76 hausted by the end of a speci? c period, the remaining funds will be accumulated to cover future medical expenses.71 Since health care costs can be so high that they exceed the ? nancial capacity of the individual, MSAs are usually offered in combination with high-risk health insurance. The bene? ts of the high-risk insurance is limited to the costs of speci? c treatments or takes effect only if a certain yearly deductible paid out of the MSA is exceeded (Nichols et al. 1997). MSAs directly address moral hazard behaviour. Since patients must ? nance parts of their health care expenses out from their own MSAs, i.e. from funds they themselves have saved, a higher degree of cost-consciousness is achieved (Schreyögg 2004, p. 8). Similar to out-of-pocket-payments, MSAs provide incentives for individuals to take responsibility for their own medical needs, resulting in a greater ef? ciency in allocating resources. MSAs seem to be appropriate for resolving, at least in part, the con? icting goals between social policy and economic rationality. From a political point of view, fundamentally transforming the pay-as-you-go oriented SHI into a funding medical care system based on MSAs would be seen as extremely unrealistic, but it can be used as a supplementary instrument, which targeted speci? c weaknesses of the SHI. The pharmaceutical sector in particular seems to be suitable for applying MSAs: ? rstly, out-patient pharmaceutical expenditure can be clearly distinguished from other disbursements for health care services. Secondly, the existence of homogeneous generics and interchangeable analoga enables more extensive price competition than in the market of general medical treatments. Thirdly, several studies show that preventive care and pharmacy are the most price elastic types among health care services.72 Danzon and Pauly (2002) ? nd that the direct moral hazard effect of insurance growth in the United States accounts for between one-fourth and one-half of the increase in drug spending. Hence it can be supposed that increasing the cost-consciousness of patients results in more price competition between pharmaceutical manufacturers. 4.5. Economic Analysis I: Price Equilibrium To my knowledge, as of this writing there has never been an attempt to model MSAs in a theoretical framework. In the same way, most of the literature on RP in pharmaceutical markets is descriptive; but there are a few noteworthy exceptions: Zweifel and Crivelli (1996) used a Bertrand duopoly model to analyse the pricing policy of pharmaceutical companies in Germany before and after the introduction of RP. Danzon and Liu (1997) applied a monopolistic competition model to consider imperfect physician agency. Using a vertical differentiation model with two ? rms, Merino-Castello (2003) 71 Depending on the institutional organization of the insurance system, savings can also be used as old age reserves, e.g. retired people can be liberated from paying the obligatory contribution when their MSA exceeds a speci? c amount. Another possibility would be to give the account holder the right to bequeath savings to his descendants (Schreyögg 2004, p. 5). 72 An overview is given in Ringel et al. (2002). 77 played both Stackelberg and Bertrand competition games. And Königbauer (2006) extended this approach with a combination of horizontal and vertical differentiation and additionally considered market entry and health risks to patients. In the following we try to compare the impact of different insurance arrangements, i.e. coinsurance, RP and MSAs, on the price-setting behaviour of pharmaceutical manufacturers. Implementing different kinds of control instruments, we are forced to use a simple basic model which can be adapted to RP as well as MSAs. In section 4.5.1 we start with a common vertically differentiated market model when sellers differ in quality, such as Metrick and Zeckhauser did it in 1996. Subsequently, like Merino- Castello (2003), we include RP in section 4.5.2. Finally, in section 4.5.3, we introduce MSAs into the theoretical framework. 4.5.1. Basic Model: Coinsurance Metrick and Zeckhauser (1996) showed in a simple vertically differentiated duopoly, the in? uence of products’ quality on prices and the sellers market share.73 With some reinterpretations of the variables used, this model can be easily adjusted to the situation where there is a brand-name producer and its generic competitor in a pharmaceutical market.74 We assume a partial equilibrium framework with two pharmaceutical producers. The ? rst ? rm produces the brand-name drug (B) whose patent protection has already expired. The competitor offers a bioequivalent generic copy (G) of the brand-name drug. There exists a continuum of consumers distributed uniformly by their valuation v on the interval [0,1]. Each consumer can buy either zero or one unit of the good, but if someone wants to buy, then he has always enough money to do so. In case of indifference between buying and not-buying, the consumer buys and he always chooses the brand-name product if he is indifferent between the two drugs. His bene? t from buying one unit from producer m (with m =B, G) is ??m, where ?m is the “perceived” quality for each producer. Although the two drugs are pharmacokineticly and pharmacodynamicly identical, consumers associate high quality with the brand-name drug and low quality with the generic version due to received brand awareness. We assume . The individuals are covered by health insurance with a proportional co-payment rate k. The price of the drug is Pm. Thus the utility function is given by (4.1) 73 Metrick and Zeckhauser (1996) used a standard model of vertical differentiation, quite similar to e.g. Gabszewicz and Thisse (1979). 74 We do this in line with Merino-Castello (2003). 78 if the consumer buys one unit from producer m, and (4.2) otherwise. The consumer buys the brand-name drug if  (4.3) and  . (4.4) On the contrary, he buys the generic version if  (4.5) and  . (4.6) Otherwise he does not buy. Since market entry is excluded, in equilibrium the ? rms can earn pro? ts. We assume ? rm B possesses a ? rst mover advantage due to the period of patent protection of the branded drug in which a high quality reputation has been established and large market shares have been achieved. Brand-awareness and popularity grant ? rm B an advantage even after patent expiry. Therefore, we play a sequential price-setting game (Stackelberg) in which ? rm B is the price leader and ? rm G is the follower. The game is solved by means of backward induction to ? nd the equilibrium level of prices and quantities. Producer m maximises his pro? t ?m, which is a simple price times quantity: ?m = PmQm. We assume no costs of production.75 Firstly, we turn to ? rm G’s pricesetting problem, using conditions (4.5) and (4.6): . (4.7) The ? rst-order condition is . (4.8) Solving for PL gives G’s reaction function . (4.9) 75 Marginal costs of production for drugs are very low, especially compared with research and development costs. Moreover, including a constant marginal-cost production function would not alter the results (Metrick and Zeckhauser 1996, FN 11). 79 Now we turn to B’s price-setting problem using conditions (4.3) and (4.4): . (4.10) Substituting (9) into (10) results in: . (4.11) Finally, maximizing B’s pro? t function delivers his optimal price: (4.12) and quantity: . (4.13) Therefore, ? rm G choose: (4.14) and . (415) 4.5.2. Alternative 1: Reference Pricing Merino-Castello (2003) assumes the reference price, PRP, as a linear function of both branded and generic drug prices: PRP = =PRPG + >PRPB . = and > are exogenous weights. Thus the model is quite ? exible and can be adapted to different price setting schemes. Additionally, a proportional co-payment k is considered. The modi? ed demand functions have the form as follows. The consumer buys the brand-name drug if  (4.16) and  . (4.17)