Abstract
This research explores the nature and determinants of consumer demand for food safety in Kenya, in particular, demand for aflatoxin-tested maize. Consumption of high levels of aflatoxin can be fatal, and chronic exposure has been linked in numerous studies to liver cancer, suppressed immune response and child stunting. In Kenya, where maize is a staple in the diet, a significant proportion of maize and maize flour samples fail to meet regulatory standards, although these numbers can vary widely by year, season and region. One study by researchers with the CDC found that 65% of maize samples collected from 20 major millers did not meet the national standard (Gathura 2011) and a study using 2013 data found that 26% of maize flour samples were above this standard (Moser and Hoffmann 2015). Because aflatoxin is unobservable and regulatory standards are imperfectly enforced, even informed consumers cannot be confident of the safety of maize or maize flour available in the market. Using a randomized controlled trial, we track the sales and prices of branded maize flour both before and after one of the brands becomes the first in Kenya to have its aflatoxin-testing procedures verified by a third party lab and to use a logo indicating this on its package.
We contribute to the literature in two important ways. First, consumer demand for food safety is not well understood. Because modern regulatory systems keep unsafe food out of the market, consumers do not directly choose safety as an attribute in developed countries. In markets where regulation is poorly in enforced or non-existent, consumers frequently do not have information on food safety. Our project takes advantage of a unique opportunity to study an emerging food safety system in Kenya. Commercial maize millers are beginning to respond to increasing consumer awareness of aflatoxin and pressure from the government by investing in aflatoxin testing. A third-party verification program, which validates a mill's aflatoxin testing processes, recently became available to millers and provides millers a way to signal to consumers that their products are safe.
In terms of preventive health technologies in general, the evidence suggest that consumers' willingness to pay for such investments is low.
Reasons likely include lack of awareness about the effectiveness or cost effectiveness of prevention measures, liquidity and credit constraints, present-biased time preferences, and attention constraints (Dupas, 2011; Spears, 2009; Kremer and Glennerster, 2011). However unlike other health investments that require discrete shifts in one’s use of time, attention or other scarce resources, households already purchase food; purchasing the brand that has been tested requires only a minor change in behavior. Further, the cost of purchasing safe food is a marginal increase to an unavoidable cash outlay; consumers are known to be less price sensitive to additional costs compared to stand-alone costs. Finally, food safety may be perceived to be correlated with other food attributes, such as taste or consistent quality.
Second, we contribute to the understanding of sustained behavior change. Although a pilot study found a high willingness-to-pay for a one-time purchase of aflatoxin-tested maize, it may be that the introduction of tested maize constituted a temporary shock to the salience of aflatoxin as a health risk, and that demand for tested maize will fall as this salience fades. Consumers may also falsely believe that tested maize will be correlated with attributes such as improved taste, and revert to untested maize when this is found not to the case. While several studies test the impact of information on health behavior at a single point in time (Madajewicz et al, 2007; Jalan and Somanathan, 2008; Luoto 2009), we are not aware of work that characterizes the dynamic impact of information. Learning and limited attention models give different predictions about the dynamic impact of information on health behavior. A learning model predicts that information leads to permanent behavior change, whereas a limited attention model, in which information increases the salience of a particular health risk temporarily, predicts that the effect of information diminishes over time (DellaVigna, 2009).