Markets for water

A serpentine stretch of the California Aqueduct in Palmdale, along mile post 327.50 on February 7, 2014. | Credit: Florence Low / California Department of Water Resources, Feb. 7, 2014
A serpentine stretch of the California Aqueduct in Palmdale, along mile post 327.50 on February 7, 2014. | Credit: Florence Low / California Department of Water Resources, Feb. 7, 2014

This is the English version of a section in Ferreira (2014) 1, a paper dedicated to overview the experiments in Industrial Organization, to appear in Revista de Economía Industrial, and is reproduced here with the permission of the journal.

The great majority of water collection and distribution in the world rely on a central authority. Economists have long suggested the introduction of market structures to improve the efficiency of water allocation, allowing voluntary transfers towards the uses where it is more valuable (see, for instance, Easter et al., 1998 2). I will summarize one of the most important works on the use of experimental economics to understand how this may work: the article by Murphy et al. (2000) 3, where the authors report the results of their experiments to study the efficiency of a closed-envelope, uniform double auction for the simultaneous allocation of water and water transport capacity among sellers, buyers and transporters. Closed-envelope is jargon for an auction in which bids are sent simultaneously by bidders (in closed envelopes that only the auctioneer can open), and cannot be changed. Double auction means that both sellers and buyers can make offers, and uniform means that once a price is established by the system, everyone pays the same.

The authors conduct their experiments to simulate the situation in the South of California, where rain is scarce and irregular, and where water must be brought by conducts from dams in several rivers or can be extracted from underground aquifers. There are also several environmental and transport restrictions, as well as prohibitions to export water from some sources. To accommodate all the main features of the water sector in South California the experiment is designed with 17 nodes in which there are 9 active experimental subjects (some agents must make decisions in more than one node.)

These are the nodes, divided according to their characteristics:

  1. Three sources of surface water, corresponding to the rivers Sacramento, San Joaquin and Colorado.
  2. Three sources of underground water, corresponding to the aquifers in the region.
  3. Five urban districts: Sacramento, the Bay Area in San Francisco, the cities of San Joaquin Valley, the metropolitan area of Los Angeles, and San Diego. These districts can only buy water.
  4. Four agricultural agents: Sacramento Valley, Northern San Joaquin Valley, Southern San Joaquin Valley, and Colorado River irrigators. These agents have rights to buy water, and to sell from their surface sources, but they cannot sell from underground aquifers. The first has rights over the Sacramento river and one of the three aquifers; the second has also rights over the Sacramento river and over a second aquifer; the third has rights over the San Joaquin river and the third aquifer; the fourth only has rights over the Colorado river. Note that each aquifer is managed by just one agent, so that there are not externality problems here.
  5. Two transport canals with active agents: the aqueducts of the Colorado and San Diego rivers. There are another two canals, but they operate automatically: the pumps in the California Delta, where the Sacramento and San Joaquin rivers meet, and the canals south of the Delta.

The amount of water that is actually transferred in California is too small to estimate supply and demand functions after the data they provide. Instead, the experiment uses an econometric model of agricultural production based on the Central Valley Production and Transfer Model, as described in U.S. Department of Interior (1997) 4. Among other things, in this model one can change the available quantities of water in a particular region and compute the shadow price of water in this region. In this way one can estimate the demand function for each agent. Howitt (1995a and 1995b, 5, 6) shows that this kind of models is robust and useful for this type of simulations. In the experiments the supply of water follows eight-period cycles with different available amounts, imitating the irregular rainfall and rivers’ flow in the region.

This experiment design has as many markets as places where water is finally consumed, with a different price in each market defined by the double auction mechanism. To be in line with the idiosyncratic characteristics of the experiment, the market clearing mechanism solves a linear programming problem that minimizes costs subject to the restriction that supply equals demand, and also subject to some minimum and maximum quantities in the transport canals. Finally, and due to the complexities of the final experiment, several previous training sessions were conducted and the experimental subjects with the worst results were discarded for the definitive round.

The market efficiency, measured as a proportion of the actual surplus respect to the competitive surplus (the maximum possible), is around 91%, which a very good result, given that the number of agents is limited and there is a monopolist in each of the transport canals. In fact, when the experiment is repeated and the design allows for two agents in each of these canals, efficiency increases to 94%. To check that the results are due to the strategic interactions by the agents and are not biased by the experimental design, the authors run a series of simulations where a computer posts random bids, and obtain efficiency percentages between 0% and 5%.

The distribution of surplus is of particular importance. According to the experiment the main problem is found in the area of San Diego, the one with the fewest connection to receive water, that gets less than half the surplus it would get under perfect competition conditions. However, by creating competition on the lines, with just two rather than one agents, the amount of water increases slightly, around 3%, but it is enough for the price of water to drop 20% and to increase the surplus up to 84% of the surplus under perfect competition conditions.

Another problem detected in the experiments is that the market is not sensitive enough to changes in the amount of available water. This seems to be due to the high volatility of traded quantities, that in turn may be hiding the volatility of the total amount of water. However, this volatility decreases with the experience of the subjects and disappears when competition is added to the transport canals.

What do the results from these laboratory microeconomies tell us about the design of market institutions in the “real world”?

In the words of the authors:

The applicability of experimental results to the understanding of similar non-laboratory situations is referred to as parallelism (Smith 1980)7. Clearly, the experiments reported in this paper are a stylized representation of California’s water network. However, the laboratory experiments allow us to learn about the water market institution in a simple environment that we can control. This permits us to build of body of evidence identifying the strengths and weaknesses of an institution, and provides an opportunity to develop modifications at a relatively low cost before implementation in the field. Thus, laboratory experiments can provide valuable insights to reduce the uncertainty inherently associated with the implementation of new institutions.”


  1. Ferreira, J.L. 2014. Investigación experimental en Economía Industrial (Experimental research in Industrial Economics). Revista de Economía Industrial 393, forthcoming.
  2. Easter, K.W.; Rosegrant, M.W. and Dinar, A., eds. 1998. Markets for Water: Potential and Performance. Norwell, MA: Kluwer Academic Publishers.
  3. Murphy, J.J.; Dinar, A.; Howitt, R.E.; Rassenti, S.J. and Smith, V.L. 2000. The design of “smart” water market institutions using laboratory experiments. Environmental and Resource Economics 17, 375-394.
  4. U.S. Department of the Interior. 1997. Central Valley Project Improvement Act Draft Programmatic Environmental Impact Statement, Technical Appendix, Vol. 8, U.S. Department of the Interior, Bureau of Reclamation.
  5. Howitt, R.E. 1995a. A calibration method for agricultural economic production models. Journal of Agricultural Economics 46, 147-159.
  6. Howitt, R.E. 1995b. Positive mathematical programming. American Journal of Agricultural Economics 77, 329-342.
  7. Smith, V.L. 1980. Relevance of Laboratory Experiments to Testing Resource Allocation Theory, in Kmenta, J. and J. Ramsey, eds., Evaluation of Econometric Models. New York: Academic Press.

Written by

Leave a Reply

Your email address will not be published.Required fields are marked *