Past trends in cohabitation at ages 18 to 59

Unlike legal marital status, there are no annual official estimates of the cohabiting population of England & Wales. The best sources of data on trends in cohabitation are therefore from surveys such as the General Household Survey (GHS) and the Labour Force Survey (LFS). Please click here for graphs showing recent trends in the proportions cohabiting amongst never-married and divorced people aged 18 to 59 as measured by both surveys. At the time the projections were prepared, data were available to 2003 from the LFS and to 2002 from the GHS. There are gaps in the GHS series in 1997 and 1999 when the survey was not carried out.

For never-married people, the LFS/GHS data show a continuation of the long-term rising trend but with evidence of a slowing down in the most recent years. For divorced people, there is no clear evidence of changing levels. The LFS does suggest a slight increasing trend, but the GHS data indicates a more constant trend (or even slightly downward in the most recent data for males). The periodic official ONS estimates of the population cohabiting take account of evidence from all sources. The latest three sets are for 1992, 1996 and 2003 and they show a slightly downward trend during this period. For separated and widowed people, the numbers cohabiting are much smaller and it is very difficult to establish reliable trends from the available data.

Based on this analysis of recent trends, it has been assumed that proportions cohabiting will continue to increase for never-married people. But for other statuses, it has been assumed that proportions will remain constant at each age.

Cohabitation assumptions for the never-married population

The assumed future proportions cohabiting for the never-married population were calculated by extrapolating the trend in the LFS time-series for the proportion cohabiting at ages 18-59 forward to 2013 but with trends gradually assumed to diminish over that period. This was similar to the approach used to prepare the assumed marriage and divorce rates in the legal marital status projections.

An age profile was then fitted which was consistent with these assumed overall proportions cohabiting at 18-59 for the never-married. Fitting an appropriate age profile for future cohabitation has always been a difficult task in cohabitation projections. It is extremely difficult to obtain reliable trend data on cohabitation for five-year age groups (say) from survey data. Nevertheless, as cohabitation becomes longer established, it is reasonable to assume that there will be a gradual ‘rectangularisation’ of the age profile. And the last three sets of official ONS cohabitation estimates do indeed show clear evidence of this rectangularisation. Click here for a graph showing estimated proportions cohabiting in 1992, 1996 and 2003 by five year age group.

For these projections, an improved method has been used for fitting appropriate age distributions. The new method derives from a cohort analysis of the proportions cohabiting in the never-married population from the last three sets of ONS population estimates. This analysis shows strong evidence that from a given age, the proportion of never-married people cohabiting has remained almost constant as the particular cohort ages. In other words, beyond a certain age, the proportion of never-married people cohabiting aged x in year y is very similar to the proportion cohabiting aged x + t in year y + t. Click here for a graph illustrating this relationship.

The cohabitation age profile has therefore been fitted by assuming that proportions cohabiting will remain constant at very young ages, but at older ages will continue to follow these cohort trends. In the years to 2013, these proportions have been controlled so that they are consistent with the extrapolation of the overall 18-59 series. In later years, the cohort trends are assumed to continue, leading to a further rise in the overall proportion cohabiting at ages 18 to 59 and also to increasing numbers cohabiting at older ages. Click here for a graph showing the overall projected proportions of never-married people cohabiting at ages 18-59 resulting from applying these assumptions.

Cohabitation assumptions for the previously-married population

For divorced, separated and widowed people, it has been assumed that proportions cohabiting will remain constant at each age.

Variant assumptions

The high and low cohabitation variants allow for some uncertainty in the base population by using initial proportions cohabiting in each age/sex/legal marital status group which differ from the initial proportions used for the principal projection by ± 2.5 per cent. Thereafter, the variants assume that the proportions cohabiting in each group will diverge further from the principal projection, so that from 2018 onwards they differ by ± 20 per cent for never-married individuals and by ± 10 per cent for others. The lower margin of uncertainty assumed for the latter reflects the greater stability of the recent cohabitation time-series for previously married groups.