# Census Definition: Meaning, Types and Examples

A census is the complete measurement of every member of a defined population. Instead of studying a subset, you collect data from all units, so the resulting counts and averages describe the population exactly, with no sampling error. This article gives the census definition, explains the types, and shows when a census beats a sample and when it does not.

## Quick Answer

- A census collects data from every unit in a population, so the population size $N$ equals the number of units measured.
- A sample collects data from a subset, so the sample size $n$ is smaller than $N$.
- Because a census covers everyone, its mean and standard deviation are parameters, not estimates.
- Censuses are used for national population counts, business registers, agriculture, and any small population you can fully enumerate.
- Samples are used when the population is large, when measurement is destructive or costly, or when speed matters more than completeness.

## What Census Means

In everyday language, a census is an official count of people in a country, state, or district, usually with details about their characteristics [1]. The United States conducts a decennial census every 10 years, mandated by Article I, Section 2 of the Constitution, and the results determine how many seats each state holds in the House of Representatives and how federal funds are distributed [2].

The precise statistical definition is broader. A census is the procedure of systematically acquiring, recording, and calculating information about the members of a given population, which is then usually presented through statistics [3]. The United Nations describes the essential features of a population and housing census as individual enumeration, universality within a defined territory, simultaneity, and defined periodicity, and recommends conducting population censuses at least every ten years [3].

The key word is every. If you measure all 10 units in a small lab batch, that is a census. If you measure 4 of them, that is a sample. The same logic scales to millions of people.

## How It Works

A census does not use a special formula. It uses the population formulas directly, because you have data on every unit.

For a population of size $N$ with values $x_1, x_2, \dots, x_N$:

$$\mu = \frac{\sum_{i=1}^{N} x_i}{N}$$

$$\sigma^2 = \frac{\sum_{i=1}^{N}(x_i - \mu)^2}{N}$$

$$\sigma = \sqrt{\sigma^2}$$

Each symbol means:

- $N$ is the population size, the total number of units.
- $x_i$ is the value for unit $i$.
- $\mu$ (mu) is the population mean, a parameter.
- $\sigma^2$ (sigma squared) is the population variance, divided by $N$.
- $\sigma$ (sigma) is the population standard deviation.

When you take a sample instead, you estimate these quantities with $\bar{x}$ and $s$, and you divide the sample variance by $n - 1$ to correct for the fact that you did not observe everyone. That correction is why sample statistics carry uncertainty and census parameters do not.

## Worked Example

A lab measured a marker in mg/dL across 10 samples. The full set of 10 is the population, and the first 4 rows form a sample drawn from it.

| lab_id | value_mg_dL |
|---|---|
| 1 | 12 |
| 2 | 15 |
| 3 | 14 |
| 4 | 10 |
| 5 | 18 |
| 6 | 13 |
| 7 | 16 |
| 8 | 11 |
| 9 | 17 |
| 10 | 14 |

Population size $N = 10$. Sample size $n = 4$.

Population values: 12, 15, 14, 10, 18, 13, 16, 11, 17, 14.

Sample values (first 4): 12, 15, 14, 10.

Census mean:

$$\mu = \frac{12 + 15 + 14 + 10 + 18 + 13 + 16 + 11 + 17 + 14}{10} = 14.0000$$

Sample mean:

$$\bar{x} = \frac{12 + 15 + 14 + 10}{4} = 12.7500$$

Population variance and standard deviation:

$$\sigma^2 = \frac{\sum (x - 14.0000)^2}{10} = 6.0000$$

$$\sigma = \sqrt{6.0000} = 2.4495$$

Sample variance and standard deviation, using $n - 1$:

$$s^2 = \frac{\sum (x - 12.7500)^2}{4 - 1} = 4.9167$$

$$s = \sqrt{4.9167} = 2.2174$$

Standard error of the sample mean:

$$SE = \frac{s}{\sqrt{n}} = \frac{2.2174}{\sqrt{4}} = 1.1087$$

Difference between the sample mean and the census mean:

$$12.7500 - 14.0000 = -1.2500$$

Here is the same computation in Python.

```python
import statistics
pop = [12, 15, 14, 10, 18, 13, 16, 11, 17, 14]
sample = pop[:4]
mu = statistics.mean(pop)          # census mean
xbar = statistics.mean(sample)     # sample mean
sigma = statistics.pstdev(pop)     # population SD
s = statistics.stdev(sample)       # sample SD (n-1)
print(f"mu = {mu:.4f}")
print(f"xbar = {xbar:.4f}")
print(f"sigma = {sigma:.4f}")
print(f"s = {s:.4f}")
```

Output:

```text
mu = 14.0000
xbar = 12.7500
sigma = 2.4495
s = 2.2174
```

The sample mean of 12.75 sits 1.25 mg/dL below the true population mean of 14.00. That gap is sampling error, and it exists purely because 4 units were measured instead of 10.

## How to Interpret It

When you run a census, the numbers you compute are the truth for that population at that moment. The mean of 14.0000 is not an estimate. It is the exact average of the 10 values, and there is no confidence interval to report because there is nothing to infer.

When you run a sample, every statistic is an estimate with uncertainty attached. The standard error of 1.1087 tells you how much the sample mean would bounce around if you repeated the sampling. A census has no equivalent quantity, because repeating it would give the same answer.

This distinction shapes how you report results. Census results are stated as facts about the population. Sample results are stated with margins of error, confidence levels, or standard errors. If you report a sample mean without uncertainty, you are hiding the most important part of the analysis.

## When to Use It (and when not to)

Use a census when the population is small enough to measure completely, when you need exact counts for legal or administrative purposes, or when the cost of missing units is high. National population counts, business registers, and agricultural censuses all fall in this category [3]. The U.S. Economic Census, for example, is the official five-year measure of American business and the economy [2].

Use a sample when the population is large, when measurement is expensive or destructive, or when you need results quickly. A carefully chosen random sample can provide more accurate information than an attempt to conduct a full census, because you can invest more effort per unit and reduce measurement error [3]. If testing a product destroys it, a census would destroy the entire population.

A practical middle ground is to use census data as a baseline for designing samples. Census counts provide sampling frames such as address registers, and they supply the population sizes needed to weight survey results so they represent the whole population [3]. If you are learning the basics of drawing units from a larger group, the [sample mean guide](/blog/data-analysis/sample-mean) walks through the estimation side.

## Census vs Sample

| Feature | Census | Sample |
|---|---|---|
| Units measured | Every unit in the population | A subset of units |
| Size notation | $N$ | $n$ |
| Mean symbol | $\mu$ (parameter) | $\bar{x}$ (estimate) |
| Variance divisor | $N$ | $n - 1$ |
| Sampling error | None | Present, measured by standard error |
| Typical cost | High | Lower |
| Best for | Small or legally mandated populations | Large, costly, or destructive measurement |

The core trade-off is completeness against practicality. A census removes sampling error but does not remove measurement error, and it is often slower and more expensive. A sample accepts sampling error in exchange for speed, lower cost, and the ability to measure each unit more carefully.

## Common Mistakes

- **Calling any large dataset a census.** A census is defined by coverage, not size. A million-row sample is still a sample. Check whether every unit in the defined population was measured.
- **Using $n - 1$ on census data.** When you have the whole population, divide the variance by $N$. The $n - 1$ correction exists to fix bias in estimates, and there is nothing to estimate in a census.
- **Reporting a confidence interval for a census mean.** Census parameters have no sampling distribution. Report the value directly.
- **Ignoring nonresponse in an official census.** People who are not counted still belong to the population. Undercoverage biases census counts even though the method is nominally complete.
- **Assuming a census is always more accurate.** Measurement error affects every unit you measure. A well-run sample with careful measurement can beat a rushed census [3].
- **Forgetting the reference date.** A census describes a population at a point in time. Populations change, so an old census count may not describe the current population.

## Limitations

A census cannot eliminate measurement error, nonresponse, or coverage gaps. If some units are missed or some answers are wrong, the "complete" count is still wrong, and there is no standard error to warn you. Official censuses handle this with post-enumeration surveys and adjustment methods, but the underlying problem never disappears.

A census also cannot tell you about anything outside its defined population and reference period. It is a snapshot, not a forecast. For large populations, the cost and time required often make a census impractical, which is why most ongoing measurement uses samples and intercensal estimates instead [3]. If you want to understand how the population being measured is defined in the first place, see [what a population is in statistics](/blog/data-analysis/population-definition-statistics).

## Frequently Asked Questions

### What is the simple definition of a census?

A census is the collection of data from every member of a population. It produces exact counts and averages for that population, with no sampling error. National population counts are the most familiar example, but any complete enumeration qualifies [3].

### What is the difference between a census and a sample?

A census measures all $N$ units in a population. A sample measures $n$ units, where $n$ is smaller than $N$, and uses those units to estimate population values. Census results are parameters, while sample results are estimates with uncertainty.

### How often is a census taken?

The United Nations recommends that population censuses be conducted at least every ten years [3]. The United States holds its decennial census every 10 years, as required by the Constitution [2]. Other censuses, such as the U.S. Economic Census, run on different schedules [2].

### Is a census always better than a sample?

No. A census removes sampling error but keeps measurement error, and it is usually slower and more expensive. When measurement is destructive or the population is very large, a well-designed sample can give more accurate information than a full census attempt [3].

### What types of censuses exist?

Beyond population and housing censuses, common types include agricultural censuses, business censuses, and traffic censuses [3]. Each applies the same principle: measure every unit in a defined group. If you are working with measured quantities like the lab values above, the [quantitative data examples](/blog/data-analysis/quantitative-data-examples) article shows how these variables are classified.

## References

1. [census | Wex | US Law | LII / Legal Information Institute](https://www.law.cornell.edu/wex/census)
2. [Our Censuses](https://www.census.gov/programs-surveys/censuses.html)
3. [Census - Wikipedia](https://en.wikipedia.org/wiki/Census)

## Further Reading

- [Census Glossary](https://www.census.gov/glossary/)
- [Subject Definitions](https://www.census.gov/programs-surveys/cps/technical-documentation/subject-definitions.html)
- [Census 2000 Subject Definitions](https://www.stats.indiana.edu/c2k/dp_profiles/2kdefinitions.html)

## Related Articles

- [What Is a Population in Statistics? Definition and Examples](/blog/data-analysis/population-definition-statistics)
- [Quantitative Data Examples: Definition and Types](/blog/data-analysis/quantitative-data-examples)
- [What Is Descriptive Statistics? Definition and Examples](/blog/data-analysis/descriptive-statistics)
- [What Is Analytics? Definition, Types and Examples](/blog/data-analysis/what-is-analytics-definition)
- [Sample Mean: Definition, Formula and Examples](/blog/data-analysis/sample-mean)