Competitiveness, Affirmative Action and Taxes

There are three treatment dimensions:

Difficulty level (between subjects): easy vs difficult

Compensation scheme (within subjects): i) piece rate ii) tournament iii) tournament with affirmative action iv) tournament with taxation

Binary choice between compensation scheme (between subjects): i) piece rate vs pure tournament ii) piece rate vs tournament with affirmative action iii) piece rate vs tournament with taxation iv) pure tournament vs tournament with affirmative action iv) pure tournament vs tournament with taxation

2024-06-13

2024-12-31

1) Preferences of compensation scheme
2) Individual productivity across compensation schemes
3) Group level productivity across compensation schemes

1) is measured as the fraction of individuals choosing each compensation scheme in the binary choice settings
2) is measured as the number of correctly solved tasks
3) is measured in terms of the sum of tasks correctly solved within a group

1) How primary outcomes vary across difficulty levels
2) How primary outcomes vary with demographics (age, gender, etc)
3) How primary outcomes vary with preferences (risk, social, competitivness etc)

We propose a lab-experiment to study the role of affirmative action and taxation on the competitiveness of individuals. Under a series of compensation schemes our subjects will have to find zeros in a series of matrices of zeros and ones. There are two different difficulty levels, easy (10x10 matrices) and difficult (large 10x15 matrices). Before the start of the actual treatment, all subjects have to perform 2 minutes of practice under both difficulty conditions. We will then randomly introduce variation in the productivity of our subjects. Half of the participants will solve the task under easy conditions and the other half under difficult conditions. Subjects will be informed about their difficulty condition and this assignment will remain constant for a given subject throughout the experiment.

In the actual experiment, all of our subjects will first encounter their assigned task under four different compensation schemes for 5 minutes and we will measure productivity (number of matrices correctly solved). We will randomize the order of these four compensation schemes (within subject). The compensation schemes we study encompass:

i) piece rate incentives,
ii) tournament incentives where two subjects with easy and two subjects with difficult tasks compete and only the two winners get paid (and subjects are aware of the group composition)
iii) tournament incentives (see ii) with affirmative action where those with the difficult task are given a compensatory boost
iv) tournament incentives (see ii) with taxation where the winners are taxed and the proceeds are redistributed.

After subjects have completed the four tasks, they each face a binary choice between two compensation schemes and complete the task under the chosen scheme in the fifth and final round. Our main treatment dimension (between subjects) is the binary choice subjects are given in the last round and consist of:

i) piece rate vs pure tournament
ii) piece rate vs tournament with affirmative action
iii) piece rate vs tournament with taxation
iv) pure tournament vs tournament with affirmative action
v) pure tournament vs tournament with taxation

Additionally, we will collect information on demographics (gender, age, etc) and use a survey to elicit preferences (risk, social, competitiveness, etc) and beliefs on their performance in the tournament conditions.

Of primary interest will be how discriminated and non-discriminated subjects decide in these binary choices and how productivity varies among different compensation schemes. Further, we are interested how these choices are related to individual characteristics (gender, age etc) and survey elicited measures of preferences (risk, competitiveness, over-confidence etc).

Not available

Randomization is performed within the software (oTree) on the fly.

The randomization is done at the subject level.

No

500 subjects

500 observations, one primary observation per subject.

Since the true randomization is performed, we expect the five treatments to have equal number of subjects, i.e. 100 subjects per treatment.