Examining the Impact of Legal Arizona Worker Act on Native Female Labor Supply in the United States

Low-skilled immigration has been argued to lower the price of services that are close substitutes for household production, reducing barriers for women to enter the labor market. Therefore, policies that reduce the number of low-skilled immigrants who work predominantly in low-skilled service occupations may have an unintended consequence of lowering women’s participation in the labor market. This article examines the labor supply impact of the Legal Arizona Workers Act (LAWA), which led to a large decline in the low-skilled immigrant workforce of the state. The analysis shows no evidence that LAWA statistically significantly affected US-born women’s labor supply in Arizona. This finding is partly explained by an increase in native workers in household service occupations due to LAWA, which offset the decline in immigrants in these occupations and caused the cost of household services to be relatively uninfluenced by the passage of LAWA.


Introduction
Between 1970 and 2000, the participation rate of female labor force in the United States increased from 43.4% to 61% (Acemoglu et al., 2004). Despite this large increase, women are still spending much more time on household works compared to men (Cortes and Tessada, 2011). 1 At the same time, a recent work underscores the role of low-skilled immigrants who work predominantly in low-skilled service occupations in lowering the price of services that are close substitutes for household production (Cortes, 2008). As such, the influx of low-skilled immigrants has been argued to increase native women's participation in the labor market The analysis yields a few main results. First, the number of low-skilled immigrant workers in the labor force shrank substantially in Arizona due to the passage of LAWA. In the absence of LAWA, I estimated that the share of low-skilled immigrants in Arizona's workforce would be higher by approximately 1.1% points or 9% of its level in 2006. Despite this large decline, I found no evidence that LAWA significantly affects the native women's labor force participation rate or average weekly work hours. This result holds even among highskilled US-born women who are most likely to be affected by LAWA because of higher opportunity costs of spending time for household works compared to low-skilled native women.
Perhaps surprisingly, I found no evidence that the implementation of LAWA led to a statistically significant increase in the average time use for housework, gardening, and caring for children among US-born women in Arizona. Further analysis shows that this result is driven partly by an increase in native workers in household service occupations due to the implementation of LAWA, which offset the decline in immigrants in these occupations and caused the cost of household services to be relatively uninfluenced by the passage of LAWA.
This increase in native workers in household service occupations is consistent with relative task redistribution argument in which low-skilled immigration nudges US workers toward occupations that require higher communication skill to reduce downward pressure on their wages (Peri and Sparber, 2009; Peri and Sparber, 2011). As LAWA shrinks the low-skilled immigrant workforce in Arizona, there is less incentive for natives to specialize in occupa- This paper is constructed as follows. Section 2 describes the background of LAWA and conceptual framework. Section 3 discusses the empirical methodology and data used in the analyses. Section 4 documents the results. Section 5 concludes.

Background and conceptual framework
The universal E-Verify program such as LAWA can be traced back to the Immigration Reform and Control Act (IRCA) of 1986, which requires new hires to present documents verifying their eligibility to work legally in the US and imposes sanctions on employers knowingly hiring unauthorized immigrants. These measures to curb unauthorized employment in IRCA, however, have been argued to be ineffective because there was no reliable, quick way to verify the authenticity of the documents used to prove identity and work authorization (Cooper and O'Neil, 2005). To address this shortcoming, the E-Verify system was rolled out to several states in 1997 under the name of Basic Pilot. Participating employers enter the new hire information from the employment eligibility form (Form I-9), and E-Verify system checks that information with Social Security Administration and Department of Homeland Security databases.
If there is a discrepancy, the employer is notified of a tentative nonconfirmation, and the new worker has 8 federal working days to contest the discrepancy. While the discrepancy is being contested, the employer is not allowed to fire the new hire because of the discrepancy. However, the employer has to terminate the employment of the new hire if the discrepancy is not resolved after that period. For authorized workers, the inaccuracy rate of E-Verify is approximately 1%, while for unauthorized workers, the error rate is approximately 54% (Westat, 2009).
The Legal Arizona Worker Act was signed into law in July 2007 and implemented on January 1, 2008. It is the first law of its kind that requires all businesses in a state to verify the employment authorization of new hires through the federal E-Verify system. The literature provides guidance on how LAWA may adversely affect the labor supply of U.S.-born women.
Recent works by Cortes (2008) and Cortes and Tessada (2011) argue that low-skilled immigration lower opportunity cost of working by reducing the price of services that are close substitutes for household production. If LAWA leads to higher cost of household services, mainly because of the decline in low-skilled immigrants predominantly working in this sector, the labor supply of US-born women would be adversely affected as it increases the opportunity cost of working.
It is worth noting that LAWA might not necessarily increase the cost of household services. Recent works have documented that immigration nudges US workers toward occupations that require higher communication skill to reduce downward pressure on their wages (Peri and Sparber, 2009; Peri and Sparber, 2011). A more recent study in Europe found that native European workers are more likely to experience upward mobility in their occupation status in response to immigrants' influx (Cattaneo et al., 2015). It follows that immigration restriction policy such as LAWA might lead to downward mobility in US-born workers' occupation status, resulting in more US-born workers filling low-status occupations such as household services. If this is the case, the cost of purchasing household services would be relatively unaffected by the passage of LAWA. I examined if this is indeed the case in the following analyses.

Empirical methodology and data
To examine the impact of LAWA, I used the SCM pioneered by Abadie and Gardeazabal (2003) and further developed by Abadie et al. states indexed by j = 0, 1,..., J. Let the value j=0 correspond to Arizona, while the rest of the states (j = 1,..., J) are candidate contributors to the control group (i.e., the donor pool). Let G 0 be a (k × 1) vector whose elements are equal to the values of the pretreatment characteristics of Arizona that we want to match as closely as possible. Similarly, let G 1 be a (k × J) matrix collecting the values of the same variable in the donor pool.
The SCM identifies the vector of weights W * = (w 1 , …, w j ), that minimizes the difference between G 0 and G 1 W:   Table 1). In terms of its share of the labor force, the share of low-skilled immigrants in Arizona's workforce declines by 1.1% points relative to its synthetic control after the passage of LAWA (Figure 2a and Panel A of Table 2).
To find out if this decline occurred simply because of chance instead of LAWA implemen-   Indeed, the implied P-values (i.e., the probability that we observe a difference-in-differences estimate that is as large -negatively, in this case -as Arizona) of the impact of LAWA on both the size of the low-skilled immigrant workforce and its share in Arizona's workforce are 0.023 and 0.045, respectively (Panel A of Table 2).
A concern is that the adoption of LAWA closely coincided with the Great Recession, and therefore, these findings may simply be driven by the economic downturn at the time. However, the SCM approach used already accounts for any changes that affect the country as a whole, and unless the Great Recession affects the Arizona labor market differently than the rest of the country, it will not threaten the validity of my findings. A concern is that one of the industries that were hit hardest by the Great Recession, construction, is a leading employer of low-skilled workers in Arizona. As noted above, however, I created synthetic Arizona that minimizes the difference in the employment share of the construction industry with actual Arizona, which should mitigate the possible bias arising from this concern. Furthermore, recent studies (Bohn There is also a concern that another controversial Arizona state bill, SB 1070, which gave local law enforcement agencies more power in enforcing immigration laws and passed in 2010, may bias the impact of LAWA. However, before the law was supposed to take effect, a federal judge issued a preliminary injunction that blocked its most controversial provisions, and by 2012, the Supreme Court had struck down many of these provisions. It is unlikely, therefore,   DD, difference in differences; ACS, American Community Survey. that Arizona state bill SB 1070 had much impact on reducing the low-skilled immigrant workforce in the state. Indeed, a recent study by Amuedo-Dorantes and Lozano (2015) found that the SB 1070 had a "minimal to null" impact on the share of noncitizen Hispanics in Arizona. 8 Since LAWA was adopted in Arizona, six other states have implemented similar universal E-Verify programs. However, none of them have led to a decline in the low-skilled immigrant workforce as observed in Arizona (Panels B-G of Table 2 and Figures A1-A12). The answer to why we only observed a significant impact of universal E-Verify program in Arizona is outside the scope of this study. Nevertheless, there are a few possible reasons. For example, the scope of LAWA is broader because it requires all employers in Arizona to run new hires through the E-Verify system, while in some states such as Georgia, only businesses that use more than   Table 3 show the impact of LAWA on the participation rate of highskilled native females with at least some college education. Because high-skilled women have higher opportunity costs for the time spent on household work compared to low-skilled women, the response to LAWA should be stronger among this group. Contrary to the expectation,      shows that there is no evidence that LAWA led to a statistically significant reduction in women's weekly work hours across the four quartiles (Panel B of Table 3) Table 3 and Figures A17 and A18).
Although the analyses so far have mainly focused on women, conceptually LAWA might also affect the labor supply of US-born men because they also consumed household services in practice. Despite the negative estimates, suggesting that LAWA reduced the labor supply of US-born men, this decline is not statistically significant at the conventional levels (Panels C and D of Table 3 and Figures A19-A22). 9 To summarize, there is no evidence that native women's participation rate and average  (Table 4). For low-skill US-born women, the result instead suggests a negative relationship between LAWA and the daily time spent on household works among this group. Qualitatively similar findings were also observed among US-born men ( Figures A23 and A24 and Table 4).
The  in 2007 using the average wage in household service occupations as a proxy. Figure 9 shows a SCM analysis on the average hourly wage of workers employed in household service occupations in Arizona before and after the adoption of LAWA. 11 There is no evidence that LAWA statistically significantly increased the average hourly wage of workers employed in these occupations. The difference-in-differences estimate shows that the average hourly wage in household service occupations increases by approximately 1.2% (Panel A of Table 5), but this increase is not statistically significant with a P-value of 0.400.
The result that the average hourly wage in household service occupations was not increased by LAWA is surprising, especially because the theory of equilibrium wages based on a standard labor demand and supply framework implies that the reduction in workers in household service occupations should increase the wages of workers in these occupations. The next step of answering why the native female labor supply is relatively unaffected by LAWA would then be to examine whether LAWA led to a significant reduction in the aggregate supply of workers in household service occupations.

LAWA and household service occupations' workforce
The analysis in the previous section shows that the average hourly wage in household services occupation was not statistically significantly increased by the passage of LAWA. One explanation is that a labor market adjustment in Arizona caused the aggregate supply of workers in household service occupations to remain at a similar level after the passage of LAWA in 2007. To examine this, I first analyzed if LAWA has indeed led to a decline in the size of the immigrant workforce in household service occupations. 12 Figure 10 shows that this is the case.
In the absence of LAWA, the number of immigrants is projected to keep increasing to a level above 60,000 workers, while this number declined to approximately 50,000 in actual Arizona.
Indeed, the permutation test shows that it is very unlikely that this decline happens simply by chance because there are no other states in which such a deviation between a state and its synthetic control was observed. The difference-in-differences estimate shows that there would be approximately 9,724 additional immigrant workers in household service occupations in the absence of LAWA (Panel B of Table 5).
If the passage of LAWA led to a significant decline in the number of immigrant workers in household service occupations, the theory predicts that LAWA must have increased the number of native workers in these occupations as to leave the average wage in these occupations unaffected by LAWA. Figure 11 shows that this is indeed the case. After LAWA was adopted in 2007, the increase in native workers in household service occupations is significantly larger relative to what was projected in the absence of it. The difference-in-differences estimate shows that the number of natives in these occupations would be lower by approximately 7,636 workers in the absence of LAWA (Panel B of Table 5). Comparing this estimate with that of immigrant

workers, a large share of the impact of LAWA (~75%) is compensated for by the increase in native workers. A rather interesting finding is that this increase is driven by US-born men.
This result mainly reflects the finding that LAWA induced more male immigrants in household service occupations to leave (and not coming to) Arizona after its passage in 2007 (Panel B of Table 5 and Figures A25-A28).
To see if this increase in native workers in household service occupations is indeed large enough to leave the size of the workforce in these occupations unaffected by the passage of LAWA, I repeated the analysis for the overall number of workers (foreign and US born) in these occupations. Figure 12 shows that the overall number of workers in household service occupations is relatively unaffected by the passage of LAWA. The difference-in-differences estimate supports the evidence from the graphical observation that the size of household service occupations workforce is not statistically significantly affected by LAWA (Panel B of Table 5).
A question remains to be answered: why LAWA increased the number of native workers in household service occupations? One possible answer to this may lie from the fact that native   Table 6 and Figures A29 and A30). 13 The analysis indeed shows evidence that occupational income score of low-skilled US-born males decreases due to LAWA, suggesting that these workers would be in a higher paid occupation in the absence of the policy. The difference-in-differences estimate shows that LAWA reduces the occupational income score of low-skilled US-born males by 0.37 (1.34% relative to pre-LAWA score). This finding is similar to that of Lee et al. (2019) who also found evidence of occupational downgrading by US natives following Mexican repatriations in the 1930s. 14

Sensitivity checks
The key finding in this paper is that LAWA reduced the number of foreign-born workers in    Another concern to the validity of the results presented in this study is the potential exis-  Table 7, and the main findings hold under this alternative specification.

Conclusion
The influx of low-skilled immigrants has been argued to reduce the price of household services and alter the optimal time allocation between household production and market work for women (e.g., Cortes, 2008; Cortes and Tessada, 2011; Barone and Mocetti, 2011). As such, policies that lead to a decline in low-skilled immigrants may have an unintended consequence of reducing women's participation in the labor market.
In this study, I examined the impact of the LAWA of 2007, which requires all employers in Arizona to verify if a worker is authorized to work in the United States through the federal

Availability of data and material
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The author declares that he has no competing interests.

Funding
Not applicable.
Author's contributions Not applicable.      Table 3 Table A6 Weights used in construction of synthetic Arizona for household service sector analysis in Table 5 Donor states Hourly wage Native supply Native male Native female