Year of Publication


Degree Name

Doctor of Philosophy (PhD)

Document Type

Doctoral Dissertation


Business and Economics



First Advisor

Dr. Christopher Bollinger

Second Advisor

Dr. Aaron Yelowitz


Considerable interest has been placed on the subject of what to do with the sizable undocumented population currently residing in the United States. Having a full understanding of the size of this population and the impact immigration policy has on them is of critical importance to policy makers. Data limitations in nationally-representative surveys have limited the analyzes of the effects of immigration policy in the academic literature. As such, this dissertation consists of three essays contributing to the literature on the impact of immigration policy and on identifying unobserved populations.

In my first essay, I examine the labor market response of undocumented youth that participated in the Deferred Action for Childhood Arrivals (DACA) program which provides them temporary deportation relief and work authorization. I use data from the U.S. Citizenship and Immigration Services to construct a probabilistic measure for unobserved DACA participation. Using the American Community Survey (ACS), I estimate a two-sample model of the effect of participating in the DACA program. I also estimate spillover effects of DACA on eligible but non-participating undocumented youth. I find that DACA significantly improved labor market and education outcomes of DACA recipients, with magnitude of the treatment-on-the-treated effects at least twice as large as the intent-to-treat estimates obtained from using only the observed eligibility indicator typically used in literature. Evidence of a negative spillover effect on eligible non-participants is documented with a decrease in labor force participation and school attendance.

My second essay considers nonsampling error due to item nonresponse in the estimates of the size and legal composition of the foreign-born population produced using the ACS. The standard practice to address item nonresponse is to impute values under the assumption that nonresponse is conditionally random. I form credible interval estimates that make no assumptions about the values of missing data by considering all uncertainty due to item nonresponse. Without this assumption, the size of the foreign-born population in the US falls somewhere between 40.4 and 59.4 million as of 2019 compared to the Census estimate of 44.9 million. Bounding estimates of the size of the undocumented population fall between 7.3 and 23.3 million compared to the widely accepted estimate of 11 million undocumented immigrants.

In my third essay, I return to analyzing the effects of the DACA program. I examine misclassification bias arising from item-nonresponse in the estimated intent-to-treat effects of the DACA program. Assigning DACA eligibility is based on the responses to specific demographic questions, any of which the individual may not respond to. If the assumption that nonresponse to these questions are conditionally missing fails, this can lead to traditional misclassification bias and attenuate the results when using imputed values. Adjusting for potential misclassification bias by removing non-respondents leads to estimates of the intent-to-treat effects of DACA on labor market outcomes that are 22% to 77% higher than when including non-respondents, depending on the outcome of interest.

Digital Object Identifier (DOI)