Introduction
Picture the following scene: young men and women line up, waiting to enrol in a lottery in Lesotho. Like people all over the world, they see the chance to win a substantial amount of money which they can use to pay off debts, buy household essentials and possibly some small personal luxuries. However, unusually in this lottery, they are not queuing up to buy a ticket. Instead they are waiting to be tested for sexually transmitted infections (STIs), as getting a clean bill of sexual health is the only way they can take part. Indeed, this lottery is a new breed of cash-transfer scheme that aims to reduce HIV infection – one that is quite different to our usual image of cash-transfer schemes, and as such is emblematic of a new behavioural economics-inspired approach to HIV risk reduction in Africa.
UNAIDS calls recent biomedical advances in the fight against HIV ‘game-changers' (UNAIDS 2011). Innovations such as male medical circumcision, treatment-as-prevention and drugs to prevent mother-to-child transmission appear to slash HIV transmission rates.1 However, HIV incidence2 is still high in many southern and eastern African countries, and the human cost in terms of ruined lives continues to build. Consequently, interventions to change sexual behaviour remain important in the HIV policy discourse, and, among them, HIV-related cash transfers have become prominent. These transfers intend to make direct or indirect changes in sexual behaviour, thereby reducing HIV incidence.
Here, I scrutinise the conceptualisation of ‘choice’ and structure underpinning HIV-related cash-transfer schemes in Africa. The success of cash transfers to reduce HIV incidence has been agreed upon by a range of commentators, as we shall see below. Despite positive reviews, cash transfers have been criticised elsewhere for their questionable ethics and uncertain sustainability. This article confirms and extends previous criticisms, arguing that the analytical foundations of cash-transfer interventions are unconvincing when viewed from a political economy lens. This article is significant in two senses. First, its detailed analysis of the project outcomes suggests success has not been as universal as proponents (and indeed some critics) claim. Second, it highlights the role of new behavioural approaches to HIV risk reduction (some of which result in the lottery style of cash transfer described in the first paragraph), and thereby deepens previous criticisms about sustainability and ethics by linking concerns to a continuing debate about the causes of high HIV incidence.
Below I discuss the various foundations underlying HIV-related cash-transfer interventions, before going on to review projects in sub-Saharan Africa. Finally, I argue that a political economy approach to HIV risk provides a broader perspective to understand high rates of HIV acquisition. Here I will use the concept of political economy based on a radical, materialist interpretation of change in African countries. I am referring to an approach that understands the social relations of production, is historically specific and addresses the issues of class, conflict and power. In particular, I will attempt to show how all variants of HIV-related cash transfers fail to acknowledge the inequities (driven by local and global forces) that have led to high rates of HIV prevalence. As well as a more critical analysis of the kinds of evidence produced by the interventions themselves, this is the new contribution of this article – as previous criticisms of HIV cash transfers have not situated HIV risk in the context of the particular kinds of neoliberal globalisation found in African countries.
This article argues that HIV-related cash transfers obscure and depoliticise the underlying drivers of the HIV epidemics in African countries. In the worst examples, they may even harm those who acquire HIV, as the ethos underlying the new wave of HIV-related projects implies that HIV acquisition is the result of poor choices. This can only add to stigma and blame. However, this scepticism and caution is not widely shared in the policy world, where HIV-related cash-transfer approaches provide attractive and tractable. This article starts by setting out how HIV-related cash transfers became so popular.
The fashion for HIV-related cash transfers in Africa
Before reviewing cash transfers in sub-Saharan African countries, it is important to note that health-related cash transfers are used in other parts of the world, most notably in the now-rich countries to assist in the prevention of disease and in treatment adherence among groups that are seen as difficult to reach through usual methods.3 In African countries, cash transfers have been used extensively with non-HIV aims, such as for boosting school enrolment and poverty reduction. Programmes have been both universal (available to all) and targeted (available to a selected group), conditional (disbursed only if the recipient follows a particular action) and unconditional (disbursed to anyone in the target group).
Indeed, this review includes examples of projects and programmes whose main aim was not HIV related, but which made HIV-related claims as part of secondary goals or findings. Nine cash-transfer projects or programmes (with HIV-related main or secondary goals) are identified below. The rate of implementation of such programmes may well have been encouraged by the wave of positive coverage from practitioners and international development agencies. For example, on the eve of the International AIDS Society summit in Vienna in 2010, the World Bank released a press statement hailing the success of two HIV-related cash-transfer projects in Malawi and Tanzania. Dr David Wilson, the Bank's Director for its Global HIV/AIDS Program, announced, ‘These two studies show the potential for using cash payments to prevent people, especially women and girls, from engaging in unsafe sex while also ensuring that they stay in school and get the full benefit of an education’ (http://www.worldbank.org, press release, Vienna/Washington, July 18, 2010). The UNAIDS 2013 Global Report suggested that:
[n]ew strategies have emerged to reduce young people's vulnerability to HIV, including social cash transfers that create incentives for safer behaviours. … There is clear potential for cash transfers to support HIV prevention for young people, and continued research on the HIV prevention role of such programmes is recommended. (UNAIDS 2013, 18)
Policy endorsements have been mirrored by academic approval, including two systematic reviews (Pettifor et al. 2012; Heise et al. 2013). Of the 10 studies reviewed by Pettifor et al. (2012), nine showed positive results on sexual behaviour and one showed a reduction in HIV prevalence among young women. The authors concluded that, based on preliminary data, cash payment interventions to reduce HIV risk were effective, especially among young women (Ibid.). Heise et al. (2013) were more circumspect as we shall see below, but even they concluded that the emerging evidence suggested that cash transfers could be an important addition to policy.
Praise for a behavioural intervention, especially by key international agencies, is not necessarily at odds with the fact that biomedical innovation currently dominates public health discourse. Cash transfers are seen as complementary to the biomedical advances. Pettifor et al. (2012, 1733) and de Walque et al. (2012a, 6) argue transfers can help with uptake of biomedical interventions, such as antiretrovirals and drugs for prevention of mother-to-child transmission. Pettifor et al. (2012, 1729) also warn that limits to the scale-up and uptake of biomedical interventions mean that behavioural-focused interventions remain important.
Despite wide admiration for cash transfers, critical views do exist. The systematic reviews mentioned above are concerned that cash transfers may not be sustainable (Pettifor et al. 2012; Heise et al. 2013), while Heise et al.’s (2013) more critical review doubts whether cash transfers can change the structural factors causing high HIV risk. Other evidence suggests that the impacts of cash-transfer projects are short-lived. Rigsby et al. (2000) studied the impact of a four-week cash-transfer project on anti retroviral drug adherence among a sample of US veterans. While the use of monetary reinforcement clearly helped treatment adherence during the project, eight weeks after its end adherence in the cash-transfer group had returned to near-baseline levels.
Some commentators raise a wider set of concerns. Harman, for example, proposes that those excluded from the project feel disincentivised to change their behaviour: ‘The question of who is left out limits sustained behaviour change … local communities have increasingly been found not to engage in activities that help themselves and their communities unless they receive financial support to do so’ (2011, 876). To provide support for this concern, Deci, Koestner, and Ryan's (1999) meta-review of 128 studies suggests external motivation, such as cash transfers, tends to undermine ‘intrinsic’ or internalised motivation. Similar concerns emerged in a review of cash interventions in tuberculosis treatment adherence, which were compounded by the observation that transfers created sharp power asymmetries between monitoring staff and clients (Eichler, Levine, and the Performance-Based Incentives Working Group 2009). Harman develops the issue of power beyond that of the project itself, to argue that the use of HIV-related cash-transfer projects, with their frequent requirement of abstinence and/or safe sex, has ‘significant implications for individual choice, the family and forms of social engineering that restrict childbirth within specific socio-economic demographics’ (2011, 876). These criticisms will be confirmed but also extended by reference to a wider political economy approach in the remainder of the article.
All cash transfers are not created equal
HIV-related cash-transfer schemes have emerged from several theoretical perspectives: social protection, social work and, more recently, behavioural economics (for a review see Heise et al. 2013). Social protection cash transfers can be conditional or unconditional. Where conditional, they are often premised on school attendance. Examples below have been implemented in Kenya and Malawi. Often the link to HIV comes from an attempt to change the structural factors that lead to high HIV risk through education, poverty reduction and women's empowerment. Elsewhere in this special issue, O'Laughlin considers how the term ‘structural' is used. Handa et al. (2014, e85473) set out these channels, arguing that ‘cash transfers may reduce the risk of HIV by addressing structural risk factors, especially poverty, vulnerability and low human capital.' School attendance might reduce HIV risk directly (as it is argued that sex is less likely to occur in school), and indirectly, as young people will receive HIV education and are more likely to have partners of the same age (rather than older partners who may have already acquired HIV). At the same time, Handa et al. suggest that HIV risk is reduced through a poverty channel, as cash transfers may postpone sexual debut by reducing poverty and making young people less likely to engage in transactional sex, with young women less dependent on male partners, reducing sex generally and unprotected sex acts in particular.
A social work approach (borrowing from the study of anti-social behaviour) is also mobilised in interventions that require school attendance through conditional grants. For example in the Zimbabwe project listed below, school attendance is seen as changing sexual behaviour as it encourages bonding with ‘prosocial’ adults, peers and institutions, especially important for orphans and other vulnerable children (OVCs) (Hallfors et al. 2011, 1083). Hallfors et al. argue that supporting OVC girls to stay in school will lead to ‘greater attendance, … higher educational aspirations, more positive expectations about the future, more equitable gender attitudes, more protective attitudes about sex, lower self-reported sexual behaviour, and lower rates of marriage and pregnancy’ (Ibid.).
In contrast, behavioural economics sees people as economic agents, some of whom are myopic, heavily discounting the future benefits of safe sex and focused on immediate reward (the pleasure of unsafe sex). Borrowing from psychologists who use cash transfers to change behaviours such as smoking and substance abuse, these interventions operate through tightly designed frameworks of conditionality and reward. Specifically, conditional cash transfers are seen as changing the balance between short- and long-run benefits so that individuals are more likely to choose safe sex (de Walque et al. 2012b). This specific behavioural economics approach is conceived within a libertarian paternalist view of changing an individual's ‘choice architecture’ (Thaler and Sunstein 2008). While individuals are seen as optimisers, some apparently make decisions using poor judgement – and this ‘sub-optimal’ behaviour is at the heart of HIV spread.
The assumption that some individuals are at risk of HIV acquisition because they cannot account for future risk adequately is the basis for a number of interventions below, but most notably the lottery-style intervention in Lesotho. This intervention, by not guaranteeing a cash payout (but only the chance of one), is quite different from other kinds of cash transfer where participants might expect a certain cash payment should they fulfil any necessary conditions. The designers of the Lesotho lottery, Björkman-Nyqvist et al. (2013), argue that certain people engage in risky sexual behaviour despite a high background level of HIV as they weigh short-term gains more highly than long-term benefits. Similarly, referring to young people involved in the Tanzania RESPECT project, de Walque et al. (2012a, 6) conclude that ‘[t]hese young people appear to understand their HIV risks and know how to behave to prevent transmission – yet they don't choose to act on that knowledge.’
These different theoretical approaches lead to varying kinds of conditionality. A poverty-focused intervention based on the social protection approach may have no conditionality and might target poor households generally, while an education or gender equality project arising from the same social protection approach might be conditional on school attendance and/or may be focused only on households with school-age girls. Projects developed with a social work approach would certainly have conditionality based on school attendance, with payments directly to the school attendee. However, interventions informed by behavioural economics are premised on a far more tailored conditionality in order to incentivise behaviour change. While other cash-transfer designs might be conditional on engaging in an easily monitored behaviour, such as school attendance, the behavioural economics approach requires participants to refrain from complex, private behaviour, such as risky sex. As such, self-reported monitoring is eschewed in favour of ‘objective’ measures (such as a sexually transmitted infection test), and monitoring and reward transfer are frequent in order maximise incentives for those with a high discount rate (Medlin and de Walque 2008, 3, 5; de Walque et al. 2012a, 9, 14). As such, behavioural economics interventions will have much smaller, more frequent cash transfers than poverty- or education-focused interventions. Indeed, some behavioural economics projects simply offer the chance of a cash reward, as in the Lesotho lottery, and so are attractive to funders because of their cheapness.
In practice (as we will see below), while some projects recorded HIV prevalence to chart impacts, only a few used it to screen recipients (i.e. requiring a negative HIV result to be a recipient). For those from a social protection or social work perspective, this eschewing of HIV status as a conditionality undoubtedly has its roots in ethical concerns about stigmatising and impoverishing individuals who have received a positive test (Heise et al. 2013). However, as Medlin and de Walque (2008, 12–13) illustrate, for behavioural economics interventions, there is an additional set of reasons why HIV is not chosen as an indicator. Background HIV rates are low in many settings (meaning that risky sexual behaviour may not always lead to HIV acquisition) and, given the incubation period of HIV, testing positive may be too distant from unsafe sex. Medlin and de Walque (12) argue that it is better to use the absence of other sexually transmitted infections as a condition, if they are prevalent within the population, linked to risky sex and can be tested cheaply and conveniently.
The new behavioural economics approaches are effectively embellishments of a standard micro-economic model of consumer choice, based around the analysis of the choices of a rational and representative consumer who attempts to maximise his/her utility. The decisions of the representative individual are then aggregated up to explain outcomes at the level of the society (Fine and Milonakis 2009, 3). This is reductionist in several ways. The first is that the issue of a ‘representative individual’ is problematic, as complex and shifting gender- and class-based social norms around sex and relationships exist for different groups within any one country (Johnston 2013).
More generally, these approaches are set up with only limited attention to the broader social context, and more specifically the way that social norms over sexual behaviour are established. For example, a number of authors have pointed to the predominance of male migration in explaining concurrency in some contexts (for example, Marks 2007; Katz 2009), while others have pointed to the delays resulting from inadequate transport infrastructure and border red tape to explain commercial sexual transactions (for example, Stillwaggon 2006). By isolating behaviour from the social norms that affect it, all three of the cash-transfer approaches (behavioural economics, social protection and social work) are treating these norms as exogenous, when in practice they are dynamic and reflect wider social and economic changes (of which more below).
In this way, the approaches above lose any sense of context or historical specificity, and then go on to make very particular assumptions about the way in which decisions are made. The first assumption is that individuals make independent decisions over unsafe sex and HIV risk. Is it really possible to isolate decisions over sex and condom use from other factors, such as pregnancy and trust? To understand this, we might instead want to know more about the role and value of condom use and sex in different kinds of relationship. Further, authors like Marais (2005) and Marks (2007) remind us that both the construction of health information and its acceptance is not a simple matter. Ideology in the form of religious or moral standpoints has an important to role to play. Those who think that education is the sole determinant of take-up of orthodox medical information about HIV transmission are, therefore, fundamentally misguided.
We are also left wondering about the concept of ‘choice’ in these approaches. Individuals are seen as having a high degree of choice about engaging in or refusing certain kinds of sexual behaviour. This ignores the significant role of coercion in sexual activity by strangers or partners, despite the evidence from a range of studies of its prevalence (Watts and Zimmerman 2002, 1235; WHO 2010, 12; Garcia-Moreno and Watts 2011). Even where coercion is not present, approaches based on free choice are far from applicable. Shula Marks (2007, 867) suggests that many young South African women are ‘choice-disabled’, being dependent to a large extent on men for economic support and status, thus engaging in risky sexual behaviour to secure their attentions and fatalistically accepting the concomitant risk of HIV and sexual violence as a necessary part of survival. In recognition of the inter-linkages of sex, money and social standing, we may then want to think not about sexual choices, with all of the concomitant understandings of ‘freedom’, but rather about options available to particular types of people. These options may be extremely narrow for some, especially for poor women in certain areas who may not always face violent compulsion to engage in sexual relationships, but whose degree of ‘choice’ is severely limited.
On these grounds we might be sceptical about the likely success of cash-transfer policies, given their construction around concepts of choice and incentives that are divorced from an understanding of the options individuals have. How do cash transfers work in practice?
The practice of HIV-related cash transfers in Africa
Table 1 distils the key characteristics of HIV-related, Africa-based, cash-transfer schemes, evaluated in terms of the evidence that they had reduced HIV risk in both the short and long term. These cases, located through web searches, are listed where either academic articles, project documentation or website information was publicly available. Subsequently, the impact of each project on HIV risk is set out. Of course, the discussion is limited by the availability of information and evidence. All of the evidence presented in the project documentation comes from randomised control trials (RCTs) – i.e. where large-scale quantitative exercises are carried out to investigate the impact on an intervention by comparing the outcomes for participants with those for a ‘control’ group. Arguably, these exercises wrongly privilege certain kinds of quantitative information, have a narrow focus and are poor at understanding causal links (Deaton 2010; Cartwright and Hardie 2012; Basu 2013), and these concerns are relevant in understanding the impact of cash transfers, as we shall see below.
| Place | Name | Target group | Aim | Conditionality | Design | Source |
|---|---|---|---|---|---|---|
| Kenya – national | Cash Transfer Programme for Orphans and Vulnerable Children (OVCs) | OVCs are 0–17 years old in poor households, with at least one deceased or chronically ill parent/carer | Improving the welfare of OVCs | None | US$20 per month to ‘the main care giver’ in each participating household. | Handa et al. (2014) |
| Kenya – two western districts | School Support Through Uniforms and HIV Education | Boys and girls enrolled in Grade 6 | Changing sexual behaviour, fertility, and reducing the risk of HIV and herpes simplex virus | School attendance | Three intervention arms which provided school uniforms, HIV education, and both uniforms and HIV education. | Duflo, Dupas, and Kremer (2012) |
| Malawi – Zomba area | Schooling, Income and HIV Risk (SIHR) | Young women, 13–22 years old, never married | Improve educational attainment, reduce HIV/STD risk, prevention of unwanted pregnancy and early marriage | Mixture of conditional and unconditional transfer, dependent on the intervention arm | Carers received an amount randomly chosen for each enumeration area (ranging from $4 to $10 per month); while the amount given to girls was determined by lottery (ranging from $1 to $5 per month). Those in secondary school had their fees paid. | Baird et al. (2012) |
| Malawi – rural areas | Malawi Incentive Programme | People, 14 years and above | Assisting individuals and couples to maintain their HIV status | Maintenance of baseline HIV status for one year. For couples, both members must maintain their HIV status | Each individual or couple randomly drew an amount to be paid 1 year later. Individuals received between zero and approx. $16; couples, between zero and $32. | Kohler and Thornton (2012) |
| Tanzania – rural southwest | The RESPECT study | People aged 18–30 | Reducing HIV acquisition | Testing negative for four curable STIs | Two intervention groups, paid either approx. $10 or $20 per four-month testing round. | de Walque et al. (2012b) |
| Lesotho – rural and peri-urban villages across 5 districts | Evaluating the impact of short-term financial incentives on HIV and STI incidence among youth in Lesotho | Peopled aged 18–30 years | Changing sexual behaviour and HIV incidence | Testing negative for two easily treated STIs | Two intervention groups, with individuals eligible to participate in a lottery, with prizes the equivalent of US$74 or US$147 per four-month testing round. In each round, four lottery winners were drawn (2 men and 2 women). | Björkman-Nyqvist et al. (2013) |
| Zimbabwe – five rural areas of Manicaland Province | Supporting Adolescent Orphan Girls to Stay in School as HIV Risk Prevention | All Grade 6 girls (ages 12–14) with one or more deceased parent in a selected set of primary schools | Reducing HIV risk | School attendance | School support included: fees, exercise books, uniforms and other school supplies. Girls who were required to board to attend secondary school had assistance with boarding costs. Female teachers were trained as helpers to monitor participants’ school attendance and to assist with absenteeism problems, for which a small fund was also available. Participant schools benefited from a feeding programme. | Hallfors et al. (2011) |
| South Africa – northwest area of Agincourt | Swa Koteka, HPTN 068 | Young women aged 13–20 in Grades 8–11 and their parents | Reducing HIV incidence | School attendance | Each month that a girl attends 80% of school days, the family and the young woman will receive a cash-transfer payment (approx. $18 to female carer and $9 to girl). There is also a community education aspect aimed at changing gender norms. | No published results as yet but some information can be found online, e.g. http://www.cpc.unc.edu/. |
| South Africa – Vulindlela in KwaZulu-Natal | Reducing HIV in Adolescents (RHIVA), CAPRISA 007 | Boys and girls aged at least 13 years in Grades 9 and 10 in selected schools | Reducing HIV incidence | School attendance and other educational milestones, such as passing exams | Cash incentives paid to learners – amount unclear at time of writing. | No published results as yet. Some information can be found online, e.g. http://www.caprisa.org/. |
Source: Author's extracts from sources listed in final column.
Before moving to a more detailed assessment of these projects, it is important to initially evaluate their impact on HIV acquisition. And here it is clear that their success has been far more limited than the policy literature would suggest.4 For those projects measuring HIV status, only a subset reported a reduction in HIV incidence over time or in HIV prevalence among the intervention group compared with the control group. In the Lesotho lottery, for which only medium-term information is available, there was a reduction in HIV incidence rates for the intervention compared with the control. Two years after the project start, HIV incidence was significantly lower among the study participants, especially for women and for the intervention arm facing the highest lottery payments (Björkman-Nyqvist et al. 2013). The Malawi Schooling, Income and HIV Risk (SIHR) project found lower HIV and herpes simplex virus (HSV) rates only for some arms of the intervention, i.e. not for girls who started out as school dropouts and then went on to receive the transfer (Baird et al. 2012).
In two other projects for which HIV data was available, the Malawi Incentive Programme and the Tanzania RESPECT project, there was no change in HIV rates. Both studies raise particular challenges for researchers. In the Malawi Incentive Programme, the authors are shocked to find that male recipients end up with a greater HIV risk. While Kohler and Thornton (2012) report no overall impact of the conditional cash transfer on remaining HIV negative, this was averaged over a reduced risk for women and a higher risk for men. Male participants were more likely to report having sex than non-participants, and, despite being slightly more likely to report using a condom, this increased their overall HIV risk. In contrast, female recipients were less likely to report having had sex, leading the authors to conclude that conditional cash transfers can be protective for women (Ibid., 169). In the Tanzania RESPECT project, while the risk of the targeted STIs was lower for the treatment group with highest payment, there was no impact on three other, far more serious STIs: HSV, syphilis and crucially HIV (de Walque et al. 2012b).
Some interventions did not measure HIV, instead relying on various indicators of HIV risk (such as sexual debut, pregnancy, early marriage etc.). However, these indicators did not move neatly together. The Kenya OVC intervention changed the overall age of sexual debut but was not able to change any other features of self-reported sexual behaviour. Young recipients (aged 15–25) had a 23% lower rate of sexual debut than others (Handa et al. 2014). However, it is unclear if there was any additional risk reduction from delayed sexual debut, and certainly the transfer had no other impacts on HIV risk, with indicators such as condom use, number of sexual partners, transactional sex etc. being unaffected. The western Kenya project had varied results (depending on which intervention arms participants were in). The uniform subsidy only reduced teen marriage, pregnancy and not HSV, while HIV education only reduced out of wedlock teen pregnancy in first three years. When the two were implemented together, fertility fell less than using the education subsidy alone but HSV was lower (Duflo, Dupas, and Kremer 2012). Certainly for the western Kenya project, it could not be argued that the results for the cash intervention were in all ways superior to those for HIV education.
Worryingly, longer-run impact studies exist for both the western Kenya and the Tanzania RESPECT projects. In western Kenya, the impacts of a uniform subsidy on marriage and school dropout were eroded as participants aged, and seven years later these had disappeared. In the Tanzanian RESPECT project, after two years the impact on STI acquisition by women vanished (Heise et al. 2013). For other projects, either there was no long-run impact data or it was not available at the time of writing.
In evaluating HIV-related cash transfers, it is important to note the limited usefulness of evidence coming from the interventions themselves. RCTs have generally been criticised for their inability to understand causal pathways, owing to their reliance on quantified end-results. This general criticism seems specifically relevant here, where few projects were able to discuss the evidence on causal mechanisms (and where those that did were open to challenge). For example, in the Malawi SIHR project, the project team hypothesised that reduced HIV prevalence for some participants relative to the control group came about as participants had younger boyfriends than non-participants (Baird et al. [2012] report that the partners of girls in the treatment group were one year younger on average than those in the control). However, some authors have doubted that lower HIV rates were due solely to the choice of older partners. Heise et al. (2013) note that only 2.5% of girls in the control group versus 0.5% of girls in the intervention groups had older sexual partners, and so the small reduction in age of partners could not explain the difference in HIV prevalence. Moreover, I would argue that the fact that girls in the conditional arm had the same teen pregnancy and marriage rates as the control group suggests that the mechanisms for impact are unclear.
With this caveat about the kinds of evidence available from RCTs, I will now turn to an investigation of the support for each of the theoretical positions that were identified as supporting cash transfers for HIV risk reduction. In the social work approach, anti-social peer-group orientation and low self-esteem lead to poor life choices and raise HIV risk. In the social protection approach, HIV risk is raised by the choices forced on poor people, or specifically on poor women. In behavioural economics, HIV risk is the result of poor discounting of time and risk, leading people to choose immediate pleasure over safe sex or abstinence.
The social work approach hypothesises that school attendance has a risk-reduction effect. Certainly, many of the projects listed above were effective in increasing school attendance for some participants, but the impact on HIV risk has been far more complex. In the Kenya OVC programme, secondary school-age participants were 8 percentage points more likely to be in school than the control and Handa et al. (2014) argue that this was the cause of the delay in sexual debut that was also found. However, as we have noted, while sexual debut was delayed, the indicators of risk for those participants who were sexually active (number of sexual partners, condom use and transactional sex) were unaffected. In the western Kenya project, those in the group receiving uniform subsidies were less likely to have dropped out of school and also less likely to have been married or borne children two years later (Duflo, Dupas, and Kremer 2012). However, there was no impact on HSV incidence and the authors suggest that the impact may have occurred through lower rates of committed partnership, while casual partnership rates remained similar. In the Zimbabwe intervention, Hallfors et al. (2011) report that two years after the start, participants were less likely to have dropped out of school and to have initiated sexual debut. The study did not collect information on HIV or other STIs, and noted that the data they collected on self-reported sexual intercourse was unreliable (Ibid.). These three interventions, then, seem to have reduced some kinds of sexual activity but cannot clearly be seen to have reduced HIV risk on all counts.
Importantly, all three made relatively large transfers to the young people they targeted, especially the Zimbabwe and Kenyan OVC projects. In projects with smaller interventions, the tensions in the relationship between school attendance and HIV risk become clearer. In the Malawi SIHR, with its lower cash transfer, setting a condition of school attendance may have had negative effects on teen marriage and pregnancy. Baird et al. (2012) argue that the cost of schooling may have led families to push girls into early marriage. So the interaction between the size of the cash transfer and the implicit cost of the conditionality is important (e.g. fees, uniform, books and transport costs are all incurred when children have to attend school).
Consequently, the social work approach (with its focus on the pro-social effect of school on HIV risk) seems to be making too simplistic an assumption about the role of schooling, without looking at its costs and the social context. This is supported by a wider set of evidence on the relationship between education and HIV. While in most African countries, studies suggest that those with higher education usually have higher HIV prevalence, there is growing evidence that this pattern is changing (Hargreaves et al. 2008). In several settings, Hargreaves et al. (2008) conclude that strong statistical associations suggesting greater HIV risk in the more educated at earlier time points were replaced by weaker associations later. Similar shifts in patterns were found for Tanzania, the only setting with national data at the time of writing (Hargreaves and Howe 2010).5 What this shows is that the relationship between education and HIV is complex. Taking the case of Zambia, as reported by Hargreaves et al. (2008), for both young, urban men and women education was becoming protective, while this was not true for rural areas, where having some education was still associated with higher HIV prevalence.
The results are more promising for those working from a social protection perspective. Results for interventions with different-size transfers for different treatment arms suggest that larger payments are more likely to have an impact. This was true for the Tanzanian RESPECT and the Lesotho lottery projects (see above). Only in the Malawi SIHR was there no apparent relationship between transfer value and HIV or HSV prevalence (Baird et al. 2012). Even where we do not have project evidence on the impact of different-size transfers, the headline results seem to suggest it is important. In the Malawi Incentive Programme, the authors blame the small size of the transfer ($0–$16 after one year), as well as the lagged reward structure, for the failure to have an impact (Kohler and Thornton 2012, 168, 169). Similarly, the Kenya OVC (with a monthly payment of $20) had a bigger impact on sexual debut than the conditional Malawi SIHR (with a monthly payment to carers of $4–$10 and to girls of $1–$5).
This may be a source of support for those who see poverty or, where funds are targeted to girls, gender inequality as the biggest HIV risk. However, it is also true that the authors of the report on the Malawi Incentive Programme suggest that higher income puts male participants at greater risk of HIV. They speculate that, among other things, the detrimental impact on male HIV risk came about because participants could spend more money on items that made them more attractive (such as new clothes), and that some men were able to use the money to purchase more sex (Ibid., 186). More than that, we have already seen that the impact on HIV risk was itself unclear for most studies, even with the higher payment arms.
This ambiguity about income fits with other studies that show that the relationship between wealth and HIV is far more complex. The vast majority of studies provide a snapshot picture of HIV prevalence and they tend to find either find a clear positive relationship between wealth and HIV status (i.e. the rich are more likely to be HIV positive) or an ‘inverted U-shaped’ relationship (i.e. the middle wealth groups have the highest HIV prevalence) (Mishra et al. 2007; Parkhurst 2010). Parkhurst (2010, 520) finds a positive correlation is common in poorer African countries, but not in richer countries. The three available longitudinal studies (i.e. where a group of people are followed over time) also produce entirely contradictory results, despite all being drawn from southern Africa. One shows that poor men have the highest incidence, another that it was the middle-income men, while the third finds no relationship between wealth and incidence of HIV (Johnston 2013).
The evidence on female income and HIV risk also challenges simple assumptions that increasing women's relative income will reduce their HIV risk. A systematic review of 35 or so studies found slightly more evidence that women with higher socio-economic status have higher HIV acquisition (Wojcicki 2005). Wojcicki concludes ‘there is some indication that access to increased funds for women may put them at increased risk for HIV infection – potentially by giving them access to more partners or opportunities for travel’ (19). When projects assume that female poverty increases HIV risk, they seem to be glossing over a complex and possibly contradictory picture.
Based on the discussion above, it is clear that the praise for HIV-related cash transfers has selectively interpreted the evidence. In the following section, we will evaluate the approaches that generated these projects more broadly, using a radical political economy approach.
A critical political economy assessment
Political economy work on health seeks to explain how poor health, or affliction – as O'Laughlin (2013) terms it – is produced, often unintentionally, by the nature of production structures, and the social and political factors that determine health care provision. To fully understand this production of affliction, however, we must differentiate between different forms of disease. With the main mode of HIV in Africa being sex, Hunter's work (2010, 218; and in this issue) convinces us that sex (and other aspects of intimacy, including affection, love, pleasure and fertility) is always linked in multiple ways to making a living. Indeed, Hunter (2010, 11–12) suggests that we must consider intimacy rather than solely sex, defining the broader term to include ‘fertility, love, marriage and genital pleasure’ (3). Analysing HIV through the political economy of affliction and intimacy places importance on the relationship between capital and labour, and between classes of labour transcribed by gender and social status. Specifically sexual norms and health provisioning reflect the economic and social relations of production, and it is important to expand on this here to explain why.
Using class analysis and linking intimacy to work and livelihoods, we can see how sexual norms are created and change over time. Eschewing a vision of simple sexual norms (of the poor, of the educated … etc.), we instead adopt an approach where norms are varied by social grouping in different settings. Thus in Tanzania, Bujra (2006) explains the creation of class-based sexual networks. In South Africa, Hunter (2010) explains the risks faced by the poor in peri-urban KwaZulu-Natal. At the same time as being complex, sexual norms respond dynamically to changes in patterns of employment and residence.
Here we see how the pattern of uneven development within and between African countries can lead to a restructuring of sexual norms in ways that promote concurrency, transactional and commercial sex. Specifically, authors point to the way in which both local and global forces have led to substantial inter- and intra-country migration, as well as growing inequality. For example, Bujra (2006) has argued that the growing wealth, power and mobility of an elite group in Tanzania led to patterns of sexual networking, including commercial and transactional sex, which raised HIV risk. Hunter (2010) considers how the dismal income-earning opportunities, especially for women, in a peri-urban area of South Africa have changed the nature of intimate relationships. Men are able to adopt a new ‘playboy’ status, where their relative economic power allows them to have several intimate partnerships with poorer, economically vulnerable women. This pattern of multiple partnership is compounded by the fact that, despite having better economic options than women, employment security for poor black men has also deteriorated in the face of a collapse of employment growth and increased casualisation. Older and more stable patterns of partnership, where a man was able to maintain his wife in a rural homestead, are no longer viable. Hickel (2012) points to similar trends, arguing that bleak income-earning opportunities for poor women in Swaziland, coupled with high inequality, have led to a rise both in transactional and commercial sexual relationships. For him, patterns of partnership can also be explained by male migrancy to South Africa, as this promotes shifting and transient sexual partnerships.
This discussion of the way that intimate partnerships have been affected by rising inequality and uneven economic development across the African region places HIV risk in a new light. While patterns will vary between different countries, classes and genders, it is clear that patterns of intimacy are related to the long run and dramatic changes to the way that individuals reproduce themselves, to the places where they live and work. No longer is it the result of poor decisions by individuals about safe sex, but instead it reflects the long integration of African economies into a globalised economy.
However, HIV risk can only be fully understood if we combine the analysis of intimacy and sexual norms with that of health provisioning. Before taking each of these in turn, we should remind ourselves of the factors that increase both the transmission of HIV and its aggressiveness. There is evidence that the presence of other sexually transmitted infections increases the likelihood both of contracting HIV, and of transmitting it (Stillwaggon 2006; see also CDC 1998). Poor nutrition, parasitic disease and malaria may raise the chances of contracting HIV, and also the onset of AIDS (Stillwaggon 2006). Male medical circumcision appears to significantly reduce HIV transmission from women to men during sex (WHO 2012b), although there are dissenting voices (Perrey et al. 2012). Access to medication is important. Taking antiretrovirals (ARVs) reduces HIV transmission from an HIV-positive person to their HIV-negative partner (WHO 2012a) and drugs are needed to reduce mother-to-child transmission (UNAIDS 2010). Finally, the safety of medical interventions is crucial in reducing medical transmission (Gisselquist 2004).
Thus, public health policy on the treatment of sexually transmitted infections, on circumcision, nutrition, drug access for mother-to-child transmission and ARVs generally will affect both adult and child transmission rates in particular settings. On each of these issues, there are reasons for concern in most African countries. Indeed, until 2003, poor countries were prevented from importing generic ARVs for almost a decade (Hickel 2012, 526) as a result of the WTO's Trade Related Aspects of Intellectual Property Rights. The costs of ARVs were kept at extremely high levels, available only to the wealthiest individuals. Hickel (Ibid.) argues that not only were a generation of AIDS sufferers condemned to death, but the lack of effective AIDS mitigation produced a strong disinclination for testing in Swaziland. Thus, barriers to low-cost generic ARVs increased HIV incidence through direct and indirect channels. At the same time, medication to reduce mother-to-child transmission has not been fully accessible to those who need it (UNAIDS 2010).
Many authors have pointed to the limited health services available in sub-Saharan Africa, in terms of both private and public provision. The scale of medical transmission of HIV in many African countries is fiercely debated, with poor historical and current medical practices possibly transmitting HIV far more frequently than usually acknowledged (see Gisselquist 2004, for example). Further, O'Laughlin (2013) suggests that the lack of employment-related health care provision (restricted to the elite of the formally employed) is a key reason for limited health care access, with casual or migrant work having few or no health benefits attached. Hickel (2012, 519) reminds us that many migrants receive particularly poor health provision, taking the case of South African migrants where curable STIs are often endemic (raising HIV risk). As Campbell (2003) and Stuckler, Basu, and McKee (2010) argue, South African mining companies have externalised the health costs of the industry, in terms of AIDS, tuberculosis and other mining-related health conditions. At the same time, the outcome of neoliberal labour market policy has been a greater casualisation and informalisation of work (Whiteside 2005; Hickel 2012; O'Laughlin 2013). Casualisation further reduces employers’ commitments to a worker's health, as well as producing extremely insecure employment, forcing women in particular into risky sexual relationships.
In the light of political economy understandings of HIV transmission, the limitations of cash-transfer projects are clear. Let us take each aspect in turn. Can cash transfers change sexual norms? This might depend on the extent to which longer-run patterns of reproduction and residence are changed. To echo other criticisms of cash transfers (such as Ghosh 2011), cash transfers are limited in their ability to fundamentally and significantly change the economic options available, when these are determined by wider financial, trade and industrial policies.
Instead we need an increase in decent employment opportunities (i.e. with stability, benefits and reasonable wages), and opportunities for poor women are crucial. Hickel (2012, 528–529) sets out how these changes could take place for Swaziland, suggesting the creation of local employment options that would slow labour migration rates, and that this would need a wide swath of support for small-scale agriculture: investment in small-scale farmers; tariff barriers to protect the domestic food market from subsidised foreign imports; the sourcing of US food aid from local farmers, instead of American agribusiness; greater integration of small-scale farmers in the sugar industry and other agricultural industries; and the introduction of legislation to allow women equal rights to land and credit. Hickel (529) also suggests that there should be widespread domestic employment creation through state- and donor-funded infrastructure projects, open to both male and female workers.
Alongside substantial changes to the means of reproduction and residence patterns, health provisioning must be improved. Can cash transfers lead to an improvement in health generally? Evaluations of cash transfers generally suggest that they appear to have positive effects on household food consumption and dietary diversity, while conditional cash transfers specifically increase awareness of health messages, immunisation rates, and use of health and nutrition services (Gaarder, Glassman, and Todd 2010; Ranganathan and Lagarde 2012; Ruel and Alderman 2013). However, evidence of effects on health and nutritional outcomes is limited, and the most comprehensive indicator, child anthropometry, suggests there have been no statistically significant effects of either conditional or unconditional transfers overall (Ruel and Alderman 2013). Information for individual programmes supports this, suggesting that only a few programmes have effects on bodily outcomes and these have tended to be for the youngest or poorest children or those on the project for longer (Ibid.). Aside from specific impacts on macro- and micro-nutrients, little data has been collected on other health outcomes and even less shows significant changes.
Several reasons have been given for the lack of impact on nutritional and health outcomes, given that health service use increased (Ibid.). There may be a statistical basis – that most evaluations use RCT data that comes only from programmes in middle-income countries while programmes in low-income countries might be more effective. More fundamentally, it has been argued that programmes might be poorly designed, too short or not generous enough. Finally, the lack of correlation between increased use of nutritional/health services and improved health and nutritional outcomes implies that health services are poor quality (Ranganathan and Lagarde 2012; Ruel and Alderman 2013).
A political economy approach would consider the issue of health provisioning differently, and as an issue not only for the public sector, but for the private. As O'Laughlin (2013) has argued for South Africa, it is important that there is wider responsibility among employers for worker health, and for ARV provision. Migrant rights to health care must also be improved, and, for international migrants, this will require policy change by host governments. An improvement in employment-based health care rights would be limited to larger employers, however, leaving the majority of African populations dependent on state-provided health care.
The menu for state economic action laid out above requires active trade and industrial policies, massive investment in health and weaker fiscal discipline. Cash-transfer programmes are extraneous to this programme. This critical evaluation confirms the concerns of earlier writers about the sustainability and ethics of HIV-related cash transfers. The intrusiveness of some interventions can only justified by their potential benefits, but we have seen how limited these are. At the same time, for example, very specific ethical concerns are raised by a project which runs a lottery with a group of people who apparently struggle to evaluate risk. The use of the lottery structure in this case seems more to do with affordability than ethical or theoretical issues. Conditionality can also have (unintended) harmful consequences. For example, the CAPRISA 007 RHIVA project is intended to be conditional not only on school attendance but also good exam performance, raising the issue of whether adjustments can be made for different academic abilities. Interventions requiring a negative STI test do not always explain what happens to those who fail the test – given that curable STIs might be chosen, are participants offered treatment if they test positive? These issues become more sensitive as target groups usually start at 14. Often teenage girls are particularly at risk of acquiring HIV, but the issues of power imbalance are never clearer than when young, poor individuals are targeted.
Conclusion: choice and structure in HIV-related cash-transfer projects
Cash transfers are the latest fashion – liked by almost everyone, seemingly effective and potentially cheap. However, above, I concluded that they are far from successful in reducing HIV risk. The behavioural versions also border on the unethical, and certainly in their focus on ‘faulty decision-making' they compound the stigma that is attached to HIV acquisition. The contribution of this article is twofold. It provides a critical analysis of existing cash-transfer schemes, providing a more detailed reading than previous assessments and one that recognises the limitations of RCT-generated data. Second, the article provides a radical political economy perspective on the conceptual basis for these interventions. Projects based on social work theory (and centred on the positive effects of education) may fail because they do not recognise the economic and social relations that govern the options that young people face. Projects based on poverty reduction or gender inequality have an equally narrow view. Of course, reductionism is clearest in schemes based on the libertarian paternalist version of behavioural economics. Here the reductionism underlying interventions is one that blames HIV acquisition on faulty discount rates (i.e. the inability to properly respond to HIV risk). These projects set out to change only the way that people make choices. However, a wider set of factors leading to high HIV prevalence is ignored: factors related to the economic and social relations of production. Not only are sexual norms influenced by economic and social factors, but HIV risk itself is related to wider health provisioning. The underlying causes of high levels of HIV are deeply political in nature.
Consequently, readers will not be surprised that cash-transfer projects have had such complex and limited impacts. They may be equally unsurprised that cash transfers remain so attractive to policy makers. After all, the oversimplified premise of these projects edits out politically difficult questions of power, distribution and class. As Harman (2011, 879) argues, ‘Cash transfers perform key roles essential to the main paradigm of global health: they are performance oriented, show measurable results, engage in social protection and can be replicated in multiple different contexts.’
A political economy approach to HIV would instead consider the broad factors that affect patterns of intimacy, illness and health provisioning. As HIV risk is produced by these factors, we must consider how those broad factors emerge and are sustained. By recognising the need for a change in the social and economic organisation of the economy, an inherently political analysis is produced, one with very different conclusions to the menu of behaviour change interventions. However, in place of a far-reaching policy agenda to change health risks, instead we indulge in a fashion for cash transfers. Individuals, not states or markets, must change their behaviour. Cash transfers try (with limited degrees of success) to change how people act in situations where deep-seated economic and social factors have led to shockingly high HIV prevalence. Lotteries, sliding-scale transfers and other incentive schemes will be designed, implemented and evaluated – all the time avoiding the bigger, more difficult and politically sensitive questions.