thing.

The census: trying to count everyone

A census attempts a complete count of a population, not a sample. Many countries run one every ten years. Even here, choices appear: who counts as a resident, how to reach people without fixed addresses, and how to adjust for those inevitably missed. The census then anchors thousands of other statistics, because it tells us the denominator, the total the rates are measured against.

Measuring unemployment: who counts?

The unemployment rate is not the share of people without a job. Under the international (ILO) standard used widely, a person is unemployed only if they are without work, currently available to work, and have actively looked for work in a recent period, typically the past four weeks. Someone who has given up searching, a discouraged worker, is counted as outside the labour force, not as unemployed. Change that definition and the rate changes, even if nobody's situation does.

Price indices and inflation

Inflation is measured with a price index, most familiarly the Consumer Price Index (CPI). Statisticians define a basket of goods and services meant to represent typical household spending, then track the total cost of that basket over time, weighting each item by how much people spend on it. The index is set to 100 in a base period; a later reading of 106 means prices rose about 6 percent. The basket is reviewed and updated as spending habits shift.

GDP: adding up an economy

Gross Domestic Product sums the value of goods and services produced in an economy over a period. What is inside the boundary is a choice: unpaid housework and much volunteer labour are excluded, while government services are valued in particular agreed ways. GDP is a powerful summary, but it was never designed to measure wellbeing.

Build a tiny CPI. Suppose households spend on just two things: bread (weight 70 percent) and petrol (weight 30 percent). In the base year both cost an index of 100. A year later bread rises 4 percent and petrol rises 20 percent. The overall index is 0.70 x 104 + 0.30 x 120 = 72.8 + 36 = 108.8, so measured inflation is 8.8 percent. Notice the weights did the heavy lifting: because bread dominates spending, its modest rise counts far more than petrol's steep one.

Imagine unemployment falls from 6 percent to 5 percent and a headline declares the job market booming. But suppose the drop happened because many discouraged people stopped searching and so left the labour force. They have no work, yet by definition they are no longer unemployed. The rate improved while the underlying situation may have worsened. The number is correct; read without its definition, it misleads.

National statistics agencies build their price indices from an explicit, published basket. The US Bureau of Labor Statistics and the UK Office for National Statistics both maintain a representative list of hundreds of goods and services, from groceries to rents to streaming subscriptions, and update it regularly as spending shifts, adding new items and retiring outdated ones. Prices are collected across many locations, then combined using expenditure weights drawn from household spending surveys. Because the basket and weights are choices, different indices such as the older UK RPI and the CPI can report different inflation figures for the same period, purely due to differing methods. Exact items and update schedules change over time, so consult the agency's current documentation for specifics.

• "The unemployment rate counts everyone without a job." No. It excludes people not actively seeking work.
• "Inflation is one true number." Different indices with different baskets and methods give different figures.
• "GDP measures how well people live." It measures market production, and deliberately omits much unpaid and non-market value.
• "A census is a perfect count." Even a full enumeration misses people and requires documented adjustments.

Given three items with spending weights and their price changes, compute the weighted price index, then change one weight and watch the headline inflation figure move without any real price changing.

Think Like a Maester: Before you trust a headline number, ask what it is defined to include, because the definition, not the world alone, decides what the number says.

1. Why is an official statistic better described as constructed than as simply observed?
2. Under the standard definition, what three conditions make a person count as unemployed?
3. In the worked example, why did bread's 4 percent rise affect the index more than petrol's 20 percent rise?
4. How can the unemployment rate fall while the labour market genuinely weakens?
5. Why can two agencies report different inflation figures for the same month without either being wrong?

The numbers a society uses to describe itself, the census, unemployment, inflation, and GDP, are constructed from deliberate definitions and methods, not read straight off reality. A price index weights a chosen basket of goods; an unemployment count depends on who is defined as actively seeking work; GDP draws a boundary around what production counts. None of this makes the figures untrustworthy, but it does mean the fine print is part of the fact. Read the definition, and you read the number correctly.

• Types of Data
• Statistics in the Age of Big Data