ساختار، رفتار و قدرت بازار در بازار کار تکاملی با جستجو تطبیقی
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|16902||2001||39 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Economic Dynamics and Control, Volume 25, Issues 3–4, March 2001, Pages 419–457
This study uses an agent-based computational labor market framework to experimentally study the relationship between job capacity, job concentration, and market power. Job capacity is measured by the ratio of potential job openings to potential work offers, and job concentration is measured by the ratio of work suppliers to employers. For each experimental treatment, work suppliers and employers repeatedly seek preferred worksite partners based on continually updated expected utility, engage in efficiency-wage worksite interactions modelled as prisoner's dilemma games, and evolve their worksite behaviors over time. The main finding is that job capacity consistently trumps job concentration when it comes to predicting the relative ability of work suppliers and employers to exercise market power.
Market power refers to the ability of sellers or buyers to exert a perceptible control over market outcomes that enables them to attain higher individual welfare levels than they would achieve under competitive market conditions. Understanding the relationship between market structure, market behavior, and market power in markets with multiple agents engaged in repeated strategic interactions has been a major focus of analytical, empirical, and human-subject experimental researchers in industrial organization since the early 1970s. To date, however, definitive conclusions have been difficult to obtain. For example, Tirole (1988, Part II) presents a unified theoretical treatment of oligopoly decision-making in terms of noncooperative game theory. He focuses on the choice of price, capacity, product positioning, research and development, and other strategic variables. Only equilibrium behavior is considered, however; and a common finding for his games with incomplete information is that multiple equilibria exist with widely differing characteristics. The question then arises: given agents with incomplete information and limited computational capabilities, under what conditions would these agents learn to coordinate on one type of equilibrium versus another, and what would be the resulting dynamic implications for market power? In a survey of empirical work on market power in industrial organization, Bresnahan (1989, pp. 1051–1055) summarizes his overall findings as follows: ‘although the (new empirical industrial organization) has had a great deal to say about measuring market power, it has had very little, as yet, to say about the causes of market power’. Holt (1995, Section VII) notes that, although the nonmonopolized double auction is widely used in experimental research with human subjects, whether market power has any efficiency effects in this context remains an open issue. For posted-offer auctions, Holt points out that capacity constraints and some forms of transactions costs have reliably produced supra-competitive prices in experiments performed by himself and several other researchers; but so far the number of experiments along these lines has been small. This lack of definitive results reflects the complex nature of market power in actual real-world markets. Given this complexity, it would seem useful to complement these previous approaches to the study of market power with controlled computational experiments. This paper investigates the evolution of market power in the context of a computational labor market framework with strategically interacting work suppliers and employers.1 As will be clarified in Section 2, the labor market framework is a flexible computational laboratory permitting experiments with a wide variety of alternative specifications for the exogenous aspects of market structure and agent attributes. The primary purpose of this study, however, is to take a first cut at the computational study of market power in relation to market structure and market behavior over time by specifying these exogenous aspects in relatively simple terms. Thus, as implemented for this study, the labor market framework comprises multiple work suppliers and employers who repeatedly participate in costly searches for worksite partners on the basis of continually updated expected utility, engage in efficiency-wage worksite interactions modelled as prisoner's dilemma games, and evolve their worksite strategies over time on the basis of the earnings secured by these strategies in past worksite interactions. All work suppliers are assumed to have the same size wq, where wq is the maximum number of potential work offers that each work supplier can make. Similarly, all employers are assumed to have the same size eq, where eq is the maximum number of job openings that each employer can provide. Moreover, there is no entry into, or exit from, the labor market; the number NW of work suppliers and the number NE of employers are both held fixed during the course of each experimental run. Market power for work suppliers is measured by the degree to which their average attained welfare level deviates from the average welfare level that they would obtain in a competitive (full employment) market outcome under the assumption of mutually cooperative worksite behavior. Market power for employers is similarly defined. Intuitively, it seems reasonable to postulate that the extent to which market power accrues to work suppliers or employers in the labor market framework depends in part on job capacity, as measured by the ratio (NE·eq)/(NW·wq) of total potential job openings to total potential work offers, and on job concentration as measured by the ratio NW/NE of the total number of work suppliers to the total number of employers. Job capacity measures the total potential availability of job openings relative to work offers, whereas job concentration measures the extent to which control over job openings is concentrated among relatively few employers. 2 By construction, for any given level of job concentration, job capacity varies inversely with the size wq of each work supplier and directly with the size eq of each employer. The joint implications of job capacity and job concentration for the exercise of market power are not easy to predict a priori. For example, consider the overall effect on welfare and market power when the labor market framework comprises 12 work suppliers, each able to make one work offer, together with six employers who each have four job openings. The excess supply of job openings favors monopoly power by work suppliers, but the concentration of job openings in relatively few hands favors monopsony power by employers. The experimental design of this study consists of the systematic variation, from high to low, of both job capacity and job concentration. For each given specification of these two market structure conditions, 20 different runs are generated using 20 different pseudo-random number seed values.3 In examining the resulting run histories, particular attention is focused on the experimental determination of correlations between market structure and the formation of networks among work suppliers and employers, and between network formations and the types of worksite behaviors, welfare outcomes, and market power outcomes that these networks support. The primary objective of this study is to test the following four hypotheses regarding the impact of job capacity and job concentration on the ability of work suppliers and employers to exercise market power:Hypotheses H3 and H4 can hold simultaneously without contradiction. However, hypotheses H1(b) and H2(a) yield contradictory predictions in conditions of excess job capacity and high or balanced job concentration, and hypotheses H1(a) and H2(b) yield contradictory predictions in conditions of tight or balanced job capacity and low job concentration. The main finding of this study at the aggregate data level is that job capacity is the dominant factor affecting the ability of work suppliers and employers to exercise market power. Aggregate market power outcomes strongly support the job capacity hypotheses H1 and H3(a) and weakly support the job capacity hypothesis H3(b), but they provide little support for either of the job concentration hypotheses H2 or H4. Surprisingly, controlling for job capacity, job concentration has only small unsystematic effects on attained market power levels. To better understand these aggregate market power findings, the complicated nonlinear relations linking market structure, market behavior, and market power are carefully examined at a more disaggregated level. This examination reveals even stronger support for the aggregate data finding that job capacity is the key variable determining the relative market power of work suppliers and employers. For example, in contrast to aggregate data indications, it is shown that the job capacity hypothesis H3(b) is strongly supported if a small number of sample economies are omitted for which complete coordination failure occurs and all agents perform extremely poorly. The disaggregated data also highlight the importance of job search costs and behavioral flexibility (provocability) for the realistic assessment of market power opportunities ex ante and the accurate measurement of market power ex post. The labor market framework is described in Section 2. In Section 3, descriptive statistics are constructed for the ex post classification of network formations, worksite behaviors, welfare outcomes, and market power outcomes. The experimental design of the study is explained in Section 4, and experimental findings are discussed in Section 5. Concluding remarks are given in Section 6.
نتیجه گیری انگلیسی
As shown in previous sections, the aggregate market power findings of this study indicate that job capacity generally has the hypothesized H1 and H3 effects: all else equal, increased job capacity increases the market power of work suppliers and reduces the market power of employers both in absolute and relative terms. Disaggregated market power findings reveal that these effects are particularly strong and clear when attention is focused on the most commonly observed network formations for each treatment setting. In contrast, neither aggregated nor disaggregated market power findings provide much support for the job concentration hypotheses H2 and H4. On the contrary, controlling for job capacity, the effects of job concentration on the ability of work suppliers and employers to exercise market power are surprisingly small and unsystematic. Hypotheses H2 and H4 seem a priori intuitive on the grounds that concentrating work offers in fewer work supplier hands should provide work suppliers with an increased opportunity to exercise monopoly power, and concentrating job openings in fewer employer hands should provide employers with an increased opportunity to exercise monopsonist power. On the other hand, it may be that too much concentration lessens the ability of work suppliers or employers as a whole to adapt their worksite strategies in a flexible manner in response to the worksite strategies used by their worksite partners. Consequently, there may be too little diversity in the pool of worksite strategies used by the concentrated agent type for evolutionary selection pressures to efficiently act upon. To test this inflexibility hypothesis, it will be necessary to introduce the absolute numbers of work suppliers and employers as treatment factors in addition to their concentration ratio. It will also be necessary to examine alternative learning algorithms calibrated more carefully to the learning behavior observed in real-world labor markets and in human-subject labor market experiments. Further work is needed to test the robustness of the findings of this study to variations in the scope and range of other parameter specifications as well. As preliminary as these findings may be, however, they do caution against the common practice of confounding capacity and concentration effects in market power studies by letting these two factors vary together in an uncontrolled manner. A potentially important by-product of the current study is a better understanding of the fundamental role played by organizational costs in sculpting and sustaining network formations and hence in determining the welfare and market power levels attainable by work suppliers and employers. For example, when job capacity is tight, high job search costs can result in widespread worker discouragement and exit from the labor force that ends up hurting employers as well as work suppliers. Consequently, although tight job capacity tends to favor employers, it also increases the risk of network coordination failure. Organizational costs are a key focus of researchers in transactions costs economics — see, for example, Williamson and Marsten (1999). Nevertheless, the significant problems posed by network organizational costs for the definition and measurement of market power have not received much attention to date. A second potentially important by-product of the current study is a better appreciation of the role played by behavioral flexibility in protecting agents against the exercise of market power by other agents. In the labor market framework, work suppliers and employers repeatedly choose and refuse their worksite partners and evolve their worksite behaviors over time. Thus, even when jobs are in excess supply, attempts by work suppliers to exert market power by repeatedly defecting in worksite interactions can provoke retaliatory defections by employers as well as firings (refusals of all future work offers from offending work suppliers). Similarly, even when jobs are tight, attempts by employers to exert market power by repeatedly defecting in worksite interactions can provoke retaliatory defections by work suppliers as well as quits (work suppliers redirecting future work offers elsewhere). Consequently, although structural asymmetries tend to favor one agent type over another, attempts by the favored agent type to exploit this advantage increase the risk of behavioral coordination failure. In summary, agent-based computational frameworks such as the labor market framework used in this study permit the systematic experimental investigation of behavioral and network formation processes that appear critical for understanding the relation between structure, behavior, and market power in real-world labor markets.