Access
This article is part of Nature's premium content.
Published online 22 April 2008 | Nature | doi:10.1038/news.2008.757
Column: Muse
What's Madonna got to do with it?
The interests of vulnerable children in institutional care are not well served by the latest bout of dodgy statistical reasoning, says Philip Ball.
When a scientific paper comes with a separate commentary on its ethics, you know you’re in a controversial arena. That’s certainly the case for institutional childcare.
To read this story in full you will need to login or make a payment (see right).
Comments
Reader comments are usually moderated after posting. If you find something offensive or inappropriate, you can speed this process by clicking 'Report this comment' (or, if that doesn't work for you, email redesign@nature.com). For more controversial topics, we reserve the right to moderate before comments are published.
FROM THE EDITOR: WE ARE POSTING A RESPONSE HERE FROM KEVIN BROWNE AND SHIHNING CHOU, AUTHORS OF THE WORK DESCRIBED IN THIS COLUMN: The biased viewpoint of Philip Ball is evident in this article, as it only refers to one of the two major findings of our paper [Chou S and Browne KD, The relationship between institutional care and international adoption of children in Europe. Adoption & Fostering, 32(1): 40-48 (2008)]. It does not mention the strong association between incoming international adoptions and high numbers of young children in institutional care without their parents in receiving countries (such as Belgium, Spain, France, Finland). The call for domestic adoptions to alleviate the suffering of children at national level has been made recently [Colombani J-M et al. ‘Rapport sur l’adoption: Mission confiee par le president de la Republique et le Premier’, Paris: La documentation Francaise (2008)]. Yet, Philip Ball’s letter only pointed out that international adoption is associated with large numbers of children in institutional care from sending countries, challenging our assumption that, over time, it is reasonable to expect a reduction in the numbers in institutions if the sending countries were attempting to prevent children coming into institutions. We acknowledge that a correlation does not prove causation. However, longitudinal data from the Government of Romania shows a reduction in the number of children in institutional care only after a moratorium on international adoption was put in place in 2001. By stopping the ‘demand’, services were allowed to develop to prevent infant abandonment and promote domestic adoption, which could not previously compete with international adopters in terms of fees paid to adoption agencies. Philip Ball claims that Romania is a special case but poor practices have been observed in other European countries with specific examples quoted by the Parliamentary Assembly of the Council of Europe (PACE) in Bulgaria, Moldova and Ukraine. Both UNICEF [UNICEF, ‘Child Protection: A Handbook for Parliamentarians’, Geneva: Inter-Parliamentary Union (2004)] and PACE (2007) have identified international adoption as a ‘market’ based on supply and demand and concluded the need to reform international adoption practices based on the principle that ‘there is no such a thing as a right to a child’ [Parliamentary Assembly of the Council of Europe, ‘Disappearance of Newborn Babies for Illegal Adoption in Europe’. Strasburg: Council of Europe; Social, Health and Family Affairs Committee (2007)]. We are not opposed to international adoption when it is in the proven best interests of the child and upholds the rights of the child but current practices rarely achieve this. In many cases, parents select children for adoption rather than parents being selected to match the child’s needs. Furthermore, the UNCRC Article 21 says that international adoption is the last resort when all other domestic alternatives have been exhausted. Countries in economic transition use it as the first resort, even when it is the child’s right for the state to intervene and support their parents in difficulty (Article 19). We have advocated [Chou S, Browne KD and Kirkaldy M, International adoption on the internet. Adoption & Fostering. 31(2): 22-31 (2007)] that international adoption requires regulations and monitoring and international regulation to prevent money being paid directly to agencies or others for providing an adoption service. We suggest they should provide services on the basis of charitable donations by the general public. --Kevin Browne, Ph.D. & Shihning Chou, MSc WHO Collaborating Centre on Child Care and Protection, UK & School of Psychology, University of Liverpool
First, I want to point out where I agree with Chou and Browne. They are right to insist that “there is no right to a childâ€, and that adoption of any kind should always be done solely in the best interests of the child. Potential adopters deserve fair and considerate treatment, but, simply by applying to adopt, they don’t acquire a right to do so. I also agree that it would be good to work towards a situation in which no profit of any kind can be made from international adoption. Chou and Browne stress the fees to adoption agencies; I believe that a potentially bigger risk of corrupting influences arises when large payments are made in sending countries. But I welcome what they say in this regard. I am afraid that otherwise, my criticisms of their paper still stand. Chou and Browne offer no response to the technical problems I raised with their paper, but instead complain that I don’t discuss other aspects of the work, namely the correlation between institutional care and overseas adoption in receiving countries. It is not clear why my neglect of this aspect of the work constitutes ‘bias’; the two aspects of their conclusions are independent, in the sense that one can imagine either being the case without the other. And I don’t see why there should be any obligation to address every issue the paper raises; I focused simply on the claim that overseas adoption raises institutional numbers in sending countries. It seems Chou and Browne therefore have no response to my critique of this aspect of the work – they simply reiterate the paper’s assertions in this regard, without explaining why my criticisms of the logic by which they are derived are invalid. But since they raise the issue, let’s look at this other conclusion. Indeed, I’m glad to have the opportunity to do so, because it is based on even shakier evidence. Whereas my criticisms of the claims about sending countries were due largely to the untenable logic of the argument, the claims about the receiving countries are flawed also because of a fallacious handling of data. Browne claims that there is a correlation between the number (per capita) of children in institutional care in receiving countries and the proportion of total adoptions that are international. Even if this were so, we’re back to the issue that correlation says nothing about causation. But is it so in the first place? The crucial evidence is in Figure 3 of the paper, which shows the alleged correlation. Here Browne seems to ignore another fundamental aspect of statistical analysis that all students are taught: correlation coefficients do not tell the whole story, and one should always inspect the actual data. When one looks at Figure 3, it is immediately obvious that the best-fit line relies on a single outlier: the statistics for Belgium. Remove this, and the data for all the other nations tell a very different story: if there is a positive correlation, it is much, much steeper: almost vertical. In other words, the span from a near zero proportion of ICA to near-100 percent happens across only a very small change in the number of children in institutions, which means that this proportion has very little effect on institutionalization. This point is illustrated very nicely in the Wikipedia entry on correlations, which shows four very different data sets that have the same Pearson correlation coefficient. Figure 3 closely resembles the case on the lower right, in which all the data points bar one lie on a vertical line: that is, x is effectively independent of y except for this single outlier. Why is Belgium an outlier? It’s a good question, and one that Browne himself has looked at previously (Browne et al., BMJ 322, 485-487; 2006). Here he shows that the five European countries with the highest rates per capita of institutionalization of young children (under 3) are Bulgaria, Latvia, Belgium, Romania and Serbia/Montenegro. Browne rightly points out the detrimental effects of institutionalization, and criticizes these countries for its overuse. He calls for more expenditure on community health and social services to prevent children from having to be separated from their birth families. I fully agree with this. But Browne et al. here make no link between these high rates of institutionalization and ICA. And that’s not surprising, since Belgium is the only one of these five countries that receives children for adoption from other countries. In other words, very evidently high rates of institutionalization can have nothing to do with ICA, but are the result of social and cultural norms in those countries. And this is evidently the case in Chou and Browne’s paper too. Yes, Belgium has a very high proportion of young children in institutions (56 per 10,000 of population) and a high proportion of adoptions that are international (86.7 percent). But Ireland, Cyprus and Italy also have high proportions of ICA, yet among the lowest numbers in institutions. (Curiously, Ireland’s proportion of ICA has somehow plummeted from 92.9 percent in Table 1 to around 70 percent in Figure 3, while the number of children in institutions remains the same. Why is this? Why was the discrepancy not spotted in peer review?). And then consider Norway: 98.6 percent of adoptions are ICA, but there are essentially NO young children in institutions. Ah, but where is Norway in Chou and Browne’s Figure 3? How strange – it is not there at all! Neither is Iceland, with 92.9 percent of adoptions being ICA and again no young children in institutions. At the other extreme, we learn above that Latvia, Bulgaria and Romania also have high numbers of institutionalized children – but no ICA at all. And they’re not in Figure 3 either! Well of course not, Chou and Browne might say – they are sending countries, not receiving countries. But this doesn’t provide any justification from omitting them from the analysis. If we’re looking for what might cause high rates of institutionalization, and investigating the possibility that ICA might have a role, then by excluding countries that have the former but do not practice the latter, we are immediately biasing the figures. We are prohibiting any data points in the lower right of the plot, which would otherwise tend to act against a positive correlation. In fact, the authors have apparently banned data points from all four corners except the top right – not only do Norway and Iceland not appear, but neither does the UK or Slovenia where both quantities are very small. They are, in other words, choosing to include only those data that create a positive correlation. Let’s look for a moment at how the logic of the claimed correlation plays out for the UK too. If high rates of ICA tend to encourage high rates of institutionalization, then we might infer that the tiny rates of ICA in the UK have helped to keep down the numbers of children in institutions. But this isn’t so at all: the low rates of institutionalization in the UK are due to specific healthcare policies that do not permit it (fortunately), while the low rates of ICA are probably due to the fact that this practice has been discouraged by social services and has not become a social norm. In other words, the two figures have nothing to do with one another. But of course they don’t – you can see as much from the contrasting cases of Norway and Latvia. In other words, Chou and Browne’s conclusions here seem to be based on selective and misleading use of the statistics and the data that support them. I am astonished that this analysis was approved by referees. And I am not alone in this. I don’t claim to be an expert in ICA demographics (which makes it all the more concerning that the above flaws are so easy to spot). But similar issues to these, along with other serious criticisms of Chou and Browne’s paper, have been raised by a group of ICA experts in a response published in the same journal (Adoption and Fostering 32(2), 63-67; 2008). Unfortunately, neither these comments nor those will have any counterbalancing effect on the impact that Chou and Browne’s paper made in the media, thanks to a neat but irrelevant catchphrase. Chou and Browne point out that organizations such as UNICEF and PACE have criticized some of the ways ICA operates. This is true and important: there are problems that should not be ignored. But I say again that the statistics in their paper, and the dogmatic and generalized conclusions they draw from them (suggesting, absurdly, that Europe stands as proxy for the whole world), make no useful contribution to this debate. The PACE report that they mention actually advocates that, in order to reduce the large numbers of children currently living in institutions, the rules for international adoption should be EASED, in complete contrast to Chou and Browne’s implication that these large numbers are somehow due to intercountry adoption in the first place!