《实验经济学》课程教学资源（参考文献）When Workers Do Not Know - The behavioral effects of minimum wage laws revisited.pdf_大学文库

When workers do not know – The behavioral effects of minimum wage laws revisited Xianghong Wang ⇑ School of Economics, Renmin University of China, #59 Zhong Guan Cun Ave., Beijing 100872, China article info Article history: Received 28 January 2012 Received in revised form 19 March 2012 Accepted 18 May 2012 Available online 23 May 2012 JEL classification: J08 J3 D03 PsycINFO classification: 3000 3600 Keywords: Minimum wage Fairness Employment Asymmetric information abstract Previous experimental results have shown that the introduction of a minimum wage increases wages in a monopsonistic labor market. The results rely on the assumption that the minimum wage laws are common knowledge among employers and workers, which is often violated in less developed labor markets. This paper examines the effect of asymmetric knowledge about the minimum wage, and its interaction with the level of the minimum wage standard. We find that, whether the workers have knowledge about the minimum wage significantly changes the behavioral impacts of the minimum wage policy. With common knowledge, most firms offer wages beyond the minimum wage level. When workers do not know about the minimum wage policy, many firms, including those who used to pay higher wages before the introduction of the policy, choose to pay wages at the minimum level or lower their offers toward the minimum level. This causes the introduction of the minimum wage not effective in increasing average wages when the minimum wage standard is low. Therefore, a low minimum wage under asymmetric knowledge may potentially have a negative spillover effect on wages. The asymmetric knowledge also affects the employment level negatively. These findings help explain why the impacts of minimum wage are different in labor markets where workers have different degree of information access. This has strong implications for the implementation and welfare impacts of minimum wage laws. 2012 Elsevier B.V. All rights reserved. 1. Introduction Minimum wage legislations are widely used in many countries as a way to raise the living standard of low income workers. In emerging market economies like China, new policies are being developed in this area. It is important to understand the implications of these changes for the labor market. This paper focuses particularly on the behavioral effects of the minimum wage policy when workers and firms have asymmetric knowledge about it. In addition to how the minimum wage affects employment and wages, economists have been particularly puzzled by the spillover effects of minimum wage, that is, when minimum wage increases in a labor market, the wage increase in the market may exceed the minimum level required by the legislature (Card & Krueger, 1994; Falk, Fehr, & Zehnder, 2006; Owens & Kagel, 2010). The experimental study conducted by Falk et al. (2006) provides some insights into these issues. The authors found that workers’ reservation wages and fairness concerns are the driving force behind these phenomena. In their experiment, they have workers and employers seek employment contracts between each other first without a minimum wage, 0167-4870/$ - see front matter 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.joep.2012.05.004 ⇑ Tel.: +86 10 82500293. E-mail address: shwang06g@gmail.com Journal of Economic Psychology 33 (2012) 951–962 Contents lists available at SciVerse ScienceDirect Journal of Economic Psychology journal homepage: www.elsevier.com/locate/joep

and then with a minimum wage policy imposed. In their context, fairness means, given a firm’s profit function, what is the appropriate offer a firm should give to a worker or the appropriate amount the worker is entitled to receive. When a minimum wage is introduced into the market, the workers’ reservation wages increase due to fairness concerns or the entitlement effect induced by the minimum wage reference, even to levels higher than the minimum wage. The firms are thus forced to pay wages higher than the minimum wage. Much more research is worthwhile to extend the study of Falk et al. (2006) beyond the scope of their design. We examine two issues that, to our knowledge, have not caught the attention of the researchers, the effect of asymmetric knowledge about the minimum wage policy between the workers and firms, and the effects of different levels of minimum wages. The motivations of this study are two folds: one is to test whether the firms’ behavior in Falk et al. is due to self-interest or to their own fairness preferences; two is to apply the experiment to conditions that are more likely to be observed in a developing country. Will firms still offer wages higher than the minimum wage (MW) level if workers have no knowledge about the minimum wage policy? If a policy introduces a minimum wage lower than what many firms have already paid, will the firm reduce its wages? While the earlier experiment by Falk et al. suggests that firms increase wages due to expected fairness concern on the part of the workers, their experiment does not rule out the possibility that firms are behaving according to self-interest, or their own fairness preference, or even altruism. If the positive spillover effect found in previous experiments are attributed mostly to the employers’ expectation of the increase in workers’ reservation wages (RWs), then employers’ beneficial offers to workers will be less or nonexistent when workers are not aware of the minimum wage (MW). While lacking of the knowledge about the minimum wage policy on the part of workers and weak enforcement of the policy may exist in any country (see Ashenfelter & Smith, 1979; Strol & Frank, 2003, for examples), it is more likely to exist in the developing countries. In China, for example, since the scope of its minimum wage policy was broadened in 2004, some studies have shown that many low-income workers, especially migrant workers in China are not well informed about the minimum wage policy and are often not paid to the minimum wage level. Du and Pan (2009) compared the minimum wage coverage of the migrant workers and the local workers using urban labor data in 2001 and 2005. While all workers are legally covered by the minimum wage policy, not being covered by the minimum wage here means that the paid wage is lower than the minimum level due to non-compliance behavior. The percentage of workers whose hourly wage is above that year’s minimum wage is 47.8% vs. 78.5% respectively for the two groups in 2001 and 21.1% vs. 44.6% in 2005. This indicates a drop of minimum wage coverage over time for both groups with the rise of the minimum wage standards, with a bigger drop for the migrant workers. They also showed that the ratio of the minimum wage to the average had dropped, from 0.44 in 1995 to 0.28 by 2006. The survey of Han and Kong (2006) in Guangdong province showed that 52.7% (out of a sample of 1760) migrant workers reported that they had never heard of the minimum wage policy, and 69.1% workers said they did not know what their minimum wage actually was. Xie’s (2010) survey in Jiangsu province reported that only 24.7% migrant workers knew their minimum wage standard. The lack of workers’ knowledge about the minimum wage policy may help explain the low coverage of the migrant workers. It may have made it easier for the employers not to comply with the standards. The extensive supply of the migrant workers offers the employers with every opportunity for lowering labor standards including wages. The low minimum wage floor may have a negative impact on workers’ wages. More of the previous studies on minimum wage policy seem to have focused on the employment effect rather than its distributional effects (Adams & Neumark, 2005). While economic theory with complete competition predicts that raising mandated wage floors will lead to some employment reductions, as stated in Flinn (2006), in the standard monopsony model, minimum wages can lead to higher wages and employment level. Falk et al. (2006) also show that the impact of minimum wage on employment is ambiguous, depending on the magnitude of its impact on reservation wages. Empirically, Wang and Gunderson (2011) studied the impact of minimum wage on employment in China and found that the effects varied for different sectors in which the degree of market pressure varied – consistent with the monopsonistic behavior. While we test the effects of minimum wage policy both on wages and on employment, our focus is more on the wage effects. The advantages of using the experimental method to study the behavioral effects of minimum wage policy have been illustrated out by a few studies (Charness & Kuhn, 2010; Falk & Fehr, 2003; Falk & Heckman, 2009). It is particularly helpful for our analysis of information asymmetry. Our study will use the similar setting and the same type of subject pool as Falk et al. (2006). We design an experiment in which we manipulate the knowledge of the employers and workers about the MW policy (the INFORM variable). Either both have common knowledge about the MW (INFORMED or INFORM = 1) or only the employers have knowledge about it (UNINFORMED or INFORM = 0). We derive hypotheses to predict that wages are positively related to the minimum wage level, and wages are lower under UNINFORMED conditions. We use a survey to the subjects after the experiment to confirm our analyses. The results of our experiment are consistent with our hypotheses. Wages were higher in the INFORMED conditions and the high-level MW conditions. In the INFORMED treatment, more firms gave offers above the minimum level, but in the UNINFORMED treatment, more firms offered wages only at the minimum level, including many who used to offer wages higher than the minimum level. As stated earlier, firms’ reactions to the minimum wage policy can be due to the concerns for fairness or for self-interest. The introduction of minimum wage may serve as a reference point that shifts the fairness perceptions of firms even without the pressure of the workers. Previous work has proven the power of reference or anchoring effect of even arbitrary numbers (Ariely, Loewenstein, & Prelec, 2003). Firms’ initial fairness perception before the introduction of the MW and the reference point of the MW may imply the coexistence of multiple standards. When multiple standards of fairness exist, self-serving bias may drive the fairness perception of the agents (Babcock, Loewenstein, Issacharoff, & Camerer, 1995; Babcock, Loewen- 952 X. Wang / Journal of Economic Psychology 33 (2012) 951–962

stein, & Wang, 1995b). Firms who used to hold fair wages that are higher than the MW may thus reduce their payment to be equal or close to the minimum wage. Or when workers are not aware of the minimum wage, firms motivated by self-interest may try to take more share of the payoff. Our study makes the following contributions: First, we confirm the findings of Falk et al. (2006) about the general impact of minimum wage on wages under common knowledge conditions. Second, the design with different levels of minimum wage helps explain the impacts of different wage floors on the labor market. Third, our behavioral study of workers and firms under asymmetric information suggests different welfare results to workers when information is complete or not. This will have important policy implications. Since firms are inclined to reduce their payment to the binding minimum wage level when workers are not aware of the legal constraint, we can imagine that if the enforcement of the minimum wage laws is not strong enough, firms are likely to deviate from the legal request and pay wages lower than the minimum level. This may help explain why Du and Pan (2009) found different coverage levels of minimum wages for migrant workers and local urban workers. The most effective compliance mechanism might lie with empowering the workers with knowledge. In the following section, we present our experimental design and explain our hypotheses. Section 3 reports the results. Section 4 concludes the paper with discussions. 2. Experimental design 2.1. Experimental game and procedures The experimental game is similar to the one used by Falk et al. (2006). All subjects were undergraduate students at the Renmin University of China, allowing us to compare our results with their student pool of Switzerland. We first describe the procedure for one experimental session before we explain the design in the next section. Each session of the experiment has sixteen subjects, with twelve assigned as workers and four as firms.1 Each subject participated in only one session and is randomly assigned as a firm or a worker throughout the session. Each session contains two games, one with no MW policy, and the other with a MW policy. Each game is played for fifteen periods (rounds). In each period of the experiment, three of the 12 workers are randomly matched to one of the four firms in the labor market, resulting to four matching groups in one session. Each firm can make wage offers to any number of the three workers they are matched with, and has to pay the hired workers the same wage. The workers can accept or reject the firm’s offer. Following the procedure of Falk et al., the decision is actually made by asking the workers to state their reservation wages after being presented with the firm’s revenue function. If the worker rejects the offer w (i.e., when the worker’s reservation wage is higher than the firm’s offer), he or she is unemployed and earns nothing in this period. If a worker accepts the offer (i.e., when the worker’s reservation wage is lower than the firm’s offer), a binding contract is formed and the worker receives a wage w, and the firms’ revenue increases according to the marginal revenue the worker generates. Each firm’s revenue is the same as shown in Table 1, and this is common knowledge to the workers. Firms’ profits are given by the total revenue minus the total wages. Each subject had to read a detailed set of instructions before the session started. Participants were given several questions about the feasible actions and the payoff consequences of different actions, and were then given answers to review. We also gave two practice periods before the actual session started. The exchange rate between experimental currency units (‘‘points’’) and real money was 450 Points = 1 RMB Yuan (US $0.15). The computerized experiment was programmed and conducted with the experimental software z-Tree (Fischbacher, 1999). At the end of the experiment, participants also answered a survey about their decisions and perceptions. A session lasted approximately one and half hours and subject earned on average 20 RMB Yuan (US $3). 2.2. Treatments Our experiment design consists of between-subject treatments and within-subject treatments. The within treatment is similar to that of Falk et al. (2006) with each subject playing two games in each session. The experiment contains totally 14 sessions that begin with 15 periods in game one without a minimum wage (NO) followed by 15 periods in game two with a minimum wage (MW). The range for permissible wage offers for firms is defined as follows: the constraint 0 6 w 6 1000 prevails in the NO treatment, while they have to obey MW 6 w 6 1000 in the MW treatment and the value of MW differs according to the between-subject design. The between-subject design affects only the MW part, or the second part, of each experimental session. It contains two treatment variables, the level of binding minimum wage and the knowledge of workers about the minimum wage, each taking two levels. This makes a two-by-two experimental design, four treatment conditions. The binding minimum wage in the MW condition is set to be either 170 or 220. The firms always have full knowledge about the minimum wage policy. The 1 We also conducted a pilot experiment with 24 subjects in a session as Falk et al. did, but that severely slowed down the random matching procedure during the experiment, probably due to our network capacity. X. Wang / Journal of Economic Psychology 33 (2012) 951–962 953

workers, however, may have full knowledge (INFORMED) or no knowledge about the minimum wage policy (UNINFORMED).2 The firms also know whether workers have the knowledge about the policy or not. In the UNINFORMED conditions, the instructions to firms for the second part of the experiment include the following passages: ‘‘The second experiment is almost the same with experiment 1, with the exception that your offered salary has to be no less than 170 points, i.e., 170 6 wage offered 6 1000.’’ ‘‘The above mentioned payment regulations apply to all firms and they all know this regulation. But workers are uninformed about it.’’ The instructions to workers for the second part of the experiment state that ‘‘The second experiment is the same with the first experiment’’. Therefore, the workers are unaware of the introduction of the minimum wage, which mocks some real labor market conditions when workers are not informed about the policy change. We consider this an important advantage of using the experimental method to test the effect of asymmetric information since in the real market it is not possible to control whether the workers have knowledge about the minimum wage policy.3 The inclusion of NO condition in all sessions of the experiment provides a comparison benchmark with the experiment of Falk et al. and allows us to examine changes of agents’ positions as the minimum wage is introduced. We run two sessions each of the uniformed and informed treatments with a minimum wage of 220 and five sessions each of the uninformed and informed treatments with a minimum wage of 170. We use more resources for the 170 level since the 220 level has been done in Falk et al. (2006). 2.3. Hypotheses We analyze how minimum wage policy affects the workers’ reservation wages, the firms’ wage offers, the contract wages, and the employment level. The contracts and employment are the results of the workers’ and firms’ decisions. Therefore, the analyses with the reservation wages and offers are important for the understanding of the behavioral effects. If we assume pure rationality and money-maximization, the predicted results will be as discussed in Falk et al. (2006). The selfish workers accept any wage offer above or equal to zero. Firms offer a wage of one (or zero) to all three workers in the NO condition. In the minimum wage conditions, firms are required to pay wages equal to or above the minimum wage. Since both 170 and 220 are less than the marginal product of the third worker, firms offer the minimum wage to all three workers. All three workers are employed in the NO condition and MW condition. Considerable evidence has suggested that the self-interest hypotheses do not predict the behavior well in the case of our experimental game. Social preferences, and the perception of fairness, in particular, play important roles. People have heterogeneous preferences for fairness and reciprocity. This will affect their reservation wages and wage offers in the absence of the minimum wage. One common allocation preference may follow the pattern found in the ultimatum game: the proposer offers on average 70–80% of the equal split to the responder (Camerer, 2003; Fehr & Gächter, 2000). As Falk et al. computed, if a firm hires three workers, an equal split of the revenue of 1000 between the firm and the three workers would be a wage of 250. An offer of 70–80% of 250 implies a wage of between 175 and 200. According to Falk et al., firms’ offers are close to the profit-maximizing wages assuming firms know the distribution of the reservation wages of workers with a fairness concern. As the minimum wage policy is introduced, the decisions of the firms who try to maximize their profit are likely to be driven by three factors: their own social preference, the legal constraint, and their expectations of worker’s fairness concerns. We analyze the impacts of different factors on the firms’ offering behavior according to their different types. Since people have heterogeneous preferences about fairness, some firms’ offers under NO condition would be lower than the minimum wage level required in the MW condition (we name these firms Type 1 firms) and some would be equal to or higher than the minimum level (Type 2 firms). When the minimum wage is introduced, a few changes are expected. First, the legal constraint will force Type 1 firms to raise their wages at least to the minimum level. Second, the firms’ social preferences and their expected fairness concern of the workers (or strategic concern) will determine whether Type 1 firms will raise the wages equal to or beyond the minimum level. Third, social preferences or strategic conTable 1 Firms’ revenue function (Falk et al., 2006). Employed workers Total revenue Marginal revenue 0 0– 1 390 390 2 740 350 3 1000 260 2 Note that each subject is randomly assigned to only one session of one condition. Therefore, it is not possible for one subject to be assigned to both INFORMED and UNINFORMED conditions. Then each subject is randomly assigned as a worker or a firm without changing the roles even though they are randomly matched to form different firm-workers groups for each of the 15 rounds in one session. 3 Some may argue that there can be a learning effect on the part of the workers about the policy from the multiple periods of the game. However, the experiment was done in a short period of time. Even if there is a learning effect in the later periods, it would have worked against our results and have weakened the significance of the test. 954 X. Wang / Journal of Economic Psychology 33 (2012) 951–962

cern will determine whether Type 2 firms will remain at the level they offered under NO condition, or reduce their offer or raise their offer. As evident in Falk et al. (2006), many workers frequently increased their reservation wage above the MW level after the introduction of a MW. This implies an increase of their perceived fair wage. Expecting the workers’ concern for fairness and entitlement, many Type 1 firms will increase their offers beyond the minimum level. Similarly, Type 2 firms are less likely to reduce their offers to the minimum. However, it is possible that Type 2 firms who used to pay higher wages are affected by the reference effect of the minimum wage standard and reduce their offers to a level closer to the minimum. Brandts and Charness (2004) found that when a MW is imposed, the highest wage was chosen only half as frequently in the minimum-wage condition as in the no-minimum condition, even though the average wage was higher in the minimum-wage condition. The higher the MW reference is, the higher the reservation wages are expected, and the higher the offered wages should be. Therefore, based on the condition that the minimum wage is common knowledge to both sides, the following fairness hypotheses will hold. Hypothesis 1. Minimum wage policy will generally increase wage levels. Firms’ offers, reservation wages, and contract wages will be higher in MW condition than in the NO condition. Hypothesis 2. Firms’ offers, workers’ reservation wages, and contract wages will be higher in the 220 MW condition than in the 170 MW condition. When workers are unaware of the MW policy, we predict different behavior on the side of the firms. In the UNINFORMED condition, the firms’ consideration of the effects of minimum wage on workers’ fairness perception is no longer as important as in the INFORMED condition. We predict the following changes for different groups: First, more Type 1 firms will increase their offer to just the MW level rather than higher; Second, more Type 2 firms that have paid wages higher than the MW are likely to reduce their offers to a wage level closer or equal to the current MW. This prediction is based on how fairness concern is affected by information access and on how fairness is driven by self-serving bias. Previous studies have shown that fairness concern is given more weight under complete information than under incomplete information (Babcock, Loewenstein, Issacharoff, et al., 1995; Babcock, Loewenstein, & Wang, 1995; Güth & Damme, 1998). Furthermore, people tend to arrive at judgments of what is fair or right that are biased in favor of their own self-interests (Babcock, Wang, & Loewenstein, 1996), especially under ambiguous conditions (Babcock, Loewenstein, Issacharoff, et al., 1995; Babcock, Loewenstein, & Wang, 1995; Haisley & Weber, 2010). The study of Babcock et al. (1996) directly showed that the choices of comparison references by the negotiators for salary contracts are influenced by their self-serving bias in fairness perceptions. Self-serving bias implies that, when the MW reference provides to the firm an argument to adjust their offer to a lower level than before, they are inclined to perceive the lower offer to be fairer. Since workers are not aware of the minimum wage in UNINFORMED, they will not induce changes in fairness perceptions from the minimum wage. In contrast, workers reservations wages will increase in INFORMED condition. The combination of the decisions by firms and workers will predict higher wages in INFORMED conditions. We arrive at the following hypothesis from above analysis. Hypothesis 3. When the minimum wage is common knowledge, firms’ offers and workers’ reservations are higher; and they will exceed the minimum level more frequently than when workers are not aware of the minimum wage. Hypothesis 4. Firms whose offers in the no-minimum-wage condition are higher than the minimum wage standard are more likely to reduce their offers toward the minimum level in the UNINFORMED condition than in the INFORMED condition. The impact of minimum wage on employment is positive or negative (Falk et al., 2006, p. 1359). If the minimum wage has a strong enough impact on reservation wages, it is possible for firms to reduce employment. The theoretical argument for non-negative impact of minimum wage on employment is the existence of monopsony in the labor market (Manning, 2003). According to this model, dominant employers in a particular labor market could actually increase their employment in response to minimum wage increases. Such firms are generally reluctant to raise their wages to attract new recruits because they have to pay the higher wage to their existing incumbent workforce to maintain internal equity. When minimum wages are imposed, they no longer face such constraint because they now have to pay all of their low-wage workers the same minimum wage. Therefore, they are no longer reluctant to hire workers that they otherwise would not have hired because they would have to raise wages to hire such workers and they would have to pay those higher wages to their existing workforce as well. 2.4. The questionnaire At the end of each experimental session, we asked the subjects to answer a questionnaire to complement our experimental analysis. The questions are mostly concerned with the subjects’ strategic concerns and fairness perceptions. X. Wang / Journal of Economic Psychology 33 (2012) 951–962 955

When the minimum wage is present, we ask the subjects what their view of fair wage is. If they switch their own role of worker or firm to the other side, what would be their offers or expected reservation wages? We will examine how these perceptions differ in different minimum wage treatments. 3. Results 3.1. Summary and specification issues We focus our analyses on the effects of minimum wage policy first on firms’ offers, workers’ reservation wages, and then on settled wages, and employment. Table 2 reports the mean values of these variables in eight cells, including four betweentreatment conditions and two within-treatment conditions. Obviously, the wage values are higher in the INFORMED conditions and the 220 conditions compared with the UNINFORMED conditions and the 170 conditions. In our analyses of the data in the following sections, we need to consider a few estimation issues. First, since observations within a session may be dependent on each other, in all our regressions, we report the robust standard errors that are clustered on sessions. Second, to control for individual specific effects, the fixed effects method or random effects method may be considered in the analyses with firms’ offers and workers’ reservations. Because the explanatory variables in our analyses are all experimental treatment dummy variables that do not vary across periods, fixed effects method cannot be used. The random effects method would be appropriate assuming that individual effects are uncorrelated with the error terms. Since the results we have obtained are almost the same for the random effects and pooled OLS method, we omit the random effects in the report. Third, since subjects play the game for multiple periods, they may experience some learning effects even though they are re-matched for different groups. This is especially possible for workers in the UNINFORMED conditions who are not informed about the minimum wage policy. There are two possible ways to mitigate this potential problem. First, we can include the lagged dependent variable in the regression analysis. This, however, will lead to the problem of endogeneity and inconsistent estimation of parameters when the sample size is small. The second solution is to include the time dummy variables in the regressions. We repeated our regressions for firm offers and workers reservations with the period dummy variables, and did not find any changes with the results of our estimates. Therefore, the reports below omit the results with time dummy variables. 3.2. Introduction of the minimum wage This section examines the within-subject effect of introducing the minimum wage (MW condition) after agents have played the game in NO condition. To compare our experiment with the results in Falk et al. (2006), we shown in Fig. 1 the histogram of contract wages before and after the introduction of the minimum wage in the 220 INFORMED treatment. The average wage in the 220 INFORMED MW condition is 232.7 with a standard deviation of 15.5, compared to the average of 199 in the NO condition with a standard deviation 23.7. While most wages in the NO condition are below the 220 level with a median of 200, most wages in the MW condition are higher than 220 with a median of 228. Only 10% of the wages in the MW condition are exactly at the minimum level of 220. To test the significance of introducing the minimum wage policy, Table 3 reports regression results with firm offers, workers’ reservation wages, paid wages, and employment as the dependent variables. The regressions are run separately for the 170 conditions and the 220 condition. The estimates for the MW dummy are all significantly positive in the 220 conditions, but not significant for firm offers and paid wages in the 170 conditions. At this point, we have not included the treatment variable INFORM in the regressions, this may have resulted in the insignificant effect of the minimum wage on firms’ offers and paid wages in the 170 conditions with INFORMED and UNINFORMED pooled together. We next run the above regressions for 170 INFORMED and 170 UNINFORMED conditions separately. The results indicate that the estimate for the MW dummy is significantly positive for firm offers in the 170 INFORMED condition (7.2667 with a p-value 0.0405), but not significant in the 170 UNINFORED condition (2.030 with a p-value 0.8314). The difference between the INFORMED and UNINFORMED conditions is consistent with our Hypothesis 3 that the minimum wage has a stronger impact in the INFORMED condition. However, the impact of the MW dummy is stronger in the 220 conditions than in the 170 conditions, which is consistent with our Hypothesis 2. We analyze the impacts of information and the level of minimum wage in more detail in the next section. 3.3. Knowledge about minimum wage policy and minimum wage levels The next result concerns how workers’ knowledge about the minimum wage and minimum wage level affects the wages and employment results. We test the hypotheses that wages are positively affected by the binding minimum wage level and workers’ full knowledge about the minimum wage policy. We first examine the pattern of firm offers. Figs. 2 and 3 indicate that firm offers that are higher than the MW levels are more frequent under INFORMED MW conditions than under UNINFORMED MW conditions. In the INFORMED MW cases, only 10–20% of the offers are exactly at the minimum level in the 170 condition or the 220 condition. In the UNINFORMED 956 X. Wang / Journal of Economic Psychology 33 (2012) 951–962

MW cases, these come to about 42% for the 170 condition and 70% for the 220 condition. Comparing the 170 conditions and 220 conditions, fewer offers in the 220 conditions are higher than the minimum wage level than in the 170 conditions. This indicates that firm offers are affected both by the reference point and the absolute values of the minimum wage. This is consistent with the empirical finding in the field by Du and Pan (2009) that the coverage by the hourly minimum wages among workers decreased from year 2001 to year 2005 as the minimum wage standards were raised. Workers’ reservations show similar changing patterns as above from NO to MW conditions when the minimum wage is introduced with workers INFROMED. When workers are not aware of the minimum wage policy, their reservation wages are little affected by the introduction of the minimum wage (histograms omitted). Table 2 Averages of the dependent variables in eight cells. Offered wage Reservation wage Paid wage Employment 170 Informed NO 186.92 142.58 192.85 2.33 [n = 300] [n = 900] [n = 697] [n = 300] (34.63) (60.82) (33.09) (0.73) MW 194.19 175.66 195.44 2.64 [n = 300] [n = 900] [n = 792] [n = 300] (19.59) (31.73) (19.76) (0.55) 170 Uninformed NO 176.99 131.48 182.13 2.45 [n = 300] [n = 900] [n = 736] [n = 300] (47.56) (63.14) (45.60) (0.69) MW 179.02 131.00 179.18 2.74 [n = 300] [n = 900] [n = 821] [n = 300] (11.37) (56.16) (11.52) (0.46) 220 Informed NO 196.53 151.44 199.02 2.50 [n = 120] [n = 360] [n = 300] [n = 120] (23.58) (54.49) (23.71) (0.64) MW 231.71 218.46 232.69 2.70 [n = 120] [n = 360] [n = 321] [n = 120] (14.97) (23.99) (15.45) (0.51) 220 Uninformed NO 178.90 129.33 184.83 2.26 [n = 120] [n = 360] [n = 271] [n = 120] (34.15) (64.97) (29.83) (0.68) MW 222.03 140.01 221.92 2.87 [n = 120] [n = 360] [n = 342] [n = 120] (5.25) (75.08) (5.05) (0.33) Note: Standard deviations are in parentheses. Fig. 1. Distribution of contract wages in the absence and the presence of a minimum wage level of 220. X. Wang / Journal of Economic Psychology 33 (2012) 951–962 957

To test the significance of the impacts of information asymmetry and minimum wage levels. Table 4 presents the regression results with wages and employment level as dependent variables.4 Workers’ knowledge about the minimum wage significantly increases wage offers, reservation wages, and paid wages. The minimum wage level also significantly increases firms’ wage offers, and paid wages. These results confirm Hypotheses 2 and 3. We also tested the interaction effect of the minimum wage and workers’ knowledge and did not find it significant. The employment level is negatively affected by INFORM and positively affected by LEVEL. Since the two factors that cause wage increases affect employment in different directions, the impacts of minimum wage on employment may be ambiguous. Since we have different number of sessions run for the 170 conditions and 220 conditions, we also examine the impact of INFORM separately for the two MW levels. The results show that the impact of INFORM is not significant on employment level in the 170 condition, and is significantly negative in the 220 condition. INFORM impacts the other dependent variables positively at the 1% significance level in both conditions. 3.4. Sources of differences between treatments In the following, we examine the behavioral pattern of workers and firms depending on their choices before and after the introduction of minimum wage. This will help us decompose the sources of positive or negative spillover effect of the minimum wage. We distinguish those firms whose initial offers were lower than the minimum wage in the No condition (Type 1 firms) from those who already gave offers equal to or higher than the minimum wage level (Type 2 firms). If firms who used to offer wages lower than the MW level under the NO condition actually increase their offers beyond the MW level under the MW condition, we can expect a positive spillover effect of minimum wage. If firms who used to offer wages higher than the MW level under the NO condition reduce their offers because of the MW reference, we can expect a negative spillover effect of MW on wages. Those who used to offer wages equal to the MW level could also be affected by the reference of the MW level and increase their offers. In the 170 conditions (including INFORMED and UNINFORMED), 31% of the firms belong to Type 1, and the rest Type 2. In the 220 conditions, 93% belong to Type 1. Table 5 reports how the two groups of firms switched their offer levels from NO condition to MW condition. We omit the 220 conditions from the table for the sake of clearer presentation since the 220 conditions show a similar pattern. In the 170 INFORMED conditions, 89% of the Type 1 firms and 90% of the Type 2 firms gave offers above the MW level when the policy was introduced (76.9% or 100% of the two types did this in the 220 conditions). This is consistent with the finding of Falk et al. (2006) that, with common knowledge, the minimum wage policy has a positive spillover effect of minimum wage. In contrast, in the UNINFORMED condition, only 46.1% of Type 1 firms and 63.5% of Type 2 firms gave offers above the MW level in the 170 conditions (only 27.8% and 20% of the two types did this in the 220 conditions). Among those who used to offer higher than or equal to the MW level (Type 2 firms), 36.5% (from the 170 condition) or 80% (from the 220 condition) of Table 3 Effects of introducing a minimum wage. Offered wage Reservation wage Paid wage Employment 170 MW dummy 4.65 16.30*** 0.18 0.30*** (5.15) (6.02) (5.65) (0.03) Constant 181.96*** 137.03*** 187.25*** 2.39*** (6.95) (7.87) (7.70) (0.03) N 1200 3600 3046 1200 Prob > F 0.0124 F <0.0001 <0.0001 <0.0001 <0.0001 R2 0.4152 0.0829 0.4077 0.1152 Note: OLS estimation with robust standard errors clustered on sessions in parentheses. The minimum wage dummy is 1 if a minimum wage exists and 0 otherwise. Random effects have obtained the same parameter estimates and do not change the significance tests. Adding period dummy variables does not change the results either. ⁄⁄ Indicates significance at the 5% level. *** Indicates significance at the 1% level. 4 This regression tests the between-subject effect, including only the observations of the second game of each session when MW is imposed. We can also include the observations of the first game and control for the within-subject effect by including the MW dummy. Then the effect of IMFORM does not change much except it is no longer significant for employment level. The MW dummy is significantly positive for all dependent variables. 958 X. Wang / Journal of Economic Psychology 33 (2012) 951–962

them now offer wages at the minimum level. In fact, in the 170 condition, 85% of those firms whose initial offers were higher than 170 have reduced from their previous offer in the NO condition (significantly higher than the 43% in the INFORMED condition). This means that, when workers are unaware of the policy, the minimum wage can have a negative spillover effect. We will test the significance of this impact further. We also do the same analysis for workers categorized as Type 1 or Type 2 according to their reservation wages in the NO conditions. For both Type 1 and Type 2 workers, higher percentage of them increases their reservations beyond the minimum wage level in the INFORMED conditions than in the UNINFORMED conditions (35% vs. 8.7% for Type 1, and 73% vs. 44% for Type 2 in the 170 conditions; 28% vs. 4.4% for Type 1, and 50% vs. 10.5% for Type 2 in the 220 conditions). To further examine the impact of the information asymmetry on firms’ behavior, we perform a logit regression analysis for the binary dependent variable HIGHER, which is equal to 1 when the firm’s offer is higher than the MW or zero when the offer is just equal to the MW when the MW policy is imposed. The results presented in columns (1) and (2) of Table 6 indicate that, when workers are informed about the MW policy, firms are more likely to offer wages that are higher than the MW level. When the MW constraint is high, firms are less likely to offer above the MW level. We also perform the above regressions for the HIGHER variable separately for the 170 conditions and 220 conditions. The impact of INFORM is still positive at the 1% significance level for both groups. In column (3) of Table 6, for firms whose previous offers in the 170 NO condition are higher than the 170 MW level, we specifically test how INFORM affects the probability that they reduce their offer from their previous position when the policy is introduced, measured by the binary dependent variable LOWER. It is shown that INFORM strongly reduces this probability. In fact, among the 327 firm observations in the 170 group whose initial offers were higher than 170, 85% of them reduced their offers in the UNINFORMED condition while only 43% of them reduced their offers in the INFORMED condition. Only 13 firm observations in the 220 group offered higher than the minimum in the NO condition, and all of those (3 counts) in the UNINFORMED condition reduced offers while only 10% (1 out of 10) in the INFORMED condition did so. This confirms our Hypothesis 4. Fig. 2. Distribution of firm offers with or without the knowledge of workers about the MW 170. Fig. 3. Distribution of firm offers with or without the knowledge of workers about the MW 220. X. Wang / Journal of Economic Psychology 33 (2012) 951–962 959

We conclude from this section that the knowledge of workers about the MW policy strongly increases the offers given by the firms. When workers do not have knowledge about the policy, low minimum wage standard may have a negative reference effect, causing some firms to reduce their offers from their previous offers. This helps explain why in Section 3.2, we find that the minimum wage policy does not increase firms’ offers in the 170 UNINFORMED condition. 3.5. The questionnaire – strategies and fairness perceptions In this part, we report the results of the survey about the agents’ strategic concerns and their fairness perceptions filled by the agents after they have played the games. Table 4 Effects of minimum wage level and worker’s knowledge about MW. Dependent variable Offered wage Reservation wage Paid wage Employment Inform 13.6016*** 54.3107*** 14.6593*** 0.1183*** (4.4443) (11.2722) (4.5541) (0.0481) Level 40.2661*** 25.9027** 40.0682*** 0.0978*** (3.3517) (11.9422) (3.4557) (0.0375) Constant 179.8025*** 126.1756*** 179.9670*** 2.7475*** (3.3861) (11.3990) (3.4784) (0.0413) Number of obs. 840 2520 2276 840 Prob > F 0.0001 0.0001 0.0001 0.0001 R2 0.6319 0.2648 0.6328 0.0226 Note: OLS estimation with robust standard errors clustered on sessions. Standard errors are in parentheses. Random effects have shown the same parameter estimates and very similar standard errors and do not change the significance tests. Adding period dummy variables does not change the results either. ** Indicates significance at the 5% level. *** Indicates significance at the 1% level. Table 5 Firm offers from NO condition to MW condition. Count (%) 170 INFORMED 170 UNINFORMED Offer 2 = MW >MW Offer 2 = MW >MW Offer 1 ChiSq) 438.5457 434.69079 212.2986 ( MW, HIGHER = 0 when Offer 2 = MW. LOWER = 1 if firms who used to offer higher than MW level have reduced their offer, LOWER = 0 otherwise. ** Indicates significance at the 5% level. *** Indicates significance at the 1% level. 960 X. Wang / Journal of Economic Psychology 33 (2012) 951–962