Economic institutions evolve. Many, like the New York Stock Exchange (NYSE), have roots going back to medieval times: princes, merchants, and guild masters would set the rules and adjust them to keep up with rivals, or to help their constituencies. By the twentieth century, the rules were typically adjusted according to committee decision (NYSE, 1988), influenced by lawyers, politicians, and representatives of various constituencies. Economists could only watch from the sidelines.
But times are changing. In the last decade or so, it became routine for committees, lawyers, and politicians to hire an economist for advice. And on occasion, economists were asked to design entirely new market institutions, especially for the Internet.
Policy advice and institutional engineering draw on theory, but are qualitatively different sorts of tasks. The goal is not to refine timeless principles but rather to get the right decision or the right design on time. Experiments are a helpful tool to provide empirical evidence, to assess the performance of different existing institutions ceteris paribus, and to finetune a new institution. Economists here have a role similar to architects and engineers, to adapt existing knowledge to the idiosyncrasies of a particular place and time.
Of course, the policy or engineering process does not always turn out well. Perhaps the most spectacular recent failure is electric power deregulation in California in the late 1990s. The state government tried to balance the wishes of power generators (especially those already present), large corporate customers, consumer groups, and taxpayers. Each group hired economists, but the final result was a political compromise. It included a price ceiling for retail customers, very inelastic demand (despite the availability of technologies that enable demand to be contingent on time of day, temperature, etc.), and concentrated supply. The result turned out to be disastrous for everyone: customers suffered blackouts as well as extraordinarily high utility bills, distributors went bankrupt, suppliers (after enjoying huge but brief windfall profits) faced scandal and poor financial prospects, and the politicians can only hope that voters forget the whole mess.
For recent analyses, see Smith (2002), Wilson (2002), and Wolak (2002). We cannot resist noting that laboratory experiments with even crude representations of supply and demand conditions would have pointed out the susceptibility to price manipulation (Holt et al., 1986; Friedman and Ostroy, 1995).
A scholarly survey of experiments in policy analysis and institutional engineering is hampered by the fact that most such work is unpublished. The findings usually remain in the hands of the organizations that commissioned the study. Still, some studies are released to the public, and we will discuss just a few of them.
Experiments to evaluate alternative policies resemble scientific experiments in that the alternative institutions are already defined. The analysis, however, seeks to provide a "good enough" answer to a specific question, rather than general results. The client—usually a governmental agency or a private company—asks the question and decides when the answer is good enough. The experimenter's task is to construct an environment that will provide the most informative results, given the client's time and budget constraints. Usually, there is no opportunity to follow-up on puzzles that emerge during the investigation.
Two classic policy experiments are reported in Hong and Plott (1982) and Grether and Plott (1984). Hong and Plott were hired by the US Department of Transportation and the Interstate Commerce Commission (ICC) to study the possible consequences of a proposal by the barge industry to require advance notice of any changes in posted price. The report was due in one month.
The proposal sounds innocuous to most people: advance notice just seems like common courtesy, and helps clients plan their affairs. But industrial organization theory, controversial at that time, suggested that advance notice might facilitate collusion on higher prices. The ICC wanted to avoid such an outcome.
The barge industry—freight transportation on inland waterways—has many complexities. It is differentiated by start and end points, it has to accommodate a variety of cargo sizes and priorities, it is subject to seasonal fluctuations, it contains large and small buyers and sellers with different short-run and long-run elasticities, etc. The art of policy analysis here (and elsewhere) is to find a simplification that gets to the root of the issue and that satisfies the client.
Hong and Plott chose to take a single representative market, and to re-create an approximate scale replica with and without the proposed change. The design had only one focus variable, the price announcement procedure, which took two values: Posted Price and Telephone. Posted Price included advance notification, while Telephone allowed private bilateral bargaining over phone lines, monitored by the experimenters for data capture and control. The authors also included as a treatment variable the nuisance most likely to be mentioned by adversaries: seasonal fluctuations in cost and demand. That is, some sequences of trading periods used cost and demand parameters representative of the "low" season and other sequences used "high" season parameters. Other control variables, like the number of subjects (eleven buyers, twenty-two sellers), and the other basic parameters were held constant. Two replications required a total of four market sessions held on successive nights using the same group of subjects. Each session included both a low and a high season, and used either Telephone or Posted Price. The authors fit these treatments into an ABBA design that neutralized the effect of experience.
The results were clear. They found that advanced price posting indeed caused higher prices, lower volumes, and reduced efficiency. It hurt the small participants and helped the large sellers (who had backed the proposal). The burden of proof was then shifted back to those advocating the change, and in the end the proposal was not adopted.
Grether and Plott (1984) conducted an experimental study for the Federal Trade Commission (FTC) to assess the claim that four domestic producers of tetraethyl lead (a gasoline additive) were colluding to maintain uncompetitive high prices by using three practices: advanced notification of price change, a guarantee to customers that nobody else could get a lower price, and quotes inclusive of transport costs.
Grether and Plott chose to focus on the first two practices. Even so, they ended up with twenty-four possible treatment combinations: three levels for price publication, two levels for price access, two levels for advanced notice, and two for the guarantee. They held constant other variables such as the exchange institution (telephone bilateral search), supply and demand parameters, but still had to reduce the number of treatments to stay within budget and time. They therefore decided against a factorial design and concentrated on the most interesting combinations: all disputed practices present and all disputed practices absent. They ended up running eight treatment combinations in eleven laboratory sessions of sixteen to twenty-five periods each with an ABBA crossover design.
The results clearly supported the conclusion that prices are near the competitive equilibrium when the disputed practices are absent, but are substantially higher when the practices are present. From an academic point of view, this study needed follow-up work to assess the separate and interactive effects of the disputed practices, but for the authors' purposes it was good enough to convincingly argue that those practices were not innocent. After the experiment, the defendants lost the case to the government in trial, but won on appeal.
These classic studies opened the way for many more, few of which have been published. A recent unpublished study that we conducted illustrates the use of field experiments. At the height of the dot.com bubble, we were contacted by a startup firm that was developing a new electronic auction format. We wrote a white paper based on existing theory and existing data that suggested some possible strengths for the Calendar™ auction, a hybrid
descending auction with some ascending features (Cassar and Friedman, 2001). Then we had the opportunity to conduct a field experiment in conjunction with fund raising for the 2001 UCSC Economics Alumni reunion. Local companies had donated items for the event, and we used those that came in pairs.
We put one of each pair in an electronic English auction and the other in an electronic Calendar auction format. To neutralize sequencing effects (the second week of auction turned out to have more traffic than the first), each pair of items was assigned randomly either to Group 1 or to Group 2. Group 1 items were sold the first week under the English (i.e. ascending) format, the second week under the Calendar format. The sequencing was reversed in Group 2. Since the goal for this auction was to raise money, we bid a third of the nominal value of each item not meeting this threshold by the third day of the auction. We considered the items unsold when our bid won the auction. (We then resold those items during the reunion at a silent or an oral ascending auction, not part of the field experiment.)
The results were unambiguous: in our environment, the English format raised higher profits. Four pairs of goods were sold under both formats and in each of these cases the English price was higher than the Calendar price. Eight items were sold only under one format, and in each case the unsold item was in the Calendar format. The remaining six pairs did not sell in either format.
In fairness to our clients, we should say that (as noted in our white paper) the field auction environment (thin trading of once-off items to be delivered later to inexperienced traders) is probably the least favorable to the Calendar auction. Had the Calendar format done well, the field experiment would have given it a tremendous boost, but as it turned out we still do not know the Calendar auction's relative performance in more favorable environments.
Another caveat is appropriate for studies of this sort. Greater revenue or efficiency do not automatically imply that a new market institution will displace a pre-existing alternative. There are at least three obstacles (Friedman, 1993). First, those who profit from the old format may be able to enlist political support to suppress the new rival (Olson, 1982). Second, a buyer or seller might actually prefer trading in an inefficient format if it reveals less of his private information. Third, transaction volume itself is a source of efficiency. Sellers prefer a format where they expect to find more buyers, and likewise buyers prefer a format where they expect to see more sellers. Thus a popular old format has a built-in advantage, called a network effect or an economy of scale. Indeed, a new format with small market share may have lower efficiency than the old format at large share, even though it would surpass it at equal share (David, 1985).
Institutional engineering hardly existed 20 years ago, but it already dominates several important areas such as the auctioning of spectrum licenses (revenues in the tens of billions of dollars) and the annual assignment of new medical doctors to US internships. In these and emerging areas such as airline landing rights and the allocation of space station resources, economists have played leading roles in creating new economic institutions.
Roth (2002) highlights three general characteristics of the task: the necessity of fast delivery; the value of existing knowledge from related markets; and the political forces affecting the final choices. Theory is important in the early stage for developing intuitions, but it cannot provide all the necessary details. Field data are not available when you consider something completely new, so laboratory experimentation becomes especially valuable.
The spectrum auctions for wireless communication devices are leading examples of institutional engineering (see Cramton, 1995; McAfee and McMillan, 1996; Plott, 1997; Milgrom, 2000). In the 1980s the Federal Communications Commission (FCC) became disenchanted with allocating spectrum bands by a political process or by lottery, and started a hearing process on auction design. The FCC and some of the larger telecommunications companies, such as Pacific Bell and Airtouch Communications, soon hired academic economists, including several experimental economists, to advise on how to allocate spectrum licenses efficiently.
The FCC initially favored sequential sealed auctions for the licenses, but the economists pointed out that sealed auctions encourage overly cautious bidding due to the winner's curse (see Chapter 9). We recommended simultaneous increasing auctions, and eventually (due in part to the results of new pilot experiments as well as existing theory and evidence) the FCC agreed.
The environment is complex because the value of one license to a particular spectrum band in a particular metropolitan area depends on the allocation of other bands in the same area (substitutes) and the same band in adjoining areas (complements). Existing theory was silent on these matters, but a number of economists (Preston McAfee, Paul Milgrom, and Bob Wilson in particular) recommended a simultaneous soft close. That is, bidding should remain open in all auctions for related licenses as long as new bids appeared in any one of them.
Intuition and pilot experiments suggested that some bidders would prefer to wait until the end to make serious bids, in order to prevent others from learning anything about their valuations. Milgrom and Wilson proposed a fix they called the "activity rule." Bidders had to maintain active bids to keep the right to bid at the end. Designers also had to deal with a variety of other problems such as possible collusion, and the ploy of using a proxy company that can declare bankruptcy right after winning. The US spectrum auction turned out to be quite successful, and economists again played leading roles in later spectrum auctions in Europe. For example, in the United Kingdom, the number of potential bidders barely exceeded the natural number of licenses, and the engineers (guided by pilot experiments in the lab) decided to append a final sealed stage to the auction (Binrnore and Klemperer, 2002).
The results were considered a major success at the time. Of course, the crash of telecom stocks in 2001-2002 removed some of the luster, even though the main reasons for the crash were unrelated to the spectrum auctions. More recently, the auction of the spectrum for high-speed Internet access posed several problems due to the large number of possible packages. Economists using simulation and experiments demonstrated that package bidding could achieve higher efficiency than single-item auctions; see Ledyard et al. (1997), Plott (1997), Cybernomics (2000), Milgrom (2000), and Ausubel and Milgrom (2001).
The first experiments in institutional design still are instructive. Grether et al. (1981) report a classic experiment on the allocation of airplane landing rights in the United States. Landing rights were allocated by committees of airline representatives certified by the Civil Aeronautics Board. With the deregulation of the US airline industry in the late 1970s, this allocation procedure was seen to be a possibly significant barrier to entry by new companies. The experiment examines the impact of various committee and market allocation processes. The authors found under the committee process that there were inefficiencies in handling interdependencies among airports, that the outcome is sensitive to the default option in case of agreement failure, and that the result does not respond to the profitability for the individual airlines. Under market process, they found no significant speculation in landing slots, that the price of landing slots was determined by the marginal value to airlines, and that outcomes were more efficient. Unfortunately, the political process did not lead to actual reform, and airport slots are still not competitively awarded.
Also in the 1980s, the Federal Energy Regulatory Commission funded a series of studies on electric power and natural gas networks. These studies are surveyed by McCabe et al. (1991). These goods have important indivisibilities and complementarities. For example, a gas distributor wanting to make a purchase from a gas producer needs to know the availability and price of transmission rights held by pipeline owners. The deregulation process continues its slow and uneven course, with occasional input from the economics laboratory.
New computer technology in the 1980s permitted the creation of "smart" computerassisted exchange institutions, such as "combinatorial auction" for natural gas, that potentially have higher efficiency than traditional bilateral contracts. The basic idea, sometimes called the Smith auction, is to ask participants to send bids and then to use the computer to compute the allocation and prices that maximize surplus with respect to the bids sent. In theory, participants might not find it in their interest to bid their true values, but in the lab it seems that strategic manipulation is usually unprofitable. As noted in Chapter 8, the efficiency may be due to a biased learning procedure.
Rassenti et al. (1982) used a combinatorial auction of this sort to allocate packages of airport takeoff and landing rights in the laboratory. Despite the repeated early attempts by inexperienced subjects to manipulate the system, they achieved overall efficiencies of 98-99 percent. The Arizona team later applied smart computer-assisted markets to a proposal to deregulate the electric power industry (McCabe et al., 1989), and a Caltech team applied the idea to trading pollution permits (Ledyard et al., 1997).
We close our unsystematic survey of institutional engineering with a recent success story about a professional labor market (Roth, 2002). Before the 1950s, the US market for medical internships (entry-level MD positions) had a timing problem. Each hospital found it advantageous to make offers before their rivals, resulting in appointments before medical students could establish clearly their interests and talents. Indeed, the market unraveled to the extent that some appointments were made to students years before graduation! Medical schools tried to prevent this "market failure," for example, by not sending official letters of recommendations before a certain date, but their efforts were not successful. After several attempts, a centralized clearinghouse was introduced and operated successfully until the 1980s. Changes in the medical profession required revisions of the matching system.
Roth led the new design effort. He showed that the historic success of earlier clearinghouses depended on a property called stability: given the submitted preferences, no pair of hospitals or interns would prefer to switch. Kagel and Roth (2000) designed an experiment to examine the effect of different matching algorithms (the stable deferred acceptance market mechanism versus the priority matching mechanism) while holding everything else constant. In the first set of periods, the subjects arranged matches in a decentralized market with enough competition and congestion to create the unraveling problem noted earlier. The subjects had then the opportunity to make early matches at a cost, or to wait and use one of the two centralized mechanisms. The stable mechanism stopped the unraveling and restored efficiency, while the unstable mechanism did not. The similarity of the lab results and the historical field results strengthens confidence that the stability property really is the key to understanding the history of the medical internship market. The Roth-Peranson design was adopted in 1997 as the new algorithm in the entry-level labor market not just by the American physicians, but in many other professions in the United States and Canada.
For further readings, see Unver (2000a, b, c) for follow-up experiments and computational studies extending this analysis to other mechanisms and features of the markets using them. See Roth (2002) for additional analyses showing that even in the presence of complementarities that could undermine stable matching (as couple going together or linked jobs) the departures from simple theory are small and rare in large markets.
Ausubel, L.M. and Milgrom, P. (2001) "Ascending auctions with package bidding," Working
Paper, University of Maryland. Binmore, K. and Klemperer, P. (2002) "The biggest auction ever: the sale of the British 3G
Telecom Licenses," Economic Journal, 112(478):C74. Cassar, A. and Friedman, D. (2001) "An electronic calendar auction," White Paper commissioned by OneDayFree.
Cramton, P. (1995) "Money out of thin air: the nationwide narrowband PCS auctions" Journal of Economics and Management Strategy, 4:267-343.
Cybernomics (2000) "An experimental comparison of the simultaneous multi-round auction and the CRA combinatorial auction," paper presented at the FCC Combinatorial Conference, May 5-7.
David, P. (1985) "Clio and the economics of QWERTY," Economic History, vol. 75, no 2, AEA Papers and Proceedings.
Friedman, D. (1993) "How trading institutions affect financial market performance: some laboratory evidence," Economic Inquiry, 31:410-435.
Friedman, D. and Ostroy, J. (1995) "Competitivity in auction markets: an experimental and theoretical investigation," Economic Journal, 105(428):22-53.
Grether, D.M. and Plott, C.R. (1984) "The effects of market practices in oligopolistic markets: an experimental examination of the ethyl case," Economic Inquiry, 22(4):479-507.
Grether, D.M., Isaac, R.M., and Plott, C.R. (1981) "The allocation of landing rights by unanimity among competitors," American Economic Review, 71(2):166-171.
Holt, C.A., Langan, L.W., and Villamil, A.P. (1986) "Market power in oral double auctions," Economic Inquiry, 24(1):107-123.
Hong, J.T. and Plott, C.R. (1982) "Rate filing policies for inland water transportation: an experimental approach," Bell Journal of Economics, 13(1):1-19.
Kagel, J.H. and Roth, A.E. (2000) "The dynamics of reorganization in matching markets: a laboratory experiment motivated by a natural experiment," Quarterly Journal of Economics, 115(1):201-235.
Ledyard, J.O., Porter, D., and Rangel, A. (1997) "Experiments testing multiobject allocation mechanisms," Journal of Economics and Management Strategy, 6(3):639-675.
McAfee, R.P. and McMillan, J. (1996)'Analyzing the airwaves auction," Journal of Economic Perspectives, 10(1):159-175.
McCabe, K.A., Rassenti, S.J., and Smith, V.L. (1989) "Designing 'smart' computerassisted markets," in V.L.Smith, ed., Papers in Experimental Economics, Cambridge: Cambridge University Press, pp. 678-702.
McCabe, K.A., Rassenti, S.J., and Smith, V.L. (1991) "Experimental research on deregulated markets for natural gas pipeline and electric power transmission networks," Research in Law and Economics, 13:161-189.
Milgrom, P. (2000) "Putting auction theory to work: the simultaneous ascending auction," Journal of Political Economy, 10, 105-114.
New York Stock Exchange, Inc. (1988) Constitution and Rules, Chicago: Commerce Clearing House.
Olson, M. (1982) The Rise and Decline of Nations: Economic Growth, Stagflation, and Social Rigidities, New Haven: Yale University.
Plott, C.R. (1997) "Laboratory experimental testbeds: application to the PCS auctions," Journal of Economics and Management Strategy, 6:605-638.
Rassenti, S.J., Smith, V.L., and Bulfin, R.L. (1982) "A combinatorial auction mechanism for airport time slot allocation," Rand Journal of Economics, 13:402-417.
Roth, A.E. (2002) "The economist as engineer: game theory, experimentation, and computation as tools for design economics," Fisher-Schultz Lecture, Econometrica, 70(4): 1341-1378.
Smith, V.L. (2002) "Power to the people," Wall Street Journal, editorial 10/16/02, p. A20. Unver, M.U. (2000a) "Backward unraveling over time: the evolution of strategic behavior in the entry level British medical labor markets," Journal of Economic Dynamics and Control, 25(6-7):1039-1080.
Unver, M.U. (2000b) "Computational and experimental analyses of two-sided matching markets," PhD Thesis, University of Pittsburgh.
Unver, M.U. (2000c) "On the survival of some unstable two-sided matching mechanisms: a laboratory investigation," Mimeo, University of Pittsburgh. Wilson, R.B. (2002)"Architecture of the power markets," Econometrica, 70:1299-1340. Wolak, F.A. (2002) Statement before the Senate Committee on Commerce, Science and Technology on Enron's role in the California Electricity Crisis, May 15, Testimony.
Was this article helpful?
We Are Not To Be Held Responsible If Your Online Trading Profits Start To Skyrocket. Always Been Interested In Online Trading? But Super-Confused And Not Sure Where To Even Start? Fret Not! Learning It Is A Cakewalk, Only If You Have The Right Guidance.