We are happy to announce that all trial registrations will now be issued DOIs (digital object identifiers). For more information, see here.
Liquidity Affects Job Choice: Evidence from Teach For America
Last registered on May 13, 2019


Trial Information
General Information
Liquidity Affects Job Choice: Evidence from Teach For America
Initial registration date
May 01, 2019
Last updated
May 13, 2019 11:40 PM EDT
Primary Investigator
Harvard University
Other Primary Investigator(s)
PI Affiliation
The Wharton School
PI Affiliation
The Wharton School
PI Affiliation
Harvard University
Additional Trial Information
Start date
End date
Secondary IDs
The main research question is whether many Americans lack the liquidity to accept a job offer. We test this by randomly varying the size of grants and loans in an existing transitional funding program provided to early career job applicants. The hypothesis is that even small increases in cash on hand -- in the form of grant or loan money -- will increase the likelihood of accepting the job. We implement the experiment with Teach For America, a teacher placement organization. The experiment takes place over three years with 7,295 admits to TFA who apply for and are eligible for all transitional funding. Transitional funding decisions are made in “batches”, about once per week from October through June. We randomize to treatment arms within each batch. In the first two years, the treatment arms provide modest increases in transitional funding by increasing either the grants or loans provided. The third year is a “self-replication” and stress test of the first two years of results. To self-replicate, for those who exhibited a treatment effect in the first two years (and those in similar, but slightly better financial shape), we maintain the same treatment assignments. To stress test the null results, for those who did not exhibit a treatment effect, we tripled the size of the main experimental variation. Across all three years, we find that for admits whose expected costs of moving exceed their estimated cash on hand, providing immediate financial assistance increases their odds of accepting their TFA offer. Further, we find no treatment effects, even for large financial assistance, when provided to admits whose cash-on-hand exceeds their expected costs. We measure “accepting the job” by whether or not the job applicant is working through Teach For America as of the first day of school on the first semester of their assignment.
External Link(s)
Registration Citation
Coffman, Lucas et al. 2019. "Liquidity Affects Job Choice: Evidence from Teach For America." AEA RCT Registry. May 13. https://doi.org/10.1257/rct.4168-1.0.
Former Citation
Coffman, Lucas et al. 2019. "Liquidity Affects Job Choice: Evidence from Teach For America." AEA RCT Registry. May 13. https://www.socialscienceregistry.org/trials/4168/history/46421.
Sponsors & Partners

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details
The experimental population is all of those admitted to TFA during the 2015--2016, 2016--2017, and 2017--2018 admissions cycles, who applied for transitional grant and loan (TGL) funding, and were eligible for grants according to TFA’s formula.
Applications to TFA and the TGL funding were done on a rolling basis, and treatment assignment was done in lockstep. TGL packages and randomization to treatment were calculated in “batches” of applicants, done roughly once per week.
In 2015--2016, admits were randomly assigned to treatments 1--3 below.
In 2016--2017, until March 15, 2016, admits were randomly assigned to treatments 1--3 below, and after March 15, 2016, admits were randomly assigned to treatments 1--4 below.
In 2017--2018, admits in the bottom quintile of TFA’s measure of “expected contribution” (see below) were assigned to treatments 1--4 below, and admits in the top four quintiles of “expected contribution” were assigned to treatments 1, 5, and 6 below.
All randomized assignment was done with equal probability for each potential treatment.

The full list of treatment groups is:
1. Control: The standard TFA TGL package based on the admit’s financials, less a uniform amount compared to the previous year to fund the experimental variation.
2. $600 Loan: $600 in extra loan money on top of the control grant and loan award
3. $600 Grant: $600 in extra grant money on top of the control grant and loan award
4. $1,200 Grant: $1,200 in extra grant money on top of the control grant and loan award
5. $1,800 Loan: $1,800 in extra loan money on top of the control grant and loan award
6. $1,800 Grant: $1,800 in extra grant money on top of the control grant and loan award
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
The primary outcome of interest is whether the admits join TFA, as measured by being a TFA teacher as of the first day of school of the first semester of the assignment.
Primary Outcomes (explanation)
We receive periodic data from TFA about the “disposition” of applicants, meaning whether they are still on track teaching in the program. The disposition variable TFA sends is a categorical variable that indicates whether the applicant is still in the program and, when they are not, the last milestone they reached before they left. We count an applicant as having joined TFA if, in the September after they were admitted to the program, TFA lists that they had begun teaching for the Fall semester. In particular, if applicants’ disposition in September includes “Y1S”, TFA’s code for having begun teaching, applicants are counted as having joined TFA.
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
Across all three years, treatment assignment was done as explained in 22. Applicants were notified of their TGL funding package via email. Experimentally provided increases to funding were not mentioned. No other interaction with the admits was made until the post-experiment survey.

The data were analyzed after year two. Year three was an intentional self-replication attempt based on those results.
Experimental Design Details
Randomization Method
Randomization was done via random number generator in Stata. Specifically, for each batch of applicants, applicants were sorted into a random order, and the treatment groups were put in a random order. Then, the first applicant was assigned the first treatment, the second applicant to the second treatment, and so on, with the order of treatments repeated until all applicants had been assigned to a treatment. Note that this ensures that, within batch, no treatment can be assigned more than one more applicant than any other treatment.

In 2017, this process was done separately for the bottom quintile of expected contribution (who were randomized between four treatment groups, including control) and for the remaining quintiles (who were randomized between three treatment groups).
Randomization Unit
Individual-level randomization
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
Sample size: planned number of observations
Sample size (or number of clusters) by treatment arms
Sample sizes (number of individual admits)
Control: 2,232
$600 Loan: 1,795
$600 Grant: 1,784
$1,200 Grant: 413
$1,800 Loan: 525
$1,800 Grant: 546
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB Name
University of Pennsylvania Office of Regulatory Affairs
IRB Approval Date
IRB Approval Number
Post Trial Information
Study Withdrawal
Is the intervention completed?
Is data collection complete?
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers