Analyzing Labor Market Outcomes and Florida DCF’s Role in ...

27
1 Florida State University Analyzing Labor Market Outcomes and Florida DCF’s Role in the Integration of Florida’s Refugee Benefit-Eligible Individuals By: Samantha Kunin A Thesis submitted to the Department of Economics in partial fulfillment of the requirement for graduation with Honors in the Major Degree Awarded: Spring, 2019

Transcript of Analyzing Labor Market Outcomes and Florida DCF’s Role in ...

1

Florida State University

Analyzing Labor Market Outcomes and Florida DCF’s Role in the Integration of Florida’s

Refugee Benefit-Eligible Individuals

By:

Samantha Kunin

A Thesis submitted to the Department of Economics

in partial fulfillment of the requirement for graduation with

Honors in the Major

Degree Awarded: Spring, 2019

2

The members of the Defense Committee approve the thesis of Samantha M. Kunin defended on April 10th, 2019. Signatures are on file with the Honors in the Major Office.

______________________________ Dr. Anastasia Semykina

Thesis Director

______________________________ Dr. Samuel Staley

Committee Member

______________________________ Prof. Mark Schlakman

Outside Committee Member

4

“Immigrants’ ability to communicate with members of the indigenous population is

probably the most important single alterable factor contributing to their social and

economic integration” (Dustmann and Van Soest 472).

Introduction

In the case of the United States, the rapidly shifting composition of the immigrant

population from countries that are sociolinguistically distant from the U.S. has motivated

a recent surge of interest in analyzing the more qualitative question of English language

proficiency across immigrant cohorts and national origin categories (Kulkarni and Hu).

These differences in language are compounded by broader trends such as the post-

industrial economy’s increasing demand for skilled employees (e.g. those with that least

technical/vocational training), changes in immigration policy, and concerns about a

relative decline in immigrant human capital in terms of language ability. Best estimates

suggest that fluency in English increases employment probabilities by about 22

percentage points. OLS estimates by Dustmann and Fabri (696), for example, show that

proficiency in English is associated with 18-20% higher earnings.

Baran (2018) states that, in the United States, employment is the cornerstone of

the refugee resettlement process because of its centrality in providing self-sufficiency.

However, there exists under- and unemployment of immigrants due to language barriers

and other factors. This is a common phenomenon referred to as “brain waste” and can

be attributed to the stringent credentialing procedures common in many industrialized

countries (Sienkiewicz 18).

This project seeks to further understand the effect of English proficiency on labor

market outcomes for refugee benefit-eligible individuals in Florida. It also analyzes and

provides recommendations for the complexities of data collection within governmental

5

agencies coordinating refugee resettlement. Based on current research, higher

recorded English Proficiency should result in a higher percent change in wages for

refugees. The empirical results – support the hypothesis that a higher background level

in English (along with being male and having a higher prior education level) result in a

positive change in their likelihood of employment as well as wages.

United States Reception & Placement Program

The United States maintains the largest resettlement program globally. In Fiscal

Year (FY) 2016, the U.S. admitted nearly 85,000 refugees, a number which has since

decreased to a cap of 45,000 refugees in FY2018. The State Department’s Reception

and Placement (R&P) Program provides refugees with a loan to travel to the United

States, which they are required to start repaying after they arrive. The R&P Program

then supplies resettlement agencies a one-time sum per refugee to finance their first 30-

60 days in the U.S. These resettlement agencies differ in financial allocation of

resources, budgets, number of staff, availability of interns, as well as geographical

responsibility. In the United States, the Department of State appoints nine voluntary

agencies (VOLSAGS) to work in either refugee referrals or resettlement: International

Organization for Migration, Church World Service, Episcopal Migration Ministries,

Ethiopian Community Development Council, Hebrew Immigrant Aid Society (HIAS),

International Rescue Committee (IRC), Lutheran Immigration and Refugee Service,

United States Conference of Catholic Bishops (USCCB), U.S. Committee for Refugees

and Immigrants, and World Relief. From the date of arrival, the Office of Refugee

Resettlement (ORR) at Health and Human Services provides short-term cash and

medical assistance to new arrivals, as well as case management services, English as a

6

Foreign Language classes, and job readiness and employment services. These

programs are designed to facilitate refugees’ successful transition in the U.S. and help

them to attain self-sufficiency.

Florida’s Refugee Resettlement Program

The Department of Children and Families’ Refugee Services Program is federally

funded by the Office of Refugee Resettlement (ORR) within the Department of Health

and Human Services to assist refugees achieve economic self-sufficiency and facilitate

social adjustment within the shortest possible time. Florida’s refugee program is the

largest in the nation, receiving more than 27,000 refugees, asylees, and Cuban/Haitian

entrants each year. However, the majority of Florida’s refugee program participants are

Cuban and Haitian entrants who do not have access to the standard resettlement

assistance that is available to newly-arrived, definitive refugees and asylees. These

individuals enter the United States through the Cuban-Haitian Entrant Program (CHEP),

a federal program administered by U.S. Citizen and Immigration Services (USCIS),

which thereby gives them “humanitarian parole.” Humanitarian parole is an immigration

status that enables the recipient to be treated the same as refugees for purposes

eligibility for public benefits including Temporary Aid for Needy Families (TANF) and

Medicaid.

USCIS coordinates the reception, processing, and community placement of

Cubans and Haitians paroled into the United States, accounting for more than 80% of

the refugees in Florida. Refugees receiving “Reception & Placement” assistance

through Voluntary Agencies under contract to the U.S. Department of State represent

less than 25% of Florida’s refugee population. For this reason, the following analysis

7

considers both individuals on humanitarian parole (e.g. Cuban and Haitian entrants) as

well as those receiving assistance through Florida’s Reception and Placement program.

While refugees from a wide variety of countries arrive to the state of Florida with

a diverse set of backgrounds and varying professional skills, and many have

transferable skills, “not all skills sets are deemed valuable.” (Sienkiewicz 18)

Employment services are thus provided to assist eligible refugees/entrants in achieving

economic self-sufficiency and effective resettlement through gainful employment.

Services primarily target refugees in their first two years in the United States, but

refugees remain eligible for up to 60 months. Employment services include pre-

employment counseling and orientation, direct job preparation and placement, 90 and

180 day follow-up, On-the-Job Training (OJT), re-credentialing/re-certification, and

career laddering services for refugees with professional backgrounds.

Within Florida, the Department of Children and Families hosts regional meetings

known as Refugee Task Force Meetings. Community Liaisons facilitate Refugee Task

Force meetings in each community with large numbers of refugees. The Task Forces

meet bi-monthly and include refugee resettlement agencies, contracted providers,

federal, state and local government agencies, refugee-led self-help organizations, and

other entities and individuals concerned with refugees.

Academic Literature Review

English language comprehension for refugees is cited most frequently as the

greatest challenge to obtaining gainful employment. Refugees with lower levels of

English proficiency are less likely to be employed (Sienkiewicz 18). Not surprisingly, the

8

wages of refugees with limited English comprehension are lower than refugees who are

more proficient in the language (Sienkiewicz 18).

“Role shock” or “role strain” occurs when the high status attributed to one’s

previous profession is lost or not valued in the resettlement country. Consequently,

many refugees decide to pursue new careers, despite the lower earning potential, as

they do not have the time or money to invest in upgrading their original credentials

(Sienkiewicz 19). Educated refugees with non-transferable qualifications then must

compete for the same jobs as unskilled workers. The loss of social and financial status

associated with their previous occupations may result in anger, stress, and frustration.

Refugee arrivals in a host community typically represent an increase to the

supply of workers, making the labor market one of the most important channels of

impact (Mayda 4). While many studies have analyzed the effects of the presence of

immigrants, little literature exists on refugee outcomes in relation to English-learning,

especially within localized areas. Using evidence from the U.S. Resettlement Program

to analyze the labor market outcomes of refugees and their outcomes did not begin until

2017.

Regarding English proficiency for refugees, estimates of language coefficients in

OLS regressions are plagued by two methodological issues: the choice to acquire

proficiency in a foreign language may be endogenous, and the language measured

reported in survey data may suffer substantially from measurement error (Dustmann

and Fabri 696). These limitations are discussed further and addressed in Methodology

and Limitations. In order to increase the robustness and significance of their results,

studies tend to increase the number of variables used in the regressions in order to see

9

if one independent variable is related to another when considering the effect on the

dependent variable. For example, simple regressions may solely consider the effect of

job position on wages (being that, in a Management position, you would expect to have

higher earnings than individual working in Agriculture). However, job location is

considered (e.g. New York City vs. Tallahassee), geography may have a greater effect

on wages than the job position, lessening the propensity that job title impacts wage.

Therefore, the following analysis considers multiple factors in addition to English

proficiency: gender, job title, county of resettlement, country of origin, and education

level.

With regards to gender, research shows that the labor market participation of

women from their countries of origin is extremely low (Dourleijn and Dagevos 2011;

Jennissen and Oudhof 2008). This may be due to practical reasons, such as family care

taking and household duties, or to traditional values on gender roles (Bakker 1779). Job

positions, even for native-born workers, has been shown to increase wage earnings and

the likelihood of employment.

Large variations by ethnicity or source country have been observed among the

children of immigrants. This heterogeneity has been interpreted as the result of

differences in vulnerability and resources between immigrant groups, in terms of

individual and family socioeconomic background—particularly education and official

language ability—and group cultural and community characteristics (Hou and

Bonikowska 6). Therefore, country of origin was important to consider in this paper to

determine if there was an additional effect.

10

Godoy (2017) notes two distinct ways in which local labor market conditions can

affect later outcomes. First, there could be effects through persistence on the individual

level, i.e., effects on early experience or individual scarring effects of unemployment. In

the first case, people who are placed in a bad labor market will gain less early

experience and accumulate less country-specific human capital. They perform worse in

the labor market in the future, regardless of the improved state of the labor market.

Second, there may be an effects persisting on the local level from extended

unemployment combined with limited geographical mobility. In this case, people who

are placed in a bad labor market will be more likely to experience difficult conditions

later, even if there are no effects on them as individuals (2015). For this reason, county

of resettlement was considered in the regressions. OLS regressions were the most

feasible manner to assess these indicators with respect to Florida refugee benefit-

eligible individuals’ earnings and likelihood of employment and coincided with the

conclusions of the Literature Review.

Data

The data was obtained from the Florida Department of Children and Families in

September 2018 and was revised until February 2019. After three meetings with DCF,

as well as continuous email correspondence, the data was as complete as DCF was

able to guarantee. Initial data collection began in November 2018 and was completed in

February 2019. This data set has a total of 254,888 unique values, and there were

120,498 that had reported wages for use. The database includes all of the variables

needed for OLS regressions. However, a significant amount of work was required to

modify the raw variables into a form appropriate for STATA manipulation. The data was

11

originally sent in Excel or in raw ZIP files, which I had to save as a different format in

STATA to be able to open them. The data included: Client ID, Provider (VOLAG) Name,

Gender, County of Resettlement, Country of Origin, Verified Employment Date,

Indicator if Individual speaks English, English Proficiency, Prior Education Level, Job

Title, Wage Amount, ESL Program Type, and ESL Program Name spanning the years

2008-2018 and these variables are explained further in the Appendix.

Observing the effect of English proficiency on percent change of wages required

each client have an assigned wage amount. For ORIG_GRADE_LEVEL, the variables

were grouped and then generated into new variables that indicated whether the Client

had received primary, secondary, post-secondary, or advanced education. For

ORIGIN_COUNTRY, the countries were grouped based on geographic region and on

education rankings from the Education Index of the 2016 Human Development Report.

For JOB_TITLE, the job positions were divided into categorical groups consistent with

those defined by 2018 Occupational Outlook Handbook from the Bureau of Labor

Statistics (i.e. an Accountant is categorized as being a position in Business and

Finance.) For COUNTY, ENGLISH_LEVEL, and PROVIDER_NAME, each of the

possible values (i.e. COUNTY = Leon, ENGLISH_LEVEL = 1, PROVIDER_NAME =

International Rescue Committee) were coded to see if differences in values affected

wages. For ENGLISH_SPEAKING, GENDER, and EMPLOYED, dummy variables were

created to indicate either one outcome or the other (i.e. affirmatively speaks English or

not, is female or not, or is employed or not.) Dummy variables, also known as binary

variables, are variables that take on two values: one or zero. For example, with

GENDER, female is a dummy variable taking on the value one for females and the

12

value zero for males. Using these variables leads to regression models where the

parameters have easy-to-understand interpretations which will be expanded upon in

Findings.

Continuous email correspondence with FDCF from August 2018 to April 2019

ensured that all possible data points were available to conduct regressions. While data

points spanned from 2008 to 2018, the Department of Children and Families had not

verified that they had collected the same panel data from the same clients. Thus, in

most years, they were not able to get further data about their current employment from

the same client. The Department of Children and Families relies on self-reporting

surveys, and the Department continues to investigate best practices to resolve these

issues for other research endeavors, including contracting with outside vendors.

Unfortunately, FDCF’s coding does not provide a feasible method for evaluating the

merits of ESL.

Originally, the data had been identified as panel data, which is a dataset in which

the behavior of, in this case individuals, are observed over time (Princeton). Panel data

allows you to control for variables you cannot observe or measure like cultural factors or

national policies, but most researchers note these factors is rarely attainable due to

non-response in the case of micro panels, as was the case with this dataset.

13

Above: the frequency in which a CLIENT_ID was found to be repeated, meaning that

time-series data was available to indicate change in wage for some clients. However,

53.33% of the clients were found to only have one year’s data and only 28.8% of clients

had two years of data. Studying change over time from 2008-2018 becomes

increasingly difficult without access to robust panel data. In addition, English proficiency

did not vary over time - meaning that the individuals did not report having a higher

English proficiency with time. However, given these limitations, the methodology was

straightforward. Regressions could still be performed with the available data to see if

expected results gleaned from national and international studies could be applied to

those living in Florida.

Methodological Approach

Multiple variable regressions were used to measure change in earnings. Using

logarithmic regressions, I showed approximate percentage change in wages for a one-

unit increase in the explanatory variable and can be modeled as:

log(𝑤𝑎𝑔𝑒) = 𝛽0 + 𝛾𝐸𝑃 + 𝛽1𝐸𝑛𝑔𝑙𝑖𝑠ℎ + 𝜇

14

Y is the dependent binary variable, β(0) is the intercept of the model, English is a

dummy variable equal to one if the individual speaks English, and zero otherwise. This

is a linear probability model, because Y = 1 or 0, and β cannot be interpreted as the

change Y given a one-unit increase in any independent variable. For EP, I created a

dummy variable equal to 1 if EP = 1 and 0 if not, another equal to 1 if EP = 2 and 0 if

not, and so on. In addition, a non-binary variable that ordinally ranked EP 0-4 was also

tested to see if there was a statistically significant difference between testing English

Proficiency using a dummy or nonbinary variable. The reference group used was EP=0,

indicating that the individual had no English proficiency. The five World-Class Instruction

Design and Assessment (WIDA) English Language Development Standards (ELD) are

used for EP. EP=1 indicates Entering language proficiency, EP=2 indicates Beginning,

EP=3 indicates Developing, EP=4 indicates Expanding, EP=5 indicates Bridging, and

EP=6 indicates Reaching. Following the explanations of this regression, the hypothesis

in notations is , where the presence of English proficiency above 0 has a positive

effect on the log earnings of refugee benefit-eligible individuals.

In reality, there were several variables that were included a more representative model

is:

log(𝑤𝑎𝑔𝑒) = 𝛽0 + 𝛾𝐸𝑃 + 𝛽1𝐸𝑛𝑔𝑙𝑖𝑠ℎ + 𝛽2𝑒𝑑𝑢𝑐 + 𝛽3𝑔𝑒𝑛𝑑𝑒𝑟 + 𝛽4𝑐𝑜𝑢𝑛𝑡𝑦 + 𝛽5𝑜𝑟𝑖𝑔𝑖𝑛 + 𝛽6𝑗𝑜𝑏 + 𝜇

Where education, county of resettlement, country of origin, and job title are all relevant

productivity characteristics, and the null hypothesis of no difference between English

Proficiency (EP) levels is: 𝐻0: 𝛿 = 0. The alternative that increased English Proficiency

has an effect is 𝐻1: 𝛿 > 0.

15

In addition, another regression estimated the likelihood of employment given their

ability to speak English and English proficiency:

𝐸𝑀𝑃𝐿𝑂𝑌𝐸𝐷 = 𝛽0 + 𝛾𝐸𝑃 + 𝛽1𝐸𝑛𝑔𝑙𝑖𝑠ℎ + 𝜇

The actual model that specifies all the variables, while omitting education since that

data was not given by DCF, is:

𝐸𝑀𝑃𝐿𝑂𝑌𝐸𝐷 = 𝛽0 + 𝛾𝐸𝑃 + 𝛽1𝐸𝑛𝑔𝑙𝑖𝑠ℎ + 𝛽2𝑔𝑒𝑛𝑑𝑒𝑟 + 𝛽3𝑐𝑜𝑢𝑛𝑡𝑦 + 𝛽4𝑜𝑟𝑖𝑔𝑖𝑛 + 𝛽5𝑗𝑜𝑏 + 𝜇

And the original model upon which this is based, the linear probability model, is:

This regression has also been simplified to explain the standard model for an

OLS regression. But the full regressions also examined the the effect of gender, prior

education levels, countries of origin, counties in which they are resettled, job positions

held in Florida, as well as their resettling VOLSAG. The education of non-working

individuals was not reported in the data, so education was not included in the

employment regression.

While the raw data was originally in Excel, the data were consolidated into one

file using STATA so that commands could be executed simultaneously. While other

methods were considered such as analyzing time-series data using fixed and random

effects and/or taking lags and differences of time-series data, the dataset was too

unbalanced due to the aforementioned gaps in DCF’s data collection.

When the Office of Refugee Resettlement (ORR) conducted an analysis on the

labor market outcomes for refugees in the United States, an issue that they attempted

to address was endogeneity, and in particular, reverse causality. As was mentioned in

Dustmann and Fabbris’ papers, methodological concerns are common. They tried to

16

address these problems by combining an instrumental variable (IV) estimator that

eliminates the bias with a matching estimator that addresses the problem of

endogenous choice of language acquisition (2005). The methodology originally tried to

follow that same approach by looking at English programs in which individuals were

enrolled, but they were classified as being in the same type of program. I use controls to

account for individual differences in education, professional qualifications, and cultural

backgrounds.

Data Analysis

The results, in their entirety, are displayed in tables in the appendix. Overall, the

results show an increasingly positive effect on wages when the Client has a higher prior

English level. Two OLS regressions examined the effect of English proficiency on

wages and the effect of English proficiency on the likelihood of the Client being

employed. For Entering English learners, wages increased by 0.5%; For Beginning

learners, wages increased by 3.2%. Developing learners experienced a 7.4% increase

in wages. Expanding learners experience a 7.6% increase in wages. The increase in

wages proportionally increases with a higher English proficiency. I also was able to test

for multicollinearity, which showed that all of the explanatory variables had a Variable

Inflaction Factor (VIF) less than 5, except for the variables for the years in which the

data was collected (e.g. 2008, 2009…) which was coded as T10, T11, etc. When also

testing to see if English Proficiency was still significant using a non-binary variable

instead of a dummy, there was still a 1.4% increase in wages and a P-Value of 0.

Yet, within certain subsets, the analysis shows that other characteristics do as

well (i.e. gender.) The difference in predicted wages between men and women was

17

about 7.2%, favoring a comparable man’s wage. Wages also increased with prior

education background. Those refugees who had completed high school experienced a

1.4% increase in wages compared to those who did not. Those with incomplete post-

secondary education had a 2.5% increase in wages compared to those without who had

completed elementary school. Those who had completed post-secondary schooling as

opposed to those without formal education had a 7.9% increase in wages. Using

management jobs as a comparison group, every job group had an associated increase

in wages except for Production workers (e.g. Butcher and Meat Cutter, Baker, Etcher).

The results for Life, Physical, and Social Science (e.g. Agriculture), Personal Care (e.g.

Clerk, Baggage Porter), and Installation (e.g. Helper - Carpenter, Helper - Electrician)

were statistically insignificant.

When considering countries of origin, the results show that those in the Middle

East and Africa tend to have lower wages, while those from Latin America and Asia

tend to have higher wages. This could be attributed to the potentially better education

systems that exist within these regions. Looking at the counties in which they were

resettled, those resettled in East Central Florida tend to have the highest increase in

wages, but the results are statistically insignificant with a P-Value of 0.351. The most

robust results can be seen in Job Title, Country of Origin, Gender, Prior Education

Level, and English Levels.

When considering the effect of English proficiency on the likelihood of being

employed, the individual with higher language facility was more likely to be employed

than not (recall, in this regression, Y = [EMPLOYED (1) or NOT EMPLOYED (0)] an

increased likelihood as English proficiency increased. Being an Entering English learner

18

increases the probability of employment by 0.066. Being a Beginning learner increases

the probability of employment by 0.061. A Developing learner increases the probability

of employment by 0.084. Expanding learners experienced a 0.047 increase in the

probability of employment. The increase in the probability of being employed

proportionally increases with a higher English proficiency.

Yet again, within certain subsets, the analysis shows that other characteristics

also influence the likelihood of employment (i.e. gender.) Women had a .055 unit lower

likelihood of being employed. When considering their countries of origin, the results

showed that all were likely to increase employment when compared to the reference

group Eastern Europe (whose countries tend to have the highest education level in

comparison to the rest of the geographic regions), and the results were statistically

significant. Looking at the counties in which they were resettled, refugees resettled in

North Central Florida tend to have the highest increase in the probability of employment,

but the results are statistically insignificant with a P-Value of 0.486, as are the other

results for County of Resettlement. Again, the most useful results can be found by

comparing English proficiency and gender.

The dependent variables in the two regressions are different, and the explanatory

variables are not the same because education was not included in the employment

regression. Therefore, the results are different, which is to be expected. There were

several variables that could account for why some results are statistically insignificant,

but the most glaring reason is missing values.

19

Limitations

Limitations in drawing conclusions from the analysis have been addressed as

recommendations for the Florida Department of Children and Families and their

associated resettlement agencies. Specifically, the recommendations include amending

the way and the frequency with which they collect information data for analysis. Given

newer and more complete datasets, instrumental variables may provide a better

estimate of the effect of English Proficiency, providing subgroup results with greater

statistical significance. While English proficiency is positively associated with a higher

percent increase in wages for these individuals, this association is also due to the

diligence and pressure that these families experience in needing to find the best

positions possible to provide for their typically large families and improve their own

outcomes.

Throughout the process of completing this research, navigating STATA, and

learning about the complexity of being able to provide a robust analysis of labor market

outcomes for refugee-benefit individuals in Florida, the most difficult component is

realizing the reformation that needs to firstly occur on a state level.

The Florida Department of Children and Families is inundated with a variety of

responsibilities, but checking the completeness of the survey data that they collect is not

necessarily a priority. Despite frequent contact with the DCF Contract Manager, the

department’s analysts had no control over what was actually missing. A portion of this

analysis should acknowledge and recommend that a restructuring of the current data

collection system to ensure that researchers can provide accurate results. Better data

could increase the information that they can include in reports to the U.S. government,

as well as help them understand which professions, educational backgrounds,

20

geographical backgrounds, and regions of Florida are performing better in terms of

successful refugee resettlement and integration.

Conclusion

The findings reinforced the expectation that higher recorded English Proficiency

will result in higher wages and higher likelihood of employment. I believe that the

analysis provided is essential to the discussion of the impacts of various characteristics:

gender, county of resettlement, education level, and job position are significant for both

increases in wages as well as likelihood of employment. Further consideration can be

taken in refugee resettlement agencies and within the U.S. Department of State for

these characteristics that likely impact their labor market success in their communities.

Most importantly, these agencies should focus on not just promoting, but ensuring the

progression of English proficiency for refugee benefit-eligible individuals in the State of

Florida. Demonstrating that there is a positive effect on percent change in wage and

likelihood of employment associated with assessed English proficiency provides

evidence for advocates of further investment into access to English training as a tool for

better job acquisition and retention.

21

References

Bakker, L., Dagevos, J., and Engbersen, G. (2017). “Explaining the refugee gap: a longitudinal

study on labour market participation of refugees in the Netherlands.” Journal of Ethnic &

Migration Studies, 43(11), 1775–1791. https://doi-

org.proxy.lib.fsu.edu/10.1080/1369183X.2016.1251835.

Baran, B. “Survival, expectations, and employment: An inquiry of refugees and immigrants to

the United States.” Journal of Vocational Behavior, vol. 105, 2018, pp. 102-115.

https://doi.org/10.1016/j.jvb.2017.10.011.

Dustmann, C. and Fabbri, F. (2003). “Language Proficiency and Labour Market Performance

of Immigrants in the UK.” The Economic Journal, vol. 113, no. 489, p. 695. EBSCOhost,

login.proxy.lib.fsu.edu/login?url=http://search.ebscohost.com/login.aspx?direct=true&db

=edsjsr&AN=edsjsr.3590195&site=eds-live&scope=site.

Dustmann, C. , Fabbri, F. and Preston, I. (2005). “The Impact of Immigration on the British

Labour Market.” The Economic Journal, 115: F324-F341. doi:10.1111/j.1468-

0297.2005.01038.x.

Dustmann, C., and Van Soest, A. (2002). “Language and the earnings of immigrants.”

Industrial and Labor Relations Review, 55: 473-92.

Godoy, A. (2017). “Local labor markets and earnings of refugee immigrants.” Empirical

Economics. https://doi-org.proxy.lib.fsu.edu/10.1007/s00181-016-1067-7.

Hou, F., and Bonikowska, A. (2016). “Educational and Labour Market Outcomes of Childhood

Immigrants by Admission Class.” Analytical Studies Branch Research Paper Series.

https://files-eric-ed-gov.proxy.lib.fsu.edu/fulltext/ED585303.pdf.

22

Kulkarni, V. S. and Hu, X. (2014). “English Language Proficiency Among the Foreign Born in

the United States.” 1980–2007: Duration, Age, Cohort Effects. Int Migr Rev, 48: 762-

800. doi: 10.1111/imre.12060.

Mayda, A.M., Parsons, C., Giovanni P., and Wagner, M. (2017). “The Labor Market Impact of

Refugees: Evidence from the U.S. Resettlement Program.” U.S. Department of State

Office of the Chief Economist. Working Paper No. 2017-04. Washington, DC.

https://www.state.gov/documents/organization/273699.pdf.

T. Bivens, personal communication, September-March 2019.

Torres-Reyna, O. (2007). “Panel Data Analysis Fixed and Random Effects using Stata.”

Princeton University. http://www.princeton.edu/~otorres/Panel101.pdf.

WIDA. (2019). “English Language Development Standards.” WIDA.

https://wida.wisc.edu/teach/standards/eld.

23

Appendix

24

25

26

27