CAPÍTULO III. RESULTADOS
3.3. Propuesta de mejora productividad
Overall, this paper highlights that assessing the effects of policies that target unauthorized immigrants, such as E-Verify, requires a good measure of unauthorized immigrants. While the H.L.E. is a subgroup of the unauthorized population that is very likely to be unautho-rized, the H.L.E. is not a representative sample of the entire unauthorized population. In the setting of population changes and universal E-Verify, the H.L.E. and “logical edits” show a reduction in the unauthorized population but subgroup analysis is inconsistent across the definitions. Misclassification error is a likely explanation for these results. When applied to the setting of employment and E-Verify mandates, the results show that the measure of unauthorized immigrants matters. Results are sensitive to the sample chosen (in labor force v. all adults). Given these inconsistent findings, it is clear that the H.L.E. do not accurately the effect of E-Verify mandates on the likelihood of employment. Moreover, the results here suggest that additional years of data are important in capturing the effect of E-Verify mandates.
The main question left to answer is what a researcher should do if data on unauthorized immigrants is not available. One potential solution is to derive the misclassification prob-abilities of H.L.E.’s relative to unauthorized immigrants as measured through the “logical edits.” These probability could then be used to correct estimates using H.L.E. Another is-sue arises for samples where the “logical edits” are not available. Specifically, in this paper, the Basic Monthly CPS files do not include enough information to replicate the “logical edits.” In this scenario, it would be possible to apply as many “logical edits” as possible to the Basic Monthly files and then apply a selection criteria to mimic the distribution of immigrants by country of origin, state of residence, and year of immigration. Lastly, the most obvious option is to make public-use files of the “logical edits” microdata.
On the accurate estimation of E-Verify mandates, a few threats to validity warrant discussion. First, it is likely that states that pass E-Verify mandates also pass other anti-immigration laws that may affect employment rates. Indeed, Arizona’s Legal Arizona Workers Act included an E-Verify mandate as one of many measures to curtail unautho-rized immigrants’ employment. Without explicitly controlling for these other measures, the estimated effects of E-Verify mandates may very well be due to these other factors.
Thus, it is necessary to find a way to control for this “policy climate.” Various authors have
developed a measure that would account for this immigration policy climate. For exam-ple, Leerkes, Leach and Bachmeier (2012) conduct factor analysis to code states into three different levels of immigration control: high, moderate and low. Using data on employers participation in E-Verify, restrictive state laws, county and city involvement in the 287(g) program, the authors are able to construct a single measure (“internal control index”) for each state by year that is then used to classify each state into the different levels of control.
By using these more refined measures by state and year, it would be possible to capture the immigration policy climate of each state.
One last threat to validity is the level of enforcement within each state of the E-Verify mandate. While the E-Verify program can track the number of cases employers process, there is no guarantee that employers are processing all potential hires. The need here is to find a measure that quantifies the enforcement of anti-immigration laws. Fortunately, various authors have conducted these studies and some would be suitable for my study.
In particular, Watson (2010) codes information on 287(g) on a year-by-year basis between 1993 and 2002. Using a dataset that “consists of counts of Immigration and Naturalization Services (INS) ‘deportable aliens located’ as the result of internal investigations, by INS internal district, country of origin, and fiscal year” (Watson, 2010). The correlation between 287(g) enforcement and E-Verify mandates is arguably strong enough for this measure to be a good proxy of enforcement. Thus, using a measure like the one presented by Watson (2010) (but extended through 2014), it would be possible to control for enforcement levels of E-Verify mandates that may be confounding the analysis conducted thus far in this paper.
Table 4.1: Overview of E-Verify mandates
North Carolina June 2011 October 2012
South Carolina June 2011 January 2012
Utah March 2010 July 2010
Universal check mandate (alternate processa)
Colorado January 2007 August 2008
West Virginia March 2012 June 2012
Source: http://www.troutmansanders.com/immigration/. Notes: (a) These al-ternate processes mandate that an employer verify the legal status of a newly hired employee but not necessarily through the E-Verify program. (b) Other mandates refer to mandates that are not universal but cover different sectors.
Typically these sectors are public employers, public contractors/subcontractors or state agencies.
137
H.L.E. Logical Edits H.L.E. Logical Edits
Years in U.S. 14.1 10.9 Age 35.4 34.4
(8.7) (6.9) (8.8) (8.7)
% >5 years 83.3 73.9 % Female 43.1 43.9
(37.3) (43.9) (49.5) (49.6)
% Between 1 and 5 years 13.3 20.2 % Employed 71.0 73.1
(34.0) (40.2) (45.4) (44.4)
% Last year 3.4 5.9 % Unemployed 6.3 5.9
(18.0) (23.5) (24.2) (23.5)
% Mexico and Central Am. 100.0 69.7 % N.I.L.F. 22.7 21.1
(0.0) (45.9) (41.9) (40.8)
% Other North Am. 0.0 0.4 % White 58.4 50.0
(0.0) (5.9) (49.3) (50.0)
% Other Latin Am. 0.0 10.6 % African American 0.6 5.4
(0.0) (30.8) (7.4) (22.6)
% Europe 0.0 3.1 % Asian 0.1 12.7
(0.0) (17.4) (3.4) (33.3)
% Asia 0.0 13.3 % Other 40.2 31.3
(0.0) (33.9) (49.0) (46.4)
% Africa 0.0 2.7 % Married 57.5 52.2
(0.0) (16.2) (49.4) (50.0)
% Oceania 0.0 0.2 % Single 33.3 37.9
(0.0) (4.2) (47.1) (48.5)
138
(46.2) (49.8) (0.0) (34.3)
% H.S. or equiv. 30.9 26.4 % College or more 0.0 14.1
(46.2) (44.1) (0.0) (34.8)
Observations 545,209 658,576 Observations 545,209 658,576
Weighted N’s 78,010,809 102,089,870 Weighted N’s 78,010,809 102,089,870
Source: ACS 2005-2015. Notes: Averages presented with standard deviations in parentheses. H.L.E. refers to the “Hispanics with Low Education” proxy.
139 unauthorized immigrants, prime-aged (20-54) sample
All Not recent
(>5 years) Recent
(1-5 years) New
(<1 year)
H.L.E. Logical
Edits H.L.E. Logical
Edits H.L.E. Logical
Edits H.L.E. Logical Edits ACS 2005 - 2014
E-Verify in -0.060⇤⇤⇤ -0.067⇤⇤⇤ -0.028 -0.050⇤⇤ -0.226⇤⇤⇤ -0.160⇤ -0.394 -0.150 current year (0.022) (0.023) (0.028) (0.021) (0.062) (0.095) (0.253) (0.173)
Observations 510 510 510 510
ACS 2005 - 2015
E-Verify in -0.075⇤⇤⇤ -0.074⇤⇤⇤ -0.038 -0.053⇤⇤ -0.276⇤⇤⇤ -0.177⇤⇤ -0.481⇤ -0.198 current year (0.023) (0.024) (0.027) (0.022) (0.083) (0.079) (0.277) (0.190)
Observations 561 561 561 561
Source: ACS 2005-2015. * p<0.1, ** p<0.05, *** p<0.01. Note: Standard errors are robust, clustered by state (shown in parentheses). H.L.E. refers to the “Hispanics with Low Education” proxy and are meant to replicate the results of Orrenius and Zavodny (2016). Their results use ACS 2005-2014: All: 0.061⇤⇤(0.023), Not recent: -0.026(0.026), Recent: 0.258⇤⇤⇤(0.071), New: 0.464⇤(0.259). The second set of results extends the analysis to include 2015.
140 unauthorized immigrants, full sample
All Not recent
(>5 years) Recent
(1-5 years) New
(<1 year)
H.L.E. Logical
Edits H.L.E. Logical
Edits H.L.E. Logical
Edits H.L.E. Logical Edits ACS 2005 - 2014
E-Verify in -0.060⇤⇤⇤ -0.083⇤⇤⇤ -0.028 -0.063⇤⇤⇤ -0.226⇤⇤⇤ -0.177⇤⇤ -0.394 -0.195 current year (0.022) (0.021) (0.028) (0.022) (0.062) (0.075) (0.253) (0.140)
Observations 510 510 510 510
ACS 2005 - 2015
E-Verify in -0.075⇤⇤⇤ -0.090⇤⇤⇤ -0.038 -0.065⇤⇤⇤ -0.276⇤⇤⇤ -0.200⇤⇤⇤ -0.481⇤ -0.228 current year (0.023) (0.021) (0.027) (0.024) (0.083) (0.064) (0.277) (0.148)
Observations 561 561 561 561
Source: ACS 2005-2015. * p<0.1, ** p<0.05, *** p<0.01. Note: Standard errors are robust, clustered by state (shown in parentheses). H.L.E. refers to the “Hispanics with Low Education” proxy and are meant to replicate the results of Orrenius and Zavodny (2016). Their results use ACS 2005-2014: All: 0.061⇤⇤(0.023), Not recent: -0.026(0.026), Recent: 0.258⇤⇤⇤(0.071), New: 0.464⇤(0.259). The second set of results extends the analysis to include 2015.
Table 4.5: Estimates of the impact of EVerify mandates on probability of employment -2002 - 2014 - Annual Social and Economic Supplement
In Labor Force Population Full CPS
All Male Female All Male Female
(1) (2) (3) (4) (5) (6)
Logical Edits
Universal 0.050⇤⇤⇤ 0.042⇤⇤⇤ 0.024⇤⇤⇤ -0.040⇤⇤⇤ -0.050⇤⇤⇤ 0.103⇤⇤⇤
(0.004) (0.005) (0.006) (0.007) (0.006) (0.010) Observations 60,266 39,338 20,928 83,335 44,552 38,783 HLE
Universal -0.005 -0.029⇤⇤⇤ 0.041⇤⇤⇤ -0.103⇤⇤⇤ -0.155⇤⇤⇤ 0.072⇤⇤⇤
(0.005) (0.005) (0.012) (0.004) (0.010) (0.009) Observations 47,993 32,596 15,397 67,650 36,817 30,833 Naturalized
Hispanic
Universal 0.047⇤⇤⇤ 0.048⇤⇤⇤ 0.107⇤⇤⇤ -0.161⇤⇤⇤ -0.407⇤⇤⇤ -0.012 (0.004) (0.008) (0.017) (0.011) (0.013) (0.020) Observations 31,193 16,712 14,481 45,748 21,184 24,564
US-Born non-Hispanic
Universal -0.029⇤⇤⇤ -0.035⇤⇤⇤ -0.021⇤⇤⇤ -0.012⇤⇤⇤ -0.026⇤⇤⇤ 0.002⇤⇤⇤
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Observations 1,007,447 515,796 491,651 1,536,977 728,680 808,297
Source: CPS-ASEC 2002-2014. * p<0.1, ** p<0.05, *** p<0.01. Notes: Controls include gender (when applicable), race, age, marital status, number of children in household, educational attainment, industry fixed effects, state fixed effects, time (year, month) fixed effects, state specific time trends, unemployment rates. Standard errors clustered at the state level. All regressions use survey weights (wtsupp).
142 Universal (alternate process)
Other non-universal mandate
Figure 4.1: States with E-Verify Mandates
Figure 4.2: Population for prime-aged unauthorized immigrants in US - H.L.E. v Logical Edits, ACS 2005-2015
2005 2007 2009 2011 2013 2015
0 million 2 million 4 million 6 million 8 million 10 million 12 million
Unauthorized Population - H.L.E.
Unauthorized Population - Logical Edits
Year
Notes: H.L.E. refers to the “Hispanics with Low Education” proxy. Estimates are weighted using ACS person weights.
Figure 4.3: Comparing H.L.E. and Logical Edits - Characteristics of likely unauthorized by different proxies - Prime-aged workers - 2015
H.L.E.
Notes: H.L.E. refers to the “Hispanics with Low Education” proxy. Estimates are weighted using ACS person weights.