|
Field
Trial Start Date
|
Before
April 18, 2023
|
After
May 15, 2023
|
|
Field
Trial End Date
|
Before
December 31, 2023
|
After
February 01, 2024
|
|
Field
Last Published
|
Before
April 08, 2023 11:31 AM
|
After
May 12, 2023 10:41 PM
|
|
Field
Intervention (Public)
|
Before
We are running a lab experiment with a setting where participants can discover how to specialize. We study if market efficiency and specialization differ between groups with different race and gender compositions. We are also interested in market size effects.
Update April 8th 2023:
We have decided to run an online experiment on mTurk and/or Prolific instead. There are two main reasons. First, a more diverse subject pool will allow us to recruit people from different ethnicities into our markets. We initially thought this was possible at the UT Dallas lab, but after examining the subject pool in more detail, we learned that the number of white/Caucasian subjects would be limited. We also plan a larger number of treatments than initially planned. Therefore, a larger online subject pool is beneficial. Second, we believe homophily effects may occur more naturally in a general population setting rather than at the laboratory where all participants are students and are used to interacting across ethnicities.
|
After
Update May 12th 2023:
We plan to run an online experiment on Prolific instead of the initially registered lab experiment. Three considerations led to this change: First, we need a diverse subject pool, and we learned that the UT Dallas lab is not sufficiently mixed. Second, we now plan to have more treatments than initially planned. A more extensive online subject pool is beneficial. Third, homophily effects may occur more naturally in a general population setting than in the laboratory where all participants are students.
|
|
Field
Intervention Start Date
|
Before
April 18, 2023
|
After
May 15, 2023
|
|
Field
Intervention End Date
|
Before
December 31, 2023
|
After
February 01, 2024
|
|
Field
Primary Outcomes (End Points)
|
Before
Efficiency
Efficiency in the last 14 periods
Degree and speed of specialization
Number of same-race (or gender) versus across-race (or gender) specialized pairs in Mixed treatment
Dropouts during the experiment conditional on race
|
After
Efficiency
Efficiency in the last 10 periods
Degree and speed of specialization
Number of same-race (or gender) versus across-race (or gender) specialized pairs in Mixed treatment
Dropouts during the experiment conditional on race
|
|
Field
Primary Outcomes (Explanation)
|
Before
Efficiency is the percentage of realized gains relative to the most efficient outcome (competitive equilibrium).
Specialization is the shift in time devoted to the good for which a participant has a comparative advantage relative to the autarky outcome.
Specialization across race or gender is the degree to which it happens between individuals of different races/genders.
Dropouts/attrition can be a problem for online experiments. For us, it's a possible outcome variable. There are 4 or 8 people in a market, and if some drop out, it hinders specialization, but the market can continue. Dropouts reflect low anticipated earnings, a dislike for the interactions, etc., or randomness. So, if, on average, white (respectively, Asian) participants are more likely to drop out in Mixed than in Homogenous, this is part of the treatment effect we want to capture.
|
After
Efficiency is the percentage of realized gains relative to the most efficient outcome (competitive equilibrium).
Specialization is the shift in time devoted to the good for which a participant has a comparative advantage relative to the autarky outcome.
Specialization across race or gender is the degree to which it happens between individuals of different races/genders.
|
|
Field
Randomization Method
|
Before
Participants of the UT Dallas LBOE lab subject pool are invited randomly using the recruitment software SONA.
Update: Participants are recruited via Prolific and/or mTurk.
|
After
Participants are recruited via Prolific
|
|
Field
Randomization Unit
|
Before
Markets (groups of 4 or 8)
|
After
Markets (groups of 4)
|
|
Field
Planned Number of Clusters
|
Before
25-30 markets per treatment. (updated from 20 as the online environment allows us to collect more data.)
|
After
15-20 markets per treatment.
|
|
Field
Planned Number of Observations
|
Before
100-120 unique participants per treatment, about 500 subjects.
|
After
ca. 500
|
|
Field
Sample size (or number of clusters) by treatment arms
|
Before
100-120 participants per treatment.
|
After
60-100 unique participants per treatment
|
|
Field
Power calculation: Minimum Detectable Effect Size for Main Outcomes
|
Before
Wilcoxon-Mann-Whitney test:
Efficiency in the last 14 periods: minimum detectable effect size is 0.74, the unit is the proportion relative to competitive equilibrium (between 0 and 1), and the expected standard deviation is 0.23. 50% increase over expected baseline efficiency (from 50% to 70%).
Specialization: same.
Wilcoxon matched-pair test:
Number of same-race versus mixed-race pairs in Mixed treatments: minimum detectable effect size is 0.37, the unit is the number of specialized pairs per market (between 0 and 2), and the expected standard deviation is 0.92 for same-race and 0.74 for mixed.
|
After
|
|
Field
Intervention (Hidden)
|
Before
We have two main treatments. Markets have 4 participants and last 35 rounds (updated from 48 initially due to moving to the online environment). A participant's task in each round is to choose how much time to devote to producing Goods Orange and Blue. They can discover possibilities for trade and specialization to generate gains from trade. Everyone chooses an image that most closely represents them. Images differ in race and gender and are selected from the Chicago Face Database.
In the Baseline (Homogenous) condition, markets are homogenous (i.e., all South-Asian, all white). In the Mixed condition, markets consist of two pairs that are superficially similar within pairs and different across pairs. For example, a market has 2 white and 2 South Asian players. Importantly, in Mixed, gains from trade are designed to be larger for cooperation between dissimilar people.
We also plan to increase the market size to 8 people to study the impact of competition and market size on specialization. We further plan to manipulate the observability of people's actions, which will increase search costs and potentially amplify homophily effects.
|
After
We have three main treatments. Markets have 4 participants and last 40 rounds. A participant's task in each round is to choose how much time to devote to producing goods Orange and Blue. They can discover possibilities for trade and specialization to generate gains from trade. Everyone chooses a photograph that most closely represents them. Images differ in race and gender and are selected from the Chicago Face Database.
In the homogenous market condition, all market participants are of the same race. In the mixed market condition, markets consist of two 2 white participants (of the same sex) and 2 black (or south Asian) participants. There are two versions of the mixed market condition. In one, the main gains from trade are available within-race, while in the other condition, they are found across-race.
We also plan to change the observability of people's actions as another treatment dimension, which will increase search costs and potentially amplify homophily effects.
|
|
Field
Secondary Outcomes (Explanation)
|
Before
We will analyze the chat people use to enter trading relationships. Specifically, we will analyze how active and sociable people are, and if the discussions are positive/negative.
Image selection is the photograph people select.
|
After
We will analyze the chat people use to enter trading relationships. Who chats with whom? And how active and sociable are people?
Image selection is the photograph people select.
|