Water Quality Violations and Avoidance Behavior: Evidence from Bottled Water Consumption

100 words) In this paper, we examine the impact of poor water quality on avoidance behavior by estimating the change in bottled water purchases in response to drinking water violations. Using data from a national grocery chain matched with water quality violations, we find an increase in bottled water sales of 22 percent from violations due to microorganisms and 17 percent from violations due to elements and chemicals. Back-of-the envelope calculations yield costs of avoidance behavior at roughly $60 million for all nationwide violations in 2005, which likely reflects a significant understatement of the total willingness to pay to eliminate violations. Water Quality Violations and Avoidance Behavior: Evidence from Bottled Water Consumption Joshua Graff Zivin, Matthew Neidell, Wolfram Schlenker The provision and public dissemination of information about health hazards has become an increasingly important part of state and federal programs designed to manage environmental and health risks. Examples include the US Environmental Protection Agency’s (EPA) Toxics Release Inventory, the US Food and Drug Administration’s fish advisories, and California’s ‘smog alerts’ program. The central idea behind each of these programs is that the provision of information allows the public to engage in behavioral responses to minimize exposure should the costs of that exposure exceed its benefits. Whether such information is a substitute or complement to environmental standards is an open question. In either case, understanding responses to such informational approaches is critical for determining both the costs and the effectiveness of these programs. One important area where informational approaches play a key role is in the management of drinking water quality. The EPA, under the auspices of the Safe Drinking Water Act (SDWA), places strict limits on roughly 90 chemicals or contaminants in community drinking water systems, which is accessed by nearly 270 million people in the United States. Despite these limits, roughly one in ten Americans is served by a drinking water system that exceeds these limits on at least one dimension (Duhigg, 2009). Such violations must be disclosed to consumers under the SDWA Amendments of 1996. This paper examines avoidance behavior in response to these disclosures regarding drinking water violations. Matching geocoded violations data for Northern California and Nevada from 2001-2005 We would like to thank Sofia Villa-Boas and Kristin Kiesel for helping us obtain the bottled water sales data. We thank Nicolas Alvear and Charlene Lee for excellent research assistance. The usual disclaimer applies. with sales data from a major supermarket chain, we estimate the change in bottled water purchases as a result of tap water violations. Since the behavioral response in this case is a market-based one, it is straightforward to calculate the costs of avoidance behavior. We find a statistically significant increase in bottled water sales of 22 percent from violations due to microorganisms and 17 percent from violations due to elements and chemicals. Combining these store-level estimates from Northern California and Nevada with national sales data on bottled water consumption, we compute back-of-the-envelope costs of avoidance behavior at roughly $60 million for all violations in 2005, noting this figure likely reflects a significant understatement of the total costs of avoidance behavior, and thus willingness to pay to eliminate violations. I. Data The SDWA Amendments of 1996 require 150,000 community water districts (CWDs) throughout the US to actively monitor contaminants levels. If contaminants exceed maximum contaminant level (MCL) standards, a CWD must notify the EPA of the violation and adhere to the Public Notification Rule. This rule requires CWDs to notify customers within 24-hours if the responsible contaminant poses an immediate health threat (primarily microorganisms and nitrates) and within 30 days for other health threats. CWDs must notify customers through various social media outlets, posting in public places, and personal delivery. Relevant for our analysis, these notifications must include a description of the violation and potential health effects, the population at risk, actions consumers can take, when the violation occurred, when a resolution is expected, and language encouraging broader distribution of information regarding the violation. Nitrates and nitrites violations result in reductions of similar magnitude, but are not statistically significant. 3 Unfortunately for our analysis, we do not have data on the exact details of the notification provided by districts. The Safe Drinking Water Information System (SDWIS) maintained by the EPA contains detailed records of all violations. After filing a Freedom of Information Act request, we obtained historical information on all violations that occurred in the US from 2001-2005. These data include the start and end date of the violation, the contaminant responsible for the violation, as well as characteristics about the CWD, including the county and population served. We contacted all water districts in California and Nevada that serve at least 100 people and obtain the zip codes they serve. If a water district serves more than one zip code, we split the population served between zip codes based on their overall population. For example, if a water district serves customers in two zip codes, where zip code 1 has twice the population of zip code 2, we assigned two-thirds of the population served to zip code 1 and one-third to zip code 2. We combine individual violations into three broadly defined groups based on their potential health effects, which is also consistent with the public notification rule. “Microorganisms” pose immediate gastrointestinal health threats to all individuals; this largely consists of coliform bacteria and can be removed by boiling tap water. “Nitrates” pose immediate threat of “blue-baby syndrome” to infants and can not be removed by boiling. “Elements/chemicals,” which includes natural occurring elements, such as arsenic, manufacturing chemicals, such as tetrachloroethylene, and disinfection byproducts from removing microorganisms, also can not be removed by boiling. The health effects from elements/chemicals, which include cancer and toxicity to various organs, typically arise from longer-term exposure and do not require immediate notification. Table 1 displays the number of violations for each group along with the mean duration of each violation for both the US and our final sample for the years 2001-2005. Our data on bottled water consumption uses weekly sales (Wednesday-Tuesday) from a national grocery chain for their stores in Northern California and Nevada for weeks starting October 31, 2001 until November 2, 2005. The data includes sales in dollars as well as quantity sold for 308 Universal Product Codes (UPC). These UPCs cover still water (e.g., Aquafina, Dasani), sparkling water (e.g., Pellegrino, Perrier), and flavored sparkling water (e.g., Calistoga lemon flavored sparkling water). Different sizes of the same product (e.g., 16oz versus 1 gallon) have distinct UPCs, although the size of a bottle is unfortunately not identified in the UPC database for most codes. Aggregating quantities is complicated by the fact that an increase in demand might be met by switching from smaller to larger bottles while holding the number of units sold constant. Thus, we aggregate sales in dollars for all 308 UPCs by store and week as our dependent variable. Store level sales were linked to water violations by matching water districts with stores that are located in the zip code that is served by the district. Our baseline model uses zip-codes, and not a distance measure, to match water districts to zip codes. Such an algorithm better captures the relevant customer base of a store because zip codes are much larger in rural areas and customers may drive further to reach a store than customers in urban areas. Table 2 displays community characteristics by violation frequency. We define “high (low) violations” as being above (below) the median number of violations for the time period studied. In areas with more violations, residents on average consume less bottled water and come from lower socio-economic status (SES). This pattern is consistent with poorer provision of public goods in lower SES areas and that bottled water consumption is a normal good. Both indicate the importance of accounting for the endogeneity of violations. II. Methods In a standard utility maximization setup, we can think of the demand for bottled water as a function of the price of bottled water, the price of tap water, the price of substitutes (all qualityadjusted), income, and both time-varying and time-invariant individual preferences. If a violation occurs, the quality-adjusted price of tap water increases, thus increasing the demand for