Understanding consumption and purchase of tap water and diverse tap-alternative drinking water sources in Detroit, Michigan

The causes and consequences of individuals' reliance on alternatives to the tap in the United States (US), including factors driving health and equity disparities, have become relatively well documented in recent years. In most cases, reliance on tap alternatives and distrust of tap water have been treated as synonymous phenomena (Teodoro et al. 2022; Jaffee 2023). Levels of tap water reliance and distrust also vary widely by the data source analyzed, as questions regarding trust, tap water and tap-alternative beverage consumption are not worded consistently across surveys.

On the other hand, nearly regardless of whether tap water reliance or distrust is measured, studies consistently show that certain socioeconomic disadvantage (income, race/ethnicity, sex, education) factors are associated with distrust (Doria et al. 2009; Pierce & Gonzalez 2017; Javidi & Pierce 2018; Rosinger & Young 2020).

Moreover, tap water reliance appears to have fallen in recent years, especially among minoritized populations since the Flint water crisis (Rosinger et al. 2022) as bottled water reliance has risen (Teodoro et al. 2022; Jaffee 2023). Past research shows that households rely on different types of water – and more broadly, all beverages – based, in part, on their perception of the risks associated with the water and other beverage sources available.

There is good reason for heightened risk perception of tap water and reliance on tap-alternatives given the growing recognition of widespread contamination in drinking water in higher income countries, especially the US, which poses real and significant health risks (Meehan et al. 2020; Scanlon et al. 2023). Endemic contaminants such as arsenic (Spaur et al. 2023) have raised concerns; but, public attention by science and media coverage regarding lead Per- and polyfluoroalkyl substances (PFAS) (Andrews & Naidenko 2020) and other emerging contaminants in drinking water have particularly worsened public perception.

At the same time, tap water disuse and distrust are often rendered incorrectly as behavioral and attitudinal symptoms of pure misperception of tap water quality (Family et al. 2019; Wilson et al. 2023). An overemphasis on information deficit models in historically marginalized communities obscures how patterns of inequality, poverty, racism, and coloniality produce rational distrust in tap water (Meehan et al. 2020). Lower reliance on the tap and distrust often results from severe tap water quality and broader public service deficiencies experienced in both the past and present by individuals subjected to economic impoverishment, discrimination, and political marginalization (Abrahams et al. 2000; Anadu & Harding 2000; Fragkou & McEvoy 2016). For these and other reasons, some scholars are arguing for a focus on ‘trustworthiness’ of water providers rather than residents' trust in tap water (Wilson et al. 2023).

Regardless of the cause, reliance on tap water alternatives has direct, adverse consequences for public health and affordability (Ogden et al. 2012; Onufrak et al. 2014; Javidi & Pierce 2018). In particular, individuals' purchase of bottled water as an alternative to tap water involves significant expenditures in terms of money, time, and overall utility. Substituting bottled water for tap water can result in thousands of dollars per year in additional household costs, according to even the most conservative estimates (Javidi & Pierce 2018). Further studies have proven that a perception of tap water as unsafe has health consequences (Ogden et al. 2012), in that individuals more commonly substitute sugary beverages for all types of water, both the tap and tap-alternatives, when they perceive the tap to be unsafe (Onufrak et al. 2014). People who primarily drink bottled water tend to drink less water (Rosinger et al. 2018), and the resulting under-hydration contributes to worse health outcomes, all else equal (Rosinger & Young 2020). Another increasingly well-documented impact of tap water distrust is on mental ill-health, including anxiety and depression (Wutich et al. 2020). For example, the presence of tap water contamination has been associated with both acute and chronic emotional distress in contexts such as Flint, Michigan (Carrera et al. 2019; Sneed et al. 2020; Reuben et al. 2022) and Texas colonias (Vandewalle & Jepson 2015).

Bottled water companies use branding and marketing to benefit from individuals' experience with tap water contamination, the stigmatization of tap water (Brewis et al. 2021) and public concerns about emerging health risks (Wilk 2006). While guidelines of the International Bottled Water Association theoretically impede companies from marketing using direct comparisons to tap water, scholarship on bottled water companies' tactics suggests otherwise (Gleick 2010; Pacheco-Vega 2019).

Sugar-sweetened beverage (SSB) companies, too, use misleading marketing, which indirectly undermines trust in tap water, often over-stating the scientific case against the safety of tap water (Cohen et al. 2015; Moran et al. 2018). SSB reliance leads to well-documented negative health outcomes. The most ubiquitous health impact of high consumption of SSB beverages is degraded dental hygiene due to a higher risk of individuals incurring dental cavities (Ogden et al. 2012). SSBs are also associated with excess weight gain or obesity, type 2 diabetes, and cardiovascular disease (von Philipsborn et al. 2019). These effects are of heightened concern among children in historically marginalized communities, given that they can have life-long impacts (Bleich & Vercammen 2018).

Despite our growing understanding of the prevalence and drivers of tap water reliance and associated distrust, several gaps in knowledge remain. These include low-income, minoritized individuals' consumption of tap-alternative water in relation to different tap-alternative sources of water, which have been studied in much more limited fashion than single-use bottled water (Pierce & Lai 2019; Teodoro et al. 2022). These tap-alternative water sources include single-use bottled water, larger bottles, and packages of water delivered to homes and bought outside the home. Different tap-alternative water sources also have varied health, affordability, and environmental impacts, with single-use bottled water reliance being the worst for household affordability and the environment.

Moreover, the relationship between tap water consumption and tap-alternative water sources, alongside non-water beverage consumption, has largely been assumed as a tradeoff rather than a complement, yet has not been empirically studied. Water source consumption is often characterized as a discrete choice between two different categories of water rather than a potentially complementary and constrained decision-making process of consuming multiple sources simultaneously (Jaffee 2023). Consumption of water from multiple sources, or the mixing and matching of different sources, has not been studied to the same extent in the US as in low- and middle-income country contexts, which has so much literature on the topic that it has undergone systematic review (Daly et al. 2021).

The context for resident perceptions of drinking water quality and reliance on different tap and tap-alternative sources is locally nuanced. Detroit's Water and Sewerage Department (DWSD), which serves our entire study area, has not reported a ‘primary’ health-related violation of the federal Safe Drinking Water Act for at least the 5 years preceding the survey (DWSD 2024). At the same time, active lead service line replacement efforts have been ongoing in the city for several years. Some ‘secondary’ contaminants, such as manganese, may be more likely than primary contaminants to lead to tap water distrust because of their impact on discoloration (Pierce et al. 2019) but may be uniquely challenging to address in Detroit (Newberry & Hughes 2024). Detroit often has levels near exceedance. These factors could contribute to residents' perceptions that drinking water quality is unsafe or a reduced preference for tap water.

The broader relationship, including trust dynamics, between DWSD and city residents is also complicated by adversarial actions related to water shut-offs, especially for low-income, minoritized individuals. Just before the city filed for bankruptcy, >16,000 homes had experienced a water shut-off, which doubled following the filing in 2014. After this period and before the COVID-19 pandemic, there were almost 100,000 shut-offs (Gaber 2019). Shut-offs in Detroit have shown to be more likely to occur in some neighborhoods, particularly those with higher Black/African American composition and lower income (Helderop et al. 2023), which comprise the majority of our individuals surveyed in the current study.

Moreover, there are broader, well-documented failures in ensuring drinking water equity and trust in de-populating, industrialized cities, especially in the region surrounding Detroit (Yang & Faust 2019; Ward et al. 2024). Part, but not all, of these dynamics are linked to the Flint water crisis, cover-up, and scandal, including cross-community organizational solidarity between Flint and Detroit (Howell et al. 2019). While shut-offs are the focus of much water insecurity research in Detroit (and they likely drive tap-alternative water consumption in some cases), other factors that affect water consumption behaviors have been less well-documented. For example, recent evidence suggests that water price hikes resulted from interest rate swaps by the DWSD (Ponder & Omstedt 2022).

Given these gaps in understanding in higher-income countries and the context of potential distrust in Detroit, in this study, we examine several possible relationships between tap water consumption, tap-alternative-alternative water consumption, and non-water beverage consumption using unique 2021 individual data from the StAND project in Detroit (reference redacted). This survey focuses on low-income, largely minoritized households and includes measures of self-reported drinking water consumption patterns. We necessarily rely on a non-representative sample focused on a socioeconomically vulnerable sub-population with expected high levels of tap-alternative water consumption. We focus on evidence regarding the consumption of tap-alternative water sources in relation to tap water, and the degree of reliance on multiple water sources.

This study draws on unique cross-sectional data involving participants in a broader, longitudinal study (reference redacted). We first conducted sampling for the parent study at the neighborhood level (n = 11 neighborhoods defined as 500 m around a park); then, we contacted all residents within the 500-m buffer for potential enrollment. Neighborhoods were selected based on their status as having high vacancy rates and containing a park no longer maintained as a conventional park. These study neighborhoods are located primarily in Detroit City Council Districts 5 and 6 and had a census-recorded median income of $21,057 and a high average percentage of Black residents, 82%.

We restricted recruitment to one English-speaking adult per household who could walk continuously for at least 15 min, as the parent study focused on people's mobility and their neighborhoods. All households within each neighborhood were eligible to participate.

For the current study, we collected data from May to October 2021 (n = 259), including a survey about individual demographics, household characteristics, water usage frequency, types of water consumption, and SSB consumption.

Surveys were conducted either on paper by participants themselves or with a research team member reading and completing the paper survey for the participant in-person, depending on participant preference. Surveys were conducted in compliance with institutional review board standards.

We collected the following measures of water usage and consumption via the survey: the degree of tap water consumption (coded as water1), the degree of overall bottled water consumption (water2), and the frequency of single-use bottled water consumption (water7). Additionally, we collected self-reported data on the frequency of the following other types of tap-alternative-alternative water purchases: a package of bottled water from a grocery or convenience store, a gallon container filled up at a vending machine, a large container from a water store, water delivered to the home in a large container (waters 8–11). Given the relatively low frequency of reported purchase of tap-alternative sources other than single-use bottled waters and their less expensive nature compared to single-use bottled water, we also combined these tap-alternative response data into a variable measuring the number of overall tap-alternative water sources purchased (compilation of variables coded as water 7–11). We then focus our analysis on three major categories: degree of tap water consumption, degree of single-use bottled water consumption, and frequency of tap-alternative water sources consumed.

We also collected data on and coded the frequency of SSB consumption. The exact question wording and response categories for each of these variables used in the study are recorded in Table 6 in Appendix 1.

Additionally, we collected demographic data information (age, sex, race/ethnicity, household size, income) and housing data (length of time lived in home, single family home) via the individual survey. Race/ethnicity categories were based on the US census designations. Single family home ownership is a strong proxy for control over housing quality and tenancy, including premise plumbing condition, which can affect tap water quality and perception, and thus was included in our regression models (Pierce et al. 2019; Newberry & Hughes 2024).

We used Stata v16.0 for all descriptive and multivariable statistical analyses. We employed ordinal logit regression for our first two outcomes of interest, categorical degrees of tap and single-use bottled water consumption, respectively. We used Poisson regression for the final modeled outcome of interest, the number of tap-alternative sources consumed. Our outcomes of interest are all accounted for in terms of the past month of recall by the individual:

• 1. Degree of daily tap water consumption (Ordinal logit model).

• 2. Degree of daily single-use bottled water consumption (Ordinal logit model).

• 3. Frequency of unique tap-alternative water sources purchased (Poisson model).

In all models, we adjusted for factors that previous studies have shown may correlate to tap water consumption and tap-alternative consumption levels, as well as tap water trust. First, we controlled for the race, income, education level, sex, and age of the respondent. We also adjusted for household size, housing type and tenure, and SSB consumption levels of the individual respondent.

Our analysis included 259 individual respondents across unique households; however, some respondents did not provide responses to all variables. Considering low levels of non-response for some key variables, this limited our regression analysis to final sample sizes between n = 158 and n = 167, depending on the regression specification.

While this presents potential concerns for biased non-response, we find that individuals from households surveyed but omitted from our regression analysis were not systematically different from included households by water source consumption, race, education level, age, or sex. Moreover, as noted above, we necessarily rely on a non-representative sample because the original survey was broadly focused on socioeconomically vulnerable populations, their trust in institutions, and levels of social cohesion and associated health behaviors.

We also tested the sensitivity of our results to model type specification. We found using ordinary least squares regression and Poisson or ordinal logit frameworks as alternatives to the preferred model type for each outcome of interest makes little to no material difference in results. Given that there is no convention in the literature, we additionally explored several other constructions of tap-alternative water source consumption as an outcome of interest, with results largely consistent with our preferred model structures.

We first present descriptive statistics for key variables, summarized in Table 5 in Appendix 1. We sort these in terms of whether individuals did or did not report consuming tap water as their primary drinking water source at home. We focus our analysis, however, on three primary outcomes of interest that measure the degree of and relationship between tap and tap-alternative water consumption: the level of daily tap water consumption, the level of daily individual bottled water consumption, and the diversity of tap-alternative water sources relied upon by individuals.

Overall as seen in Table 1, individual tap water and tap-alternative water consumption patterns are mixed in terms of primary source reliance, and high and low degrees of reliance. Fifty-seven percent of all survey respondents report primarily relying on tap water for drinking water at home. In terms of degrees of consumption reliance on different drinking water sources, at the high end, about 32% of all respondents reported drinking tap water five or more times at home daily, versus 31% drinking any type of bottled water five or more times daily. At the low end of consumption, 26% of individuals surveyed reported drinking tap water zero times daily, versus 21% reporting that they do not drink any type of bottled water daily.

Respondents also reported a fair degree of unique tap-alternative water sources purchased. Two-thirds (67%) of respondents reported that they purchased single-use bottled water, the most expensive kind of tap-alternative source, more than once a month, with 13% reporting that they relied upon it every day.

Generally, however, the reported diversity in water source consumption was high. Some respondents reported up to five different tap-alternative water sources they routinely used, and the average number of tap-alternative sources relied on was 1.81 (range = 0–5). A substantial portion of respondents, 21%, indicated that they never bought tap-alternative water within the past month, whereas 11% relied upon all five tap-alternative sources at least once in the past month.

Given these trends, we used multivariable regression analyses to examine factors correlated with tap and tap-alternative water consumption. We tested the relationship between tap and bottled water consumption, including single-use bottled water consumption and multiple tap-alternative water source consumption, while holding other factors associated with tap trust and water sources constant. While we explored other model outputs as variants of the above specifications by changing dependent variable definitions slightly, we report only our models of the three primary outcomes of interest here. Other regression results are available upon request.

Overall, multivariable regression results suggest drinking tap water and drinking other tap-alternative water sources were positively correlated. This relationship does not appear to be an artifact of modeling choice or construction given that we also found tap water consumption and bottled water consumption to be positively associated in binary tests, albeit not at statistically significant levels. The tap water consumption model has low explanatory power overall (Pseudo R² = 0.033), but this level is not unusually low for individual decision-making models.

First, when we model tap water consumption levels (Table 2), we find that overall bottled water consumption is positively associated with higher tap water consumption. However, the statistical significance level is just outside the 90% confidence level. In other words, we find that if an individual drinks tap water, that individual may also rely on other tap-alternative water sources. Black/African American respondents and female respondents were significantly less likely to consume tap water overall. A longer tenure in the neighborhood was positively associated with tap water consumption. Other control variables found to be statistically insignificant were education, housing type, income, and SSB consumption.

Second, we examined single-use bottled water consumption frequency as an outcome of interest (Table 3). We found that overall bottled water consumption frequency was significantly positively correlated with single-use bottle consumption but negatively correlated with tap water use. However, this relationship was not statistically significant. The lowest-income residents and females surveyed were significantly more likely to consume single-use bottled water than other respondents.

We also found that household size was positively correlated and age was negatively correlated to single-use bottled water consumption frequency, the latter perhaps suggesting a broader generational trend in bottled versus tap water consumption. Other control variables were found to be statistically insignificant in relation to single-use bottled water consumption levels.

Finally, we explored the use of multiple tap-alternative water sources simultaneously as an outcome of interest (see Table 4). Again, we found that bottled water consumption frequency was positively associated with the number of tap-alternative sources, and that consuming multiple tap-alternative sources was negatively correlated with tap water consumption frequency. We also found that Black/African American and lower-income respondents were more likely to rely on more tap-alternative sources. Interestingly, we found that females were less likely to use multiple tap-alternative water sources simultaneously, whereas single-family households are more likely to do so. Other control variables did not have meaningful explanatory power in this model specification.

Overall, our study's findings corroborate some existing knowledge in the literature on tap water consumption and its relationship to tap-alternative water sources and broader beverage consumption, while challenging other common wisdom. Our work illustrates the complexity in understanding tap water consumption's relationships to tap-alternative water consumption for low-income minoritized individuals, as well as other beverage consumption outcomes. First, many individuals consume up to five tap-alternative water sources daily. In other words, they drink a diversity of tap-alternative water sources, rather than there being a binary divide between tap water and single-use bottled water (Rosinger et al. 2018). Contrasting with conventional assumptions and research designs in the most-cited drinking water source reliance and trust scholarship, we find that tap and overall bottled water consumption levels are positively related (Doria 2006; Teillet et al. 2010). Also, contrary to assumptions in much of the literature, SSB consumption was not found to substitute for or tradeoff with water source consumption of any kind (Leung et al. 2018). Yet, consistent with other studies' assumptions, however, tap water consumption was negatively correlated with both single-use bottled water consumption and multiple tap-alternative water source consumption (Rosinger & Young 2020).¹

We found that female sex was consistently negatively associated with each of our outcomes of interest: tap water consumption, single-use bottled water consumption, and multiple tap-alternative water source consumption. Female respondents were also less likely to use the tap as their primary source of drinking water and to consume fewer tap-alternative water sources and SSBs. This may suggest that females drink fewer beverages overall, in addition to drinking less tap water due to higher distrust (Javidi & Pierce 2018), but this relationship should be explored further.

Identifying as Black or African American in the survey was negatively associated with tap water consumption and single-use bottled water consumption but positively related to the degree of unique, tap-alternative water source consumption. On the other hand, very low-income (less than $10,000 per year) individuals reported higher levels of single-use bottled water consumption and the number of tap-alternative water sources consumed (Jaffee 2023). By contrast, several other correlates of tap-alternative water source consumption, tap water distrust, and bottled water consumption previously identified in the literature, such as education and housing type, did not have meaningful explanatory power in any of our analyses (Pierce et al. 2019).

All tap-alternative sources cost several multiples of the per-unit price of tap water to purchase. However, single-use bottled water is the most expensive per-unit, often hundreds of times more so than tap water (Gleick 2010; Teodoro et al. 2022). Considering our study population had low annual incomes on average, those individuals relying on tap-alternatives exhibited severe out-of-pocket affordability concerns which need further exploration. Moreover, the environmental impact of all tap-alternative water sources is worse than tap water due to the pollution caused by production but especially the incomplete recycling of packaging (Wang et al. 2019) and microplastic dispersion (Gambino et al. 2022). For these reasons, from the environmental dimension, single-used bottled water has the worst impact (Gleick 2010; Pacheco-Vega 2019).

The health impacts of reliance on tap-alternative water sources are more nuanced. As noted above, there is good reason and often necessity for individuals to rely on tap-water alternatives. People who primarily drink bottled water tend to drink less water (Rosinger et al. 2018), and the resulting under-hydration leads to negative health outcomes (Rosinger & Young 2020). That being said, tap-water alternatives are not necessarily safe, or safer than the tap, including single-used bottled water (Raj 2005). Packaged and large-vended water is particularly prone to microbial contamination and rare disease outbreaks (Pierce & Lai 2019; Hile et al. 2023).

Our most concerning finding from a social policy perspective is the high levels of reported single-use bottled consumption among very low-income individuals, which are likely to cause outsized financial burdens (Javidi & Pierce 2018). Any future policy interventions must avoid information deficit approaches which are insensitive to the experiences and decision-making rationales of individuals relying on tap-alternatives and are likely to prove ineffective. More research is needed to inform how to target effective support to sub-groups reliant on tap-alternatives, including very low-income, Black or African American, and female individuals. Overall, we stress that messaging interventions must better account for the experiences, knowledge, values, priorities, goals, rationales, and visions of the target populations.

In addition to affordability, from a public health perspective, it is essential to better understand the drivers of low overall water and beverage consumption among females. Learning more can, in some cases, inform the implementation of ways to support safe tap reliance and thus enable less expensive and environmentally-damaging sources of consumption. An experiment in Detroit to provide detailed, user-friendly at-the-tap water quality information regarding individuals' tap water using point-of-use sensors is promising in influencing consumption (Schubert et al. 2024). Especially in cases where there is a high likelihood of a lack of safe tap water provision, increased monthly financial support such as the Supplemental Nutrition Assistance Program to purchase bottled water may also be an option. This approach has been tried regionally in a pilot program in the San Joaquin Valley of California (Nourish California n.d.). These types of programs may also further consider point-of-use tap water filtration devices. However, such devices present maintenance and cost concerns unless funding and training on filter cartridge replacement are provided (Gilliland et al. 2024).

There are also several limitations of our study which deserve future exploration. First, using surveys is inherently a limited approach to analyze phenomena as complex as water-use practices and water-related beliefs. For example, the question ‘Do you use tap water at your residence as your main source of drinking water?’ does not leave room for a detailed explanation about if and when respondents use tap water as their primary source of drinking water. We thus acknowledge the limitations of cross-sectional survey research for understanding people's water-use practices and note that the likely context-specific variability informing such practices over time would be important in future studies. The survey also has a relatively small sample size and is not a representative sample of the broader population, so generalizability is limited.

Moreover, categorical rather than linear outcomes in survey questions for different beverage source consumption levels were somewhat limiting. These response constructions were designed and chosen based on previous experience with respondents having trouble recalling exact purchase and consumption levels, but they still yielded imperfect responses. The lengthy recall period of 1 month for the purchase of tap-water source alternatives also clearly had limitations. However, it was again based on our previous experience and experimentation with recall periods in previous surveys (Young et al. 2019).

Finally, and crucially for understanding and designing policies to address household financial burdens in more detail, we attempted to collect detailed beverage expenditure data based on consumption levels. However, respondents did not respond to these questions often or consistently; thus, we could not include the results in our analysis, and quantifying exact levels of beverage expenditures across tap-alternative sources remains an avenue for future research.

This study was funded by the National Cancer Institute (NCI) of the National Institutes of Health (NIH) 1R01CA239187–01, the Detroit Medical Center, Michigan State University's Clinical Translational Science Initiative, the Vice President for Research and Graduate Studies and the Provost Undergraduate Research Initiative.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.