A Cross-Cultural Real-Effort Experiment on Wage-Inequality Information and Performance

Burchett and Willoughby (2004) is an exception. Their experiment was conducted in the United Arab Emirates.

The exchange rate between Yuan RMB and Euro was around 9–10 at the time when the experiment was conducted.

For example, in task A the computer screen presented a series of numbers: “12, 10, 18, 54, 52, 60, 180, ___”. Subjects were required to calculate the number after “180” according to the pattern revealed by the previous numbers. The relationship between the consecutive numbers is as follows: 12−2 = 10, 10+8 = 18, 18×3 = 54, 54–2 = 52, 52+8 = 60, 60×3 = 180. The algorithm repeated here is subtracting 2, adding 8, and multiplying by 3. Therefore, the right answer is 178 (180–2 = 178). In task B, the computer screen presented a series of letters, such as “H C D E K Q”, and subjects were required to arrange these letters in alphabetical order. Thus, the correct answer is “C D E H K Q”.

Instructions are available from the authors upon request.

In an appendix, available from the authors upon request, we analyze this model more formally in a game-theoretic framework. We show there that when a Nash equilibrium such that and exists, that equilibrium is unique. However, if there is a Nash equilibrium such that and , other Nash equilibria may also exist for some parameter values. In those Nash equilibria, and such that Note that in any such equilibrium . In the appendix, we also show that any such equilibrium is Pareto inferior to the equilibrium where and .

In an appendix, available from the authors upon request, we analyze more formally a game-theoretic model with both disadvantageous and advantageous inequities in the utility functions of both the h and the l players. We show there that if, for a set of parameter values, a Nash equilibrium in which l’s monetary income is below h’s monetary income exists, that equilibrium is unique with . However, if, for a set of parameter values, a Nash equilibrium in which l’s monetary income equals h’s monetary income exists, there will in general be multiple equilibria. In all of those equilibria, l’s monetary income will equal h’s monetary income. Thus, in such equilibria, neither player is receiving disutility from any form of inequity. Any such equilibria for which the incomes of l and h are between l’s self-interested maximum level of income, , and h’s self-interested maximum level of income, , are not Pareto-rankable. This is because being closer to is better for l, while being closer to is better for h. In all such equilibria, , qualitatively the same prediction as is given by the equilibrium where . However, it is also possible that equilibria for which the incomes of l and h are equal exist outside of this range. We show in the appendix that any equilibrium in which the equal income levels of l and h are below will be Pareto-dominated by an equilibrium where the incomes of both l and h are equal to . This is because there are no inequity considerations in either equilibrium since incomes are equal in both cases. However, at , both l and h are closer to their self-interested maxima. Analogously, any equilibrium at an income level above will be Pareto-dominated by an equilibrium where the incomes of both l and h are equal to . We thank an anonymous referee for suggesting we analyze the multiple equilibria that can arise in the case of both disadvantageous and advantageous inequities.

As in Bartling (2011), if the cost of effort is partially but not completely accounted for in assessing disadvantageous inequity, low-piece-rate participants would increase their effort levels relative to uninformed participants but not by as much as if the sole concern was a monetary-income comparison.

We also examine differences in average performance between high-piece-rate and low-piece-rate participants in a panel-data regression framework with clustered errors for each participant, analogous to the framework introduced below for comparing the performance of low-piece-rate participants among the different treatments. Despite a lack of statistical power owing to the small number of high-piece-rate participants, the results are consistent with the Mann–Whitney results. For Germany, they indicate no performance differences between high- and low-piece-rate participants. For China, they show a significantly lower increase in performance over time for the high- than for the low-piece-rate participants. These results are not reported here to save space, but are available from the authors upon request.

We thank two anonymous referees for this observation.

As a robustness check, we also ran the same regressions using two slightly different methods to account for the lack of independence across periods for each participant: in one, we used random effects for each participant and in the other we used a combination of both random effects and error clustering by participant. The coefficients are identical for each method, while the standard errors differ slightly. There are no qualitative differences in any of the statistical inferences obtained by any of the three methods.

The tests on both countries combined reported in Table 4 are done by first running a regression for both countries combined, clustering the errors by participant. A dummy variable for country is used and interacted with each of the independent variables. To save space, the results of the regression itself are not reported here, because they are identical to the reported results of the regressions run separately for Germany and China. However, the hypothesis tests reported in Table 4 are based on this regression for both countries combined.

In the first five periods, the intercept for the UPRI treatment is Constant + Infotrmt + Upri. In the last five periods, the intercept for the UPRI treatment is Constant + Infotrmt + Upri + Lasthalf + Infotrmt × Lasthalf + Upri × Lasthalf. The difference between the intercept in the last versus the first five periods is Lasthalf + Infotrmt × Lasthalf + Upri × Lasthalf = 2.05 − 1.24 + 3.41 = 4.22.

In the first five periods, the intercept for the EPRI treatment is Constant + Infotrmt. In the last five periods, the intercept for the EPRI treatment is Constant + Infotrmt + Lasthalf + Infotrmt × Lasthalf. The difference between the intercept in the last versus the first five periods is Lasthalf + Infotrmt × Lasthalf = 2.05 − 1.24 = 0.81.

In the first five periods, the slope of the trend line for the UPRI treatment is equal to Period + Infotrmt × Period + Upri × Period. In the last five periods, the slope of the trend line for the UPRI treatment is equal to Period + Infotrmt × Period + Upri × Period + Lasthalf × Period + Info × Lasthalf × Period + Upri × Lasthalf × Period. The difference between the slope of the trend line in the last versus the first five periods is Lasthalf × Period + Info × Lasthalf × Period + Upri × Lasthalf × Period = −0.37 + 0.11 − 0.49 = −0.75.

In the first five periods, the slope of the trend line for the EPRI treatment is equal to Period + Infotrmt × Period. In the last five periods, the slope of the trend line for the UPRI treatment is equal to Period + Infotrmt × Period + Lasthalf × Period + Info × Lasthalf × Period. The difference between the slope of the trend line in the last versus the first five periods is Lasthalf × Period + Info × Lasthalf × Period = −0.37 + 0.11 = −0.26.

As in Footnote 12, the difference between the intercept in the last versus the first five periods for the UPRI treatment is Lasthalf + Infotrmt × Lasthalf + Upri × Lasthalf. For China, this equals 2.71 + 0.48 + 3.45 = 6.64.

As in Footnote 13, the difference between the intercept in the last versus the first five periods for the EPRI treatment is Lasthalf + Infotrmt × Lasthalf. For China, this equals 2.71 + 0.48 = 3.19.

As in Footnote 14, the difference between the slope of the trend line in the last versus the first five periods for the UPRI treatment is Lasthalf × Period + Info × Lasthalf × Period + Upri × Lasthalf × Period. For China, this equals –0.59–0.16–0.50 = –1.25.

As in Footnote 15, the difference between the slope of the trend line in the last versus the first five periods for the EPRI treatment is Lasthalf × Period + Info × Lasthalf × Period. For China, this equals –0.59–0.16 = –0.75.