Gaming against Managers in Incentive Systems: Experimental Results with Chinese Students and Chinese Managers David J. Cooper; John. Kagel; Wei Lo; Qing Liang Gu The American Economic Review, Vol. 89, No.(Sep., 1999), 781-804. Stable URL: hup:/links. jstor. org/sici? sici=0002-8282%28199909%2989%3A4%3C781%3AGAMIIS%3E2.0.C0%3B2-V The American Economic Review is currently published by American Economic Association. Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/ab gabout/terms.html. ISTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles,and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/aca. html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is an independent not-for-profit organization dedicated to creating and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact support @jstor.org. http://www.jstor.org/ Sat May2211:35:02004
Gaming Against Managers in Incentive Systems: Experimental Results with Chinese Students and chinese managers By DAVID J. COoPER, JOHN H. KAGEL, WEI LO, AND QING LIANG GU* We examine strategic interactions between firms and planners in China, comparing behavior between:(i)students and managers with field experience with this situa ion,(ii)standard versus increased monetary incentives, and (iii) sessions con ducted"in context, "making explicit reference to interactions between planners and managers, and those without any such references. The dynamics of play are similar across treatments with play only gradually, and incompletely, converging on a pooling equilibrium. A fivefold increase in incentives significantly increases initial levels of strategic play. Games played in context generated greater levels of strategic play for managers, with minimal impact on students. (JEL D23, D8, C92 Students of centrally planned economies ers, but cannot identify, a priori, which firms have long posited that these systems are prone are of which type. As a result, the central a"ratchet effect"(Joseph S. Berliner, 1976: planner sets initial targets for input usage and Martin L. Weitzman, 1980). Planners are output levels that are of broad applicability never as well informed about the production and uses the information gathered from the capacity of any individual firm as are the firm's performance to identify what is feasi- managers of the firm. The central planner ble in future planning periods. If firms re- knows that firms have different production spond naively to this initial incentive plan capacities, with some firms having higher high-productivity firms will outperform low- productivity (lower marginal costs)than oth- productivity firms. The central planner will then "reward"these high-productivity firms y reducing their input allocations and/or in- Cooper: Department of Economics, Case Westem Re- creasing their required output levels. Hov serve University, Cleveland, OH 44106; Kagel: Departme of Economics, Arps Hall, Ohio State University, 1945 Nort ever, firms should anticipate this ratcheting High Street, Columbus, OH 43210; Lo up of targets, cutting their ir production in an land China Studies, National Dong Hwa University, 1, Sec. attempt to mimic low-productivity types. This 2, Da-Hsueh Road, Hualien 97441, Taiwan; Gu: Glorious reduction of output with high-productivity University, 1882 West Yan-An Road, Shanghai 200051 firms imitating low-productivity types is China. This research was Chiang Ching-Kuo Foundation to Lo, the National Science Potentially, the ratchet effect can seriously re- uce production in centrally planned Foundation to Cooper. We have benefitted from disct Moreover, the ratchet effect is not just limited to with Patty Beeson, Robert Glaser, John Ham, Richmond planned economies. Similar strategic interactions Harbaugh,Thomas Rawski, Jean-Francois Richard, Alvin arise in any principal-agent interaction where the referees, and comments of seminar participants at Brandeis principal is uninformed about the agent's type and University,the joint Carnegie Mellon-University of Pitts- cannot (or will not) commit to a long-term con urgh applied microeconomy tract. Examples include eny rsity, Washington Universi Dennis A. Yao, 1988), job evaluation and incentive usiness, York University, and participants at the chemes(Barry W. Ickes and Larry Samuelson etings (Tucson), the Italian Experimental Economic 1987; James Dearden et al., 1990), piece-work wage meetings(Washington inter Econometric Society meeting(Chicago). We Cang Ping and the many other fine research The Chinese have a rather colorful term for this- from China textile univers “ whipping the fast
THE AMERICAN ECONOMIC REVIEW SEPTEMBER /999 rate schedules(Robert Gibbons, 1987), regulation of Several conclusions can be drawn from our natural monopolies(Michael A. Crew, 1994), and experimental data, At a general level, in all procurement contracting (Jean-Jacques Laffont and versions of the game, and with all subject pop- Jean Tirole, 1993) ulations, there is a strong initial tendency for In this paper, we report an experiment which high-productivity firms to ignore the strategic addresses several open questions about the de- implications of the ratchet effect. Only gradu velopment of the ratchet effect in centrally ally(and incompletely)does play converge to planned economies and, more generally, about pooling equilibrium. This replicates results the evolution of strategic play in games. We from other signaling game experiments Jord study a game that captures the essential features Brandts and Charles A. holt 1992; Zeinab Pa of the strategic interactions between planners tow and Andrew Schotter, 1993; Cooper et al and firms in a centrally planned economy. In all 1997a, b), indicating that this general pattern of equilibria of this game, a ratchet effect emerges play is unaffected by culture, subject popul ith high-productiy tion and incentiv productivity types, thereby resulting in a pool This general pattern of play also suggests ar answer to an important question about the We compare the dynamics of adjustment to ratchet effect: Why should the ratchet effect equilibrium between: (1)standard"versus sub- ever be observed in field settings? If high stantially increased monetary incentives, (ii)the productivity firms anticipate that planners will standard" experimental subject population ratchet up their production targets and therefore students) and an"expert"subject population imitate low-productivity types, central planners with extensive field experience with the ratchet should also anticipate firms' mimicry and select effect (managers and white-collar workers in incentive schemes that do not elicit this strate state enterprises in the Peoples Republic of (for example, precommiting to a long-term con- China), and (ii) experimental sessions con- tract). Our results show that incentive schemes ducted"in context"versus a"generic"setting. that eventually produce the ratchet effect may The latter is no doubt the most novel of our actually work quite well initially for the central experimental treatment conditions, and there- planners, with inefficiencies arising only grad fore warrants further explanation ually over time as high-productivity firms learn For sessions played in context, the language to imitate low-productivity types. It has been is deliberately designed to relate to subjects suggested that, One possibility [why a planne field experi vith similar games For exam- might not precommit to a long-term contract], ple, we refer to subjects as"planners"and of course, is that the central planner is not aware managers"with"planners deciding what kind of the benefits of commitment"(Xavier Freixas of production target to assign to managers- et al., 1985). Our results suggest a different easy or tough. To the extent that the laboratory explanation: the planner may not precommit game and its equilibria are similar to experi- because initially there are minimal benefits to ences outside the laboratory, the use of context commitment may make subjects more sensitive to the strate ing at the detailed pattern of play, sev- gic implications of the game, facilitating con- eral interesting observations emerge. First, in vergence to the pooling equilibrium. In contrast, creasing payoffs for People's Republic of China in generic sessions, the language is kept delib-(PRC)students induced substantially higher erately abstract with firm managers referred to pooling rates by high-productivity firms in early as"A"players and planners referred to as"B" plays of the game(students in standard-pay players, with B players deciding between sessions only caught up after 20 or more repe- and"Y"( tough or easy contracts)in response to titions of the game). That is, more money in- A players"choices. The generic language is duced heightened initial levels of strategic play intended to minimize any relation to subjects suggesting that money can serve as a partial field experiences and may be expected to reduce substitute for experience. This provides some of consistent with the intuition of most theorists
VOL 89 NO 4 COOPER ET AL: GAMING AGAINST MANAGERS IN INCENTIVE SYSTEMS and a result of fundamental importance for the underpinnings of our treatments in Section m study of learning in games This section also outlines the procedures used Second, context facilitated the development running the experiments, Section Ill presents of strategic play among PRC managers, but had our experimental results. Section IV summa- minimal effect on PRC students. PRC managers rizes our results and offers concluding remarks were affected both in their roles as planners and as firms, with the impact on their role as plan- L. The Ratchet Effect Game ore pronounced than as firms. This result strong(albeit, indirect) evide Payoff tables were generated to capture the PRC firms have been subject to a ratcheting up essential strategic interactions of the ratchet ef- of production targets in response to superior fect game in the simplest possible setting. Gen erformance. The fact that context had a much erating the ratchet effect requires two separate stronger effect on PRC managers than on stu- insights from players. First, central planners dents suggests that context must be eliciting (CPs)must realize that they can use the infor something from managers'experiences as man- mation revealed by firms'(Fs,) choices in re- agers. It is difficult to see what this experience setting targets. Second, Fs must realize that CPs could be other than experience with the ratchet will use any information they reveal and that effect they should use a pooling strategy to avoid Finally, contrary to our expectations, PRc revealing harmful information. Our game there- students in their role as firms exhibited signifi- fore concentrates on Fs'output decisions and cantly higher initial levels of strategic play than the CPs responses. The Appendix to our work older, more experienced PRC managers Subse- ing paper( Cooper et al., 1998) gives full details quent manipulations indicate that this surprising on the underlying model of the ratchet effect, its result was due to the age and/or lower educa- relationship to Chinese institutions, its relation tional levels of older managers. This serves as a ship to the laboratory game, and how the payoff warning to experimenters as they expand their tables were generated from the model studies to an ever larger array of subject popu- The payoff tables are based on a simplified lations: although experimenters are used to as- version of the model in Freixas et al. There are suming that subjects can make elementary two types of Fs, high productivity(FH) and low calculations, follow simple chains of productivity(FL). The game begins with the facility with such"test-taking"skills may not firm learning its type. The firm then chooses its extend to older less-educated subject first-period output, while facing a preexisting target. After observing first-period output(but The plani oe the phper is and frlvonwg se rtichet tht teeoidm e tode, the alane er s w at the t effect. Drawing heavily on the cognitive psy- choosing its second-period output. The firm at- hology literature, we present the theoretical tempts to maximize the discounted sum of its profits. The planner's objective function is the 2 Not withstanding theorists'concern with the ratchet penalty proportional to the firms profits.5a total surplus generated by the firm minus In our laboratory game, the firm,'s second ersy(see the exchange between David Granick[1980, period output decision is taken as given-this is 1983] and Michael Keren [1982]) a straight maximization problem with no inter esting strategic elements. The remaining related problem unless subjects have been is a standard signaling game with the firm acting explicitly told to consider the similarities between the tw as the sender (the first-period output problems, andor the problems are quite similar(Mary L. Gick and Keith J. Holyoak, 1980; David N. Perkins an 4 In freixas et al. the begins with the period 1 Gabriel Salomon, 1988; Salomon and Perkins, 1989). Given target being set by the CP. We eliminate this step to simplify that the problem PRC managers face in the ratchet effect e game while concentrating on the ratchet effect. game must be quite similar to their experiences in the field. ratchet effect literature
THE AMERICAN ECONOMIC REVIEW SEPTEMBER 999 TABLE 1-FIRM PAYOFF AND PLAYER PAYOFF High-productivity firm EASY TOUGH 23456 ..2332 Planner payoff low-productivity firm high-productivity firm TOUGH 528 (the second-period target is its action). The when facing an FL type and greater for timing of the laboratory game is as follows choosing TOUGH than EASY when facing an FH type. I. Prior to the start of the game Fs learn their 4. Fs learn CPs choices and CPs learn Fs type productivity level, FH or FL (it was common Payoffs are then determined. 7 knowledge that the prior probability of each The game has three possible pure strategy 2. Fs choose among several different output sequential equilibria which are pooling at out levels. An F's payoff is a function of F's put levels 1, 2, or 3. All of these equilibria have type, the output level chosen, and the CP's a ratchet effect, with FHs imitating FLs in ef- actions(see Table 1). For any given output forts to receive an easy target. To see why these level, payoffs are higher if the CP responds are equilibria, consider the pooling equilibrium with an EASY rather than a toUGH pro- at output level 2. Since both types of F choose duction target. If an F did not anticipate that the same output level, the CP must believe each his choices affected the CPs response, he type is equally likely following play of 2. Given would choose 2 as an FL and 5 as an FH; we these beliefs, the Cp maximizes her expected refer to these as Fs' full information output payoff by choosing EASY (expected value of levels 704.5 versus 674 for TOUGH given the 50-50 3. CPs see the output level of the F with whom prior distribution of F's type). For all other they have been paired, but not F's type. CPs output levels, there are no constraints on the choose between a TOUGHor EASY produc- CP's possible beliefs. The out-of-equilibrium tion target. The CPs payoffs are a function beliefs that support pooling at 2 are that any of F's type and the target chosen Payoffs are other output level must come from an FH type so that the CP assigns a TOUGH contract Given that the CP assigns EASY following 2 nanish Chaudhuri (1996) and ToUGH following all other output level o-pcriod ratchet effect ga ns had 10-12 plays of the gar 7As noted above, we eliminate F's end-period response ady-state results response. even in a onc-period signalling game. adjusting the payoffs accordingly