In the past few weeks a new Infographic called The Stats on Prostitution has been circulating online. It was produced by the website http://www.onlineschools.org/ which carries similar “Infographics” on a diverse range of topics including infidelity, breasts, money and robots.
The Stats on Prostitution gives global figures for numbers of prostitutes, percentage of men buying sex, average prices, and a host of other related topics. If you scroll to the end of you will see the references, many of which are news sources rather than original studies.
Is it possible to accurately represent prostitution and related issues using such a short summary?
Take the first figure given in the graphic: “At this given moment there are 40 million prostitutes at work”. Getting a global figure for something like the number of prostitutes or sex workers is beset with difficulties. You obviously have to start by getting figures for each country. There are probably three main techniques of calculating such numbers at a national level: censuses (where all sex workers are counted in a given country); multiplier techniques (where figures are extrapolated based on specific studies or data from a limited number of locations); or expert opinion. Using any of these techniques rests on a number of assumptions. A good introduction to them can be found here. It is often the case that initiatives to estimate numbers of sex workers will use a combination of these techniques.
Whichever of these techniques is being applied, there are quite a few major problems. The first one is to do with definitions. Most of the terms and concepts being dealt with are fairly hard to define. What is a sex worker? What is sex, for that matter? What constitutes selling or receiving a reward for sex? Do your data sources cover male, female and transgender sex workers? How you define the basic concepts has a major impact on any estimates.
If you can sort these problems out – so that you are fairly sure that all of the data sources you are using in your country are using more or less the same definition – you have another set of problems. Although sex work tends to be thought of as a more or less “permanent” or full time occupation, this is not always the case. Many sex workers have other sources of income, or do sex work for short periods of time or on an irregular basis. Mobility and migration are very common in sex work, and potentially have a major impact on population size estimates. In many countries, the volume of sex work has a seasonal character, with numbers working expanding and decreasing over the course of a month or a year. How do you know how representative your data are of different types of sex work? Is there a risk that they over- or under- represent people doing sex work in different ways?
Because of the often highly stigmatised nature of sex work, it is fair to assume that enumeration exercises don’t always capture the most “hidden” or discreet groups. Censuses may over estimate brothel-based sex work. If data from a small number of locations are being used as a basis for “multiplier” calculations, how do we know how representative these locations are – the character of the sex industry tends to vary a lot from location to location, as it is influenced by a range of factors including local economies, demand, law enforcement practices, and culture.
All of these definitional problems increase the risk of systematic errors in population size estimates. Sampling error can also be a problem, especially if estimates are based on surveys with small samples. Although confidence intervals can help in assessing the possible scale of sampling error, they are no help at all in resolving the errors that stem from bias rather than random effects – so the fact that there are confidence intervals shouldn’t in itself be taken to mean the figures are accurate. What is more, it is hard to know which direction the biases are taking the estimates – so it isn’t even that easy to be confident about whether an estimate is a minimum or a maximum.
Once you have satisfactory national estimates then you need to compile them, again confronting definitional issues – do all the data from all of the countries meet the same definitional requirements? And do they meet the same quality requirements – some estimates may be based on fair lower levels of rigour than others. Different researchers are happy with different levels of rigour. In a recent report to the UN on its AIDS programming efforts, the UK did not report any data on sex workers. This isn’t because there are no data, nor because nothing has been achieved in this area – it is simply because the only data available that strictly met the definitions of this indicator were from one study, of one sample, in one town. The people filling in the report didn’t think it was appropriate to claim that these data are representative of all of the UK. If you look at other reports on the same page you will see that many other countries do give figures on sex work – and often the figures are based on just one study, of one sample, in one town. On the other hand, the most commonly quoted figure for the sex worker population in the UK – 80,000 sex workers – was based on a highly approximated multiplier calculation and the researcher who made the estimate has repeatedly expressed surprise that it is used as a basis for policy, saying “…I was never myself at all confident about it. I felt it could be higher, but it also could have been lower”. The reality is that many countries don’t really have the sort of data required to allow for a remotely accurate global estimate, as this exhaustive review of global data shows.
Invariably then, any figures like those presented in this graphic will be built on assumptions which are built on assumptions, which are in turn built on other assumptions. This is true for all of the “Stats on Prostitution”. Many of the findings that are presented as global can be sourced to one study in one town in one country. The US dollar figures set against the size of the sex industry or of sex trafficking can similarly be shown to be based on very imprecise estimates. Nearly all trafficking figures are (I talked about trafficking estimates for the South Africa World Cup in a recent post) – and while this doesn’t mean trafficking does not exist, it means that there is a real risk that policy becomes based on assumptions. If you are interested in the issues involved in evaluating the economics of sex work, take a look at this annotated bibliography on the subject.
What other problems might there be behind the “Stats on prostitution”? Well, the graphic provides data on 11 different statistics, with some country level illustrations for contrast and impact. Clearly the authors have made choices about which information to display and which not. There are no figures presented on family life, or on the impacts of stigma against sex workers. Although the stats on murder for sex workers in the US are compared with those for other occupations, we don’t get information on sex workers’ access to justice or protection when they are victims of violence. It states that sex work is “completely legal & regulated within 22 countries”, but does not discuss whether and how regulation or criminalisation regimes affect numbers of sex workers, and their vulnerability to violence and HIV – it simply talks about the cost to the justice system of people breaking laws on prostitution. The thing is, it is quite hard to generalise about these things because in many countries the impacts of laws often have less to do with rates of arrest and prosecution and more to do with the effects laws have on the behaviours of individuals who are trying to avoid falling foul of them – as I talk about briefly here and as this fascinating study on the effects of brothel closures in Bangladesh on levels of violence and risk experienced by sex workers shows. Given these nuances it is almost impossible to make meaningful generalisations about these issues.
The language used in the “Stats” also represents a choice. Many sex workers would reject the notion that clients “purchase a prostitute” – which is loaded with connotations of coercion or slavery – rather talking about “buying sex” or “buying a sexual service”. Similarly the graphic ironically talks about the “perks” of sex work in terms of risks of violence and HIV infection. Sex workers do often face higher risks, but they don’t always see these risks as being inherent to their work. Sex workers also often talk about the benefits of working flexibly and for incomes that are considerably higher than they could earn otherwise.
Why is it important?
Clearly, any presentation of figures such as those shown in the “Stats” should be surrounded with caveats. But what difference does it make? Very often, we find that inaccurate numbers as well as a reductionist understanding of the issues are used as a basis for legislative changes, as the UK’s recent change in legislation on prostitution and trafficking shows.
In my own field, HIV prevention in developing countries*, having reliable data – and avoiding misleading or biased data – is essential for the development and delivery of effective programming with sex workers. Here also, the needs are quite nuanced. For a local NGO or health facility, having a national estimate of the number of sex workers in the country is actually not that useful. Because such national figures are necessarily based on generalisations, they are unlikely to help frontline service providers in understanding the specific needs of the people they are trying to reach. Indeed, because national figures require an a priori definition of what sex work is, using these figures as a basis for local service provision can be misleading, since the service providers may know of groups or individuals who should be targeted even if they don’t fit the national definitions. Moreover, counting people as being sex workers often implies that they recognise or define themselves as sex workers – but as local service providers and outreach workers around the world know, it is not always useful or helpful to impose or assume that beneficiaries accept that identity. So, frontline service providers do need numbers – but numbers that are relevant to their own contexts and situations. Not only that but they also need to know what else is relevant to the people they are working with – the extent to which violence, poverty, and discrimination are factors putting them at higher risk of HIV, for instance, so that services can address these issues too.
What’s the point in national numbers then? In many countries, for quite some time, HIV and AIDS programmes have all but ignored some of the groups that are most at risk for HIV infection, including sex workers. Ironically, this neglect may in part be due to a fear of stigmatising sex workers by targeting them, a hangover from the early response to HIV and AIDS and indeed from sexual health programming pre-dating AIDS, that treated sex workers as “vectors” of disease. The neglect is almost certainly also a consequence of stigma against sex workers and denial of the relevance of sex work on the part of decision makers. Having some sort of aggregated national estimate can be extremely useful for epidemiologists and advocates who are advising national AIDS programmes on how they should allocate resources, and in identifying the potential need for policy and legislative changes that can effectively help to reduce the HIV risk (and other dangers) faced by sex workers.
Organisations like UNAIDS are developing tools to help national AIDS programmes model the relevance of sex work and other “risk” factors to the dynamics of HIV in their countries. But using these models requires denominators – estimates of the size and extent of risk of each population group. Many countries have so little good information about sex work that they end up using figures from neighbouring countries to populate their epidemiological models. Even if there is agreement on the overall numbers and relevance of sex workers to epidemic dynamics, it is also important to have data that shows what exactly determines risk and vulnerability, so that decision makers understand that it is not enough to assume sex workers will be reached by general population based programmes, and so that they allocate sufficient resources to address problems such as violence, lack of access to social welfare, and the impact of legislation against sex work on vulnerability and risk. So we have a conundrum: at national level, it is important to have summary, reductionist numbers, but it is also important to have detailed, qualitative data to ensure that policy and practice respond to real needs.
The “Stats on prostitution” is a global summary, and in HIV and AIDS programming, some sort of global picture can also be useful as it helps to pinpoint the major gaps in the global response – although as this graphic from UNAIDS shows, very few countries seem to report data at all. Indeed, UNAIDS has recently started to reemphasise not only the epidemiological reasons but also the human rights rationales for paying more attention to sex work in HIV and AIDS programmes. But it is essential that any “big pictures” be presented with all the relevant nuances, and with as much objectivity as possible. It is also essential that at local level practitioners have the skills and resources needed to understand in much more detail who they are working with.
*I recognise that there are reasons other than HIV to talk about sex work, but I’m focussing on what I know best for illustrative purposes. Also, the reality in many low-income countries is that HIV and AIDS is one of the only issues that brings sex work onto the agenda of decision-makers.