Commentary: Family Life in Lockdown

Leo Van Hove^*

• Vrije Universiteit Brussel, Brussel, Belgium

Biroli et al. (2021) analyze how COVID-19 lockdowns affected patterns of activity within households in Italy, the UK, and the US. They do so based on data from self-administered surveys among adults who indicated that they were cohabiting with their partner. Biroli et al. find that sharing of most duties increased, in particular for childcare.As the authors report on their results in terms of which gender does more, or less, of a certain task, the reader gets the impression that their samples consist solely of different-sex couples. However, upon analyzing the questionnaire and inquiring with the authors, it became clear that they only asked respondents about their own gender. In the analysis, the authors simply attributed the opposite gender to the partner. Biroli et al. thus risk having the gender of partners wrong, as their samples may contain same-sex couples. Given that 5.6% of US adults identified as lesbian, gay, bisexual or transgender (LGBT) in 2020 (Gallup, 2021a), at first sight the potential bias resulting from not filtering out same-sex couples would seem substantial. Below I try to come up with more precise estimates. But first I briefly discuss the set-up of Biroli et al.'s research. Biroli et al. ran their surveys in April 2020, during the first COVID-19 wave. For the surveys in the UK and the US, they used an online tool that stratifies samples across age, sex, and ethnicity. The Italian sample is less representative, as participants were recruited mainly through social media.Crucially, only one member of each couple filled out the questionnaire (Biroli et al., 2021, p. 4, footnote 10), and reported both about their own situation as well as that of their partner. The problem is that, as indicated in the Introduction, respondents were not asked about their partner's gender. This was confirmed to me by the corresponding author, who explained that when they did the analysis, they attributed the opposite gender to the respondent's partner. This said, on closer scrutiny, it would appear that the bias resulting from the inclusion of same-sex couples is, all in limited. To start with the US, as mentioned, according to Gallup data based on 37 more than 15,000 interviews, 5.6% of US adults (aged 18 and older) identified as LGBT in 2020 38 (Gallup, 2021a). This almost perfectly with the 5.5% estimate proffered by the Williams 39Institute for 2020-2021 (Flores and Conron, 2023, p. 1). 40As can be seen in The data also show that 23.7% of LGB individuals were either married or in a civil partnership. 55Adding the "other" category increases the share to 24.6%. Unfortunately, the survey does not contain 56 data on LGB(T) individuals who are cohabiting with a same-sex partner without being married or 57 being in a civil partnership. Hence, when it comes to the measurement error in Biroli et al.'s 58 assessment of the gender of the partners of their UK respondents, one can calculate only a lower 59 limit; namely 0.73%-0.92% (23.7% times 3.1%-24.6% times 2.7%). 60The Census 2021 for England and Wales (Office for National Statistics, 2023b) does specify how 61 many individuals are cohabiting with a same-sex partner without being married or being in a civil 62 partnership. This source (thus) yields a slightly higher estimate than the 0.73% of the Annual 63Population Survey 2020, namely 1.00% (0.33% married or in a civil partnership + 0.67% cohabiting). 64 (The censuses for Scotland and Northern Ireland do not distinguish between same-and opposite-sex 65 marriages; Scotland's Census, 2023; NISRA, 2023). 66 Finally, for Italy there would appear to be even fewer statistics (ISTAT, 2022(ISTAT, , 2023)). But an 67 international survey by Ipsos (2021, p. 9) suggests lower measurement errors: in Italy 6% of adults 68 identified as non-heterosexual in 2021, compared to 10% in the UK and 11% in the US. 69 Upon scrutiny, it would appear that the bias resulting from the non-exclusion of same-sex couples is, 71 all in all, contained. I show that Biroli et al.'s heteronormative approach risks getting the gender of 72 the respondent's partner wrong in 0.94% of their US households, in upwards of 0.73%-1.00% of their 73 UK couples, and in probably a lower proportion of their Italian sample. If one argues that Biroli et al. 74 should have left out of their analysis not only LGBT individuals who are living with a same-sex 75 partner but also those who are cohabiting with a partner of the opposite sex, then the share of US 76 couples that are potentially problematic amounts to 2.09%. 77Biases of this magnitude are unlikely to substantially alter Biroli et al.'s conclusions. Still,it is a 78 of the data that should be (made) aware of.