Skip to content
As the demand increases for statistics and data to measure the impact of the COVID-19 pandemic, we have had to change our data gathering and release practices, focussing efforts on priority analysis and statistics.
Data Provider: Welsh Government National Statistics Annual Population Survey - Ability to speak Welsh by age, sex and year
[Collapse]Speak Welsh[Filtered]
Speak Welsh 1[Filtered]
Sex 1
Age 1
[Collapse]TotalClick here to sortTotal
Click here to sortFemaleClick here to sortMale
TotalAge 3-15136,700133,500270,200
Age 16-2468,80054,700123,500
Age 25-3447,50043,90091,400
Age 35-4449,50033,10082,600
Age 45-5453,50038,80092,400
Age 55-6443,50040,40083,900
Age 65+73,70054,500128,200



People aged 3 or more who say they can speak Welsh, by age and sex

Last update

12 January 2021 12 January 2021

Next update

March 2021

Publishing organisation

Welsh Government

Source 1

Annual Population Survey, Office for National Statistics

Contact email


National Statistics

Lowest level of geographical disaggregation

Local authorities

Geographical coverage


Languages covered

English and Welsh

Data licensing

You may use and re-use this data free of charge in any format or medium, under the terms of the Open Government License - see

General description

This dataset provides information for people aged 3 and over who say they can speak Welsh, by age and sex.

Data collection and calculation

These data are taken from the ANNUAL datasets from the Annual Population Survey (APS) for 2005 onwards and the Welsh Local Labour Force Survey (WLLFS) prior to that. These surveys are carried out by the Office for National Statistics (ONS). The data for Wales are based on an enhanced sample (around 350 per cent larger) compared to earlier years. APS data are collected throughout the year and are published for calendar years. WLLFS data was published for the year ended February, each year i.e. 2001 WLLFS data relates to year ended February 2002. The data do NOT exactly match annual averages derived from the 4 QUARTERLY datasets in each year due to differences in the sampling structure.
The local authority and Wales figures for 2001, 2002 and 2003 in these tables may not be the same as published elsewhere, as the numbers here are estimated using Welsh specific weights. These weights better reflect the population estimates for Welsh local authorities in these years.
Nomis is the ONS's official portal for labour market statistics. Note that some estimates from Nomis for the APS may differ slightly from those presented here due to differences in how local authority geographies are constructed.

Frequency of publication


Data reference periods

2004 to 2020

Rounding applied

Figures are rounded to the nearest 100 and so there may be some apparent slight discrepancies between the sum of constituent items and the totals as shown.

Revisions information

Since the end of March 2020, the APS has been conducted through telephone interviews instead of face-to-face interviews as a result of the COVID-19 pandemic. The ONS has been monitoring the impact this change has had on the survey and as a result have re-weighted the survey for January to June 2020 (i.e. quarters 1 and 2 of 2020). They found that a change in the survey mode resulted in a higher proportion of owner-occupiers participating in the survey and a lower proportion of renters responding to the survey than before the pandemic.

In March 2019, the Annual Population Survey data has been revised back to 2012, due to taking on board the latest population estimates.
When preparing the publication of this table for the APS results for September 2018, two coding errors were identified.
1. ‘All Persons’ previously included people of all ages – this has now been revised to only include all people aged three and over.
2. The percentage of people who say they can speak Welsh was previously calculated as a percentage of all persons - this has now been revised so that it is based only on those who responded to the question.
The data for all previous waves have been revised. Revised data is marked with an (r).

Statistical quality

As the data come from a survey, the results are sample-based estimates and therefore subject to differing degrees of sampling variability, i.e. the true value for any measure lies in a differing range about the estimated value. This range or sampling variability increases as the detail in the data increases, for example local authority data are subject to higher variability than regional data.


Welsh speakers language