The high price of many pharmaceuticals relative to the rest of the world strikes a raw nerve in the US. The title of a recent New York Times article (Rosenberg 2018) asks, “HIV Drugs Cost $75 in Africa and $39,000 in the US. Does It Matter?” A Council of Economic Advisers white paper (US Council of Economic Advisers 2018) complains that the US “conducts and finances much of the biopharmaceutical innovation that the world depends on”. while “other nations are free-riding”.
Economists generally favour low prices, but not always. Allowing firms to price discriminate – charging higher prices to consumers identified as having a higher willingness to pay – can sometimes increase output and social welfare relative to uniform pricing. Forcing the firm to charge the same price for HIV pharmaceuticals in rich and poor countries may lead to the perverse outcome in which the firm targets rich countries with a high uniform price, completely cutting out the poor. Possibly even more important than the ex post effect of expanding sales of an existing product is the effect on ex ante incentives – the anticipation of higher profits from price discrimination may lead firms to ramp up R&D investments. A ban on price discrimination could lead to a reduction in R&D and consequently a reduction in the chance that life-saving pharmaceuticals are invented.
In a recent paper, we study the welfare effects of a ban on price discrimination (among other policies such as subsidies and reference pricing) (Kremer and Snyder 2018). As a case study, we focus on the global market for HIV pharmaceuticals.
The typical approach taken by industrial organisation economists is to structurally estimate consumer demand and compare welfare from various policy counterfactuals. A different approach is needed for HIV drugs and vaccines. The market for HIV drugs involves enough government and insurer intervention that it can hardly be called private, while an HIV vaccine has yet to be invented. Our alternative approach was to input readily available historical data on population, per-capita income, and disease prevalence into a simple model of country demand and then build up from country to global demand.
We assume consumers are homogeneous within each country (Kremer and Snyder 2018). The number of potential consumers in country i equals its population, Ni. Each contracts the disease with probability equal to the disease prevalence in the country, Xi. Each is willing to pay up to per-capita income, Yi, to avoid losing a disability adjusted life year to HIV, equivalent to setting the elasticity of healthcare expenditure with respect to income to one. We explored a range of different elasticities in robustness checks.
Modelling consumers as homogeneous within countries would seem to be a gross oversimplification. After all, health and wealth can vary quite widely within a country. However, the model may not be too unrealistic if between-country variance of these variables is greater than within-country. The model could fit quite well if governments, insurers, or other large intermediaries bargain over bulk purchases of pharmaceuticals on behalf of a cross section of a country’s population.
To capture demand in the ‘state of nature,’ before widespread use of antiretroviral therapies (ARTs) and other interventions checked the epidemic, we went back to 2003 data. As Figure 1 shows, the local peak in global HIV prevalence was reached in 2003; ART coverage was still negligible then. Figures 2 and 3 present the global distribution of HIV prevalence and GDP per capita (our measure of income) in 2003. The figures underscore the devastating concentration of both HIV and poverty in sub-Saharan Africa. The US exhibits a unique combination of high HIV prevalence and high income. (Figures 1-3 are based on UNAIDS and World Bank data; see Kremer and Snyder 2018 for details).
Figure 1 Trends in global HIV prevalence and treatment
Figure 2 Global distribution of HIV prevalence in 2003
Figure 3 Global distribution of GDP per capita in 2003
The model accommodates general values for marginal production costs, side effects, and failure rates; but we will set them to zero for simplicity here.
The market for an HIV drug in country i consists of individuals who contract the disease. The number of HIV cases in country i equals its population, Ni, times disease risk, Xi. The demand curve is thus a step function with Ni Xi consumers each willing to pay Yi. Combining individual country demands yields the global demand for an HIV drug shown in the top panel of Figure 4.
To compute the profit-maximising uniform price charged by a monopolist, the revenue (also profit given costless production) earned at every possible price point can be compared and the maximum selected. The profit-maximising price turns out to be $27,400 (all monetary values in 2003 US dollars). At this price, consumers in Italy and 17 other higher-income countries purchase the drug. Though the combined population of these countries is substantial, they have relatively low HIV prevalence rates, so only 1.4 million consumers end up being served – only 4% of the global infected population. Profit, the area of the shaded rectangle inscribed below the equilibrium point, is $38 billion.
Figure 4 Global demand calibrations
Goodness of fit
Though simple, the calibration matches actual ART price and quantity quite closely. Freedberg et al. (2001) estimate that actual expenditure per disability adjusted life year in developed countries was around $23,000 at that time. Accounting for price concessions in response to public pressure described by Reich and Bery (2005), our $27,400 estimate could be quite close to the counterfactual expenditure in the absence of public pressure.
The calibrated quantity of 1.4 million, computed from Figure 1 as the reported ART coverage in 2003 (4%) times HIV prevalence (0.5%) times global population (6.4 billion), is quite close to the actual quantity of 1.3 million.
The model can likewise be used to construct a demand curve for a yet-to-be invented HIV vaccine. Individuals in country i are willing to pay up to the expected value of avoided harm, Xi Yi, for an HIV vaccine (assuming they are risk neutral; risk-averse consumers would pay more). The demand curve is thus a step function with Ni consumers each willing to pay Xi Yi. Combining individual country demands yields the global demand for an HIV vaccine shown in the lower panel of Figure 4. (Axes have been scaled so a unit of area represents the same surplus as in the top panel.)
The profit-maximising price is $130, inducing consumers in the US as well as Botswana, South Africa, Swaziland, Bahamas, Namibia, Trinidad and Tobago, and Gabon to purchase. While these other countries are poorer than the US, their extremely high HIV rates generate higher vaccine demand there. The vaccine is only purchased by 6% of the global population. Profit is $45 billion—slightly higher than that for an HIV drug. Given that there is no HIV vaccine, there are no actual outcomes to assess the calibration’s goodness of fit.
The calibrations can be used to assess the welfare consequences of various government policies. We study many policies (Kremer and Snyder 2018), but we will consider just two here: a ban on price discrimination and a government subsidy.
Banning price discrimination
Start by considering a ban on price discrimination in the market for an HIV drug. We already saw that a monopolist forced to charge a uniform price would charge $130, generating $45 billion in profit and social welfare.
By contrast, a price-discriminating monopolist would charge the price equal to consumers’ willingness to pay in each country, appropriating the whole area under the demand curve as profit. Price discrimination generates monopoly profit of $87 billion. This is also the level of social welfare generated since the firm appropriates all consumer surplus. Banning price discrimination increases consumer surplus in the market for an HIV drug by $14 billion but reduces social welfare by $35 billion.
Banning price discrimination creates an additional distortion in the market for an HIV vaccine because the product does not yet exist. The market becomes less lucrative, dulling R&D incentives. The deadweight loss from this source depends on R&D costs but according to our calculations may be as much as $60 billion.
Consider a second policy—a government subsidy in the market for an HIV drug. Given the peculiar shape of the drug demand curve, inscribed rectangles at various price points generate similar areas and similar profits, which leads the monopolist to be nearly indifferent among a variety of prices. In this situation, a small subsidy may be enough to tip this indifference in the direction of dramatically higher quantities and lower prices. For example, it can be shown that a $950 subsidy would lead the monopolist to cut its profit-maximising uniform price for a drug from $27,400 to $265.
While a $950 subsidy may seem high, it is only a fraction of the initial drug price. It results in the share of consumers served increasing from 6% of the infected population to 89%. Such a subsidy would be socially efficient even if the tax used to fund the subsidy were quite distortionary.
Lacking structural demand estimates (because the product has not been invented yet, among other reasons), we turn to demand calibrations to analyse deadweight loss and the potential welfare effects of various policies in the market for HIV pharmaceuticals. The peculiar shape of the global demand curve generates potentially large deadweight loss that may be exacerbated by a ban on price discrimination but ameliorated by a small government subsidy.
Freedberg K A, et al. (2001), “The cost effectiveness of combination antiretroviral therapy for HIV disease,” New England Journal of Medicine 344: 824–831.
Kremer M and C M Snyder (2018), “Preventives versus treatments redux: Tighter bounds on distortions in innovation incentives with an application to the global demand for HIV pharmaceuticals,” Review of Industrial Organization 53: 235–273.
Reich, M R and P Bery (2005), “Expanding global access to ARVs: The challenges of prices and patents,” in K H Mayer and H F Pizer (eds), The AIDS Pandemic: Impact on Science and Society, New York: Academic Press.
Rosenberg, T (2018), “HIV drugs cost $75 in Africa and $39,000 in the US. Does it matter?” New York Times, 18 September.
US Council of Economic Advisors (2018), “Reforming biopharmaceutical pricing at home and abroad,” Washington, DC: Executive Office of the President of the United States.