The role of packaged water in meeting global targets on improved water access

Packaged water (as either refill, bottled, or sachet water) has become an important element of water security in many low- and middle-income countries, owing to poor reliability and lack of piped water infrastructure. However, over time and across countries, the Demographic and Health Surveys monitoring program has inconsistently classified packaged water components as either improved or unimproved. Using data collected as part of the Performance Monitoring and Accountability 2020 (PMA2020) surveys on water options in nine study geographies across eight countries, we identified five geographies where packaged water constituted one of several options for 5% or more of users. In this study, four scenarios were designed in which packaged water components were variously classified as either improved or unimproved. Unimproved water use was highest in scenarios where sachet or refill water was classified as an unimproved source. Across the four scenarios, the difference in the use of unimproved water as the main option was highest (65%) in Nigeria (Lagos). That difference increased to 78% when considering all regular options. The development of these scenarios highlights the importance of classifying a source as improved or unimproved in the overall metric that indicates progress at national and international levels.


INTRODUCTION
Lack of access to direct piped water supply in many low-and middle-income countries (LMICs) has led residents to seek alternative sources (Bakker et al. ). Among sources that are actively promoted by governments, businesses, and local entrepreneurs to meet this shortcoming is water packaged in disposable plastic bottles and small sachets, as well as in large refillable containers. In the rapidly growing cities of LMICs, packaged water (an umbrella term that includes bottled, sachet, and refill water) bridges the needs unmet by public infrastructure, and has seen an enormous Nigeria (NG) and Ghana (GH). In a majority of GH's ten regions, reliance on sachet water (i.e., small ∼500 mL bagged water sold individually or in packs) as the primary drinking source increased by about 5% between 1998 and 2008 (Stoler et al. b). The largest increase, however, occurred in the Greater Accra region where sachet water use rose from almost zero to 35% over the ten-year period.
A competitive market and easy availability of packaged water has made it a popular choice, especially among the highly mobile urban population (Figure 1).
Even as packaged water use has grown, its safety has remained in question however. Samples collected in highincome countries have shown elevated bacterial levels compared to tap water samples (Raj ). There is an extensive literature on the microbiological quality of packaged water at various points in the distribution channel, but is heavily focused on point-of-use quality. Packaged water sold in GH and NG is of varying quality, depending on the type of packaging. While bottled water in these two countries was generally found to show no or lower rates of contamination The worst performers were the hand-filled hand-tied polythene bags. Fecal or total coliforms were found in nearly half (Obiri-Danso et al. ) to all of the samples tested (Okioga ; Oyelude & Ahenkorah ).
Although most studies have evaluated the quality of bottled or sachet water that is inside the packaging, there is evidence to suggest that an equally important component of water quality might be what is outside the packaging, especially during distribution and point-of-use. Egwari et al.
() identified enteric pathogens and Escherichia coli in samples collected from cooling receptacles used in the sale of sachet water, the surface of sachets, and from melted water used to cool those sachets. In light of these results, it is no comfort to know that sachet water is more commonly sold as 'pure water' in GH and NG (Akunyili ).
Of course, not all studies point to sub-standard sachet water quality. Ahimah & Ofosu () report complete compliance with national standards when evaluating sachet water found in the streets of a large city in southeastern GH.
Sachet water use is also associated with a lower likelihood of diarrhea in children and higher levels of self-reported health in women (Stoler et al. a). A recent study reported that packaged water (sachet or bottled) provides protection against point-of-consumption E. coli contamination as compared to piped water, providing public health benefits (Wright et al. ). The ambiguous safety of packaged water perhaps played a role in how it has been classified according to the Demographic and Health Surveys (DHS), a USAIDled initiative that has set global benchmarks for drinking water and sanitation standards. DHS results, in turn, feed into the WHO/UNICEF's Joint Monitoring Programme (JMP) that tracks progress on water and sanitation access.
DHS classifies water sources as improved or unimproved based on a ladder that hierarchically places each water source based on the likelihood of the source being contaminated. In addition to the main source, respondents are asked to list a backup source as well. Placing bottled and sachet water on this ladder presented a unique problem, which was resolved via the backup water source. Our survey of DHS country reports revealed that until about 2008, bottled and sachet water were classified as either improved or unimproved, depending on how the backup water source was classified.
However, starting in 2010, the backup source question was dropped from the DHS questionnaire. (The backup source  Furthermore, we focus on one of the study geographies to identify factors associated with packaged water use.

METHODS
We relied on data collected in nine study geographies across Downloaded from https://iwaponline.com/washdev/article-pdf/7/3/369/158959/washdev0070369.pdf by guest regional level, ensuring a high quality of source water and effective monitoring of the packaging and distribution process.
2. Backup-dependent scenario: Determination of improved/ unimproved is based on the backup source. This scenario is the pre-2008 DHS paradigm, and presumes that it is difficult to assess the quality of the source, and hence relies on the backup source to classify packaged water as improved or unimproved.
3. Hybrid scenario: Bottled water is improved; sachet/refill water is unimproved. This is a hybrid scenario, where only bottled water is assumed to be an improved source. This might be plausible because (i) bottled water is regulated under stricter laws in several countries, (ii) multinational companies in the bottled industry may follow uniform sourcing and production checks, and (iii) higher pricing of bottled water compared to sachet/refill water allows the manufacturers to spend more on quality control.
4. Worst-case scenario: Both bottled water and sachet/refill water are unimproved. This is the opposite of the ideal scenario, and assumes a poorly regulated packaged water industry, where the source water quality is not guaranteed, and the packaging and distribution is not monitored.
DHS classification of packaged water in GH follows the ideal scenario, while that in NG follows the hybrid scenario.
The backup-dependent scenario was employed by DHS in every country until 2008. The worst-case scenario is the only truly hypothetical scenario, but one which might be applicable in countries with poor bottling regulations and monitoring.
Following development of the scenarios, we utilized household data from one of the selected study geographies to build binary logistic (logit) regression models to study the socio-economic and structural factors associated with packaged water use, both as the main option and as one of the regular options. The dependent variable in each of the models was the use of packaged water. Independent explanatory variables included socio-economic characteristics, structural factors, and a geographic regional control.

RESULTS
In the nine PMA2020 study geographies, use of packaged water varied widely. Packaged water use as the main option ranged from 0.1% in ET to 65% in NGL (Figure 2).
When considering all regular options, packaged water use ranged from 0.4% in ET to 75% in NGL. Across all study geographies, packaged water use as a regular option exceeded its use as the main optionby as little as 0.4% in ET and as much as 15% in GH.
Based on these results, we selected geographies where More than half of Lagos state residents consumed packaged water as their main option, of which, 7% were bottled water users. Kaduna residents, on the other hand, consumed packaged water at a much lower rate, and bottled water was used by less than 1% of the residents. ID was also unique in its Downloaded from https://iwaponline.com/washdev/article-pdf/7/3/369/158959/washdev0070369.pdf by guest notable consumption of refill water. Sachet water was not presented as an option to ID survey respondents.
We next explored the relationships between the use of packaged water and socio-economic characteristics such as location and wealth. In the three study geographies that had both urban and rural samples, packaged water use as the main option was consistently higher in the urban sample ( Figure 3). A locally weighted polynomial regression of packaged water use as the main option against normalized wealth score shows consumption increasing over wealth (Figure 4).
In all geographies except NGL, packaged water consumption was extremely low in poor households and increased gradually with rising wealth. Packaged water consumption started high in poor NGL households and rose only marginally across the wealth spectrum, revealing the important role played by packaged water in all NGL households, regardless of wealth.
The plot of packaged water use as a regular option against normalized wealth score differed in magnitude but was characteristically similar to the one observed in Figure 4 (see Supplementary information, Figure S1, available with the online version of this paper).
Presented in Figure 5 is the percent of respondents in the five selected study geographies that consume unimproved water as their main option under each of the four scenarios described in the Methods section. Overall, estimates for unimproved water consumption were markedly higher in the hybrid and worst-case scenarios, as compared to ideal and backupdependent scenarios. The magnitude of difference between the two groups of scenarios was based on the fraction of respondents reliant on sachet/refill water as their main option.
Estimates for unimproved water consumption as a regular option were similar in pattern, although not in magnitude, to those seen in Figure 5 (see Supplementary information, Figure S2, available with the online version of this paper).
We next focused on one of the study geographies, GH, to investigate the role of socio-economic factors in the use   Downloaded from https://iwaponline.com/washdev/article-pdf/7/3/369/158959/washdev0070369.pdf by guest of packaged water. We already observed that packaged water use was higher in urban and wealthier households (Figures 3 and 4). However, urban households in LMICs are typically also wealthier, so a regression model was used to isolate the effect of each explanatory variable. Variables and their summary statistics are described in Table 1. Table 2  were most likely to regularly rely on packaged water.

Panels A and B in
Household location and source reliability, however, were no longer drivers of packaged water use.

DISCUSSION
The popularity and widespread usage of packaged water in many LMICs may be attributed to convenience, safety, and attractive marketing of the product itself, or to the poor public infrastructure. In the PMA2020 survey of nine study geographies, packaged water consumption varied but constituted a regular source for more than 5% of the residents in five of those geographies. NGL's urban sampling frame could be a reason for the large packaged water use when compared to the nationally representative samples of GH and ID.
However, NGL points to the outsize role packaged water often plays in an urban water landscape, where a highly mobile population can afford to pay for easy access to water, yet chronically suffers from lack of a reliable water service.
PMA2020's inclusion of multiple water sources in its survey is a departure from other accepted global monitoring programs such as DHS and JMP that only assess the main source. As a result, estimating the regular use of packaged water is a better indicator of overall packaged water consumption than merely estimating its use as the main option. In every study geography, regular users of packaged water outnumbered the main users, in some cases by a significant amount. In study geographies that had both rural and urban sampling frames (GH, ID, and NGK), urban consumption of packaged water was far higher than that in rural locations, suggesting packaged water use as, while not exclusively, a predominantly urban phenomenon. We also observed packaged water use was influenced by wealth. In all five study geographies, the proportion of packaged water users increased along with rising wealth, although there were differences in the profile of users along the wealth spectrum.
The hypothetical classification of packaged water (or its components) as either improved or unimproved sources provided meaningful insights into the role of packaged water in the drinking water landscape in the selected geographies.  Regression analysis of packaged water users in GH revealed that they are more urban and wealthier than users of other drinking water sources, even after controlling for other variables. This suggests that even as packaged water use has become popular in GH, it is mostly used by the wealthy and urban section of the population, leaving the rest of the country with poor, often unsafe alternatives.
Smaller household size and access to multiple water sources also resulted in a higher likelihood of packaged water use.
Location and source reliability were the only variables where regular users of packaged water differed from the main users. The counter-intuitive role played by reliability of the main water source in the 'main option' model and the lack of significance in the 'regular option' model suggests that users do not always rely on packaged water as an alternative to poor choices but rather to supplement existing reliable water sources. This finding may warrant further investigation as reliability may be closely correlated with other underlying factors that are not captured in our models.

CONCLUSIONS
Packaged water use has increased in many LMICs where piped water availability is at best, unreliable and at worst, non-existent. Packaged water, but more so, sachet water, has filled the unmet need for an easy and accessible water source. The PMA2020 surveys identified high packaged water use in five study areas with NGL recording the highest use. Analysis of packaged water use in study geographies with both urban and rural samples portrays packaged water as primarily an urban and wealthy consumer product. Based on an analysis of Ghanaian respondents, this relationship held even after controlling for other socio-economic factors.
Even though safety of any water source, including piped supply, is hard to guarantee without strict controls and a monitoring program, packaged water is especially ripe for contamination during production and distribution. A more comprehensive look at the role of sachet water in the larger water supply landscape suggests possible public health benefits. The mixed evidence points to the quandary of whether to place sachet water on the water ladder as an improved source or not. The development of four hypothetical scenarios played on this ambiguity, and demonstrated how classifying a certain source as improved or not plays a critical role in the overall metric that indicates progress at the national and international level.
The decision of DHS to classify sachet water as improved in GH and unimproved in NG signifies the need for country-level decisions on progress indicators that better capture the quality of water infrastructure provided.
At the same time, the DHS decision also suggests that moving sources from one category to another ad hoc can lead to a certain desired outcome. DHS and JMP metrics drive national and institutional donor policies; consequently, care must be taken to ensure that the metrics and classification schema reflect the needs of the people they are designed to help.