The Indie’s ‘Lost Girls’ may not be what it appears

Okay, this is a bit of rush job but even in the absence of any data to work with, the Independent’s claim to have uncovered evidence in the 2011 census data that gender selective abortion is being ‘widely used’ by some UK ethnic groups requires n immediate response:

The lost girls: Illegal abortion widely used by some UK ethnic groups to avoid daughters ‘has reduced female population by between 1,500 and 4,700′

The illegal abortion of female foetuses solely to ensure that families have sons is widely practised within some ethnic communities in Britain and has resulted in significant shortfalls in the proportion of girls, according to an investigation by The Independent.

The practice of sex-selective abortion is now so commonplace that it has affected the natural 50:50 balance of boys to girls within some immigrant groups and has led to the “disappearance” of between 1,400 and 4,700 females from the national census records of England and Wales, we can reveal.

A government investigation last year found no evidence that women living in the UK, but born abroad, were preferentially aborting girls. However, our deeper statistical analysis of data from the 2011 National Census has shown widespread discrepancies in the sex ratio of children in some immigrant families, which can only be easily explained by women choosing to abort female foetuses in the hope of becoming quickly pregnant again with a boy. The findings will reignite the debate over whether pregnant women should be legally allowed to know the sex of their babies following ultrasound scans at 13 weeks.

The first thing to treat extremely sceptically is the Indie’s claim to have carried out a ‘deeper statistical analysis’ than that carried out by the Department of Health’s analytical team in producing it’s 2013 report on gender birth ratios in the UK [pdf].

For one thing, the DoH report is based on birth registration data spanning the period from 2007 to 2011, i.e. it looks at the birth ratios in children as they being born to women/families living in the UK.

The census data that the Independent commissioned from the Office for National Statistics covers “the numbers of families with dependent children who were registered in the March 2011 census, broken down by country of birth of both the mother and father” and, of course, comes from a a survey of households in which respondents are asked to provide information only those persons who are usually resident in that household.

As such the data set requested by the Indie will only provide data on children born to a particular family only if those children are classed as dependants and usually reside with their family, which means it will include students under the age of 20 in further, but not higher education, and schoolchildren who live away from home during term times but will not include any non-dependant adult offspring born to a family, irrespective of whether they still live in the same household or not, nor will it include or any children born to the family that may be living permanently with other family members, or any that have been taken in care or that have been adopted, or any children that may have died prior to the date on which the census was taken or, in the specific case of migrant families, any children who may be living, or indeed have died, overseas.

There are, therefore – and I am just riffing off the top of my head – are several reasons why some of the ‘second-born’ children in the census data set supplied by the Office for National Statistics might very well not be the second child born to the families included in Indie’s analysis  and each is potential source of confounding that might introduce an element of bias into the ‘second-born’ gender ratios that the newspaper has calculated, bias that is not in least bit attributable to gender selective abortion.

At least some of the Indie’s ‘lost girls’ may not be lost at all, they’re just not living with the family into which they were born at the time the census was taken and so don’t show up in the census and if some are genuine lost, as in not alive, then that loss may well be due to infant mortality and not abortion. One thing you do have to be very careful of when looking at family structures in first generation migrants is the question of when and in what circumstances a family arrived in the UK.

One of the three minority groups in which the Indie claims to have found a discrepancy in the gender ratio in ‘second born’ children are migrants from Afghanistan most of whom, one would expect, would have arrived in the UK as refugees from the Afghan conflict and it’s therefore entirely possible that some of those who arrived has adult may have children that were left behind when they escaped the country that they haven’t been able to bring to the UK, or children that they lost prior to their arrival in the UK. In 2001, at the beginning of the current Afghan conflict,  the under-5s mortality in the country was estimated by UNICEF at 134 per 1000 live births, which means that 1 child in every 76 born in the country would not live to see it’s fifth birthday and although things have improved somewhat over the last 12-13 year, even today 1 in every 100 children born in Afghanistan will die before it reaches the age of five.  Over the same period of time, the under-5s mortality rate in the UK has fallen from 7 per 1,000 live births to just 5 per 1000, which is 1 in 200; and if that seems a worrying high figure for wealthy, developed country like the UK it’s as well to remember that the majority of those deaths are in children under 1 year of age and a sizeable proportion are attributable to thing like severe prematurity, congenital abnormalities, complications arising during birth, etc.

I’ve kind of skipped on a bit there is noting that Afghans are one of the three minorities in which the Indie claims to have found anomalies in the ‘second-born’ gender ratio, the other two being the Bangladeshi and Pakistani communities, with the core of the Indie’s claim being set out in this section of the article:

We found that in two-child families of some first-generation immigrants, having elder daughters significantly increases the chances of the second child being male – an imbalance in the sex ratio that should not occur naturally.

To double-check our analysis, we asked professional statisticians to analyse the data in more detail. They confirmed that the effect is statistically significant and that there are only two plausible explanations, which are not mutually exclusive – either gender-based abortion or the practice of women continuing to have children until a son is born.

The latter phenomenon might explain most of the gender imbalances we observed in two-child families, said Christoforos Anagnostopoulos, a lecturer in statistics at Imperial College London. However, it could not explain some sex-ratio anomalies that persisted across families of all sizes, notably for mothers who were born in Pakistan, Bangladesh and Afghanistan.

So, according to the Indie and the statisticians they consulted there are only ‘two plausible explanations’ for the apparent discrepancies in the census-derived gender ratios in these communities, one of which is the widespread use of gender-selective abortion

Bullshit!

There is, in fact, a third possible explanation relating to a known problem which is likely to affect the data set used by the Indie, which covers all dependant children recorded as living in a household by the census, and which could quite easily create the appearance of gender imbalance favouring male children in the second recorded child in those households, one which would not affect the data set used by the Department of Health, which looks only at births. It is perfectly possible that the Indie’s ‘lost girls’ are, indeed, actually missing from the census data but not because they were aborted and were, therefore, never actually born. Rather it is possible, if not likely in fact, that these missing girls actually went missing after they were born, well after in fact.

To be absolutely clear, the birth data compiled by the Department of Health does not indicate a problem with any of the three communities highlighted by the Indie. The average male-female ratio for the UK in that data set is 105.2*, i.e. 105.2 male births for every 100 female, with an accepted natural range of between 103 and 108 and only one of the three has a male-female ratio outside that range; the Bangladeshi community where the current male-female ratio amongst first generation migrants is actually 102.6, which indicates a shortfall in male rather than female births, the exact opposite of what the Indie claims – while the ratios for first generation migrants from Afghanistan and Pakistan are 105.4 and 104.3.

* And you’ll notice that this is not the ‘natural 50:50 balance’ referred to the the Indie because there is no such thing. The natural balance in birth is a touch over 51:49 in favour of males.

What this tends to suggest is that if any girls have been lost from these communities, for whatever reason, these losses have not occurred in children born since 2007. It could be, therefore, that the Indie has uncovered evidence of a historical problem which has since abated but it is actually more likely that what the Indie has actually found in evidence of a completely different problem.

What if some of the two child families in the data obtained by the Indie are actually families that had three or more children:

The government should keep records of how many young people fail to return to school after the summer holidays, a charity which helps children escape forced marriage has said.

Aneeta Prem from Freedom Charity argued the information is needed to prosecute people under new laws on forced marriage being brought in next year.

However, the Department for Education says it has no plans for a register.

Last year, the government’s Forced Marriage Unit dealt with 1,485 cases.

Thirteen percent of the victims were younger than 15 years old, with the youngest aged just two.

The cases ranged across 60 different countries, with nearly two-thirds occurring in South Asia – mainly Pakistan.

BBC News – 4 October 2013

This is not a new problem of course. Similar concerns were raised, and reported, back in 2008 due an inquiry by the Home Affairs Select Committee, which was told that 30% of cases dealt with by the government’s Forced Marriage Unit involved minors (i.e. under 18s).

Given that we know that this particular problem exists – but not the scale of the problem because it is know to be subject to substantial levels under reporting and because the government still hasn’t put in place any mechanisms to track pupils who simply disappear from school registers – and that it disproportionately affect women (men account for 15% of the FMU’s annual caseload) and that it affects at least two of the three groups in which the Indie found an apparent discrepancy in the male-female ratio in families with two dependent children, according to the census, it is entirely possible that this may account for at least some, if not most of the 1,500 – 4,700 ‘lost girls’ in the data obtained by the Indie, and yet there is nothing in the Indie’s report to indicate that this was even considered to be a possibility, let alone that any attempt was made to control for it as yet another potential source of confounding.

Of course, what the Indie noticeably hasn’t done – as yet – is either publish its full analysis or make it’s data set available so that it can be examined by other researchers to see whether the newspaper’s claimed finding actually stack up or whether there are, indeed, too many potential sources of confounding for their estimates to be considered reliable. What I will say, however, is that based on limited information given thus far I am extremely sceptical of the Indie’s methodology.