Outdoor cooking prevalence in developing countries and its implication for clean cooking policies

In recent years, the promotion of clean cookstoves to reduce smoke exposure has received much attention in both academic and policy discussions. Indeed, much is at stake: more than 3 billion people in developing countries rely on firewood and charcoal for their daily cooking purposes. According to the World Health Organisation (WHO), the emitted smoke from cooking kills 4.3 million people every year—more deaths than are caused by malaria, tuberculosis and HIV combined—making it one of the most lethal environmental health risks (WHO 2014a, Martin et al 2011).

Under the auspices of the United Nations initiative Sustainable Energy for All (SE4All) and spearheaded by the Global Alliance for Clean Cookstoves (GACC), the international development community is currently embarking on a massive effort to spur universal adoption of clean cookstoves and fuels (GACC 2011, SE4All 2015). Achieving universal adoption is a laudable outcome, but one that faces substantial organizational and financial constraints. This raises the question of whether policies should concentrate on technologies and fuels that qualify as absolutely clean from a public health perspective, such as electricity, LPG, or advanced gasifier biomass stoves, or whether intermediate technologies such as simple improved biomass stoves should also be promoted under certain conditions (Simon et al 2014). Notwithstanding their considerably higher costs and often fragmented supply chains, a recent WHO report advocates the usage of primarily LPG and electricity or advanced gasifier biomass stoves. The justification is that these stoves have proven themselves in applied settings to substantially reduce fuel consumption and to improve health outcomes for cooks and accompanying children (WHO 2016).

In the present paper, we argue for an alternative prioritization that takes into account how smoke exposure is impacted by the interaction of cookstove technologies and cooking behaviors (Jeuland et al 2015). In this regard, where people cook—whether indoors or outdoors—has important implications for the particulate matter area concentration levels, ventilation, and thus smoke exposure (see Bensch and Peters 2015, DasGupta et al 2006, Yu 2011), but has nonetheless been widely neglected in debates about clean stove distribution. Impact potentials of stoves are higher if meals are prepared indoors. Conversely, if meals are prepared outdoors, natural ventilation reduces concentration levels considerably, with an associated reduction in the beneficial impact of the clean cookstove.⁵ Scarce public resources should consequently concentrate on distributing the most advanced cookstoves among households where indoor cooking prevails and hence exposure is highest. In areas where outdoor cooking dominates, much simpler—and cheaper—improved biomass stoves are potentially more cost effective in reducing the adverse effects of biomass cooking. Donors should of course continue to invest in and promote the distribution of clean cookstoves. But given that this process depends on LPG supply chains and the electricity grid, it is likely to take decades before all regions are covered. Hence, the targeted distribution of improved biomass cookstoves, coupled with the promotion of outdoor cooking and awareness raising campaigns to encourage better ventilation practices, can act as 'bridging interventions' while supply side bottlenecks are removed.

We develop this argument in two steps. Drawing on data from the Demographic and Health Surveys (DHS), we first document cooking behavior by country, which reveals a sizeable incidence of outdoor cooking. Next, we calculate hypothetical concentration level reductions for different stove types and ventilation scenarios and then categorize the stoves into different internationally recognized emissions categories, or tiers, that are used by SE4All to define which stove type qualifies as clean. Our exercise demonstrates that this multi-tier categorization heavily depends on where the cooking is done and the ventilation in the kitchen. Based on the documented heterogeneity in cooking patterns, we suggest that the dissemination of cheaper improved biomass stoves should be given serious consideration as a cost-effective instrument to bring down exposure levels among households that cook outdoors.

2.1. Health effects of household air pollution and cooking ventilation

2.2. Policy background

We use data from the latest waves of the nationally representative DHS. The data have been regularly collected in around 90 low-and middle-income countries since 1984. For our purpose, we only included low and lower middle income countries in Africa, Latin America and South-East Asia as defined by the World Bank, thereby excluding Brazil and the Maldives. Due to data regulations, not all countries that fit this classification could be included in the analysis.⁸

Information on the cooking location is only available for those countries where the latest available wave (wave 6) or the second latest available wave (wave 5) of the standard DHS questionnaire was conducted.⁹ If information in two waves were available for one country, we used the latest wave. This leaves us with a sample of 40 countries and 650 723 household observations for the years 2006 to 2014. Most of the included countries are situated in Africa (30), followed by Asia (6) and Latin America (4).¹⁰

The DHS questionnaires contain questions regarding cooking behavior, including stove usage, cooking fuels, and cooking location. We restrict our interest to the question on the cooking location, which asks households whether they usually cook in the house, in a separate building, or outside.¹¹ It was not possible to give multiple answers, as may be relevant, for example, if a household changes cooking location according to the season.

We divide the sample between rural and urban areas, since we expect different outdoor cooking patterns for these two groups. All results are furthermore weighted to ensure nationally representative results, with the weights provided by the DHS.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As seen from figures 1 and 2, outside cooking is prevalent in both the urban and rural areas of many developing countries, reaching a high of nearly 80% in rural Niger. Notwithstanding substantial heterogeneity, a few patterns in the data are evident. Out of the 20 countries with the highest outdoor cooking rates, 18 are located in Africa. Further differentiating within the African continent shows that West African countries have the highest share of outdoor cooking. Among the ten countries with the highest outdoor cooking rates, seven are in West Africa. At the other end of the spectrum, the four countries with the lowest outdoor cooking rates are spread across South America, the Caribbean, South East Asia, and Asia, with Pakistan registering the lowest rate of about 1%.

Large differences between urban and rural outside cooking patterns are seen in some countries. We take a closer look at only those countries with more than 15 percentage points difference in rural and urban outdoor cooking patterns. This yields two different types of countries, all based in Africa: those in which more households cook outside in rural areas than in urban areas (Benin, Gabon, Lesotho and Namibia) and those in which more households cook less outside in rural areas than in urban areas (Burundi, Republic of the Congo, Democratic Republic of the Congo, Guinea, Liberia, Madagascar, Malawi and Uganda). For all other countries, no major difference between household cooking patterns in rural and urban areas is observed.

The variation in cooking location has considerable implications for the emission-concentration-exposure nexus of cooking induced smoke. In this section, we provide a back-of-the-envelope calculation of particulate matter levels for different stove types according to whether the stove is used indoors or outdoors. The aim is to show that the effective cleanliness of a stove is profoundly impacted by this distinction. We use as cleanliness categories the tiers as defined in the SE4All Global Tracking Framework (see table 1). Our analysis includes stoves from tiers zero to three. Tier four stoves are mostly those that run on electricity and LPG, so virtually free of smoke emissions and thus not relevant to this analysis. All stoves included in our analysis have in common that they are non-traditional, portable, household biomass stoves without a chimney and not used for commercial purposes.¹²

For the cookstoves examined, we rely on emissions figures from Jetter et al (2012), who analyzes the emission of 22 cookstoves in a controlled laboratory environment. The selection of stoves in Jetter et al (2012) is based on availability, which excludes a large number of other non-standard stoves and chimney stoves, but covers those most widely disseminated. The authors measure the emission (in mg min⁻¹) from a low power and a high power scenario as defined in the Water Boiling Test (WBT) protocol. Although WBTs undoubtedly diverge from actual field use, they have the virtue of allowing comparison of many cookstoves under identical circumstances.¹³

We focus on the high power scenario results, since emissions tend to be higher during this phase. Results for the low power scenario are presented in supplementary table 3 available at stacks.iop.org/ERL/12/115008/mmedia. The two scenarios are complementary: while the high power scenario simulates the actual high power use of the cookstove, such as quickly boiling water, the low power scenario simulates the long simmering of legumes or pulses (GACC 2014). The scenarios are nonetheless just a proxy indicator for actual field use, as they do not capture behavioral confounders such as stove stacking, misuse and inappropriate cooking stoves for local practices (Beltramo and Levine 2013).

The first four columns of table 2 show the cooking device, associated cooking fuel, indoor emissions and their tiers for the high power scenario. Among the cooking devices, values are presented for both a minimally tended and carefully tended three stone fire, as this is the most prevalent cooking technology in developing countries. Indoor emission rates vary considerably for the high power scenario, as can be seen in column 3 of table 2.

Zoom In Zoom Out Reset image size

Based on the cookstove and their respective indoor emissions (measured in mg min⁻¹) depicted in table 2, we calculate average PM_2.5 concentration levels (measured in μg m⁻³) in the kitchen during cooking time. To this end, we apply a variant of the single zone box model developed by Johnson et al (2011), which was refined for easier implementation by the Aprovecho Research Center (2016a) in the form of a spreadsheet tool.¹⁴ The model abstracts from different concentration levels in different parts of a room or house and has been used in the analysis of biomass cooking concentration (e.g. WHO 2014a). In line with Johnson et al (2011), further assumptions are four hours of cooking per day (one hour in the morning, one hour at lunch time, and two hours for dinner) as well as a kitchen volume of 30 m³ and 25 air exchanges per hour.¹⁵ The actual time used for cooking in the field will of course vary depending on the location.¹⁶ Plugging the values of the respective indoor emissions from each cookstove into the spreadsheet yields the indoor PM_2.5 concentration levels (µg m⁻³), presented in column 5 of table 2.

As the box model cannot calculate outdoor concentration levels, in order to approximate the reductions from moving the cooking location outdoors we rely on the evidence in the literature (Albalak et al 1999, Balakrishnan et al 2002, Rosa et al 2014, and van Vliet et al 2013, as discussed in section 2.1). As those effects occur over a broad range, we account for this variability by taking area concentration reductions as linear scalars from three scenarios that span the estimates from this literature: 40%, 60%, and 80% (see table 2, columns 6, 7, and 8).

Since the PM_2.5 concentration levels (µg m⁻³) are directly proportional to the emissions (mg min⁻¹), we can convert the concentration figures for tiers 0–4 from table 1 into the emission levels, thereby yielding the yardstick used in the SE4All-Tier system. Figure 3 shows that there is a strong effect of outdoor cooking on how the stove should be categorized. Most stoves would improve by one tier in the 40% and 60% reduction scenarios and by two tiers for the 80% reduction scenario. Similar results are obtained in a low power scenario, demonstrating a robust relation (see table A3 in the appendix).

Importantly, the difference in tier categorization applies even to very simple and inexpensive cooking devices, such as the KJC Standard, a charcoal stove that costs USD$6. This device advances up the scale when used outdoors, from tier 1 to tier 3 under the 80% reduction scenario. As an example for a fuelwood driven stove that costs USD$25, the Berkeley Darfur stove is within tier 1 with indoor emissions, but tier 2 assuming an outdoor reduction of 40% and tier 3 in case of a reduction of 60% or 80%.

These examples illustrate that when scarce public resources constrain the coverage of an intervention to disseminate clean cookstoves, which can cost upwards of USD$90, consideration of cooking location should be at least one of the factors that bears on the decision of which region is targeted, prioritizing those regions where indoor cooking predominates. This prioritization applies equally to instances where stoves are not disseminated gratis but in exchange for partial or full payment, given that affordability is one of the documented barriers to adoption of lower-cost cooking technologies using market based dissemination (Beltramo et al 2016, Bensch et al 2015, Mobarak et al 2012, Lewis and Pattanayak 2012, Pattanayak et al 2016).

The above analysis provides an indicative estimate of the potential for ICSs to improve health, but it is subject to several technical and behavioral simplifications that should be relaxed in future research. Perhaps most importantly, research is needed on area concentration and smoke exposure under different ventilation conditions as well as cooking locations using rigorous evaluation methods that rely on real-world data rather than laboratory tests. For example, negative health effects may also result from disseminating bricked stoves installed in kitchens because people switch from outside to inside cooking.¹⁷ Furthermore, there may also be a negative impact from ambient air pollution to those cooking outside that would increase if everyone cooked outside, though this effect is likely to be small in comparison to indoor air pollution, particularly in rural areas. Seasonal variation is another area warranting study. In this regard, the extent to which households that otherwise cook outdoors move indoors during inclement weather cannot be discerned from the DHS data.

Although large cookstove initiatives are currently slated for implementation, reaching the target of universal adoption of clean cookstoves and fuels is a long-term endeavor that will require massive investments extending well beyond current commitments. Given the urgency and breadth of the challenges, it behooves development agencies to chart a course of improved cookstove distribution that accounts for the interaction of this new technology with cooking behaviors. This paper has argued that the cooking location—whether indoors or outdoors—is a key mediating factor on the effectiveness of clean cookstove adoption. We further document that outdoor cooking rates are high but vary tremendously between countries and continents as well as between rural and urban areas. Our main conclusion is that while clean cookstoves are the best option to reduce the exposure to air pollution among households that cook indoors, simple improved biomass stoves are potentially the more cost-efficient policy intervention in regions where outdoor cooking prevails. While we do not claim that ventilation and outdoor cooking brings exposure levels down to what is optimal from a public health perspective, it is the absence of affordable solutions in most of rural Africa that makes our finding very pertinent. For these parts of the world, it is unlikely that in the years to come LPG-supply chain bottlenecks will be solved or the electricity grid will be rolled out. Given these constraints, simple improved biomass stoves and better ventilation thus appear as a second-best solution. Behavioral change interventions, such as health education that sensitizes to ventilation, and the coupling of those interventions with cookstove distribution, is another promising avenue for reducing smoke exposure (e.g. Barnes 2014, Grabow et al 2013, Zhou et al 2006).

The publication of this article was funded by the Open Access Fund of the Leibniz Association.