Universal service obligation (USO) refers to the situation in which an incumbent must provide the same service to all consumers in a market at the same price. Examples include postal services or even health insurance.
The problem is of course that there are consumers in the market who are expensive to serve. This includes inhabiants of rural towns for a postal company or people with pre-existing conditions for a health insurance company. If the government doesn’t want anyone in the country to have no access to the service, it can require the supplier(s) to give the same service to everyone at the same price, this is universal service obligation.
In European countries, there used to be state monopolies on many network industries, specifically think of postal services. But the European Union did not like this and required countries to liberalize some of these markets. The result was often that the original state-owned company remained (either as private or state-owned, but definitely not subsidized), and entry into the market was made possible. The incumbent at first dominated the market and it still often has USO. New entrants, however, do not.
Many countries for instance have the remnants of old state-owned postal services that are mainly monopolistic when it comes to delvierying mail, but that have to compete with a bunch of courier companies on package delivery. Now USO does not apply to these courier companies, whereas it does to the incumbent.
So in a sense these markets have a dual structure: the EU introduced competition, which can only improve things. But at the same time, they do not want anyone to go without postal services. Therefore, the USO of the incumbent has to stay intact.
Is this good or bad for the market share and the profits of the incumbent? Should the incumbent be necessarily compensated for USO? This is what Bakhtieva and Kiljanski (2013) ask in their paper. They use agent-based modeling to answer the question. They set up a country with urban and rural areas where there is one incumbent with USO and one new entrant without it. The main difference is that the incumbent is obliged to serve both urban and rural areas for a price x. The new entrant, however, can charge different prices, which are assumed to be u < x for the urban area and r > x for the rural area such that (1-m)u + mr = x, where m is the “market share” of rural areas.
The dynamics of the model are as follows: firms adjust their prices once every five period using trial and error. I.e. if a change decreases their profits, they reverse the direction of the change. Consumers meanwhile are assumed to be loyal consumers. They calculate the mean utility they derived from each firm in the past y periods. Then the difference of these two utilities determines a probability p. This probability p tells us the likelihood that the consumer will switch to the other supplier. Consumers are therefore less likely to switch when their current choice supplies them with a high utility and vice versa.
Given y = 10 and m = 0.5, the situation not so surprisingly is that both the incumbent and the entrant tend to a market share of 50%. Since m = 0.5, the market share of the rural area is 50%. Most of the rural market is therefore captured by the incumbent (who’s cheaper there), whereas the urban market by the entrant. Since the ratio of urban-to-rural is 1, market shares will be roughly equal.
With m = 0.8, the incumbent clearly enjoys an advantage in market share. Namely, there is a 60-40 split in market share. The incumbent starts out as having significantly lower profits at first but after 20 periods, it overtakes the entrant and has consistently higher profits for the rest of the simulation. The profit difference is relatively low and variable. The incumbent has profits of 600-700, the entrant 400-550 (these are arbitrary numbers). The difference usually fluctuates between 0 and 200.
Setting m = 0.2 reverses market share: the entrant now has 60%. The incumbent, however, does not seem to lose any profits. Their profits are still around 500-600. The entrant has much larger profits, varying between 800 and 1000.
Up until now, the incumbents price x was assumed to be higher than the marginal cost of serving both urban and rural areas. Therefore, while the incumbent didn’t make as much profit off of rural consumers as off of urban consumers, it still made profit off of them. What happens if we assume that the incumbent’s price is the average of the two marginal costs, implying that the incumbent has to make losses on rural consumers (an arguably more realsitic scenario)?
In this scenario, using m = 0.8, the incumbent will capture 80% of the market and will always make losses, mostly between 0 and -100. The entrant will have moderate profits around 100-200. With m = 0.55, the incumbent will again have 80% of the market, but now as the rural area is not too big, it will be able to make profits in the 200-300 range.
The two graphs below plot the incumbent’s market share and profit as a function of m (the share of the rural area).
As the authors note “even under restrictive pricing conditions, the loss-making segment does not constitute for the Incumbent a competitive disadvantage a priori and, for that reason, does not call for compensatory regulatory intervention merely because of its existence.”
In other words, for the most part the incumbent can be profitable even under this more restrictive form of pricing making compensation for the USO unnecessary in most cases. Of course the numbers in the above graph (and hence the level of m at which the incumbent becomes loss-making) are based on the arbitrary numbers chosen in the simulation. So for real-world regulatory applications one must estimate acutal empirical relationships. The simulation nevertheless shows what the general picture of such a relationship can be expected to be; at least if the sizes of the arbitrary numbers of the simulation relative to each other are realistic (the authors say they are).
What is really interesting here is that one would assume that as the share of the loss-making market segment (i.e. rural area) increases, the incumbent’s profits will strictly decrease due to USO. Such is not the case in this model because of consumer loyalty. Consumers highly value the low price of rural service thus increasing their satisfaction with the incumbent. This makes them more likely to choose the incumbent over the entrant. As long as these consumers do not use the incumbent for rural services too much (i.e. as long as the size of the rural area is not too large), the effects of increased loyalty can offset the losses from the rural area.
A nice, clear application of agent-based modeling can be found in this paper. Further research can lie in trying to calibrate the model with real-world values or extending the framework to other industries. I am definitely interested in the latter.
Naggy note: the graphs in the paper with the little I’s and E’s are truly a terrible choice, barely readable. They should have sticked with the classic colored lines.