Population Projections: Applications in Immigration Research Jennifer Van Hook Penn State University June 2019
Why use Projections in immigration research? Data Limitations Unauthorized foreign-born population Population Lever Immigration Impacts Generational Change Starting points & mobility
Unauthorized foreign-born population With Randy Capps, Julia Gelatt, and Anne Morse
…Fazel-Zarandi, Feinstein and Kaplan’s approach is very sensitive to underlying assumptions…their estimates range widely and are too uncertain to be of value for policy purposes. …the knowledge base about key inputs for their model…is insufficiently developed to support their approach.
questions What accounts for the variation in estimates? Point estimates are interpreted literally: What is the margin of error? Tension: uncertainty, decision making
Unauthorized FB = Total FB – Legal FB residual Method Unauthorized FB = Total FB – Legal FB
Converts Flows to Stocks Ut = Unauthorized FB, post-1980 foreign-born arrivals Ct = ACS estimate, post-1980 arrivals – legal nonimmigrants Lt = LPRs (time t) = Admissions since 1980 + Refugees who have not yet adjusted – Deaths – Emigrants etL = coverage error, legal foreign-born etU = coverage error, unauthorized foreign-born Converts Flows to Stocks
Cohort component method
Mexican Males Admitted and entered US in 1995 at age 20: Admissions = 1,191 Year time in US age Cohort size, Jan 1 deaths emigrants Cohort size, Dec 31 1996 1 21 1,191 2 16 1,173
Mexican Males Admitted and entered US in 1995 at age 20: Admissions = 1,191 Year time in US age Cohort size, Jan 1 deaths emigrants Cohort size, Dec 31 1996 1 21 1,191 2 16 1,173 1997 22 18 1,154
Mexican Males Admitted and entered US in 1995 at age 20: Admissions = 1,191 Year time in US age Cohort size, Jan 1 deaths emigrants Cohort size, Dec 31 1996 1 21 1,191 2 16 1,173 1997 22 18 1,154 1998 3 23 19 1,133
Mexican Males Admitted and entered US in 1995 at age 20: Admissions = 1,191 Repeat for ~500,000 admission cohorts (sex x age at arrival x year of entry x year of admission x country of birth x refugee status)
Ut = Unauthorized FB, post-1980 foreign-born arrivals Ct = ACS estimate, post-1980 arrivals – legal nonimmigrants Lt = LPRs (time t) = Admissions since 1980 + Refugees who have not yet adjusted – Deaths – Emigrants etL = coverage error, legal foreign-born etU = coverage error, unauthorized foreign-born
Uncertainty and variation in Mortality
Probability distribution for Hazard Ratio, Hispanic men, 0-4 years in U.S.
Uncertainty and variation in Emigration
Probability distribution for FB Emigration (0-4 years in U.S.)
Uncertainty and variation in Coverage Error OIS: 10% Pew: 13% Van Hook et al (2015): Mexican-born, 1995- 2010 3 methods Death Registration Birth Registration Net Migration
Probability distribution for UNAUTHORIZED COVERAGE ERROR, men 2016
Variation in ESTIMATES
Possible constraints Internal Consistency undercount cannot be negative undercount for LPRs must be less than unauthorized Should assumptions be correlated (across groups, across assumptions)? External Consistency with other estimates direct measurement, net migration from sending countries border apprehension/Visa overstay trends
Ethical quandaries Transparency ------------------------------------Certainty Policy makers need to make decisions, want certainty Scientific disputes undermine public confidence. Yet transparency is essential for scientific advancements How can we discuss uncertainty, while still assuring people that our estimates are better than guessing?