Sin Papeles y Licencia: Access to Drivers’ Licenses and Participation in Early Care and Education

Attending high-quality early education or care (ECE) is positively associated with short-term academic outcomes for children (Yoshikawa et al., 2013) as well as longer-term social and behavioral (Heckman, Moon, Pinto, Savelyev & Yavitz, 2010). Despite the demonstrated benefits of early care and education (ECE) for children with immigrant parents (Puma et al., 2010), participation rates are lower among children of immigrants than children with native-born or citizen parents among threeto fiveyear-olds (Hernandez, Denton, & McCartney, 2007). There are many barriers to accessing ECE programs for immigrants, which may be exacerbated for undocumented immigrants. Specific immigration policies may either restrict or facilitate access to ECE (Hanson, Adams, & Koball, 2016; Greenberg, Adams & Michie, 2016). One such state-level immigrant policy that may be associated with increased ECE participation is access to a driver’s license (DL) for undocumented immigrants (Hacker, Chu, Lueng, Marra, Pirie, Brahimi & Marlin, 2011; Hanson, Adams & Koball, 2016). The present study examined the association between state-level, legal opportunities to access DLs for undocumented immigrants and rates of participation in ECE among their threeto five-year-old children using a cross-sectional, quasiexperimental design. The three states examined in this study were used were: (1) New Mexico, which implemented a DL policy in 2003, (2) Utah, which implemented a DL policy in 2005, and (3) California, which passed a DL policy in 2004, but never enacted the policy. Panel data from the 1996, 2001, 2004 and 2008 Survey of Income and Program Participation (SIPP) administrations were accessed to determine legal immigration status of household members and participation in ECE. The sample was comprised of 4,709 households with undocumented members that included children between threeand five-years-old. Synthetic control models were used to compare ECE participation rates in states that had enacted DL policies to ECE participation rates estimated for synthetic controls. Results indicated that access to a DL for undocumented immigrants was associated with small but consistent differences in ECE participation, where participation rates were higher among children of undocumented immigrants in states with DL access than in control states in the post-policy period by three to five percentage points. Although future research is needed to further explore this association and to examine access to DLs at the household level, results provide preliminary evidence that driver’s licenses facilitate participation in ECE, and this may be regardless of the stringency of the policy. Results have implications for assessing DL policies as a way to integrate undocumented immigrants and their households into society and improve later outcomes for their children.

and Program Participation (SIPP) administrations were accessed to determine legal immigration status of household members and participation in ECE. The sample was comprised of 4,709 households with undocumented members that included children between three-and five-years-old. Synthetic control models were used to compare ECE participation rates in states that had enacted DL policies to ECE participation rates estimated for synthetic controls.
Results indicated that access to a DL for undocumented immigrants was associated with small but consistent differences in ECE participation, where participation rates were higher among children of undocumented immigrants in states with DL access than in control states in the post-policy period by three to five percentage points. Although future research is needed to further explore this association and to examine access to DLs at the household level, results provide preliminary evidence that driver's licenses facilitate participation in ECE, and this may be regardless of the stringency of the policy. Results have implications for assessing DL policies as a way to integrate undocumented immigrants and their households into society and improve later outcomes for their children.
iv ACKNOWLEDGMENTS This project was a journey, a science, and an art, and there are many who supported me through the end; before the start.
First of all and foremost, and before all of the rest, to my one and only major professor, who brought out in me the best. It can't be understated; all that you've done for me -from our first class together, right through finishing the D. You stuck it out and pushed me, too and stayed right by my side…through tantrums about outcomes and when the census I decried. Your expectations lofty; but your dedication higher. I know that you would burn yourself to hold my feet to the fire. I feel that with your guidance, your support and honesty, I have grown into a researcher of the utmost integrity. You've been more than a mentor: you're my swimmer and my friend (though I'm sure that you and HP hope that I don't move again). To Dr. Kathleen Gorman: I'm perpetually inspired and I only hope you know that you will always be admired.
I would also like to thank the other members of my committee for their crucial points of feedback and their genuine belief that in me, is a person who is competent and ready to succeed. Your belief in me is such an honor and humbling, indeed.
To Dr. Margaret Rogers for your passion and your drive to teach that differences and diversity are that which make us thrive. Thank you for your guidance and your support along the way, your interest in the project and the role that you did play.
v To Dr. Bob Laforge for serving as more than a statistician. For your patience and your guidance and your perfectly unique vision. 532 is still one of the best courses that I took, because it operated like real life, rather than being so sterile and "textbook." To Dr. Shanna Pearson-Merkowitz for bringing in a new perspective. For providing me with so much more than sources for my reference. I know that without such experience and also without your great feedback, in many ways my knowledge and this project would surely lack.
To Dr. de los Heros for your enthusiasm to serve as chair, and to the entire department of Spanish ---I always felt welcome there. Mil gracias a todos. Yo he aprendido un montón. Y aqui en este momento, puedo ver que tuve razón. Por hablar con los inmigrantes, es la manera en la que, el enfoque de mi investigación, yo lo encontré.
To Dr. Amy Laurent, the best work wife there could be; for reminding me that in all ways I should always just be me. For bringing back the music and making my world rock (and maybe for that ring…and for that brush…oh, and that sock). You showed me the way when I was your RA; the spirit of NADACE. You set the bar so high and then pushed me to keep my pace. You have been amazing and I could not ask for more. In sum, one of my best decisions was to tell YOU about my door.
And of course, I never could forget; the most deserving of a verse, is my extraordinary mother, Dr. Fede…the first. Exercise science and psychologywe conquered the body, and now the brain. You helped me all along my path in ways I can't explain. I don't know where I'd be without you, and all of your support. You vi did it first, you blazed the trail, then threw the ball into my court. I guess you knew, that as I do, I'd catch it and run and play, and make the most of every challenge, but all in my own way. So thank you, Mom, I love youmore than I can say. Who knew that eccentric, little kid would be DOCTOR Mush E. Mouse one day?
To my counterpart, Dr. Clark, for taking on the lead. For navigating this wild world before I knew I could be freed. You got me through, you took the heat, enduring (in)humanity, but now there's just one thing to do for both our sanity.
It's time for us to both fly free now that I've learned I can. I don't have the words to thank you that could make you understand the gratitude I have that you were always there for me, so capable and well-rehearsed that nobody could see...that behind that carefully crafted script and façade of "cool, calm and collected," was a hidden, amazing, perfect ME that for too long went neglected. I needed you, Clark, like I needed help, like I needed to lose my way, because all those things, including you, have let me know myself today.
My whole life makes more sense to me -my strengths and struggles, too, the things that I prefer, all that I see and hear and do. You were always under tremendous stress, inevitably bound to break and I was so explosive from all the restraint I could not take.
It's not that I don't like you, Clark, I love you -this is true. But I have a lot to offer, and now I know I love me, too.
Clark, YOU are phenomenal -you had no easy tasks, and there are others just like you, feeling exhausted as full-time masks. So, I hope you'll stay in touch and that we can still hang out together, but I think this new arrangement will suit both our vii needs much better. Because you need to recover, and I need to be me, for you, for me, and our well-being and for the world to maybe see that there are many strengths, many skills...an infinite range of perspectives, and it will be a much more vibrant world when perfect "MEs" are not neglected or suppressed or pushed aside, or suspiciously inspected, or compliant, silenced, made to change, and whose differences are rejected.
So, Dr. Clark, one more request: that you'll support our visibility and use your new degree of freedom in defense of diversity.
viii DEDICATION To immigrants everywhere who risk it all for their children, families, beliefs, and values. To those who leave all they know for greater opportunity and contribute to and better the countries to which they arrive.
For the courage and resilience of those without documentation who must live in the shadows. For the bravery of children who, at times, must live day to day not knowing if their parents will still be there when they get home. For the tenacity and true patriotism of DREAMers. To the resolve and persistence of those who advocate for immigrants within states and nationally.
Finally, to all those who have yet to see and embrace the beauty and strength that comes with diversity and difference, for they are losing out the most.  Early care and education (ECE) refers to a wide range of early childhood settings (e.g., pre-school, pre-k, center-based care, daycare, Head Start or other developmentally appropriate learning or care arrangements) for children who are not yet of age to enter kindergarten or elementary school (Yoshikawa, Weiland, Brooks-Gunn, et al., 2013). Research has consistently found that attending high-quality early education or care (ECE) is positively associated with short-term academic outcomes  as well as longer-term social and behavioral outcomes for children (Heckman, Moon, Pinto, Savelyev & Yavitz, 2010). Notably, effects of ECE programs are as large or larger for children from socially disadvantaged backgrounds such as low-income and minority children, including the children of immigrants (Magnuson, Lahaie, & Waldfogel, 2006;Puma, Bell, Cook, Heid, Shapiro, Broene et al., 2010).

LIST OF TABLES
Immigrants are defined by the federal government as all foreign-born, non-U.S. citizens who have migrated voluntarily to the U.S., and who do not fall into special categories (e.g., career diplomats, ambassadors, students) (United States and access resources available to their citizen and resident neighbors (Androff, Ayón, Becerra, et al., 2011).
Many state-level policies have been restrictive in nature (e.g., English-only laws, laws that require all immigrants to carry documentation of legal status at all times, laws that expand immigration enforcement to local police). Others have been expansive and have increased the rights of undocumented immigrants, such as permission is not against the law. Therefore, 'undocumented' is the most accurate and legally appropriate term and is used throughout this dissertation. extending legal opportunities for driver's licenses to undocumented immigrants. Over the past two decades, 12 states (Washington, New Mexico, Utah, Nevada, Maryland, Illinois, Connecticut, Colorado, California, Vermont, Delaware, and Hawaii) and the District of Columbia have passed legislation making state-recognized drivers' licenses (DL) available to undocumented immigrants. It has been reported by community members and leaders of immigrants' rights groups that the legal opportunity to obtain a valid DL will facilitate the ability of undocumented parents to transport children to school or other forms of childcare (Greenberg, Adams, & Michie, 2016)

Statement of the Problem
Despite the demonstrated benefits of early care and education (ECE) for children with immigrant parents (Puma et al., 2010), participation rates are lower among children of immigrants than children with native-born or citizen parents among three-to five-year-olds (Hernandez, Denton, & McCartney, 2007).

Significance of the Study
At present, more than 25% of children under age 18 currently have at least one parent who was born outside of the country (Capps, Fix & Zong, 2016), and immigration is projected to be the primary source of population growth through 2050 (U.S. Census Bureau, 2014). Immigrant parents are more likely to have lower incomes, lower levels of education and less access to resources, all of which may put their children at a disadvantage (Fortuny, Capps, Simms, & Chaudry, 2009;Jiang, Ekono, & Skinner, 2015). The sociodemographic indicators identified above, namely low income, low levels of parental education and less access to resources have all been cited as significant barriers to ECE participation (Hanson, Adams & Koball, 2016).
Other barriers include language, knowledge of programs and enrollment procedures, and transportation (Greenberg, Adams & Michie, 2016). In addition, immigrants are subject to a unique and complex set of federal and state policies to which native-born families are not. Specific immigration 2 or immigrant 3 policies, many of which vary by state, may either restrict or facilitate access to ECE. This is especially true for undocumented immigrants, who are highly limited by their legal status and often subject to restrictive federal and state immigrant policies. One such state-level immigrant policy that may be associated with ECE participation is access to a driver's license (DL) for undocumented immigrants (Hacker, Chu, Lueng, Marra, Pirie, Brahimi & Marlin, 2011;Hanson, Adams & Koball, 2016).
In the United States, undocumented immigrants are not eligible to obtain a federally recognized DL. However, over the past two decades, 12 states and the District of Columbia have passed legislation making state-recognized DLs available to undocumented immigrants. Although DL access for undocumented immigrants has been associated with other outcomes such as attending health appointments (Ayón, 2008) and enrollment in higher education (Hacker et al., 2011), the association between DL access and ECE enrollment has not been examined. The current study aimed to investigate whether legal opportunities to access DLs for undocumented immigrants via statewide policies were associated with increased participation in ECE among the children of undocumented immigrants.
2 Immigration policies refer to policies that aim to control the flow of migrants entering the country. Immigration policies include numerical caps and quotas, priorities for admittances, and country-specific bans or limitations, for example. 3 Immigrant policies refer to policies that aim to restrict or expand rights, services, access, etc. for immigrants once they have arrived in the U.S.

Purpose of the Study
Lacking transportation is a common barrier to ECE participation for lowincome and immigrant families (Hanson, Adams & Koball, 2016). Lack of access to a driver's license for undocumented immigrants has been cited as potential barrier to ECE participation in interviews with community members and leaders (Greenberg, Adams, & Michie, 2016;Karoly & Gonzalez, 2011). However, the association between DL policies and ECE enrollment for children living in households with undocumented immigrants has not been empirically assessed. Therefore, the aim of the current study was to examine whether access to DLs for undocumented immigrants was associated with enrollment in ECE among children of undocumented immigrants. National data from the Survey of Income and Program Participation (SIPP) on ECE enrollment and legal immigration status collected over the period of time during which state DL policies have been implemented was used to assess the following hypotheses. It was expected that: H1: There would be higher rates of ECE participation in states that enacted policies expanding legal opportunities for DLs to undocumented immigrants compared to states that did not enact DL policies following policy enactment.
H2: Undocumented households that had access to a DL would be more likely to enroll children in ECE than undocumented households that did not have access to a DL in their state of residence above and beyond other important predictors of enrollment (e.g., household income, years in the US, whether the head of household is employed).

Introduction
Research has consistently found that attending high-quality ECE is positively associated with short-term academic outcomes  as well as longer-term social and behavioral outcomes for children (Heckman, Moon, Pinto, Savelyev & Yavitz, 2010). Notably, effects of ECE programs are as large or larger for children from socially disadvantaged backgrounds such as low-income and minority children, and including the children of immigrants (Magnuson, Lahaie, & Waldfogel, 2006;Puma, Bell, Cook, Heid, Shapiro, Broene et al., 2010). Despite the demonstrated benefits of early care and education (ECE) for children with immigrant parents (Puma et al., 2010), participation rates are lower among children of immigrants than children with native-born or citizen parents among three-to five- year-olds (Hernandez, Denton, & McCartney, 2007).
At present, more than 25% of children under age 18 currently have at least one parent who was born outside of the country (Capps, Fix & Zong, 2016), and immigration is projected to be the primary source of population growth through 2050 (U.S. Census Bureau, 2014). Immigrant parents are more likely to have lower incomes, lower levels of education and less access to resources, all of which may put their children at a disadvantage (Fortuny, Capps, Simms, & Chaudry, 2009;Jiang, Ekono, & Skinner, 2015). Immigrants are subject to a unique and complex set of federal and state policies to which native-born families are not. Specific immigration 4 or immigrant 5 policies, many of which vary by state, may either restrict or facilitate access to ECE. Other barriers include language, knowledge of programs and enrollment procedures, and transportation (Hanson, Adams, & Koball, 2016;Greenberg, Adams & Michie, 2016).
Barriers may be especially salient for undocumented immigrants, who are highly limited by their legal status and often subject to restrictive federal and state immigrant policies. One such state-level immigrant policy that may be associated with ECE participation is access to a driver's license (DL) for undocumented immigrants (Hacker, Chu, Lueng, Marra, Pirie, Brahimi & Marlin, 2011;Hanson, Adams & Koball, 2016). The aim of the current study was to examine the association between state-level legal opportunities to access DLs for undocumented immigrants and rates of participation in ECE among their three-to five-year-old children.
Over the past two decades, 12 states and the District of Columbia have passed legislation making state-recognized DLs available to undocumented immigrants.
Although DL access for undocumented immigrants has been associated with other outcomes such as attending health appointments (Ayón, 2008) and enrollment in higher education (Hacker et al., 2011), the association between DL access and ECE enrollment has not been examined. The current study aimed to investigate whether legal opportunities to access DLs for undocumented immigrants via statewide policies 4 Immigration policies refer to policies that aim to control the flow of migrants entering the country. Immigration policies include numerical caps and quotas, priorities for admittances, and country-specific bans or limitations, for example. 5 Immigrant policies refer to policies that aim to restrict or expand rights, services, access, etc. for immigrants once they have arrived in the U.S.
were associated with increased participation in ECE among the children of undocumented immigrants.
In the following sections, background information is provided regarding immigrants in the United States including definitions of different types of immigrants, and characteristics of immigrant households with most information focused on undocumented immigrants, the focus of the current study. In order to put in context the current state of U.S. immigration policy and its relationship to the undocumented population, a historical account of immigration policy in the U.S. is also provided. The historical context of U.S. immigration policy is also intended to give insight into the increase in state-level immigrant policies over the past decades and why drivers' licenses for undocumented immigrants has been a topic of debate in many state legislatures across the country. Driver's license access will also be discussed in terms of the benefits that have been observed among immigrant populations and the negative outcomes associated with lack of access to DLs. The association between access to a DL and ECE participation rates of children of undocumented immigrants has not been assessed, despite the well-established benefits of ECE. ECE and its associated outcomes in general, and among children of immigrants will be presented, along with the potential pathways by which legal opportunities for DLs could facilitate access to ECE for children of undocumented immigrants.

Immigrants in the U.S.
There are 42.4 million immigrants in the United States representing 13.3% of the total population (Zong & Batalova, 2016 Although just 3 million children under 18 (4%) are undocumented (Massey & Pren, 2012), there are approximately 4.5 million U.S. born children with undocumented parents (Passel & Cohn, 2012) and it is estimated that nearly half (47%) of households with undocumented adults (i.e., persons over the age of 15) also include children   (Capps, Fix & Zong, 2016). Households with citizen children and undocumented members are commonly referred to as mixed-status households (i.e., those households in which at least one member is a U.S. citizen or is lawfully present in the U.S. and at least one other member is undocumented (Matthews, 2010). It is estimated that there are approximately 6.6 million mixed-status households in the U.S. .
In general, rates of economic hardship and poverty are higher among children living in households that include foreign-born members compared to children living with native-born individuals (Fortuny, Capps, Simms, & Chaudry, 2009). Poverty rates are higher still among children in mixed-status or undocumented households (Capps, Bachmeier, Fix & Van Hook, 2013), the focus of the current study. It is estimated that 51% percent of children under age 18 living with undocumented immigrants live at or below the Federal Poverty Line (FPL) compared to 24% of children of (all) immigrants and 18% of children living with native-born parents (Capps et al., 2013;Passel & Cohn, 2012). Undocumented household members also tend to have lower levels of education, with 47% having less than a high school education compared to eight percent of U.S. born residents and 12% of all immigrants regardless of legal status (Jiang, Ekono, & Skinner, 2015;Passel & Cohn, 2012).
Even when they are born in the U.S. and hold U.S. citizenship, children of undocumented immigrants are more likely to lack health insurance coverage than children of LPRs and children with native-born parents (Capps et al., 2013;Gonzales, Suárez-Orozco, & Dedios-Sanguineti, 2013).

Why don't they just "do it the right way?"
U.S. policy further ensures that being undocumented limits access to resources compared to immigrants with legal permission to be present in the U.S. and U.S. citizens (Pew Research Center, 2013). Processes do exist to immigrate to the United States in good legal standing, yet not all immigrants begin or complete these processes. It is not uncommon in political and media discourse for the blame to fall upon undocumented immigrants: "they come here to take advantage of the system," "they want to be undocumented so that they do not have to pay taxes," "they want to come here and live off our healthcare system and government even though they have not earned it," "they 're criminals, rapists, drug lords, gang members, terrorists (etc.) coming to harm the U.S.," (Ortega, 2015).
The factors that push immigrants out of their countries of origin and pull them towards a receiving country, and the motivations and the processes involved to migrate from the country of origin to the receiving country are diverse, complex, and in most cases, quite different from the perspectives offered above. Research has demonstrated that one of the strongest predictors of the size of the undocumented population in the country at any given time is U.S. immigration policy (Chomsky, 2014;Immigration Policy Center, 2009;Massey, Durand & Malone, 2002;Massey & Pren, 2012). Both historically and at present, U.S. immigration policy has tended not to consider nor to address the major economic, social, and historical contexts that have driven migration to the U.S. (Massey, Durand, & Malone, 2002). Instead, U.S.
immigration policy has tended to be driven by public opinion, isolated incidents (e.g., terrorism, war), and changing U.S. demography (Ewing, 2012).

U.S. Immigration Policy
At present, the U.S. system of immigration policies is not equipped to accommodate current immigration needs, nor does it accommodate the needs of countries with high demand for out-migration to the U.S., and the economic, social and labor demands in the U.S.   (Ewing, 2012;USCISHOL, 2012).
Immigration was encouraged, as the U.S. desperately needed immigrants to grow the population, develop land, contribute skills and trades to the growing market and to input money into the economy (Ewing, 2012). At the same time, industrialization, overpopulation, poverty, and political oppression in Europe provided an influential push for immigrants to leave their countries of origin for the "new world." This represents one point in U.S. history where a match between U.S.
immigration policy and bilateral immigration needs truly existed (Massey & Pren, 2012, Vialet, 1991. Policies pertaining to migrant flows were created, enacted, and enforced at the state level, particularly in port and border states such as Massachusetts (Guarnizo & Smith, 1998). Border and port states began passing restrictive laws that aimed to curb or halt the flow of "undesirable immigrants" into their cities, and towns (Boushey & Luedtke, 2011). Often, such laws were passed with the goal of banning immigrants who were determined by the state to be a potential "burden to society" (e.g., the poor, sick, disabled, ethnically diverse, enslaved populations, criminals, and anyone who was deemed likely to become a public charge) (Neuman, 1993 Neuman, 1993).
In 1862, the development of railways further increased the demand for the Western migration of immigrants to provide labor (USCISHOL, 2012 (Ewing, 2012).

Federal Control of Immigration and Exclusion (1875 -1920)
The immigrant population was rapidly expanding and immigration was still encouraged by the federal government. The presence of a large foreign-born population was also met with increased nativism inspired by anti-Catholicism, fear for U.S. born workers, and the linking of immigration with crime and poverty (Neuman, 1993). Thus, states continued to enact restrictive policies barring poor, disabled, or diseased immigrants from entering the U.S. (Neuman, 1993 "The passage of laws which concern admission of citizens…of foreign nations to our shores belongs to congress and not to the states. It has the power to regulate commerce with foreign nations…if it be otherwise, a single state can at her pleasure embroil us in disastrous quarrel with other nations" (Chy Lung v. Freeman, 1875).
For the first time, the Supreme Court legitimized federal control of immigration policy (Boushey & Luedtke, 2011;Smith, 1998). For many decades after, immigration policymaking at the state level ceased as states yielded to the federal government.
The first act of the United States federal government, with their newly legitimized control of immigration, was to exclude certain classes of immigrants as "undesirable" (Boushey & Luedtke, 2011 (Vialet, 1991).

Numerical Limits by National Origin (1921 -1964)
With the changing demography of the U.S. and rapidly increasing immigrant population, immigration policy became more selective (Norgarrd, 2008). Policies were passed that favored individuals from certain countries over others and explicitly banned some immigrants solely based on their country of origin (Massey & Pren, 2012;Vialet, 1991 The same year, the total number of immigrants accepted for entrance was reduced to 165,000 per year and the per-country limit was decreased to 2% of the number of people of that ancestry living in the U.S. as of the 1890 census (Vialet, 1991. This system remained in place until 1952, with immigrants from the Western Hemisphere still exempt from quotas and caps, along with wives and unmarried, minor children of (male) U.S. citizens.
During World War II and through the beginning of the Cold War, immigration policy continued to be selective based on the nationality of potential immigrant arrivals. For example, World War II prompted exclusionary policies and practices aimed at people of Japanese descent, but there was also an increase in humanitarian refugee policies and a lessening of the restrictive policies that had been placed upon other Asian immigrants previously (Ewing, 2012 (Mize, 2016).
The success of the Bracero Program is one more example of bilateral immigration policy that worked because it considered and met the needs of both the sending and receiving countries involved. During the same time period, however, the McCarran-Walter Act (1952) aimed to consolidate immigration laws of past years into one concise statute, and while it officially eliminated race as a basis of exclusion, the national origins quota system, that favored immigrants from the United Kingdom, Ireland and Germany by basing quotas on the 1920 census, was kept in place. Most of the immigration slots that were allocated for these favored countries went unused as there was little demand for migration from northwestern Europe during this time (Massey, Durand, & Malone, 2002).

Civil Rights and the "Rise" of Undocumented Immigration (1965 -1985)
The fourth and current wave of immigration began in 1965, a year that is also considered a significant time point in the context of undocumented immigration. One year after the Civil Rights Act (1964) was passed, the Immigration and Nationality  (Hoefer, Rytina & Baker, 2012;Wasem, 2011). The events and processes that led to increased migration from Latin America were never intended to do so, and in fact, some later actions were intended to deter it, yet immigration from Latin America surged (U.S. Census Bureau, 2014). Evidence suggests this was the result of immigration policies and the sociopolitical context in the U.S.
When the INA was passed, immigration from Latin America was not a pressing concern among the general public and even among the most conservative politicians of the time (Wright, Levy & Citrin, 2015). In general, there was bipartisan agreement that termination of the National Origins Quota system was necessary.
However, there was also agreement that the Bracero Program, which had come to be viewed as an exploitive labor system, should be terminated as well (Massey & Pren, 2012 (Ewing, 2012). Neither the demand for seasonal, agricultural labor in the U.S. nor the demand for employment among Mexican farm workers had changed, but the legal avenues for meeting these needs had been very suddenly taken away. As a result, the migratory flows did not end, rather they continued without contracts, permission, paperwork and documentation (Boushey & Leudkte, 2011 (Massey & Pren, 2012). Growth in the undocumented population decelerated quickly after the circular pattern of migration had been re-established without authorization. However, the initial inflow of newly "illegal," mostly Mexican workers played a critical role in the political agenda, public perception, and immigration policy-making going forward.
The undocumented population expanded from 1970 and through the 1990's.
At the same time, income inequality in the United States was increasing (Abramitzky & Boustan, 2017). Unemployment also increased from 5.6% in 1974 to 10.8% in 1982, a historic high (United States Bureau of Labor Statistics, 2018). Economic depression, regression, and inequality have always been strong predictors of antiimmigrant sentiment (Almeida, Biello, Pedraza & Wintner, 2016). It was in under these circumstances that immigration from Latin America began to be framed by the media as a "crisis" or a "tidal wave" that would drown U.S. culture with a flood of foreign-born immigrants (Santa Ana, 2002). Later, this 'Latino threat narrative' became more intense, often describing immigration as an "invasion" of "aliens" attacking "outgunned" border patrol agents (Chavez, 2008). In 1976, the commissioner of Immigration and Naturalization Services (INS) even published an article in which he described immigration as having the potential to become a national disaster (Chapman, 1988).

Granting Amnesty, but Limiting Immigrant Rights (1986 -2000)
The Latino threat narrative was further promulgated by not only the media, but by government officials, municipalities, and politicians as they found advantages to be gained by scapegoating Latino immigrants and undocumented immigration for economic conditions in the U.S. (Massey & Pren, 2012). In 1986, Ronald Regan stated that illegal immigration was a question of national security and that "terrorists and subversives" are just a few days' drive from the southern border (Kamen, 1990).
In 1992, the Chief of the San Diego unit of the U.S. Border Patrol released "Border Under Siege," a video that showed Latino immigrants running through highway traffic to enter the U.S. without inspection (Rotella, 1998). Further media portrayals deemed Latino immigration to be a "war on the middle class," or even a plot by Mexico to take back the lands they had ceded to the U.S. in 1848, and that immigration, then, would result in a total loss of the southwestern United States (Massey & Pren, 2012).
Within just a few decades, Mexican migration had gone from a legal, effectively invisible, circular flow of workers to a highly visible, publicized flow that was framed as an attack on U.S. economy, culture, safety, and territory being carried out by criminal, alien invaders (Mize, 2016). The loss of legal avenues for migration without the loss of demand for labor resulting in unauthorized flows, plus the relentless framing of Latino migration as a "crisis" or a "cultural inundation flooding North America" in a time of income inequality influenced public opinion and shifted it towards conservatism on immigration (Massey & Pren, 2012). Subsequently, immigration legislation became more restrictive and more heavily reliant upon enforcement efforts. With increased enforcement, came an increased number of border apprehensions and deportations, despite declining migrant flows following 1970 (Ewing, 2012). However, to the general public, an increased number of border apprehensions and deportations was equivalent to an increase in the number of illegal border crossings. This further led to a belief that more enforcement was needed to subdue the growing "crisis" that was undocumented immigration (that, in reality, had occurred legally for decades, had leveled off in the 1970's once the established seasonal flow resumed without authorization, and had actually decreased through the 80's and 90's).
In 1986, the federal government implemented the Immigration Reform and Control Act (IRCA). IRCA did the following: (1) legalized nearly 2.7 million undocumented immigrants who had been living in the United States before January 1, 1982, who had no unpaid back taxes, had not been found guilty of any crimes, and had at least minimal knowledge of United States government, history and the English language (USCISHOL, 2012), or who fell into a special agricultural worker category (Kerwin, 2010), (2) increased border enforcement (Chishti & Kamasaki, 2014), and (3) made it a federal crime to knowingly hire or employ undocumented workers (Chishti & Kamasaki, 2014). IRCA policies were developed based partly on the assumption that immigrants who entered the country without permission came to the United States for existing job opportunities and because they were able to enter without inspection (American Immigration Council, 2017). Although IRCA initiated the H-2A visa category for temporary seasonal agricultural workers with very limited annual slots, it did not raise limits on legal immigration to match the growing demand for immigrant labor in the U.S. (Popat, 2014 retroactively, meaning that even non-violent offenses that had been committed years prior to the passage of the law obligated detention. IIRIRA also created an expedited removal process for "criminal aliens" that did not grant them the right to a formal hearing, implemented three-and ten-year bans to re-entry for immigrants who had been present for at least 180 days in the U.S., and increased border enforcement . Second, the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) made most LPRs ineligible for means-tested public benefit programs for five years after obtaining a green card . Undocumented immigrants were explicitly barred from all public benefit programs by the federal government. Finally, the Anti-Terrorism and Effective Death Penalty Act expedited the removal processes for foreigners suspected of being associated with terrorism, allowed the detention and deportation of non-U.S. citizens on the basis of evidence that neither the accused immigrant nor their lawyer were permitted to see, and made it more difficult to be granted asylum (Massey & Pren, 2012). These policies were all enacted to restrict access for immigrants, and especially to "crack down" on undocumented immigration, which had been proposed to be part of the cause of poor economic conditions in the U.S. for decades (Citrin, Green, Muste & Wong, 1997).
Little, if any, consideration was given to U.S. labor demands, bilateral immigration needs or migratory trends. With policy mismatched to reality, there is more potential for increased undocumented immigration (Espenshade, 1995;Massey & Espinosa, 1997).

Immigration as a Matter of National Security (9/11 -Present)
The 1990's saw unprecedented enforcement efforts at the border and internally (Massey & Pren, 2012 According to the Department of Homeland Security (2017) removals increased to approximately 400,000 annually by 2009. Though none of the accused involved in 9/11 were of Mexican descent, nor had they entered through Mexico, and though all involved had come to the U.S. on legal visas, the vast majority of persons removed as part of increased enforcement efforts after 9/11 were of Mexican descent (72%) (Massey & Pren, 2012). In 2002, the U.S. government also introduced a special registration system and a "voluntary" interview program that targeted foreign-born Muslims, Arabs, and South Asians (Ewing, 2012 (Ewing, 2012).

Consequences of U.S. Immigration Policy
Even before the United States became a sovereign nation, it struggled with its identity as a refuge for immigrants and a country that relies upon immigrants, but also one that has consistently stereotyped and feared the newest arrivals to its shores. U.S. Enforcement efforts have not only failed to decrease undocumented immigration from Latin America, but they have contributed to increased immigration into to the United States. For example, the first large increase in the undocumented population from Latin America stemmed from well-established, legal, circular flows of 400,000 Mexican farm workers being unaccounted for, and, thus "illegal" with the end of the Bracero Program. In 1964, the undocumented population was estimated to be less than 300,000, but in 1968 (i.e., the year the Bracero Program was fully phased out), the undocumented population grew to over one million (USDHS, 2012). In the following decades, as legislation continually called for increased enforcement efforts, Mexican workers stopped returning home once they made it into the U.S., increasing the net inflow of undocumented immigrants by decreasing the overall outflow (Redburn, Reuter & Majmundar, 2011).
Through the 1990's and 2000's, with the rise of the Latino threat narrative, anti-immigrant sentiment, and immigration policy-making that focused on border patrol and internal enforcement, the population of legal immigrants grew as well from less than 10% admitted outside country quotas as relatives of U.S. citizens to over 40% after 199640% after and 65% after 200140% after (USDHS, 2012. When U.S. policies have been more restrictive, or are perceived to be discriminatory, legal permanent residents who may have otherwise retained their LPR status, tend to be more likely to initiate the naturalization process to protect themselves (i.e., defensive naturalization) (Redburn, Reuter & Majmundar, 2011). This also affords them the opportunity to be able to petition for family members, as the immediate relatives of U.S. citizens are exempt from quotas and caps, a process known as defensive naturalization (USCIS, 2017).
Not only have enforcement policies tended to result in less outmigration among undocumented immigrants (increasing overall net inflow), but fear among the immigrant population stemming from stringent U.S. immigration legislation, increased enforcement efforts and negative discourse around immigration have also contributed to increasing the size of the legal immigrant population. With the rise of the Latino threat narrative, in the mid 1980's, negative words were paired with "Mexican" or "immigrant" across the U.S.'s four leading newspapers over 35% of the time. Since the 1965 legislation, there has existed a growing need not only for a functional immigration system (Silva, 2015), but for CIR that addresses the current undocumented population, considers bilateral immigration needs, and that is driven by these bilateral needs as opposed to fear or extreme incidents. In the absence of CIR, Although many of these state policies have mirrored federal policy by opting to increase enforcement efforts, (e.g., enter into 287(g) memoranda of agreement with Immigration and Citizenship Enforcement, or require that all immigrants carry their papers at all times), many states have also opted to implement policies that expand immigrant rights and access. Among a wide range of bills dealing with many aspects of immigration, one topic that has been debated in many states is the issuance of driver's licenses to undocumented immigrants. Of the 208 policies enacted in 2010, 32 (15%) were related to drivers' licenses or identification cards for immigrants (regardless of legal immigration status) (NCSL, 2010). In 2005, of the 45 bills enacted, nine (20%) were regarding identification cards or drivers' licenses for immigrants across legal status (NCSL, 2005). Bills pertaining to DLs and ID cards for immigrants vary in the degree to which they are expansive versus restrictive. They also vary in terms of the target population (e.g., LPRs, undocumented immigrants, undocumented immigrants under 18).

Why Driver's Licenses?
The distribution and processes associated with the issuance of driver's licenses loans, rental homes and apartments, social and educational programs and assistance, and to prove residency for various state institutions (e.g., state colleges) (Johnson, 2004).
The federal government has not imposed upon states' rights to issue drivers' licenses (Mounts, 2003) until the passage of the REAL ID Act. Immediately after September 11 th , the licensing of undocumented immigrants became a tense national security issue as it was discovered that those who were involved in 9/11 had obtained state-issued driver's licenses with which airplanes could be boarded (Harberson & Doherty, 2002). What is often forgotten is that none of the individuals involved in the 9/11 attacks actually needed a driver's license to board an airplane. All had foreign passports that were considered valid forms of identification for this purpose at the time (Mounts, 2003). Additionally, all those involved were in the United States with legal permission and were not undocumented (United States Congress, 2002; National Commission on Terrorist Attacks, 2004). Nevertheless, the topics of identity fraud and the licensing of "potentially dangerous, criminal aliens" began to dominate the new debate over immigrant eligibility for driver's licenses (Zhao, 2003). Consistent with the strong links that had been formed between national security, terrorism and immigration throughout U.S., the REAL ID Act was enacted with the intent to "crack down" on fraudulent documents and forms of identification.

Access to a Driver's License
Issues of national security, fraud, and "rewarding" undocumented immigrants with the privilege of a driver's license are common arguments against state-issued drivers' licenses for undocumented immigrants (Johnson, 2004;Lopez, 2004).
Evidence may suggest positive rather than negative outcomes associated with increased access to a DL. In general, a greater number of drivers with a valid license in a state is associated with fewer fatal accidents (Kerry & Kumazawa, 2011). In California, it was found that drivers without valid licenses were three times more likely to cause a fatal accident than were drivers with valid DLs (Brar, 2012). In addition, though little data exist to measure the economic impact of licensing undocumented immigrants, employers in North Carolina reported that industry would benefit if undocumented workers were able to legally drive to work (North Carolina Department of Public Safety, 2014). In addition, in 2008, Oregon passed a law that required all license applicants to prove legal immigration status. Prior to its enactment, it was estimated that in response to implementation of the law, the undocumented labor force would decrease and the loss in state gross domestic product would range from $134 million to $201.9 million annually (King, 2011). Enrollment in higher education among undocumented immigrants has also been linked to DL access with an increased likelihood of participation when a state DL is made available to undocumented immigrants (Drachman, 2006).
Negative outcomes associated with fear of driving, or driving without a DL have also been observed for the general public, undocumented immigrants and oftentimes their citizen children. Evidence suggests that there are risks associated with lack of access to DLs, specifically for the undocumented population and their households. Lack of access to DLs for undocumented parents is associated with higher psychological distress, increased economic hardship among parents and lower levels of cognitive ability among their children (Yoshikawa, Godfrey, & Rivera, 2008). Furthermore, not having a DL has been associated with a fear of reporting crimes to police (Nguyen & Hill, 2016) and heightened anxiety around driving for tasks of daily living (e.g., shopping) in communities with increased immigration enforcement (White Yeager, Menachemi & Scarinci, 2014). Fear of driving without a DL has been associated with a greater number of missed appointments among undocumented pregnant mothers, a decreased use of social service programs among undocumented immigrants and also among native-born Hispanics (Toomey, Umaña-Taylor, Williams, Harvey-Mendoza, Jahromi & Updegraff, 2014), missed doctors' appointments among undocumented immigrants and their children (Hacker, Chu, Leung, et al., 2011), and as a barrier to completing child welfare mandated services (Ayón, 2008).
The association between DL access for undocumented immigrants and participation in early care and education among their children has not been empirically assessed, despite the fact that transportation to and from these programs is rarely provided. It is true that the quality of early care and education programs matters in examining later outcomes (Jeon, Langill & Peterson et al., 2010;Tietze, Cryer, Bairrao, Palacios & Wetzel, 1996), but access to a program is the first step. In the following sections, early care and education will be defined, and the outcomes with which it is associated, and outcomes specifically for children of immigrants will also be presented.

Early Care and Education
ECE refers to a wide range of early childhood settings (e.g., pre-school, pre-k,  (Duncan & Magnuson, 2013), depending on the quality of care.
Although children who do not participate in ECE tend to catch up to their peers who did attend in terms of immediate academic outcomes (Yoshikawa, Weiland, Brooks-Gunn, et al., 2013), ECE participation remains associated with important outcomes later in life. For example, having participated in early education is associated with higher educational attainment and higher earnings in adulthood (Deming, 2009;Heckman, Moon, Pinto, Savelyev, & Yavitz 2010), lower rates of pregnancy among teenagers (Sawhill, Winship, & Grannis, 2013), and lower rates of criminal behavior and incarceration (Deming, 2009). It has been estimated that for every dollar spent on early education, the return on investment ranges from $3.97 and $10.83 (Reynolds, Temple, White, Ou & Robertson, 2011) and that nearly 60 to 70% of benefits to society have been attributed to reduced criminal behavior (Belfield, Nores, Barnett, & Schweinhart, 2006). Associations between ECE and later adult outcomes demonstrate that both ECE participants and society continue to benefit long after immediate post-program gains are observed.
Positive short-and long-term effects of ECE participation have been observed among both low-and middle-income children. However, research has consistently demonstrated that children from low-income families, minority children, children who are dual-language learners and children of immigrants benefit more from participation in ECE (U.S. DHHS, 2010). Although fewer studies have focused specifically on the children of immigrants, evidence suggests that they benefit as much, or more than their peers with native-born parents (Magnuson, Lahaie, & Waldfogel, 2006).
Children with at least one immigrant parent (regardless of legal status) participate in ECE at lower rates than children with native-born parents (Karoly & Gonzalez, 2011).
Children of undocumented immigrants make up a growing share of students in grades K -12 (between 7 and 10% in 2009) (Passel & Cohn, 2012

Barriers to Participation in ECE
General barriers to accessing and participating in ECE among low-income families such as knowledge of programs and enrollment procedures, cost and affordability of programs, and complexity of enrollment paperwork or processes (Hanson, Adams, & Koball, 2016) are all likely to influence immigrant families, as their access to income, resources and programs is often explicitly restricted (Crosnoe, Pedroza, Purtell, et al., 2012;Fortuny, Capps, Simms & Chaudry, 2009;Jiang, Ekono, & Skinner, 2015;Krivo, Washington, Peterson, Browning, Calder & Kwan, 2013; Perreia, Crosnoe, Fortuny, et al., 2012). In addition, immigrant households may face additional barriers specific to their legal status.
Barriers specific to immigrants, such as language, knowledge of programs and lack of connectedness to social networks that may be able to provide information, can deter their access to programs and services including healthcare (Carson & Staley, 2016) public assistance and social safety net programs (Perreia, Crosnoe, Fortuny, et al., 2012), and to ECE programs (Karoly & Gonalez, 2011). Immigrants may lack information about the different types of ECE programs available or they may have trouble gaining information due to language barriers or complex regulations (Karoly & Gonzalez, 2011). Immigrants may be unfamiliar with rules and regulations regarding qualifying for child care subsidies and find ECE beyond their budgets (Greenberg, Adams, & Michie, 2016). Further, cultural norms may favor in-home care provided by family members rather than placement of children with unfamiliar adults (Hanson, Adams & Koball, 2016). Finally, lack of transportation has been identified as a common barrier to participation in ECE among low-income and immigrant families (Greenberg, Adams, & Michie, 2016;Mendez, Crosby & Helms, 2016). Families who lack access to reliable transportation (e.g., vehicle ownership, dependable rides/carpools, access to high-functioning public transportation) are also less likely to be able to enroll their children in ECE (Mendez, Crosby & Helms, 2016 (2013), undocumented immigrants were identified by retaining those who reported that they were not U.S. citizens, that they held some "other" status than family, employment or residential visas and non-immigrant classifications, and that they had not adjusted their status since arriving to the U.S.
Once all households including undocumented members were identified, households that reported having at least one child between three-and five-years-old (i.e., the age when children typically attend ECE) were retained for analyses. For each state at each time point (1996, 2001, 2004, and 2008), an enrollment rate in ECE (occurring outside of the home) among undocumented households was calculated as the number of undocumented households in the state reporting that at least one child was enrolled in ECE outside of the home (i.e., family daycare, a childcare or daycare center, a nursery or preschool, a federally supported Head Start program) divided by the total number of undocumented households with three-to five-

Enrollment in ECE (Dependent Variable
year-old children in the state. Factors associated with ECE participation.

Legal Access to a DL (Independent Variable
Ten covariates found in past researcher to be associated with ECE participation were used in analyses. Nine were obtained from the SIPP and one was a composite representing the accessibility of childcare assistance in each state created using data from the NWLC. All of the covariates were coded at a household level and then converted into state averages. Nine of the covariates were coded for undocumented households with three-to five-year-old children only. One covariate, the percentage of non-citizens in the state, was coded for all households.  (1996, 2001, 2004, and 2008). Access scores ranged from one through five across states and years. The mean score across all years was 3.17 (SD = 0.40).

Analytic Strategy
The final sample of undocumented households with children between three and five was identified. Data were checked for completeness before computing the outcome variable and all covariates. Prior to other analyses, descriptive statistics were examined to assess for normality of the dependent variable (DV; ECE participation rate), and covariates. The aim of the present study was to examine the association between state policies expanding DL access to undocumented immigrants (IV) and ECE participation rates among children of undocumented immigrants (DV). ECE participation rates were assessed between states with and without DL policies and within-states that enacted DL policies over time using synthetic control methods. The association between policies expanding DL access and ECE participation rates was also examined post-hoc using regression analyses and at the level of the household using a binary logistic regression. Each analysis is described in the sections that follow.

Synthetic Control Methods for Assessing State Policies
Synthetic control methods (SCM) are used in research to estimate the potential effectiveness of policy interventions or historical events occurring at an aggregate level such as the community, state or country (Abadie, Diamond, & Hainmueller, 2011). Typically, in matched-control analyses, a target is compared to a control considered to be the best match based on some number of relevant indicators. The SCM improves upon these matched-control methods by allowing for the creation of a synthetic version of the target that consists of a weighted combination of all available controls. The SCM compiles pre-policy data from multiple controls on a set of covariates theorized to be associated with the DV. The SCM uses as many or as few of the available controls as is appropriate to best simulate pre-policy explanatory factors associated with the outcome of interest observed in the target (Abadie & Gardeazabal, 2003). The rationale of using multiple controls to create a synthetic control is that the most relevant and similar characteristics of several controls weighted based on their degree of similarity to the target will produce a better match to the target than any one control state could on its own (McClelland & Gault, 2017).
The selected covariates are also weighted based on the strength of their relationship to the DV.
Once the synthetic control has been created through the optimal weighting of all available controls and covariates, pre-policy means of the observed levels of the covariates are used to assess the goodness of fit between the synthetic control and the target. A goodness of fit between the synthetic and the target, as indicated by a small mean square prediction error (MSPE), is an important assumption of the SCM. In general, an MSPE of less than 0.05 is acceptable (Abadie & Gardeazabal, 2003;McClelland & Gault, 2017). If the assumption is met, the levels of the DV observed in the target can be compared to those estimated for the synthetic control pre-and post-policy enactment. The DV is plotted over time as observed in the target and as it is estimated for the synthetic control. Percentage point gaps between levels of the DV observed in the target and estimated for the synthetic control are calculated at each timepoint by subtracting the value estimated for the synthetic from the value observed in the target.
Statistical inference is carried out through a series of placebo tests. In SCM, a placebo test refers to running the same model specified for the target but coding a control that did not experience a change to the policy as if the policy had been enacted (i.e., using the same coding parameters as used with the target; see Appendix F for full SCM code). Placebo tests can be run using one control, such as the highest weighted control in the synthetic. Ideally, however, placebo testing is conducted with multiple controls, and this process was carried out in the present study. It would be expected that any differences observed between the target and its synthetic would be unique when compared to placebo states and their synthetic controls. A complete description of synthetic control methods is provided in Appendix D.

SCM to Examine DL Policies in Relation to ECE
To assess the association between expanding access to DLs to undocumented immigrants and state rates of ECE participation among their children, target states and a pool of potential control states were identified. The three target states for which SCM was used were: (1) New Mexico, which implemented a DL policy in 2003, (2) Utah, which implemented a DL policy in 2005, and (3)  excluded from SCM because pre-policy data were not available, but it was assessed and this analysis is discussed in a following section.
Synthetic control models were conducted separately for New Mexico, Utah and California (i.e., target states). For each of these states, the pre-and post-policy time points were defined based on the year the DL policy was enacted in the state.
Then, an optimization process that determined the best combination of potential control states was run. All potential control states were weighted from 0.000 to 0.999.

Washington State
Pre-policy data were not available to assess Washington using synthetic control methods. Therefore, in order to assess the association between DL access and ECE participation in Washington, regression analysis was used. Specifically, a dummy variable was created to compare Washington to all other states that did not enact DL policies between 1996 and 2008 (Washington = 1; all other control states = 0). Time, as a continuous variable, was also entered into the model with the rate of participation in ECE as the outcome variable. Variables found to be associated with ECE participation across the study period were entered into the regression.

Household-Level ECE Participation: Binary Logistic Regression
In order to assess whether access to a DL is associated with an increased likelihood of ECE enrollment at a household level, a logistic regression was employed using ECE enrollment (1=yes, 0=no) as the outcome, DL availability (1=yes, 0=no) as the main predictor, and the same set of covariates known to be associated to ECE enrollment as used in prior analyses. The purpose of this analysis was to explore if DL availability is associated with ECE enrollment at a household rather than state level and if DL availability is predictive of ECE enrollment above and beyond variables indicative of the socioeconomic and demographic composition of households typically predictive of enrollment in ECE programs outside the home.

DL Access and ECE Rates: Post-Hoc Regression Analyses
In order to further explore the potential association between DL access and ECE participation rates, post-hoc regression analyses using a difference-in-differences framework were conducted. Specifically, time (pre-policy = 0; post-policy = 1), DL access (yes = 1; no = 0), and the interaction between time and DL access were entered into a regression model with rate of ECE participation as the outcome variable.
Variables found to be associated with ECE participation across the study period were entered into regressions. Separate regressions were run for New Mexico and Utah, the two states in which DL policies were enacted during the study timeframe.

Descriptive Statistics
Demographic information and characteristics of households included in the sample are presented in Table 1

Participation in ECE among Children of Undocumented Immigrants
Information about use of ECE and other care is reported in demonstrated that assumptions of normality were met.

Association between DL Access and ECE Rates
The purpose of the study was to examine the association between state-level policies that expanded legal opportunities for DLs to undocumented immigrants and the rate of participation in ECE among their children. It was expected that: H1: There would be higher rates of ECE participation in states that enacted policies expanding legal opportunities for DLs to undocumented immigrants compared to states that did not enact DL policies following policy enactment.
In order to examine this hypothesis, synthetic control models were used. The optimal synthetic control that was determined for each target state is presented first, followed by the synthetic control models for New Mexico, Utah and California separately. ECE participation rates over time were compared between each target state and its synthetic control, including the percent gap in ECE participation between rates in the target state and rates in its synthetic control, and results of placebo testing.

Determination of Synthetic Controls
The weights determined for all states that made up the synthetic control for each target state are presented in Georgia. All other states were weighted at 0.00. Table 3 also provides the mean square prediction error (MSPE) calculated in order to assess the goodness of fit between each target state and its synthetic control.
The MSPE calculated for each comparative case was below 0.05 (range: 0.0001 -0.006), indicating a good fit between target states and their synthetic controls.
Comparisons of the mean levels of covariates observed in each target state and estimated for synthetic controls are provided in Appendix I. The MSPE was also calculated to assess the goodness of fit between each placebo state and its synthetic control in the pre-policy period and these values are provided in Appendix J.  Table 7 provides the corresponding numeric values of ECE participation in Arizona and Nevada and those estimated for their synthetic controls at each time point.

California: A Comparative Case
In order to examine a counterfactual case to New Mexico and Utah, in which conditions were in place to pass a DL policy, but the policy never went into effect, California was examined. Figure

DL Access and ECE Rates in Washington State
Washington state enacted its DL policy in 1993, regression analysis was used to examine the influence of access to a DL for undocumented immigrants on rates of ECE participation in Washington compared to rates in untreated states controlling for significant predictors of ECE. Results are presented in Table 10. In addition, ECE rates observed in Washington and plotted in comparison to the average of all other control states are illustrated in Figure 5.

Post-Hoc Analyses of State-level DL Access and ECE Participation
In order to evaluate the association between DL access and ECE participation using a more conventional analysis that included significance testing, regression analyses using a difference-in-differences framework were conducted. DL access (yes/no), time (pre/post policy) and the interaction between DL access and time were entered into the regression model along with the percentage of Hispanic or Latino households and the percentage non-citizens in the state. Data from New Mexico and Utah were examined compared to untreated states controlling for significant predictors of ECE.

DISCUSSION
The aim of the current study was to examine of the association between state policies that expand legal opportunities for drivers' licenses to undocumented immigrants and ECE participation among their three-to five-year-old children. Using several different methodologies, the present study assessed rates of participation in ECE between states with and without policy changes, within states that passed DL policies before and after policy changes, and in one state in which policy changes were never enacted. Access to DLs and ECE participation was also examined across all states at the level of the household. Overall, results revealed small but consistent differences between the patterns of ECE participation in states that enacted DL policies (i.e., New Mexico, Utah, and Washington), and those that did not, with ECE participation rates higher in states with DL access as compared to those without postpolicy. Each of these findings will be discussed in further detail below.

State-Level DL Access and ECE Participation
In line with hypotheses, ECE participation rates in the states under study all point to increases relative to control states post policy implementation. In interpreting these findings, it is important to understand the context in which the DL policies were adopted/enacted. One argument suggests that the states that pass DL policies favorable to immigrants may in fact be states that are much more likely to favor other "pro" immigrant policies (Filindra, 2013). In that case, one might argue that it was the overall environment within a state that passed favorable DL policies that is responsible for the increased participation rates observed, rather than the DL policy per se. and aptitude, and also increases the likelihood of being insured (Johnson, 2004;Mounts, 2003). In addition, others have pointed to the fact that a DL provides an official documentation of identity that serves a wide range of purposes beyond the legal privilege to drive, such as proving residency or home address (Johnson, 2004

Contextualizing the findings: 1996 -2008
The results of the current study point to DL access as a facilitator of ECE participation for children. Though ECE participation rates were observed to be higher in states with DL access compared to states without, differences observed were small and should be interpreted within the time frame under study. With the terrorist attacks of 9/11, the U.S. government implemented additional law enforcement measures, some of which explicitly targeted certain nationalities (Ewing, 2012). Specifically, around this time, states began entering into 287 (g) memoranda of agreement with the federal government which allowed state and local municipalities to enforce federal immigration law at the local level. This increase in internal enforcement was widely considered to have the most dramatic effects on the immigrant community in terms of decreased trust in police, neighbors, agencies and less willingness to leave the house when unnecessary (Johnson, 2002). The vast majority of persons removed as part of increased enforcement efforts after 9/11 were of Mexican descent (72%) (Massey & Pren, 2012), which was generally associated with an increased fear of government, a greater number of family separations, and general "hunkering down" of immigrant families in the U.S. (Marcuse, 2006). Specifically, qualitative research carried out in immigrant communities has suggested that government responses after 9/11 were associated with increased fear of deportation, heightened stress, decreased trust in community institutions, concerns about furnishing paperwork for health insurance, and restricted access to healthcare services (Hacker et al., 2011). It would not be surprising then, to find an association between these same responses and reductions in ECE participation among immigrant households.
Anti-immigrant sentiment not only increased due to enforcement-focused policy that framed immigration as a matter of national security, but also in the wake of Alternatively, it is important to note that increased ECE rates in states with DL access were observed despite high levels of anti-immigrant sentiment, worsening economic conditions and nationally declining ECE rates as compared to matched control states exposed to the same national trends. What is certain is that a range of sociopolitical and contextual factors contribute to ECE rates and the effectiveness of any single policy cannot be evaluated separate from this larger context.

Household-Level DL Access and ECE Participation
The primary aim of the present study was to examine DL access and ECE participation in aggregate at the state level and findings provided support for initial hypotheses. The secondary analysis examined ECE participation at the level of the household and was not supported. DL access, above and beyond characteristics of households and household members, was not associated with ECE participation. The absence of an association is not surprising, especially in light of the overall decline in ECE participation over time, and the well-established associations between household characteristics such as income, employment status, parental education and identification as Hispanic or Latino and ECE participation. Additionally, the SIPP does not ask respondents to report on whether household members actually have a DL.
The DL access variable used in this analysis was based on the state in which the household was located and the year of data collection. The exploration of alternative sources of data and analytical methods might allow for more nuanced examination of DL access and ECE participation. Additionally, household level data could be collected to more directly link DL issuance to ECE participation of children.
Households that did not apply for a DL could also be surveyed to understand barriers, potential policy specifics that deter or restrict undocumented immigrants from accessing DLs, and ideas they may have for making the process more feasible.

Strengths of the Present Study
A major strength of the present study was the use of nationally representative, publicly available data from the SIPP. The SIPP is the only nationally representative survey that specifically asks respondents about the legal immigration status of all household members over the age of 15 (U.S. Census Bureau, 1998;2016). As a result, identification of the undocumented population in the present study was more accurate than other studies which have had to rely on more distal proxies of legal status (e.g., using all foreign-born, Hispanic, males within a certain age range, who qualify for specific government programs or benefits but have never participated) (Capps, Bachmeier, Fix & Van Hook, 2013). While the sample identified in the present study likely underestimated the total number of undocumented households, due to the highly sensitive nature of such disclosures, the SIPP provided a higher degree of reliability than methods used in past studies.
A second strength of the current study was its use of synthetic controls Studies utilizing synthetic control methods often focus on just one target state or unit.
The present study was able to assess four target states. The policy specifics for each of the target states assessed varied and provided for more nuanced interpretations to be made.
Another strength of the present study is its timeliness. In the absence of comprehensive immigration reform at the federal level for over 30 years, it has fallen upon the states to determine how to address large and growing immigrant populations of all legal statuses including undocumented immigrants. The issuance of driver's licenses has long been in the purview of states and with 12 states having already implemented such policies, enactment would appear to be highly feasible in a wide range of settings. Results of the present study can serve as impetus for future studies to validate the association between DL access and ECE participation with the goal of informing state-level policy decisions. In our own state of Rhode Island, a similar policy has been brought to public hearing for debate several times over the past few years. It is possible that these findings could be used in support of passing such a policy in Rhode Island and in other states in the U.S. currently engaged in the same debate.

Limitations of the Current Study
The present study was not without limitations. One limitation is that between 1996 and 2008, new SIPP panels were initiated every three to five years, limiting the number of time points that could be used for analyses to four. In synthetic control methods, the use of a greater number of time points is preferred, particularly the inclusion of pre-policy time points (McClelland & Gault, 2017

Future Directions
Results from the present study provide preliminary evidence that expanding legal opportunities for DLs to undocumented immigrants may have the potential to increase ECE participation among children of undocumented immigrants and invites several lines of future research. Evaluating more states across a wider span of time would provide more information in determining the association between access to DLs for undocumented immigrants and enrollment in ECE programs among their children.
Since 2008, nine states have enacted policies making DLs available to undocumented immigrants. It would also be useful to explore outcomes within a state that has removed access to a DL. Colorado, for example, is one state that would be very informative to analyze as it enacted a DL policy in 2014, but since then, the program has not received adequate funding and DLs that were issued initially have expired without mechanisms for undocumented immigrants to renew (Escalante, 2015).
Research that aims to explore the influence of DL policies within state contexts would be especially helpful in informing future state legislation. Future studies may also aim to explore how DL access influences interactions with local police, as not have a DL has been associated with increased fear of reporting crimes to police (Nguyen & Hill, 2016). Based on past research that fear of driving without a license is associated with missed healthcare appointments for parents and children (Ayon, 2008;Hacker, Chu, Leung et al., 2011;Toomey, Umana-Taylor, Williams, Harvey-Mendoza, Jahromi & Updegraff, 2014), future research should aim to explore the association of access to a DL for undocumented immigrants and health outcomes among undocumented immigrants and their children. Health outcomes could include attending annual or biannual appointments as well as indicators of physical health. Importantly, mental health outcomes should not be neglected, as fear of driving without a DL has been associated with psychological distress (Yoshikawa, Godfrey & Rivera, 2008) and heightened anxiety around driving for tasks of daily living (e.g., food shopping) in communities with increased immigration enforcement (White Yeager, Menachemi & Scarinci, 2014).

Exploration of the mechanisms by which DL access facilitates participation in
ECE and how the association between DL access and ECE enrollment may be moderated by the state, other active policies, DL policy specifics, or the ethnicity of undocumented immigrants, for example, should be explored. Understanding if it is through the legal ability to drive, an increased trust in government, increased rates of employment, simply having a valid form of identification that proves residency, or by other mechanisms will be important information in developing policy, and in further understanding the challenges and barriers associated with being undocumented.
Characteristics of the undocumented population are also important to consider. Asian immigrants are the fastest growing group of immigrants in the U.S., it will be important to explore the association of DL access and ECE participation among different ethnic groups of immigrants.
Finally, research has shown that the benefits of ECE extend well into the future. Children who participate in ECE have better outcomes in terms of educational achievement but also lower rates of special education placements, fewer teen pregnancies, lower rates of incarceration, and higher earnings in adulthood (Pinto, Heckman & Moon, 2010). Implicit in these findings are cost savings associated with these long-term outcomes. Future research on DL access could examine the potential costs savings as a function of increased participation rates in early care and education using predictive modeling to assess the economic benefit associated with the cost of enacting a DL policy. Using a cost-benefit analysis of implementing a DL policy to increase rates of ECE participation would provide more accurate measures of the social and educational outcomes for children of undocumented immigrants. Lacking federal reform, states may look to invest in cost-effective programs and policies that address immigration and immigrants in their communities. Research has demonstrated that access to transportation is particularly salient among undocumented immigrants (Hanson, Adams & Koball, 2016), and that it has implications for access to resources, services and programs across multiple domains, including access to ECE (Greenberg, Adams & Michie, 2016;Puma et al., 2010 It is well established that children of immigrants participate in ECE at lower rates than their peers with citizen caregivers (Karoly & Gonzalez, 2011), and data used for the present study demonstrated similar patterns in participation. We also know that participation in ECE is associated with short-term benefits such as increased school readiness, language and mathematics skills (Karoly & Gonzalez, 2011), and longer term benefits including higher educational attainment and earnings and lower likelihood of incarceration (Deming, 2009;Winship & Grannis, 2013). These associations hold true for all children, but are even stronger among children from socially disadvantaged backgrounds, including children of immigrants (Puma et al., 2010. Thus, the promotion of policies that have the potential to reduce barriers and improve access to ECE for vulnerable populations, such as DL access for undocumented immigrants, merits ongoing evaluation. Households linguistically isolated (%) 9

TABLES AND FIGURES
Note. All data presented in Table 1

ECE Participation Rates in Utah and Synthetic UT Plotted from 1996-2008 11
Figures 3a and 3b.

ECE Participation Rates in Washington Plotted in Comparison to the Average of All
Control States (1996States ( -2008 Notes. Controls included all 26 states that were used in synthetic control models and excluded New Mexico and Utah. California was also included in controls for Washington.   Any non-citizen, non-national of the United States who enters the country without permission and/or inspection or has fallen "out of status" (e.g., has overstayed a visa, has let necessary paperwork expire). Undocumented immigrants are deportable if apprehended.

Nonimmigrant*
Any person who is not a citizen or national of the United States that has been granted the right to reside in the United States temporarily, such as workers and students.
Note: Table produced  This law creates a driver's authorization card and allows applicants, regardless of legal status, to provide birth certificates or passports issued by a foreign country as proof of identity. This law also prohibits the release of information related to legal status for purposes relating to the enforcement of immigration laws. The Supreme Court rules that a child of Chinese descent born in the U.S. -whose parents were subjects of the emperor of China, but who were domiciled in the U.S. as permanent residents at the time of the child's birth, and who were not employed in any diplomatic capacity to the emperor of China -is a citizen of the United States.
1903 Immigration Act of 1903 Added four inadmissible classes: Anarchists, people with epilepsy, beggars, and importers of prostitutes.
1906 Naturalization Act of 1906 Standardized the naturalization process and made some knowledge of English a requirement of citizenship. It also established a Bureau of Immigration and Naturalization.
1907 Immigration Act of 1907 Restricted immigration for certain classes of disabled and diseased persons.

Barred Zone Act
Restricted immigration from Asia by creating an Asiatic barred zone and introduced a reading test for all immigrants over 14 years of age with exceptions for children, wives, and elderly family members.
1918 Immigration Act of 1918 Expanded provisions of anarchist exclusion.

Emergency Quota Act
Limited the number of immigrants from any country to 3% of those already in the U.S. according to the 1910 census. (An unintentional consequence of this legislation was increases in illegal immigration. People who did not make the quotas migrated to Mexico or Canada and subsequently crossed the borders in to the U.S. illegally).

Cable Act
Reversed the former immigration laws regarding marriage whereby a women lost her citizenship if she married a foreign man. (The same restriction was never put in place for U.S. men who married foreign women).

1924
Johnson-Reed Act Imposed the first permanent numerical limit on immigration and began a national origin quota system.

National Origins Formula
Capped total annual immigration at 150,000. Immigrants fell into two categories: those from quota nations and those from non-quota nations. Immigrant visas from quota nations were restricted to the same ratio of residents from the country of origin out of 150,000 as the ratio of foreign-born nationals in the U.S. from that country. The percentage of 150,000 was the relative number of visas a particular nation received. Contiguous countries to the U.S. were considered non-quota countries. Laborers from Asiatic nations were excluded but exceptions existed for certain professionals, clergy, and students.

1934
Equal Nationality Act of 1934 Allowed for foreign-born children of American mothers and immigrant fathers who had entered the U.S. before age 18 and lived in the U.S. for at least 5 years to apply for citizenship for the first time.

1930's
Federal officials deported tens of thousands (estimated more than 400,000) Mexicans and Mexican-Americans. Many, mostly children, were U.S. citizens. Applications for legal admission increased following WWII as did illegal immigration. The main source of illegal immigration was fraudulent marriage. Japanese immigration became disproportionately female as more women left Japan betrothed to emigrant men they had never met.

1942
The Bracero Program A series of bi-lateral agreements between the U.S. and Mexico that allowed millions of Mexican men to come to the United States to work on short-term, mostly agricultural contracts. Between 1942 and1964, 4.6 million contracts were signed making it the largest U.S. contract labor program.

Chinese Exclusion Repeal
Repealed the Chinese Exclusion Act and permitted Chinese nationals in the country to become citizens.

McCarran-Walter Act
Set a quota for immigrants with skills needed in the U.S. and increased the power of the government to deport illegal immigrants suspected of communist sympathies.

1953
Kwong Hai Chew vs. Colding The Supreme Court rules that once an immigrant lawfully enters and resides in this country s/he becomes invested with the rights guaranteed by the Constitution to all people within our borders.

Operation Wetback
Immigration and Naturalization Services rounded up and deported undocumented immigrants in selected areas of California, Arizona and Texas along the border. U.S. border patrol estimated that more than 1.3 million either left voluntarily or were deported.

1965
Hart-Cellar Act Repealed national origin quotas and initiated the family visa system as well as a visa system for workers (skills). The act set a quota for western hemisphere immigration and a 20,000 per country limit for eastern hemisphere immigrants.

1970's
The United States estimates that the total number of undocumented immigrants in the country is 1.1 million or half of one percent of the total U.S. population.
1980's An additional 1.3 million undocumented immigrants enter the U.S.

1982
Plyler vs. Doe The Supreme Court rules that undocumented immigrants are also within the jurisdiction of states and therefore receive equal protection of law and due process of law.
1986 Immigration Reform and Control Act (IRCA) Created sanctions for knowingly hiring undocumented workers and increased border enforcement. Amnesty was granted to undocumented immigrants already living in the U.S.

1990's
Over 5.8 million undocumented immigrants entered the U.S. in the 1990's. Mexico rose to the head of the list of sending countries followed by the Philippines, Vietnam, the Dominican Republic and China.

1990
Immigration Act of 1990 Increased ceilings for immigration and created a diversity admissions category. The number of visas for priority workers and for foreign professionals with U.S. job offers was tripled.

1990
The United States vs. Verdigo-Urquidez The Supreme Court reiterates their ruling on Kwong Hai Chew vs. Colding that undocumented immigrants who are in the U.S. are invested with the same rights guaranteed by the U.S. Constitution.

1996
Illegal Immigration Reform and Immigrant Responsibility Act of 1996 (IIRIRA) Put in place phone verification for worker authentication by employers and further increased border enforcement. The act made access to welfare more difficult for legal permanent residents and legal immigrants in the United States. In addition, the Reed Amendment attempted to deny visas to former U.S. citizens but was never enforced. Placed more agents at the southern border. Required schools to report foreign students attending classes to the federal government and stipulated that foreign nationals in the U.S. will be required to carry identification with biometric technology.

REAL ID Act
Required the use of IDs meeting certain security standards to enter government buildings, board planes, and to open bank accounts. Created more restrictions to claim political asylum. Severely curtailed Habeas Corpus1 relief for immigrants and increased immigration enforcement mechanisms. The act established standards for state-issued drivers' licenses and cleared the way for the building of barriers at U.S. borders.
Note. Table produced using information from the United States Department of Homeland Security (dhs.gov), and the United States Citizenship and Immigration Services (USCIS) History Office and Library.

Appendix D: Political Contexts in states analyzed in the current study
Prior to the 1990's when undocumented immigration, especially from Latin America, increased rapidly, most states did not have policies that either banned or allowed undocumented immigrants to access DLs. In the 1990's, a time of high nativism (Pan, 1993), many states began implementing policies that required proof of legal status in order to obtain a DL, or required that an SSN be provided (Johnson, 2004). In 2005, the federal government passed the REAL ID Act, obligating states to require proof of citizenship or legal immigration status. To date, 12 states and the District of Columbia have expanded DL privileges to undocumented immigrants.
There were two states that enacted policies between 1996 and 2008, the time period for which data existed to analyze DL policies for undocumented immigrants for the purposes of this study. Those states, referred to throughout as target states, were New Mexico and Utah. Washington enacted the first DL policy in 1993. Pre-policy data did not exist to analyze Washington before and after its DL policy, but it was used solely for the purposes of a between states comparison. California passed a similar policy in 2004, but never enacted it. California was used as a comparative control case in this study. Tables 1 and 2 provide an overview of the political climate in each target state (Table D1) and in demographically similar states that were used as placebo states in this study (Table D2)  Mexico's Republican governor has tried several times to repeal the DL policy, claiming that criminals from other states will try to come to New Mexico if they know they can obtain a DL (Crawford, 2013;Washington Post, 2011;2016). With the support of a large share of the population, the Democratic-controlled state legislature has not passed any of the governor's proposed repeals to the DL policy, ant it has not been altered since its enactment in 2003. It was estimated in 2016 that around 90,000 New Mexicans were eligible annually to obtain the DL (New Mexico Department of Motor Vehicles, 2016). In addition, it was reported that when the law was first enacted, the rate of uninsured driver's dropped from 33% to 9% (Attanasio, 2015).

California
In 2004, California passed a policy that expanded access to DLs to undocumented immigrants after explicitly barring undocumented immigrants from obtaining DLs in 1993, however the 2004 policy was never enacted (Clark, 2011).
The policy was abandoned when the federal government passed the REAL ID Act in

2005.
For purposes of the current study, California was analyzed as a control case and provided a comparison state in which all conditions for passing the DL policy were in place, but the policy never became effective. Therefore, California provided a way to explore whether it was the DL policy or the prerequisite conditions of passing such a policy in a state that influenced ECE enrollment rates. California had a consistent, Democratic-controlled legislature during this time period. The Latino citizen-age voting population (i.e., eligible Latino voters, with a larger population making it more likely that politicians will consider potential backlash) in California was estimated to be over 5,000,000 and nearly one-third of the total population was Hispanic/Latino. Furthermore, California's state legislature is considered to have high legislative professionalism (NCSL, 2015) meaning that the demographic makeup of the electorate is more likely to be tracked and reported to congressmen and considered in policy making decisions (CITE).

Utah
In 2005, Utah enacted a policy that expanded DL access to undocumented immigrants. The strictest of the DL policies reviewed for this study, Utah's policy (S 227) allows undocumented immigrants to obtain a one-year driving privilege card if they can prove Utah residency for at least six months and provide a tax identification number (NCSL, 2016). Tax identification numbers are available to nonresident and resident immigrants, their spouses and dependents who cannot obtain a social security number (Internal Revenue Service, 2017). In addition, the law explicitly prohibits any government entity to accept the DL as a valid form of identification. The DL given to undocumented immigrants displays a large, red "P" on the card to indicate that it is strictly a driving privilege card. Utah was consistently a Republican-controlled state with a very small Latino citizen-age voting population and with very low legislative professionalism (NCSL, 2015). Expanding legal access for DLs to undocumented immigrants was endorsed as an issue of public safety and in an effort to ensure that all drivers were insured (Katz, 2007). The large, red "P" on the DL was seen as a compromise to appease the most conservative among the legislature. Although immigrant rights groups in the state opposed the issuance of a markedly different DL to undocumented immigrants, leaders of such groups and the state attorney general have stated publicly that discrimination from police has not been widespread (Katz, 2007). In 2014, it was estimated that more than 35,000 undocumented immigrants held a state DL (Davidson, 2014). were shortened and administered at each of the 4 (1-year) waves (SIPP User's Guide, 2016). The following will provide detailed information on (1) the SIPP sampling design, (2) SIPP core and topical module content, and (3) interviewing procedures.

SIPP Sampling Design
The Census Bureau uses a multi-stage, stratified integrated sample design, based primarily on the decennial census of population to provide samples for all of its major household surveys, including the SIPP.  Guide, 2001Guide, , 2008Guide, , 2016.

Selection of Primary Sampling Units
The SIPP utilizes a multistage, stratified sample of housing units. In the first stage, the frame for selecting primary sampling units (PSUs) consists of a listing of all U.S. counties and independent cities along with their corresponding population counts.
Single counties are used as long as they have a population of at least 7,500. When this population threshold is not met, adjacent counties are combined to form a PSU. PSUs containing 100,000 or more housing units are considered self-representing (SR) and are always included in the SIPP sample (i.e., housing units will be sampled from these PSUs

Selection of Addresses in Sample PSUs
The survey population for the SIPP consists of adults (persons age 15 and older) in the civilian, non-institutionalized population of the U.S. To represent this population, a sample of addresses is selected from each of the sample PSUs (i.e., all SR PSUs and two NRS PSUs from each stratum) using five, non-overlapping sampling frames. Around 90% of the addresses included in the SIPP come from the largest two sampling frames -Address Enumeration Districts (EDs) and Area Enumeration Districts (Enumeration Districts are similar to Census tracts and created using political or geographical boundaries) (SIPP User's Guide, 2001Guide, , 2008Guide, , 2016.
The Address ED framework consists of a list of addresses in all EDs which were located in permit-issuing areas, and for which at least 96% of the addresses in the ED are complete, having a street name and house number. In each sample PSU, addresses in these EDs are divided into clusters, each containing two neighboring housing units. Samples of those housing clusters are then selected and assigned to the SIPP panel (Shapiro, 1983c(Shapiro, , 1984

Determination of Persons and Households and Interview Procedures
Although sampling ends at addresses, the unit of observation in the SIPP is not addresses, but persons age 15 and older and units such as families and households.
Therefore, once the final sample of all addresses is selected for the first wave of interviewing, a specific set of rules is followed in order to establish who is part of the household. In addition, rules have been established to determine who the Census Bureau continues to follow and interview after the first wave and when changes occur in household composition.
At wave one, the interviewer visits the sampled address, compiles a household roster, and attempts to interview all members of the household who are age 15 and older. The interviewer's first task is to determine whether each person present at the address is a household member. In most cases, a household member is defined as someone who sleeps in the household a majority of the time. Once the household has been determined, interviews take place with each member. The preference of the Census Bureau is that all household members over age 15 answer for themselves, however proxy interviews on behalf of another member are permitted. Based on answers to given by each household member over 15, the interviewer also identifies a reference person for the household. The reference person is usually the owner or renter of the housing unit.
Beginning at the second wave, and in all subsequent waves, the interviewer updates the household roster by listing all people living or staying at the address, including anyone who has joined the household (e.g., new spouse or baby The SIPP interview process, rules and procedures have been consistent over the years, however significant changes began in 2004 with the use of feedback or dependent interviewing. In this method, information from the previous interviews carries forward into the current survey instrument to streamline interviewing procedures and maximize data quality. This process helps to limit a common problem with longitudinal surveys, "seam bias." Seam bias is the phenomenon by which life events and changes are disproportionately reported at the seam (i.e., between waves).
They key to countering seam bias is having overlapping periods. Dependent interviewing automatically includes a portion of the next reference period. The 2014 SIPP panel uses this method for over 500 items.

SIPP Survey Content
All years and versions of the SIPP have covered the same content, though the actual arrangement of questions and sections have been structured differently. For example, prior to 1996, the core content was presented in four sections rather than two sections as has been used since 1996. In all panels prior to 2014, a core survey was administered at all waves as well as rotating topical modules that varied in content from wave to wave. In 2014, all survey content was administered at each wave including most of the original topical module content.

Core Content
Core SIPP content is asked at the start of every interview at every wave. property, and mutual funds. The smaller sections of the core content include items on health insurance ownership and coverage, education and attainment, energy assistance, and school lunch and school breakfast program participation.

Topical Module Content
Topical modules were designed to gather specific information on a wide variety of subjects. Topical modules are not repeated in each wave. Certain modules are administered at each wave following the core content of the SIPP survey. There are around 20 topical modules used throughout the duration of almost all SIPP panels.
For the purposes of the proposed study, the Child Care and Migration History modules will be used. The frequency and timing of these modules varies by SIPP panel (see Table 1). Each module is described below. respondents are also asked to report that country where they were born, the countries in which they have lived, the country from which they moved to the U.S., and in what year they came to the U.S.

The Current Study
The current study used core content, the migration history module, and the child care module of the SIPP from 1996, 2001, 2004, and 2008. The reason for using these panels is that they are inclusive of all SIPP panels administered during the time period during which most state polices pertaining to driver's license access for undocumented immigrants have become effective (see Table 1c).

Appendix F: Synthetic Control Methods Introduction
Often, social scientists are interested in research questions around the influence of historical events, widespread societal changes, or policy interventions on aggregate units such as neighborhoods, states or countries (Abadie, Diamond & Hainmueller, 2011). Traditionally, methods such as comparative case studies have been used in order to assess questions of this nature. In comparative case studies, outcomes for the treated group are compared to outcomes for groups that were not exposed to the intervention of interest (i.e., control groups) (Abadie, Diamond & Hainmueller, 2010).
The reasoning behind this is to use the outcome observed for the control group to estimate the outcome one would have observed for the treatment group if the treatment group had not been exposed to the intervention of interest. One limitation of comparative case studies has been that the selection of the control group is left up to the analyst, leaving room for error and a lack of confidence in the degree to which the control group can act as a reasonable proxy for the treated group (Abadie, Diamond & Hainmueller, 2011).
In order to address these limitations, synthetic control methods were developed (Abadie & Gardeazabal, 2003;Abadie et al., 2010). Synthetic control methods are best suited to use linear panel data and employ a difference-in-differences framework whereby effects of any unobserved variables on the outcome of interest are allowed to vary with time (Abadie et al., 2010). Synthetic control methods also allow for the creation of a data-driven control-group called a synthetic control unit (SCU). The SCU is defined as "a weighted average of all available control units that approximates the most relevant characteristics of the treated unit prior to the treatment" (Abadie & Gardeazabal, 2003). It is often quite difficult in practice to find a single unexposed/untreated unit that sufficiently approximates the most important characteristics of the treated unit, and the motivation behind the use of a SCU is that a combination of units is better able to approximate the unit exposed to the event of interest (Abadie, Diamond & Hainmueller, 2010). Synthetic control methods determine the most appropriate combination of available controls and also make clear the relative contribution of each of the potential control units entered into the model, forcing researchers to acknowledge and attend to the relative degree of similarity between the SCU and the treated unit (Abadie et al., 2010).
In order to implement synthetic control methods in R, the package, Synth can be installed (R Development Core Team, 2011; http://CRAN.Rproject.org/package=Synth). The package Synth (1) constructs the synthetic control unit, (2) allows the user to organize the data in a format needed to run the synthetic control analysis, and (3) produces tables and figures that summarize the data and illustrate results. Each step will be discussed in more detail in the following sections.

Synthetic Control Unit Procedure
Each treated unit must be analyzed separately. For each analysis, then, there is one treated unit and a "donor pool" of potential control units that were not exposed to the treatment. Therefore there are J + 1 units with one treated unit (unit i) and J remaining control units that can contribute to the SCU. In addition, the intervention of interest occurs at time T0 + 1 with the pre-intervention time period denoted as 1, 2, …, T0 and the post-intervention period denoted as T0 + 1, T0 + 2… T. The outcome that would be observed for unit i at time t in absence of the treatment is denoted as Yit N .
The outcome that would be observed if unit i is exposed to the intervention is denoted as Yit I . The goal of the analysis is to estimate the effect of the target policy, historical event, or societal change for the treated unit in the post-intervention time period (i.e., the difference between the two potential outcomes defined above; α1t = Yit I -Yit N ).
Importantly, the outcome Yit N is not observed for the treated unit in the postintervention period. This is the outcome that will be estimated using the synthetic control unit that is derived from the data by solving an optimization equation.
In order to construct a synthetic control that resembles the treated unit as much as possible in all relevant pre-intervention characteristics, a vector Ui, is created using all observed covariates for each of the potential control units. The covariates used are typically a set of predictors known to be associated with the outcome variable of interest (Abadie, Diamond & Hainmueller, 2010). In addition, a (T0 x 1) vector, K = (k1, …, kT0)' is used to represent linear combinations of pre-intervention outcomes where: These linear combinations, M, of pre-intervention outcomes are used to control for variables that are likely to change over time and are also likely to influence the outcome of interest (Abadie et al., 2010). Once the vector of observed preintervention predictors, Ui, has been established, a J x 1 vector of weights W = (w2, …, wj+1)' is created, where each W represents one weighted average of control units (i.e., one potential synthetic control unit). Weights are selected by package Synth with respect to the outcome predictors (Ui) and linear combinations of pre-intervention outcome values (M) so that the ultimate synthetic control unit best approximates the unit of interest that was actually exposed to the intervention (Abadie & Gardeazabal, 2003). This is formally expressed as: Package Synth selects weights W* = w2 * + … + w*j+1 Such that: And such that: Essentially this is an attempt to ensure the pre-intervention characteristics of the treated unit and the available untreated units match as closely as possible. Then an estimator is yielded for each of the time periods, T0 + 1, T0 + 2, …T: The estimator, [â1t], is the value assigned as the projected outcome of interest of the SCU at each time point examined. To use the estimator in the synthetic control method, a distance between the treated unit and the SCU must be determined. In order to determine the distance (i.e., similarity) between the SCU and the target state, we combine the demographic and descriptive characteristics of the treated unit in the K x 1 matrix, X1, and the values of the same set of demographics of the control units in the K x J matrix, X0. In order to create the most similar synthetic control unit, the synth( ) function chooses the vector W* to minimize this distance between characteristics of the treated unit (X1) and those estimated for the SCU (X0) subject to the weighted constraints (i.e., ||X1 -X0W||). Specifically, the synth( ) function solves for a W* that minimizes the mean square prediction error (MSPE) of the synthetic control estimator: V is defined as a (k x k) symmetric, positive matrix which is introduced to allow different weights to be applied to the variables in X0 and X1 depending on their predictive power in terms of the outcome of interest. The goal for V, then, is to assign weights such that the MSPE of the synthetic control estimator is minimized. Although the user can opt to define V, the process is implemented by default in the Synth( ) function, and utilizing the default data-driven procedure is recommended (Abadie et al., 2010). In this procedure, V* is selected among all positive definite and diagonal matrices such that the MSPE of the outcome variable is minimized over the userdefined pre-intervention period. In other words, Z1 is the (Tp x 1) vector with the values of the outcome variable for the treated unit over the course of the preintervention period, and Z0 is the (Tp x J) matrix for the control units, where Tp (1 ≤ Tp ≤ T0) is the number of pre-intervention periods over which MSPE is minimized. Then V* is chosen to minimize: where V is the set of all positive definite and diagonal matrices and the weights for the synthetic control are given by W*. Synth( ) then solves a nested optimization problem that minimizes equation (2) for W*(V) given by equation (1).
Currently, a test of statistical significance does not exist for synthetic control methods. However, synthetic control methods facilitate inferential techniques. It is suggested that researchers using synthetic control methods proceed by conducting placebo tests. This is carried out by randomly reassigning the intervention across units (i.e., to a control unit in which the same intervention did not occur). Comparing the primary analysis with the placebo test informs researchers about the rarity and magnitude of the treatment effect that was observed in the treated unit (Abadie et al., 2010).

Application to the Current Study
For purposes of the current study, the treatment units were Washington, New Mexico, Utah, and California, all of which experienced a change in policy regarding the issuance of driver's licenses to undocumented immigrants during the time period for which SIPP data were obtained (1996 -2008). The one exception was Washington, which has issued driver's licenses to undocumented immigrants since 1993. Therefore, no change took place during the time period for which data were available. In California, a policy was passed and signed by the governor in 2004 that The data were obtained from the SIPP, whereby household-level responses from each panel year were used to create a state-level database compatible with synthetic control methods. This database consisted of the outcome variable of interest (i.e., ECE enrollment rate among undocumented households with three-to five-yearold children), the covariates used to match potential control states to each target state (i.e., state-level access to childcare assistance, average weekly cost of ECE, the proportion of the state that identifies as Hispanic/Latino, the proportion of the state that speaks a language other than English at home, the average age of the head of household, the proportion of the state that identified as non-citizens, the average monthly household income, , the proportion of the population with at least a high school diploma/GED, and the proportion of households that are linguistically isolated). A value for each one of these covariates was estimated for each state at each of the four time points (1996, 2001, 2004 and 2008). These values were used to match target states to potential control states and to subsequently create an optimal synthetic control unit for each of the target states examined through the methods described above.
Once the SCUs were created, synthetic control methods were conducted using each target state and its corresponding SCU determined by the Synth( ) function. In addition, placebo tests were conducted in which the intervention was randomly assigned to the highest weighted control state in each target state's SCU (e.g., Arizona for Utah). Inferences were made using the comparisons between each target state and its SCU and the placebo test along with consideration of the social and historical context within each state during the time period examined.