This is the twelfth in our series of posts by students on the job market this year
Recent work has suggested that as many as one-third of antimalarial drugs in sub-Saharan Africa are of low-quality, a catch-all term ranging from effective counterfeit medicines to dangerous “fakes” (Nayyar et al., 2012). The persistence of low drug quality may be attributable to asymmetric information (Akerlof, 1970). Patients do not know their need for treatment, or the drug quality at the time of the purchase. In order to maximize profits, providers may then substitute cheaper, lower quality drugs. Bjorkman et al., (2012) find that fake drugs are particularly common in areas with low levels of customer knowledge about malaria transmission, where customers are potentially easier to deceive. However, the only intervention shown to reduce counterfeit drugs is the introduction of a high-quality competitor (Bjorkman et al., 2012; Bennett and Yin, 2014). Might increased customer information about purchases cause suppliers to improve their drug quality?
I address this question in my job market paper. I implement a randomized audit study in Uganda to measure how suppliers adjust price and quality if customers knew what disease the patient had (i.e., “diagnosis”) or knew the particular drug to buy. I contrast the response of drug quality with service quality, which is also low in developing countries (Das and Hammer, 2014). I find that price falls when customers present more information. Counter-intuitively, I find that both service and drug quality fall when the customer presents more information.
Design and Empirical Strategy
The study takes place in 5 districts of Uganda (including Kampala). I randomly select parishes within study districts and conduct a census of all private sector drug outlets within the parish (N=540). Drug outlets consist primarily of drug shops (53%), clinics (39%), and pharmacies (7%). These establishments are small, informal microenterprises; 74 percent have 2 or fewer workers.
Pairs of covert shoppers purchase antimalarial drugs at each outlet according to randomly assigned scripts. The scripts vary knowledge a customer could present during the shopping experience: 1) the disease the patient had and/or 2) the product they requested for the disease. In this set-up, the “control” script asked for both a diagnosis and for a drug recommendation. The three “treatment” scripts identify the effect of a customer who presents information of only the disease (malaria); only the drug they want (artemether-lumefantrine, or “AL”); or both. I standardize shopper appearance, answers to common questions, etc. All purchased drugs were tested for quality using a handheld spectrometer (N=879). The shopper data is then linked with surveys of vendors and real customers at the same outlets.
Results
Do providers respond differently to customers who know the ailment (malaria) versus those who know the appropriate drug (AL)? In general, no. Stating the patient has malaria causes the same response as requesting a specific drug. Therefore, I combine treatment groups, and present results contrasting the Any Information “Treatment” versus the “Control” group (the group who asks for both a diagnosis and a drug recommendation).
First, let’s discuss the results on price. I find that the drug price falls by 5 percent ($0.18) when customers present any knowledge about appropriate diagnosis or treatment, although some estimates are noisy.
Next, I measure the effects on drug quality in two ways, “counterfeit” and “substandard”. A counterfeit drug is chemically different from what brand the drug is labeled as, but still may be effective. In contrast, a substandard drug is chemically different from the brand it is labeled as and all other brands. Therefore, substandard drugs are less likely to be effective. Overall, 17.1 percent of purchases have at least one counterfeit tablet; 3.6 percent of purchases have at least one substandard tablet. Providers are 3.4 percentage points more likely to sell a substandard drug when customers reveal knowledge of correct diagnosis and treatment of malaria (p=0.004).
Finally, I turn to the effect of improved customer information on service quality. I measure service quality using a “checklist” following official guidelines, such as whether the customer was advised to have the patient take a malaria test (Leonard, Masatu, and Vialou, 2005; Das and Hammer, 2014). The likelihood the provider advises a malaria test falls by 6.8 percentage points (from a base of 41 percent) when customers know either the ailment or the correct drug. Although only 53 percent of outlets sell malaria tests, the likelihood of advising a malaria test does not differ according to whether the outlet actually sells tests.
Interpretation
I interpret results through a stylized model of price discrimination. Providers optimally choose price, effort (“service quality”), and whether to sell a high-quality (“good”) drug or a low-quality (“bad”) drug. Providers consider both the current and future profits of their choices. The quality of the drug is unknown at the time of purchase, but is (perfectly) revealed after the customer uses the drug – the patient either gets better or not. In contrast, service quality is observable when the customer chooses to purchase the drug.
There are two types of customers: informed and uninformed. Customer type is common knowledge. Customers choose to either buy the drug, or reject the drug offered and seek treatment elsewhere. If a customer gets a bad drug then they never return; if the customer gets a good drug, then they return with some likelihood that differs by type. In this model, “information” does not help customers identify good drugs from bad drugs. However, information is correlated with other characteristics of demand: specifically the likelihood of return business. It’s the provider’s beliefs about the customer returning which ultimately drive the dynamics of the model.
Providers optimally increase both price and service quality to just keep customers indifferent between agreeing to buy the drug and buying the drug elsewhere. Thus, increases in service quality follow directly from increases in price.
The key mechanism driving drug quality is different. The main penalty from a selling a bad drug is that the customer won’t return. Providers trade off the current benefits of selling a bad drug with the lost future profits. Therefore, customers who pay lower prices and generally are less likely to return are the “low penalty” customers. Data from real customers suggest that antimalarial customers with relatively more information do differ in their likelihood of returning to a given provider: more informed customers shop at more outlets (i.e., are less loyal customers), and are less likely to seek healthcare overall, therefore the “penalty” an outlet faces from selling counterfeit or sub-standard drugs to these customers is relatively low.
Limitations
Results are specific to the setting and type of information in this study. For example, none of my shoppers visited the same store twice or were known to the vendor. In addition, the effect of a widespread information campaign may be different from the marginal effect that I identify.
Conclusion and policy implications
My results do not suggest that customers should be prevented from learning information. On the margin, information decreases prices of antimalarial drugs: a policy goal to increase access through lower prices. However, customers face a trade-off. Information appears to reduce prices, but also to reduce quality. My framework implies that while service quality is “priced” into the drug, drug quality is not. In line with existing work by Das and Hammer (2014), service quality falls in part because providers decrease effort when prices fall. However, the provider choice of drug quality is driven by the penalties if the behavior was detected (implying that lower prices may not result in lower drug quality if customers could better identify fake drugs). I conclude that empowering customers with information relevant to their purchases should be used with other policies to improve drug and service quality in developing countries and may not be sufficient on its own.
Anne Fitzpatrick is a PhD Candidate at the University of Michigan. She is currently on the job market.
Join the Conversation