Mosquito nets and the poor: can social marketing redress inequities in access?


Rose Nathan (corresponding author), Hassan Mshinda and Honorati Masanja, Ifakara Health Research and Development Centre, PO Box 53, Ifakara, Tanzania. E-mail:,,
Joanna R.M. Armstrong Schellenberg, London School of Hygiene and Tropical Medicine, Keppel Street, London WC 1E 7HT, UK. E-mail:
Don de Savigny, Christian Lengeler and Marcel Tanner, Swiss Tropical Institute, Postfach. 4200 Basel, Switzerland. E-mail:,,
Cesar G. Victora, Federal University of Pelotas, CP 464, 96001–970, RS, Brazil. E-mail:


Treated mosquito nets are a practical malaria control tool. However, implementation of efficient delivery mechanisms remains a challenge. We investigated whether social marketing of treated mosquito nets results in decreased equity in rural Tanzania, through household surveys before the start of a social marketing programme and 3 years later. About 12 000 household heads were asked about ownership of nets and other assets including a tin roof, radio, or bicycle. A socio-economic status score was developed for each household. Net ownership was calculated for households in each quintile of this score, from poorest to least poor. In 1997, about 20% of the poorest households and over 60% of the least poor households owned a mosquito net. Three years later, more than half of the poorest households owned a net, as did over 90% of the least poor: the ratio of net ownership among the poorest to least poor increased from 0.3 in 1997 to 0.6 in 2000. Social marketing in the presence of an active private sector for nets was associated with increased equity.


Malaria has been described as both a cause and a consequence of poverty (Gallup & Sachs 2001), and there is little doubt that the consequences of the disease are likely to be worse for the poor and marginalized than for the better off. Mosquito nets treated with insecticide are one of only a few malaria control tools with a proven effect on child survival. A Cochrane review based on four large randomized controlled trials of treated nets in Sub-Saharan Africa gave an estimated 19% efficacy for all-cause child mortality (Lengeler 2001). This efficacy has also been shown to translate to effectiveness. A recent programme evaluation in Tanzania showed a 27% reduction in the risk of death in children aged 1–59 months among those using treated nets, compared with children not using any net (Armstrong Schellenberg et al. 2001). Despite this evidence, implementation of treated nets across Africa remains alarmingly slow (National Bureau of Statistics 2000; Roll Back Malaria 2002) making the coverage target of 60% by 2005 set at the Abuja summit (Roll Back Malaria 2000) in April 2000 seem increasingly unattainable.

One reason for the lack of action on net implementation is the problem of insecticide treatment and re-treatment. There is no doubt that treated nets save lives: the evidence on untreated nets is somewhat less clear. Treating a net with insecticide is simple, inexpensive, and reduces nuisance insects as well as malaria. Yet despite major efforts uptake has remained poor (Chavasse et al. 1999; Armstrong Schellenberg et al. 2002). This problem may have increased inertia among implementing groups. However, now that long-lasting net treatment is becoming a reality, and increasing awareness that untreated nets are also associated with an increase in child survival, net treatment should not be an issue that slows down implementation of mosquito net programmes.

A second reason for the slow pace of ‘going to scale’ with nets is a perceived lack of evidence of how to distribute nets in poor rural areas of Africa. Nets are generally thought to be too expensive for the majority of those most at risk of malaria to afford, and hence Ministries of Health, local NGOs and international agencies continue to seek ways to reduce the prices through total or partial subsidy (Armstrong Schellenberg et al. 1999; Hanson & Jones 2000; Guyatt et al. 2002). Social marketing programmes offer a way to increase demand through promotion at the same time supplying nets at subsidized prices. Yet subsidies are equated with a lack of sustainability, and social marketing may be a relatively inefficient use of resources. Some have argued that nets should be seen as a ‘global public good’ and should be given out at no cost to young children and pregnant women who compose the group at most risk of life-threatening malaria in most of Sub-Saharan Africa (Curtis & Maxwell 2002; Guyatt et al. 2002). Meanwhile the private sector continues to sell nets to those who can afford unsubsidized prices in the few areas of Sub-Saharan Africa which are within the reach of existing distribution systems.

Equity considerations are crucial when deciding whether to subsidise nets, how, and by what amount. The term ‘health inequity’ is used here to mean health inequalities which are unjust according to some theory of social justice (Whitehead 1992). In the current context, inequity could be summarized by the gap in net ownership between the poorest and the least poor. Social marketing – where commercial marketing methods are applied for a health gain and without a profit motive – generally involves selling products at subsidised prices. Good social marketing also uses intense and well-informed promotion and often improves access through a subsidized distribution network. Yet because nets are sold rather than given away for free, social marketing might be expected to decrease equity. Here we present evidence of the effect of a social marketing programme of treated nets on equity over a 3-year period in a rural area of Tanzania subject to intense perennial malaria transmission.


The study was carried out in a rural population of about 60 000 people living in 12 000 households in 25 villages of Kilombero and Ulanga districts in Morogoro region, southern Tanzania. The population is ethnically heterogeneous. Subsistence farming is the main activity followed by fishing and small-scale trading. A detailed description of the study area is given elsewhere (INDEPTH 2002). Social marketing of treated nets and insecticide started in May 1997, following several months of sensitization meetings and formative research on householders’ perceptions of causes of child death, mosquito nets, net treatment, and malaria (Minja et al. 2001). Details of the social marketing programme are given elsewhere (Armstrong Schellenberg et al. 1999; Hanson et al. 2003). Briefly, treated nets and insecticide for net treatment were packaged and branded according to local preferences. Village-based sales agents included health workers, shopkeepers, religious leaders and village government members, with at least one agent in each village. Successful agents were generally shopkeepers, with a few exceptional health personnel. A comprehensive information, education and communication campaign was developed and implemented. Retail prices were set at around US$5 for a treated net and $0·42 for insecticide treatment kits. The private sector sold ordinary mosquito nets throughout the period of the study. In Kilombero and Ulanga median monthly household expenditure in 1997 in a representative sample of local households varied from $77 to $96, depending on the season, of which approximately 75% was expenditure on food (INDEPTH 2002).

All households in the study area were included in a demographic surveillance system (INDEPTH 2002) whereby every household was visited every 4 months. At these visits, an adult member of the household – usually the household head – was interviewed about births, deaths, in and out migrations and pregnancies arising in his/her household since the previous visit. Interviewers completed a brief questionnaire on net ownership and proxy markers of socio-economic status in every household in 1997 (10 291 households) and again in 2000 (11 970 households). In both 1997 and 2000, information was collected on whether or not the household owned a bicycle, a radio, any animals, ducks or chickens and a tin roof. In 2000, additional information was collected on the occupation of the household head – whether he or she was involved in subsistence farming, fishing, business, driving, masonry or other employment.

Socio-economic status was assessed by constructing a household ‘wealth index’ based on household asset ownership as proposed and validated by (Filmer & Pritchett 2001). A weighted sum of the assets was calculated for each household using principal components analysis. For the 2000 data, the first three eigenvalues were 2.1, 1.4 and 1.1 and explained 15%, 10% and 8% of the variation respectively.

We used the first principal component to construct our index, which gave the greatest weight in both 1997 and 2000 to owning a radio, a bicycle or a tin roof. The weights for these three assets were as follows: for a radio, 0.62 in 1997 and 0.52 in 2000, for a bicycle, 0.60 in 1997 and 0.49 in 2000, and for a tin roof, 0.39 in 1997 and 0.43 in 2000. Both subsistence farming, and fishing as the occupation of the household head, had negative scoring factors on the first principal component. Based on the constructed index, households were classified into wealth quintiles, rated as poorest, very poor, poor, less poor and least poor. Note that there are no households in the study area which could be classified as ‘rich’ and thus we use the relative terms ‘poorest’ and ‘least poor’ as opposed to ‘poor’ and ‘rich’. To summarize the ability of the combined efforts of the social marketing programme and the private sector to reach the poorest households we report coverage in the poorest group. To summarize equity we report the ratio of coverage in the poorest households to that in the least poor (Gwatkin et al. 2000). We assessed the statistical significance of the relationship between household net ownership and socio-economic status using logistic regression of household net ownership on socio-economic status quintile, testing for linear trend using a likelihood ratio chi square test (LRC) with 1 d.f. All analysis was carried out using stata software (Stata Corporation 2001).

In order to be able to check whether our results were consistent when socio-economic status was measured using exactly the same score in both 1997 and 2000, we repeated the analysis using the index constructed for year 2000 in the households which could be linked in the two datasets.


Household net ownership rose over the 3 years of the study from 3820 of 10 291 households (37%) with at least one net in 1997, to 8745 of 11 970 households (73%) in 2000. Table 1 shows asset ownership in each quintile in 1997 and 2000. Almost none of the poorest households owned a radio, bicycle or a tin roof. Among the least poor households, over 90% had a bicycle and a radio, and almost half had a tin roof. Because of the small number of assets in 1997, the five quintiles had unequal numbers of households, with about one-quarter of the households in each of the poorest two quintiles and only one-tenth in the middle quintile.

Table 1.  Asset ownership for households in each socio-economic status quintile
YearSocio-economic status quintileNo. households (%)Mean SES scoreHousehold assets
Radio (%)Tin roof (%)Bicycle (%)Rent the house they occupy (%)Animals (%)Chickens/ducks (%)
1997Poorest2564 (25)−1.3100000
Very poor2553 (25)−0.810000100
Poor1069 (10)−0.0533264200
Less poor2290 (22)0.69531856172
Least poor1818 (18)2.15944592586
2000Poorest2127 (18)−1.231012035
Very poor2534 (21)−1.221011148
Poor2538 (21)−0.272215456149
Less poor2070 (17)0.567026599347
Least poor2637 (22)1.9895499313473

In 1997 household net ownership had a strong positive relationship with the household wealth index, with only 518 (20%, 95% CI: 19–22) of the poorest households owning at least one net and 1151 (63%, 95% CI: 61–65) of the least poor households owning a net (figure: LRC = 1011.9, P < 0.0001). The ratio of net ownership in the poorest to the least poor households was 0.3. In 2000, 1324 (54%, 95% CI: 52–56) of the poorest households owned at least one net, as did 2053 (92%, 95% CI: 91–93) of the least poor households. The association of net ownership and wealth quintiles remained statistically significant (LRC = 1100.8, P < 0.0001) but the poorest to least poor ratio had improved from 0.3 to 0.6 (Figure 1). The slope of the regression line linking net coverage and wealth decreased significantly over the 3 years (LRC for interaction between net coverage and wealth = 11.0, P = 0.0009).

Figure 1.

Household net ownership by socio-economic status in 1997 and 2000.

When the analysis was restricted to the 6919 households that were involved in both 1997 and 2000, we found very similar results. For example, in 1997 household net ownership had a strong positive relationship with the household wealth index, with only 328 (24%) of the poorest households owning at least one net and 797 (60%) of the least poor households owning a net (P < 0.0001). The ratio of net ownership in the poorest to the least poor households was 0.4. In 2000, 734 (53%) of the poorest households owned at least one net, as did 1234 (93%) of the least poor households. The association of net ownership and wealth score remained statistically significant (P < 0.0001) but the poorest to least poor ratio had improved from 0.4 to 0.6.


Our results do not support the hypothesis that social marketing decreases equity. Different measures of equity reflect the trend in different ways, but our findings suggest that equity has increased over the 3 years of the programme: the ratio of net ownership in the poorest households compared with the least poor was 0.6 in 2000 compared with 0.3 in 1997. Expressed as a difference rather than a ratio, coverage among the poorest increased by 34 percentage points, compared with 29 for the least poor. The least poor are now close to ‘saturation’, the maximum achievable coverage, and equity must increase if nets become more common (Victora et al. 2000). Although equity has improved, there is still a long way to go and the rate of improvement needs to be sustained. This calls for continued efforts in the ongoing social marketing campaign and further tailoring of the messages accompanying it, together with more research on non-users. More generally, health programmes which have attempted to target the ‘poorest of the poor’ individually have had limited success, with some positive experiences from Asia and Latin America but very few from Africa (Gwatkin 2002). Despite the failures, there is some evidence to suggest that the approach may be able to increase the percentage of service benefits that accrue to the poor.

We assessed socio-economic status through an index of wealth based on a weighted sum of household assets. Although this approach gives only a relative measure of wealth, which is not easy to compare with more conventional approaches to poverty or wealth assessment, the information was simple to collect and was applied consistently before and after social marketing. The index revealed large inequities in mosquito net ownership even in an area with apparently homogeneous socio-economic status. Validation studies in Indonesia, Pakistan and Nepal concluded that principal components analysis provided an index that was at least as good in predicting school enrolment differentials as more conventional approaches based on expenditure (Filmer & Pritchett 2001).

We focussed here on household net ownership rather than net use among children. Nets are relatively costly items, and for those considering whether or not to subsidize nets it is usually the mechanism of getting nets into households which is the biggest challenge. In our study area there is evidence to suggest that parents and the youngest children get preference for use of a single net within a household (Minja 2001). Household net ownership may be seen as necessary step in the process by which the Abuja targets may be reached, and is an additional indicator rather than a substitute for net coverage in children.

The rapid increase in net ownership between 1997 and 2000, coupled with the decrease in inequity, is likely to be the result of a combination of factors, only one of which is social marketing. Without a concurrent control group we are unable to be sure precisely what other issues may have led to the change, but one of the most important factors is likely to have been the private sector. The private sector for mosquito nets in Kilombero was extremely active; with roughly one-third of the market share of nets sold in 1998 (Kikumbih 2002). The market for mosquito nets in Tanzania was expanding rapidly during the period of the study, with three large net manufacturers within the country and extremely competitive prices.

In Tanzania there is a dynamic market for mosquito nets, which is progressively reaching further into rural communities. A social marketing programme of the type studied here can build on these opportunities for large-scale distribution. Our findings suggest that demand is high in all socio-economic groups, but it is also clear that cost remains an obstacle. In Tanzania, a higher-value voucher is being developed to address this issue. In other settings, untargeted nets distributed to all free of charge might need to be further explored. However, it should be kept in mind that there is no evidence to date on the extent to which the distribution of free nets is feasible, efficient or sustainable at a national level.

There are many different strategies of social marketing (Andreasen 1995), and it should not be assumed that other social marketing models, different to that used in our study area (Armstrong Schellenberg et al. 1999) would necessarily lead to the same effect, particularly in the absence of two enabling factors: the existing demand for mosquito nets, which was extremely high probably because of perceived mosquito nuisance, and the existing active private sector for nets.

In conclusion, social marketing was associated with rapid overall improvements in net coverage and the pace of change was higher among the poorest than the least poor. Our findings have informed plans for going to a national scale with treated nets in Tanzania, where social marketing is to be used to trail-blaze for long-term sustainable commercial marketing.


We are grateful to the Ifakara DSS field and computing teams for their invaluable contribution in data collection and processing. We thank Mr Haji Mponda for implementing the social marketing activities, and the community in the DSS area who agreed to be interviewed.

The social marketing programme was funded by the Governments of the United Republic of Tanzania and Switzerland, through the Swiss Agency for Development and Co-operation (SDC). The demographic surveillance system was partly funded by the Centres for Disease Control and Prevention (CDC); the Swiss Agency for Development and Co-operation (SDC), the Swiss National Science Foundation (SNSF) and the Multi-Country Evaluation of IMCI Effectiveness, Cost and Impact (MCE). The MCE is arranged, coordinated and funded by the Department of Child and Adolescent Health and Development of the World Health Organization, and with the financial support of the Bill and Melinda Gates Foundation and the US Agency for International Development.