Persons-per-household is the average number of people who live in a single household. Statistical agencies calculate it by dividing the population living in households by the number of households. The figure exists because much of what governments, utilities and advertisers count arrives at the level of the home - a postal address, a broadband connection, a television set - while the decisions that matter are made by people. Persons-per-household is the bridge between the two.

In the United States the Census Bureau publishes the figure as both "persons per household" and "average household size". Its QuickFacts service puts it at 2.53 for 2020 to 2024, across 129,227,496 households, using American Community Survey (ACS) five-year estimates. In advertising, the statistic sits underneath television ratings, connected TV (CTV) reach estimates and household-level identity graphs, usually without being named.

How the statistic is built

Two definitions do most of the work. According to the Census Bureau, "a household includes all the people who occupy a housing unit", whether a family, a person living alone or unrelated people sharing a flat. People in group quarters - prisons, care homes, student halls, barracks - fall outside the household population and therefore outside the average.

The ACS supplies the American figure. According to the Census Bureau, about 3.5 million housing unit addresses are interviewed each year, and the five-year series pools that sample to produce estimates down to small towns. The Current Population Survey (CPS) produces a separate annual national series, which is why American figures quoted in the press can differ in the second decimal place.

Europe uses a similar ratio from a different survey. Eurostat's EU Statistics on Income and Living Conditions (EU-SILC) put the EU-27 average at 2.3 persons in every year from 2019 to 2024, according to Eurostat data reproduced by Insee. The spread inside the bloc is wide: 3.1 in Slovakia and 2.9 in Poland in 2024, against 1.9 in Finland and Lithuania. In the United Kingdom, the Office for National Statistics (ONS) reported 2.35 residents per household in 2024 across 28.6 million households, in a bulletin published on July 23, 2025.

From homes to people in television

Television ratings began as household measures. Early meters recorded whether a set was on and what it was tuned to, which says nothing about who was watching. A household rating of 5 meant 5% of television homes were tuned in. Converting that into people requires two inputs: how many people live in those homes, and how many of them were in front of the screen.

The first input comes from universe estimates. Nielsen builds them by combining Census Bureau data with information from its national TV panel, according to the company. For the 2020-21 season it estimated 121 million TV homes containing 307.9 million people aged two and over, an implied 2.54 persons per TV home - very close to the census average for the same period.

The second input is viewers per viewing household (VPVH), a trading figure that counts only people actually watching. PPC Land's co-viewing explainer describes the mechanism: multiply households by VPVH and a household rating becomes a person rating. VPVH is usually quoted per thousand viewing homes and broken out by demographic. A worked example shows the arithmetic. A programme with a 5.0 rating in a 121 million-home universe reaches 6.05 million households; if the adults 18 to 49 VPVH is 600, that equals 3.63 million viewers in that demographic.

Once people meters arrived nationally in the United States in 1987, each panel member pressed a button to log in, so persons ratings could be counted rather than modelled. Persons-per-household remained essential for weighting: the panel has to resemble the population in household size as well as age, sex and region, or larger families end up over or under represented.

Persons-per-household in CTV and digital

Streaming reintroduced the old problem at scale. A smart TV or streaming stick usually reports one device or account, not the number of people on the sofa.

Platforms answer with co-viewing factors rather than a raw census average, since persons-per-household describes who lives in a home, not who is watching at a given moment. Google announced co-viewing metrics for Google Ads and Display & Video 360 in April 2022, citing a Nielsen study in which multiple adults watched YouTube together on a TV 26% of the time. On June 2, 2026, Google applied a "Total Co-view" definition to seven reach metrics in the Google Ads API, using a model that varies by day, time, country and demographics instead of a single global multiplier. YouTube went further in September 2026 with a "views co-viewed" metric in which a household of three watching one video counts as three views.

The Media Rating Council (MRC) set the ground rules for such conversions in its Cross-Media Audience Measurement Standards (Phase I Video) of September 2019. Methods that convert household data or universe estimates into persons "shall be empirically supported", the standard says, and "household and persons level measurement shall not be mixed" in the same reported figure.

Programmatic standards handle a parallel case differently. OpenRTB 2.6, published by IAB Tech Lab in 2022, added a Qty object whose multiplier field lets a seller declare that one ad play counts as several billable impressions - the specification's example is a digital out-of-home screen deemed to deliver 14.2 impressions - with a vendor field naming the measurement source. No equivalent standard field exists for converting a CTV household impression into persons, so those adjustments happen in measurement and planning tools rather than in the bid request.

Households as an identity unit

The household is also the default unit for addressable TV and IP-based targeting. A home's IP address, set-top box or smart TV links naturally to an address, not to an individual. PPC Land's device graph explainer notes that The Trade Desk places streaming and CTV identifiers at household level, while mobile advertising IDs and hashed emails sit at individual level. Google changed its policies from February 16, 2025 to allow IP-based targeting of households in Display & Video 360, and extended IP use to EEA, UK and Swiss publishers from August 3, 2026.

Persons-per-household then turns those household counts into claims about people. In December 2025, the Texas attorney general's petition against Hisense over automatic content recognition estimated 1.27 million affected residents by multiplying about 470,000 households by 2.7 persons per household.

Limitations and disputes

The first limitation is that an average hides the distribution. One-person households made up 29% of American households in 2025, up from 20% in 1975, according to the Census Bureau. A campaign reaching single-person homes reaches far fewer people per household than the 2.53 average suggests.

The figure also moves. Average American household size fell from 5.79 in 1790 to 2.58 in 2010, according to Pew Research Center, then rose to 2.63 in 2018, the first increase in more than 160 years. Richard Fry, a senior researcher at Pew, attributed the reversal to household growth trailing population growth, pointing to multigenerational living, which rose from 12% of Americans in 1980 to 20% in 2016. Multiplier models calibrated on older data drift.

Household linkage itself is unreliable. A Truthset study for the Coalition for Innovative Media Measurement (CIMM) and Go Addressable found IP-to-postal address links accurate 13% of the time, which FreeWheel framed as missing 87% of households. PPC Land's identity graph explainer adds that providers agreed on the same IP-to-household link only 6.4% of the time. Multiplying an inaccurate household count by persons-per-household compounds the error.

Co-viewing multipliers are contested too. A 2025 TVision paper, cited in the co-viewing explainer, found the same household counted as 1.5 viewers by one provider and 1.2 by another. Platforms that model their own co-viewing also report the larger audiences they sell, a point raised in coverage of the VAB and TVision study, which put premium CTV co-viewing at 60% of impressions and YouTube at 45%.

Privacy is the final dispute. The device graph explainer records the Federal Trade Commission's warning that household linking can expose activity on a private device through ads on a shared screen.

Not the same as

Household income is an economic attribute of a home, used for targeting and segmentation. Persons-per-household is a count of occupants and says nothing about spending power.

Household ID is a pseudonymous identifier assigned to a cluster of devices or records believed to share a home. It labels a household; persons-per-household estimates how many people the label stands for. Identity resolution covers how those clusters are built.

Reach is the number or share of distinct people or homes exposed to a campaign. Persons-per-household can be an input to reach when converting household reach into person reach, but it is not a reach metric.

Co-viewing measures how many people watch a screen together at a given moment. Persons-per-household counts residents, including those who are asleep, out, or watching on another device.

Recent developments

Measurement is moving from assumed household composition towards direct observation. Nielsen piloted wrist-worn wearables with Super Bowl LX on February 8, 2026, and on August 31, 2026 deployed seven currency changes, among them wearable co-viewing capture, a household demographic assignment model and householding for one big data provider. Roku passed 100 million streaming households in April 2026, defined as distinct accounts streaming in a 30-day period, which cannot be converted into people from the disclosed data. The Video Advertising Bureau's June 2026 addressable TV guide put 92% of pay TV households as addressable-enabled, keeping the household as the unit that most TV targeting is built on.

Timeline

  • 1790 - The first US census records an average household size of 5.79 persons, according to Pew Research Center.
  • 1950 - Nielsen begins recording television tuning for the Nielsen Television Index, initially at household level.
  • 1980 - 12% of Americans live in multigenerational households, according to Pew.
  • August 31, 1987 - Nielsen's people meter replaces diaries and set meters for US national ratings, enabling direct persons ratings.
  • 2010 - US average household size reaches 2.58, according to Pew.
  • 2018 - US average household size rises to 2.63, the first increase in more than 160 years.
  • September 2019 - The MRC issues its Cross-Media Audience Measurement Standards (Phase I Video), requiring empirical support for household-to-person conversions.
  • 2020 - Nielsen estimates 121 million TV homes and 307.9 million people aged two and over for the 2020-21 season.
  • April 2022 - Google announces co-viewing metrics for Google Ads and Display & Video 360.
  • 2022 - IAB Tech Lab publishes OpenRTB 2.6, adding the Qty object and impression multiplier.
  • February 16, 2025 - Google's policy permitting IP-based household targeting in CTV takes effect.
  • July 23, 2025 - ONS reports 2.35 residents per UK household for 2024.
  • November 2025 - Truthset study for CIMM and Go Addressable finds 13% IP-to-postal accuracy.
  • December 2, 2025 - Census Bureau reports one-person households at 29% of US households.
  • December 15, 2025 - Texas sues Hisense, estimating affected residents from household counts multiplied by 2.7.
  • February 8, 2026 - Nielsen begins its wearable co-viewing pilot with Super Bowl LX.
  • June 2, 2026 - Google applies Total Co-view to seven Google Ads API reach metrics.
  • August 31, 2026 - Nielsen deploys seven currency changes, including wearable co-viewing and household demographic assignment.
  • September 2026 - YouTube introduces the views co-viewed metric in Advanced Analytics.

Summary

Who: National statistics agencies such as the US Census Bureau, Eurostat and the ONS publish the figure; measurement firms such as Nielsen use it in universe estimates and panel weighting; platforms including Google and YouTube publish their own co-viewing models; the MRC sets rules for household-to-person conversion.

What: The average number of people living in one household, calculated as the household population divided by the number of households, and used in advertising to convert home-level counts into estimates of people.

When: US census data reach back to 1790, when the average was 5.79. Television ratings began at household level in 1950, people meters enabled direct persons ratings from 1987, and CTV co-viewing models have spread since 2022.

Where: In census and survey statistics across the US, EU and UK, and in advertising across linear TV ratings, CTV reach reporting, addressable TV and IP-based household targeting.

Why: Screens, addresses and IP addresses are tied to homes, but advertising is bought and judged on people. Persons-per-household supplies the conversion, though falling and rising household sizes, one-person homes, weak IP linkage and contested co-viewing multipliers all limit how far an average can stand in for the people actually present.