|
Field
Primary Outcomes (End Points)
|
Before
Our primary outcome of interest is whether an individual worker chooses to remain in a 'simple' job or to transition to a 'harder' job (binary choice).
We will stratify our analyses by race and gender as well as by skill level, as our initial pilot and prior work suggests that treatment effects will be heterogeneous by skill.
|
After
Our primary outcome of interest is whether an individual worker chooses to remain in a 'simple' job or to transition to a 'harder' job (binary choice).
We will perform pooled as well as separate analyses by ethnicity and gender -- we do not expect treatment effects to look the same when comparing men to women as when comparing URM to non-URM, and there may also be heterogeneity within those groups e.g. white versus Asian, Black versus Hispanic.
We will also look at treatment effects stratified by skill level, as our initial pilot and prior work suggests that treatment effects will be heterogeneous based on skill.
|
|
Field
Experimental Design (Public)
|
Before
Our treatment changes the descriptions of the jobs workers are considering, thus changing the stakes and consequences of failure. We have two cross-randomized dimensions of treatment:
1) Demographic Reveal: Under Demographic Reveal, the worker’s demographic information (gender and race) is revealed to employer subjects along with their score and their Prolific work history. Under No Demographic Reveal, only the worker’s anonymized quantitative attributes are revealed.
2) Separation Risk: Under Separation Risk, the worker might get “fired” (i.e. not be invited back to complete the second round of puzzles). Under No Separation Risk, the worker will always be invited back to complete the second round and their performance will be used to train employers in a “practice round” of evaluation.
Our final design consists of four treatment cells.
|
After
Our treatment changes the descriptions of the jobs workers are considering, thus changing the stakes and consequences of failure. We have two cross-randomized dimensions of treatment:
1) Demographic Reveal: Under Demographic Reveal, the worker’s demographic information (gender and race) is revealed to employer subjects along with their score and their Prolific work history. Under No Demographic Reveal, only the worker’s anonymized quantitative attributes are revealed.
2) Separation Risk: Under Separation Risk, the worker might get “fired” (i.e. not be invited back to complete the second round of puzzles). Under No Separation Risk, the worker will always be invited back to complete the second round and their performance will be used to train employers in a “practice round” of evaluation.
Our final design consists of four treatment cells. Treatment will be stratified by gender (Men and Women) and race (URM versus non-URM). URM in our context means a reported ethnicity of "Black" or "Other" (as opposed to "White" or "Asian") on the platform where we're recruiting. Note: "Other" predominantly includes those who identify exclusively as Hispanic, Latino, American Indian, or Alaska Native.
|