Return to the Homeland? The Impact of the Great Recession on Employment Outcomes and Labor Mobility for Native Americans

Journal of Economics, Race, and Policy (2018) 1:60 74 https://doi.org/10.1007/s41996-018-0008-6 ORIGINAL ARTICLE Return to the Homeland? The Impact of the Great Recession on Employment Outcomes and Labor Mobility for Native Americans Donna Feir 1 Rob Gillezeau 1 Received: 5 February 2018 / Accepted: 22 June 2018 / Published online: 6 August 2018 Springer International Publishing AG, part of Springer Nature 2018 Abstract We chart the impact of the Great Recession on the employment outcomes of Native Americans, in and outside of their traditional homelands. While increases in unemployment during the Great Recession for Native Americans are comparable to African Americans, employment changes are not. It appears that this difference is the result of Native Americans moving to traditional homelands and withdrawing from the labor force. Net of this move to traditional homelands, the employment response in reservations to the Great Recession was muted relative to outside of homelands. To further investigate this phenomenon, we develop actual versus predicted employment rates based on the occupational and industrial structure of employment, differentiating Native Americans by whether they live in a traditional homeland. Keywords Great Recession Native American Indigenous peoples Reservations Tribal homelands Employment Unemployment Introduction It is well established that the Great Recession generated substantial reductions in employment, earnings, and income over the short- and medium-run for workers in the USA (Elsby et al. 2010; Goodman and Mance 2011; Sierminska and Takhtamanova 2011). However, that picture largely focuses on white Americans and, to a lesser extent, larger minority populations, including African Americans and Hispanics (Hoynes et al. 2012). In this work, we seek to better understand the impacts of the crisis on Native Americans. Native Americans face many individual and institutional challenges that are distinct from other minority populations and therefore may have been differentially affected by the crisis. We consider the Great Recession s impacts on Native American employment and unemployment outcomes and put these results in the broader context by considering changes relative to other recessionary periods. Donna Feir dfeir@uvic.ca Rob Gillezeau gillezr@uvic.ca 1 Department of Economics, University of Victoria, Victoria, Canada We additionally investigate a unique cultural and institutional feature for Native Americans: the presence of traditional homelands and reservations. In particular, we consider the differential impact of the Great Recession on employment outcomes for those living inside and outside of traditional Native American homelands and whether the recession drove population movement between the two. It is unclear ex ante as to how economy-wide shocks may differentially impact Native Americans living outside of or on traditional homelands or the decision to move between them. To our knowledge, this has yet to be explored in the literature. The per capita income of Native Americans on reservations has been less than half of the US average, consistently falling below that of Hispanics, African Americans, and Native Americans living elsewhere (Akee and Taylor 2014). Reservations have been characterized as on average poorly integrated into the broader American economy with unique labor market conditions (Massey 2004), and as perpetuating poverty among Native Americans (Gitter and Reagan 2002). On the other hand, Native homelands and reservations have also seen substantial income growth for several decades (Mollica and Parker 2017) and been centers for the assertion of tribal self-government, and cultural preservation (Jorgensen 2007). Furthermore, returning to tribal lands may have provided some Native Americans with access to a broader social safety net, employment with non-traditional corporate

J Econ Race Policy (2018) 1:60 74 61 entities, or informal economic opportunities and supports. Both of these narratives about reservations have unclear implications as to how employment would be impacted by recessions. In addition, Native Americans living outside of homelands may be subject to greater degrees of discrimination with less access to certain social services than their counterparts who reside on homelands. Further, they may be exposed to broader market forces than those residing in homelands and have experiences that more closely mirror those of African Americans or Hispanics. Using data from the Current Population Survey (CPS) and the American Community Survey (ACS), we show that the Great Recession had substantial impacts on employment outcomes for Native Americans with substantial declines in prime-age male employment, potentially out-pacing those of African Americans. In contrast, Native Americans generally witnessed smaller increases in unemployment than other ethnicities, indicating that these workers were more likely to withdraw from the labor force than other groups considered. The negative employment shock to Native Americans appears to have occurred later and have had a more persistent effect than the employment shocks to other groups. There is also evidence of substantial migration among Native Americans, with populations migrating back to traditional territories. This movement back to traditional homelands may have helped individuals cope with the impacts of the recession. To our knowledge, while qualitative work has documented job loss as a motivation for moving back to reservations (Massey 2004), this is the first work to identify the quantitative significance of these flows in response to a large aggregate shock. 1 Together, the evidence points to the Great Recession having a disproportionate impact on Native Americans just as it did on African Americans and Hispanics, although with greater persistence and a uniquely large discouraged worker effect. Within the Native American population, there has been a persistent decline in the aggregate employment outcomes for Native Americans living on their traditional homelands, although the initial change in employment on homelands was smaller than for other populations. The lagged, and then persistent, decline in employment appears to be driven, at least in part, by individuals who return to their traditional homeland in response to the downturn and subsequently withdraw from the labor force. Next, we discuss the current literature on the impact of the Great Recession on Native Americans. This is followed by a discussion of the data that we put to use, the analysis of those data, and some concluding thoughts. Native Americans, the Great Recession, and Market Integration Native Americans face several distinct institutional challenges. In particular, Native Americans are perhaps the least well-integrated sub-population within the broader American economy, due in part to the legacy of colonialism (Feir et al. 2017; Dippel 2014), spatial isolation 2, and historic and contemporary institutional and individual racism (Trosper 1978; Dunbar-Ortiz 2014). The result is a population that has historically witnessed worse employment, income, and mortality outcomes than white, African American, or Hispanic populations in the USA (Akee and Taylor 2014). Given that part of the isolation of Native Americans from the American economy is rooted in the reservation system, Native Americans living off-reserve tend to show greater integration with the American economy. 3 This unique history and lack of integration could cause the Great Recession to have differential labor market impacts on the Native American population. There has been a limited amount of formal research on the impacts of the Great Recession 4 on Native Americans living either on or outside of their traditional homelands. More broadly, Native Americans have been largely overlooked as a distinct population for consideration with respect to economic cycles. Anecdotal evidence from the popular press (Peralta 2014) and from think tanks (Austin 2013) suggests Native Americans fared relatively poorly during the crisis. The most prominent academic work on the importance of the Great Recession for Native Americans comes from Burnette (2017) who considers changes in the unemployment 1 While the work of Snipp (1996, 1997) studies Native American mobility during the expansionary period between 1985 and 1990, he does not explicitly consider variation in macroeconomic trends. The work of Larriviere and Kroncke (2004) studies the return to human capital as a motivation for leaving traditional homelands using the 1990 Census, but again does not focus on aggregate fluctuations as compounding stressors on social services and available human capital on reservations. 2 It is important to acknowledge colonial policies that placed reservations on less valuable fringe and isolated lands (Dippel 2014). 3 That said, outcomes still remain lower for Native Americans living off-reserve. 4 According to the National Bureau of Economic Research (NBER), the Great Recession started in December 2007 and ended in June 2009 (National Bureau of Economic Research (NBER) 2011).

62 J Econ Race Policy (2018) 1:60 74 rate, employment-to-population ratio, and real hourly wage rates to better understand the recession s impacts on Native American women. Using data from the Consumer Population Survey, Merged Outgoing Rotation Group, Burnette finds that Native American men s unemployment rates increased more during the Great Recession than Native American women s unemployment rates, but Native women s unemployment rates increased much more than white women s. He also shows that real wages increased for the upper percentile of white women (75th percentile and above), but wages at nearly all percentiles have not increased for Native American women and have fallen at some percentiles for Native American men. Burnette (2017) is the most thorough existing analysis and we significantly expand on his work by considering an earlier recession, more detailed demographic comparisons and by considering the importance of mobility and living on or off of a traditional homeland. While the Great Recession s impacts looked very different in the USA than in many other countries, the Canadian literature provides useful insights, in particular for considering the impacts on Indigenous peoples. Lamb (2015) uses the Canadian Labour Force Survey (LFS) to show that individuals identifying as Aboriginal were substantially more likely to become unemployed and withdraw from the labor force during the Great Recession. Specifically, for those identifying as Aboriginal, there was a 15% increase in the portion of people not in the labor force between 2007 and 2010, while there was only a 3% increase for people not identifying as Aboriginal. Only about half of this difference is attributable to observable characteristics such as age, education, location, occupation, and industry. Lamb s work supports our findings, although the study only considers Aboriginal peoples living off-reserve and is unable to address some of the mobility dynamics that we are interested in considering. Finally, it is worthwhile to put the literature specific to Native Americans in the broader context of labor market outcomes during and after the Great Recession. Hoynes et al. (2012) examines how recessions differentially impact distinct demographic populations and compares the Great Recession to earlier recessions in terms of its severity, duration, and length of time before recovery using the Current Population Survey, Merged Outgoing Rotation Group (CPS-MORG) from Jan 1979 to December 2011. They focus on differences among age groups, educational attainment, race, and gender. However, they do not provide any results for Native Americans in their sample. Hoynes et al. find that the impacts of the Great Recession were felt most strongly by young workers, men, African American and Hispanic workers, and workers with low levels of educational attainment. Building off the work of Hoynes et al., we investigate the impacts on Native Americans, and explicitly compare them to other minority populations that were disproportionately impacted by the Great Recession. Data We draw on a number of data sources in order to build a broad picture of the impact of the Great Recession on Native Americans. Our primary data sources are the Current Population Survey, Merged Outgoing Rotation Group (CPS- MORG) and the American Community Survey 2005 2015 yearly 1% samples. 5 These datasets include information on employment, unemployment, labor force participation, and race among other variables. We identify Native Americans, African Americans, and non-hispanic whites using the race indicator in both the CPS and ACS and only label those with one racial identification as belonging to each group. We have made this choice as it is consistent with the work of Hoynes et al. (2012), which we refer to throughout the paper and because it makes the Native American population outside of tribal homelands more comparable to those outside tribal homelands. The majority of the literature on cyclical labor market fluctuations considering demographic differences uses the CPS-MORG. The main advantage of the CPS is that it is conducted monthly with a 6-month rotating panel. We use this data to consider employment and unemployment trends for Native Americans relative to other groups during recessions across time. We use it to mirror work by Hoynes et al. (2012) over a shorter time horizon, by including Native Americans as an individual group for consideration. The results from this analysis are more comparable to the existing literature than our second approach and allows us to put them into a broader context. There are two substantial disadvantages in using the CPS-MORG to study the impact of the Great Recession on Native Americans. The first disadvantage of the CPS is the sample size of the Native American population, which limits the degree of subgroup analysis possible. The sample of Native Americans is much larger in the American Community Survey. 6 The second disadvantage is that the CPS does not contain information on whether an individual lives in a Native homeland. Since 2005, the public use American Community Survey has included information on 5 The ACS samples were retrieved from the Integrated Public Use Microdata Series USA (IPUMS-USA) database (Sobek et al. 2010). 6 The ACS sample was on average 1.9 million households per year through 2007 to 2009 with a response rate of 97.9%, while the CPS has a monthly sample of about 60,000 individuals with an average response rate of 92.2%. In 2007, the sample of single ancestry Native Americans in the ACS was 16,365 people, while the total sample size of single ancestry Native Americans in the CPS (summing over all months) was 2,153.

J Econ Race Policy (2018) 1:60 74 63 whether an individual resided in a Public Use Microdata Area (PUMA) that includes any Census block that is designated as an American Indian, Alaskan Native, or Native Hawaiian homeland area. 7 This turns out to be an important component of the response of Native Americans to the Great Recession. Given the relative advantages of the CPS and ACS, we use both data sets to produce a more complete understanding of the experiences of Native Americans post-crisis. However, it is important to note that the data from the CPS and the ACS are not strictly comparable, as the sampling, design, and intent of the two surveys differ. First, while both surveys are collected monthly, the ACS data is reported annually while the CPS data is reported monthly. The ACS is primarily designed to produce nationally representative, single-year estimates of the socio-economic characteristics of both the institutionalized and noninstitutionalized population. The CPS is primarily designed to produce monthly estimates of the labor force characteristics of the civilian non-institutionalized population 16 years of age and older (Krorner and Howard 2011). The ACS is a general purpose demographic survey, while the CPS s primary purpose is to form the official estimates of unemployment and employment for the USA. As such, the CPS uses a more detailed battery of questions to determine employment status relative to the ACS. 8 The inclusion of institutionalized population in the ACS has an important effect on the estimates of the employment-population 7 The measure of homeland status noted above is created by IPUMS using the following procedure: For Census 2000, the Census Bureau indicated whether or not each census block was within a homeland area. To create the IPUMS homeland variable, we have applied the Census Bureau s block-level measure of homeland status to the PUMAs available in the public use data from Census 2000, Census 2010, and the ACS samples from 2005 onwards. These homeland areas can be either legal or statistical entities. Although they do not have the legal statuses of other areas, the statistical areas included in this measure are judged by local agencies to be substantially meaningful to the local native populations. Specifically, the legal entities consist of federally recognized American Indian reservations and off-reservation trust land areas, the tribal subdivisions that can divide these entities, state recognized American Indian reservations, Alaska Native Regional Corporations, and Hawaiian home lands. The statistical entities are Alaska Native village statistical areas, Oklahoma tribal statistical areas, tribal designated statistical areas, and state designated American Indian statistical areas. Tribal subdivisions can exist within the statistical Oklahoma tribal statistical areas (US Census Bureau 2000). For ease of interpretation of homeland area, we have dropped Alaska and Hawaii. For purposes of consistency in definitions in PUMAs with a homeland, we restrict our time frame to the 2006 to 2011 period. We exclude 2005 because of the inclusion of those in institutions and collective dwellings in 2006, which results in non-trivial changes in population sizes on homelands between these years. 8 For a complete comparison of the employment status questions in the CPS and ACS in 2007, 2008 and 2009, see Krorner and Howard (2011). ratios for certain subgroups (such as African Americans and Native Americans). Another important point regarding the ACS is that the employment status questions underwent a substantial change in 2008 in order to make its measurement of employment and unemployment more comparable to the CPS. Prior to 2008, differences in wording meant that the CPS was more likely to capture persons with irregular work arrangements as employed and was less likely to count those as passively looking for work as unemployed than the ACS (Krorner and Howard 2011). The question change in 2008 in the ACS implies that estimated employment rates would be lower in 2007 than in 2008, while unemployment rates would be higher, holding economic conditions constant. This implies that employment and unemployment responses to the Great Recession will be muted in the ACS relative to the CPS. This also implies that the extent to which irregular work arrangements or the propensity to passively look for work differ among population groups, differences in employment and unemployment rates between 2007 and 2008 will be difficult to interpret. However, differences in the magnitude of employment changes between 2007 2009 and 2007 2010 cannot be attributed to changes in the employment status question nor can differences in changes between population groups. Given the strengths and limitations of the CPS and ACS, we use them both when possible and compare the findings. We use the individual CPS and ACS datasets to engage in a series of empirical exercises described below to better understand the experiences of Native Americans during the Great Recession. In performing these exercises, the strengths and limitations of these datasets should be kept in mind. Putting Native American Labor Fluctuations into Context This analysis builds on the work of Hoynes et al. (2012) who systematically document which demographic groups suffer the largest changes in employment during recessions. Their work, and ours, contributes to the literature that documents the severity of the Great Recession (Elsby et al. 2010; Goodman and Mance 2011; Sierminska and Takhtamanova 2011). WhileHoynesetal. are ableto extend their analysis back to 1979, that is not possible given data constraints when explicitly considering outcomes for Native Americans, particularly if we are interested in differential outcomes for those living on versus off their traditional homelands. As such, our analysis begins in 1994 when Native Americans are distinctly identified in the CPS. 9 9 Since 1979, there have been five recessions in the USA, occurring in 1980, 1981 82, 1990 1991, 2001, and 2007 2009.

64 J Econ Race Policy (2018) 1:60 74 Percentage Change in Unemployment Rate 0 50 100 150 2007 2001 0 20 40 60 80 Month Since NBER Start of Recession Notes: This figure shows the seasonally adjusted unemployment rate. The data comes from the Current Population Survey Merged Out-Going Rotation Group. Labor market peaks come from NBER (2011). Fig. 1 US seasonally adjusted unemployment rate, months since peak While seasonally adjusted similarly to the work by Hoynes et al., the results vary slightly as a result for the different time periods. Our sample period includes two recessions, the 2001 recession and the 2007 2009 recession with the former following a more typical downturn and recovery while the latter, as has been well established, is deeper, and more persistent. Figures 1 and 4 in the Appendix show the percentage change in employment and unemployment respectively since the NBER identified beginning of each recession. As these figures show, the Great Recession started with a similar magnitude over the first several months relative to the 2001 recession, but the general labor market continued to worsen until nearly 40 months after the recession start date with a severity unmatched by other recessions in recent decades. The difference in the magnitude of these recessions is an important consideration in particular when understanding the impact on Native Americans. Below, we add to this story by considering how Native Americans fared in the Great Recession and the 2001 recession relative to whites and other visible minorities. Raw Employment Changes for Native Americans: CPS-MORG According to the CPS-MORG and as indicated in Table 1, at the outset of the recession, Native American men were in a unique labor market situation relative to other minority groups. Employment rates for Native American men were similar to that for African Americans 10 and well 10 Research by Sierminska and Takhtamanova (2011) and others have identified African Americans suffering some of the most pronounced effects of the Great Recession. below that of Hispanic Americans, but the unemployment rate was substantially lower, indicating lower labor force participation. For women, employment rates mirror those of Hispanic women with somewhat lower unemployment figures also pointing to lower labor force participation. Earnings for Native American women fall squarely between that of African American and Hispanic women, while earnings for Native American men were higher than those of African American and Hispanic men. As such, it is not appropriate to suggest that Native Americans were in a better or worse position than other minority groups at the outset of the recession, rather they were in a unique position with outcomes substantially below those of white Americans. From this starting point, we investigate the peak-totrough changes in these outcomes as derived from the CPS- MORG dataset in Table 2. One feature is immediately clear in the CPS-MORG sample: the 2001 recession hit Native American men much harder with respect to employment and unemployment than any other ethnic group. In fact, the effects are so large in the 2001 recession for Native American men that they resemble the size of the shock felt by most groups during the Great Recession. Native American women, however, follow a more typical path with much smaller impacts in 2001. If we instead focus on differences between group outcomes during the Great Recession, increases in the unemployment rate are smaller for Native American men than for African American and Hispanic men, but the decreases in the employment-population ratio drastically exceed all other ethnicity-gender groupings. Native women only see a modest drop in unemployment, and witness employment declines that mirror the declines of other minority women. As such, from the CPS-MORG, there is a clear picture in which recessions that exhibit more typical characteristics, like in 2001, hit Native American men more harshly than other groups while the Great Recession had a similar magnitude across groups. It did, however, drive much larger employment and labor force participation declines for Native Americans than other populations. While the CPS-MORG is a valuable data source because of its frequency and easy comparability to other studies regarding the Great Recession, it contains a relatively small sample of Native Americans and contains no geographic information regarding homelands. Next, we turn to the ACS to fill these holes in the analysis. Raw Employment Changes for Native Americans: ACS With the ACS, we begin by turning to the raw data regarding male labor market outcomes for Native Americans

J Econ Race Policy (2018) 1:60 74 65 Table 1 Labor market outcomes by race, gender, education, and age, May 2007 Employment Unemployment Usual weekly Hours rate (%) rate (%) earnings (2007$) last week White men 82 3.2 930 42 White women 69 2.7 648 36 Black men 66 7.1 660 40 Black women 64 5.1 584 38 Hispanic men 80 4.2 591 41 Hispanic women 56 3.2 487 37 Native men 65 4.3 789 41 Native women 57 2.8 548 37 Age 16 to 19 34 6.7 215 24 Age 20 to 24 69 5.6 419 36 Age 25 to 44 81 3.0 778 40 Age 45 to 60 76 2.4 879 41 Less than high school 50 5.3 399 35 High school graduate 74 4.0 600 40 Some college 74 3.6 646 38 College graduate 86 1.7 1033 41 Author s tabulations of Current Population Survey Merged Outgoing Rotation Group (CPS-MORG) data. Usual weekly earnings and hours last week are only calculated for those currently employed. Earnings weights from the CPS-MORG are used to form tabulations. May 2007 was the eve of the Great Recession. This table is similar in spirit to Hoynes et al. (2012) Table1 Italicized entries indicate the key contributions Table 2 Peak-to-trough percentage point changes in unemployment and employment rates by group, 2001 and 2007 recessions from the CPS Peak-to-trough changes Peak-to-trough changes in unemployment rate in employment rate Mar. 2001 to May 2007 to Mar. 2001 to May 2007 to Nov. 2001 Oct. 2009 Nov. 2001 Oct. 2009 White men 0.64 5.1 1.43 8.34 White women 1.19 2.97 2.46 3.49 Black men 0.38 6.29 1.30 9.42 Black women 1.83 4.43 3.18 7.38 Hispanic men 2.23 5.80 7.51 6.22 Hispanic women 2.10 5.65 3.59 7.83 Native American men 6.00 4.63 12.29 15.05 Native American women 5.95 0.50 4.96 8.14 Age 16 to 19 1.10 2.80 1.82 7.91 Age 20 to 24 0.66 5.73 1.85 7.80 Age 25 to 44 1.00 4.94 1.34 5.34 Age 45 to 60 0.82 3.25 2.18 2.83 Less than high school 0.62 3.91 0.09 8.30 High school graduate 1.10 6.01 2.06 7.63 Some college 1.23 3.93 2.33 4.91 College graduate 1.59 2.87 1.64 2.63 Author s tabulations of Current Population Survey Merged Outgoing Rotation Group (CPS-MORG) data. Peak-to-trough dates were taken from Hoynes et al. (2012) and this table is similar in spirit to Hoynes et al. (2012) and presented are differences in seasonally adjusted unemployment and employment rates. Earnings weights from the CPS-MORG are used to form tabulations Italicized entries indicate the key contributions

66 J Econ Race Policy (2018) 1:60 74 Table 3 Peak-to-trough percentage point changes in unemployment and employment rates by group, 2007 recessions from ACS ACS 2007, rate (%) Changes in unemployment rate Changes in employment rate Employment Unemployment 2001 2002 2007 2009 2007 2010 2001 2002 2007 2009 2007 2010 White men 78 4.6 1.02 3.46 3.86 1.30 3.54 4.97 White women 68 3.8 0.71 2.02 2.52 0.56 0.77 2.00 Black men 60 9.0 0.97 3.92 4.99 1.64 4.57 6.82 Black women 63 8.1 1.70 2.52 3.65 1.30 1.06 2.78 Native men 58 10.3 0.35 2.74 3.22 2.60 1.21 4.87 Native women 56 7.90 2.27 0.64 2.31 2.56 0.94 2.13 Age 16 to 19 35 9.3 0.97 1.55 1.77 1.89 5.32 8.29 Age 20 to 24 67 8.1 1.17 3.76 4.09 1.36 3.53 5.36 Age 25 to 44 77 4.5 1.04 2.98 3.62 1.12 1.65 3.01 Age 45 to 60 73 3.2 0.92 2.65 3.3 0.68 1.33 2.45 Less than high school 45 7.6 0.82 2.75 3.08 1.17 3.84 6.23 High school graduate 70 6.0 1.19 3.66 4.42 1.17 3.31 4.85 Some college 75 4.4 1.07 3.09 3.76 1.18 2.35 4.15 College graduate 84 2.4 0.91 1.72 2.08 0.82 0.39 0.98 Author s calculations from the American Community Survey (ACS) retrieved from the Integrated Public Use Microdata Series (Sobek et al. 2010) Italicized entries indicate the key contributions relative to other groups. Mirroring the results from the CPS data, at the immediate onset of the recession as displayed in Columns 1 and 2 of Table 3, Native American men and women saw employment and unemployment rates substantially worse than that of white Americans, but similar to labor market conditions for the African American population. While the ACS results do not paint as clear a picture as the information from the CPS, the 2001 recession similarly shows much larger declines for Native American men and women than was experienced by other groups. This suggests that more typical recessions do potentially hit Native Americans harder than other groups. The evidence around the severity of the Great Recession is somewhat less clear and is sensitive to whether we compare peakto-trough running from 2007 to 2009 or 2007 to 2010. In general, the ACS results point to the Great Recession having substantial impacts on Native Americans, 11 but somewhat smaller than those felt by the African American population. The effect appears to be delayed in that there are substantial declines in employment and increases in unemployment in 2010 versus 2009. It is important to note that while changes to the employment status question may differentially mute estimated responses between ethnic groups to the Great Recession, they will not affect the difference in the size of the effect between 2009 and 2010 apart from through actual changes in employment conditions. 11 Part of the relatively large employment impact may be driven by the relatively young age profile of the Native American population. We can draw further insights by looking at employment trends based on whether an individual lives in a traditional homeland. And, as indicated in Fig. 2, there is differentiation within the Native American population with those living on their homelands experiencing meaningfully lower employment rates than other Native Americans or African American populations. The employment and unemployment effects of the recession on those living on their traditional homelands appear to be smaller or comparable to their off-homeland counterparts initially between 2008 and 2009, but then amplified in 2010. This is particularly true for employment. Further, the employment effects are much more persistent showing little recovery in the years since the Great Recession. Given the unique labor force participation declines hitting Native Americans and this differential impact on traditional homelands, we further investigate whether there is evidence of individuals moving to homelands after experiencing a negative economic shock. In Fig. 3, we present the total population of men and women who live on a homeland by year. 12 The results in this figure support the narrative that migration to reservations 12 It should be noted that the proportion of Native Americans residing on homelands is estimated to be higher than would be expected for two reasons: first, we are focusing on single racial identity Native Americans, and second, the homeland indicator provided by the Integrated Public Use Microdata Series is at the Public Use Microarea level, which may include both homeland and non-homeland areas.

J Econ Race Policy (2018) 1:60 74 67 Fig. 2 Labor market indicators Men Employment/Population Unemployment Rate.5.6.7.8.04.06.08.1.12 2006 2007 2008 2009 2010 2011 2006 2007 2008 2009 2010 2011 Women.55.6.65.7 2006 2007 2008 2009 2010 2011.02.04.06.08.1 2006 2007 2008 2009 2010 2011 Homeland AIAN Black AIAN outside homeland White Notes: This figure shows the employment-per-population ratio and unemployment rate by race calculated from the 2006 to 2011 American Community Survey for men aged 25 to 64. The dashed line indicates after 2008. AIAN indicates American Indian or Alaska Native. All residents of Hawaii and Alaska are excluded from the sample. may be an important factor in explaining observed changes in employment rates, at least for men. The total population counts suggest that there is a notable increase in Native Americans on a homeland following the recession; this corresponds to a sharp decline in Native Americans living outside homelands. 13 To examine this possibility further, we construct weighted population counts of in- and out-migration from PUMA areas that are Native homelands by race and present 13 There is an increase in the homeland population in 2010 that does not correspond to a decrease in the outside of homeland population because of an overall population increase in 2010. these in Table 4. While out-migration from areas with a homeland are generally larger than those returning or moving to a homeland over each period, there is a substantial change in the number flowing into homelands and those flowing out that differs by race. For Native American men, there is an increase in those moving back to homelands and a slowdown of those moving away, which results in an increase in the otherwise projected population. In contrast, there is a decline in the population of white men over the same time period. A similar narrative holds for Native and white women, while the flows are smaller for Native women. We confirm that these results are statistically significant and not accounted for by differences in age, education, or distribution across states in Table 7. InTable8, weshow

68 J Econ Race Policy (2018) 1:60 74 Fig. 3 Total population residing on and outside of homelands Population 260000 280000 300000 320000 340000 Men 2006 2007 2008 2009 2010 2011 Population 260000 280000 300000 320000 340000 360000 Women 2006 2007 2008 2009 2010 2011 AIAN in homeland AIAN outside Homeland Notes: This figure shows population by race and whether individuals reside within a Public Use Microarea that includes a homeland area. The figure is calculated from the 2006 to 2011 American Community Survey for men aged aged 25 to 64, using the provided person weights. The dashed line indicates data points after the onset of the Great Recession. Note that this figure over-estimates the population living on reservation as anyone with a Public Use Microarea that includes a homeland is classified as living in a homeland area. For example, in 2006, approximately 18 percent of White Americans and 12 percent of African Americans lived in a homeland area by this measure. 55 percent of single race Native Americans are classified as living in a homeland area. AIAN indicates American Indian or Alaska Native. All residents of Hawaii and Alaska are excluded from the sample. Table 4 Migration flows to and from Public Use Microareas that include a Native homeland 2006 2008 2009 2011 Difference ( ) Native men Gross in-migration 6,715 7,730 1,015 Gross out-migration 15,425 14,560 869 in-migration out-migration 1,880 White men Gross in-migration 433,488 412,965 20,523 Gross out-migration 694,962 680,716 14,246 in-migration out-migration 6,280 Native women Gross in-migration 4,813 4,468 345 Gross out-migration 10,739 9,467 1,272 in-migration out-migration 930 White women Gross in-migration 377,415 348,665 28,751 Gross out-migration 586,470 562,165 24,307 in-migration out-migration 4,445 Author calculations from the 2006 to 2011 American Community Survey (ACS) retrieved from the Integrated Public Use Microdata Series (Sobek et al. 2010). These are the frequencies for either those who lived outside a homeland area in the prior year and in the current sample period now reside in one ( Gross in-migration ) or those who lived in a homeland area last year and now reside outside of one ( Gross outmigration ). These results are calculated from the 2006 to 2011 American Community Survey for men aged aged 25 to 64 using the provided person weights Most relevant entries were indicated in italics

J Econ Race Policy (2018) 1:60 74 69 that these findings are largely driven by those between the ages of 25 and 35. Despite substantial out-migration of Native homelands, there was still a net increase in single ethnicity Americans in homeland areas after the start of the Great Recession. This population increase is due too both differential increase in the likelihood of those under the age of 25 and over the age of 64 to remain in homelands, and possible greater increases in Native American racial identification in areas that include a homeland. To get a better sense as to the nature of the apparent movement back to homelands following the recession, we reproduce peak-to-trough changes within states that include Native American homelands and varying by homeland status in Table 5. The results provide confirmation of mobility to traditional homelands after the great recession. In particular, Native Americans living outside homelands appear to be hit by an immediate surge in unemployment in the early stages of the recession, while those on traditional homelands are relatively sheltered in this early period. However, by 2010, unemployment outside of homeland levels off while there is a substantial increase in unemployment and decrease in employment within homeland areas. While part of this is the local effect of the recession, the net result is influenced by asymmetries in mobility, and these differences are muted when movers are excluded. Figure 3 and Table 5 suggest a unique dynamic in which many Native Americans return to their traditional homelands during the recession and withdraw from the labor force. We discuss the implications of this finding in more detail in the conclusion. Overall, the data from the ACS confirms the story gleaned in the data from the CPS and allows us to consider an additional margin of adjustment. The Great Recession had a substantial impact on Native American employment outcomes. This is particularly evident with employment declining by much more than unemployment increases relative to other ethnic groups. The fall in labor force participation may be linked to the decision to return to traditional homelands. Table 5 Peak-to-trough percentage point changes in unemployment and employment rates by group, 2007 recessions from ACS, only for states with at least one reservation greater than 250 people in 2000 PUMA includes ACS 2007, rate (%) Changes in unemployment rate Changes in employment rate homeland? Employed Unemployed 2007 2009 2007 2010 2007 2009 2007 2010 White men No 79 4.52 3.86 4.63 3.89 5.83 Yes 78 4.87 3.52 3.74 4.44 5.72 Yes (no mover) 79 4.34 3.45 3.74 4.21 5.50 Yes (only movers) 72 7.13 3.88 3.84 5.60 6.98 White women No 68 3.64 2.55 3.19 1.17 2.48 Yes 68 3.84 1.95 2.54 1.08 2.54 Yes (no mover) 70 3.24 1.90 2.48 1.30 2.53 Yes (only movers) 63 6.62 2.16 2.83 0.03 2.67 Native men No 67 7.97 6.72 5.77 6.85 10.08 Yes 49 13.38 0.73 2.78 0.87 6.31 Yes (no mover) 48 13.44 0.10 1.93 3.52 5.02 Yes (only movers) 50 0.13 4.26 6.15 10.37 11.35 Native women No 58 7.35 2.66 3.58 1.77 2.87 Yes 53 9.78 0.25 3.22 2.65 4.48 Yes (no mover) 52 8.84 0.60 2.55 1.02 2.40 Yes (only movers) 56 13.63 0.59 6.65 10.33 13.72 These are author calculations from the 2007 to 2010 American Community Survey (ACS) retrieved from the Integrated Public Use Microdata Series (Sobek et al. 2010). Column 2 indicates whether the Public Use Microdata Area (PUMA) the individual resides in includes a Native American Homeland with the third row for each demographic group presenting statistics only for individuals who resided in the same PUMA the year before. The fourth row for each demographic group presenting statistics only for individuals who resided in a different PUMA the year before. This indicates the degree to which mobility factors into larger changes between 2009 and 2010

70 J Econ Race Policy (2018) 1:60 74 Table 6 Actual and predicted percentage point change in employment rate, by group (predictions based on industry-occupation mix) Data from the Consumer Survey Merged Outgoing Rotation Group Actual Predicted Actual Predicted Employment Change in employment Change in employment Q2 2007 Q2 2007 Q3 2008 Q2 2007 Q3 2009 White 75 1.08 1.34 4.24 5.09 Black 65 1.74 1.39 6.79 5.18 Native 64 3.48 1.29 14.9 5.40 Data from the American Community Survey 2007 2007 2009 2007 2010 White 73 2.16 2.02 3.51 2.98 Black 62 2.74 1.87 4.71 2.88 Native 57 1.09 2.71 3.50 4.03 Native in homeland 51 0.91 2.81 5.33 4.32 These are author calculations from the 2007 to 2010 American Community Survey (ACS) retrieved from the Integrated Public Use Microdata Series (Sobek et al. 2010) and the Consumer Population Survey Merged Outgoing Rotation group. Native in homeland indicates whether the Public Use Microdata Area (PUMA) the individual resides in includes a Native American Homeland for American Indians. We create predicted changes in employment level for each demographic group by multiplying the group s share of total employment in 34 industry-occupational cells in 2007 by US-wide changes in total industry-occupation employment and summing across industry-occupation. The difference between the actual and predicted changes can be interpreted as the group-specific component of employment that operates differently than at the national level. The 36 cells are defined by 12 industries and 3 occupations. This follows Hoynes et al. (2012) Predicted Employment Changes: CPS and ACS As a final exercise to better understand the dynamics facing Native Americans, we follow the example of Bartik (1991) and Hoynes et al. (2012) and create a predicted peak-to-through change in the employment rate for each ethnicity. This measure is based on the allocation of these ethnicities across industry-occupation 14 groupings at the pre-recession peak and then multiplying by the US peakto-trough change in each industry-occupation grouping and summing across these groups. The difference between real world changes and predicted changes is typically interpreted as a group-specific shock beyond any standard cyclical component. The occupational-industrial composition of Native American employment is substantially different than for white Americans (Wise et al. 2016) and this exercise allows us to assess the extent to which these differences are driving differences in the response to the recession. We perform this exercise for the Great Recession over several different time periods using the CPS-MORG and the ACS. These results are presented in Table 6. Using the CPS data, the actual decline in employment for whites is either equal to, or smaller than the decline in employment that would be predicted based on national industrialoccupational trends. Using the ACS, a similar pattern emerges for whites between 2007 and 2009, and in 2007 14 This includes 3 occupation levels of managerial, services, and blue collar and 12 industry groupings, creating a total of 36 cells. to 2010 the actual change is somewhat larger than the predicted change from national trends. The actual declines in employment for African Americans roughly equates to the predicted declines in the CPS-MORG data, but using the ACS data, the actual declines are notably larger than the predicted declines. This suggests that African American employment has a group component that is more responsive to economic shocks than the national average and the industrial-occupational composition would suggest. However, the differences between the CPS-MORG results and the ACS are most marked for the Native American sample. In all time periods examined using the CPS- MORG, the actual declines in employment are larger than the predicted declines in employment suggesting a strong group-dependent cyclical component that makes Native American employment more responsive to economic shocks than the national average. While this may be, in part, driven by the relative age and educational attainment of the Native American population, the estimates in Table 9 suggests this does not fully explain the group cyclical component. On the other hand, according to the ACS analysis, as of 2009, Native Americans actually outperform their predicted values between 2007 and 2009 in other words, the actual employment change is smaller than the predicted changes. The actual changes in employment are the smallest on Native homelands though the industrial and occupational nature of employment would have predicted larger declines in employment than for whites or African Americans. However, that picture shifts substantially if one considers the period between 2007 and 2010. During this period,

J Econ Race Policy (2018) 1:60 74 71 Native Americans living on their traditional homelands experienced employment rate declines greater than would be predicted given the industrial-occupational composition of employment in 2007, suggesting differential shocks leading to discouraged workers. However, for Native Americans living outside traditional homelands, actual declines in employment are less than would be predicted given the industrial-occupational composition of employment in 2007. The unexplained increase in the gap for Native Americans on their homelands between 2009 and 2010 is again consistent with Native Americans who were negatively impacted by the Great Recession returning to traditional homelands. The results from the ACS sample also suggest that the industrial-occupational structure of employment in reservations cannot explain the smaller impacts on employment observed in the short run, specifically in Native homelands. This may suggest that there is something unique about the way employers in traditional homeland areas adjust to economic shocks compared to employers outside these homelands. This could be consistent with tribal-owned corporate entities or member opportunities in benefit agreements. 15 Given the relatively small sample of Native Americans in the CPS-MORG, we favor the ACS estimates. 16 Conclusion The results in this work paint a picture of the commonalities of the Native American labor market experience with African Americans and Hispanics with respect to the severity of the shocks felt during the Great Recession. As suggested by the popular press, Native Americans have been disproportionately affected by the Great Recession with particularly enduring shocks relative to some other minority populations. Future work is necessary to discern what generates the relatively large responses of employment among Native Americans as it does not appear to be explained by observable characteristics. The extent to which structural and direct racism cause Native Americans to have more precarious employment is an important question to be addressed. Despite the similarities in the data to African American experiences, we demonstrate that there are parts of the labor market experience during the Great Recession that appear to be unique to Native Americans. The most important finding from our work is the unique mobility for Native Americans that appears to have occurred during the Great Recession. Native American workers who lost employment or were unable to find employment during the recession 15 This may also be consistent with more informal working arrangements within homeland areas for Native Americans given the changes in employment status question wording in 2008 in the ACS. 16 This is despite the shortcomings of the ACS discussed in Data. appear to have returned to their traditional homelands in substantial numbers or they did not leave when they otherwise may have done so. Given the government-togovernment relationship between the federal government and tribal nations, this mobility has notable implications for policy design in recessionary periods. This movement to return to or remain on traditional homelands suggests that homeland areas offer greater private or social security nets than non-homeland areas. Future research is necessary to understand whether access to this safety net is more or less likely to lead to long-term welfare improvements for Native Americans as individuals and for Native Nations more broadly. If the social safety net improves long-run welfare, a direct response would be for governments to ensure there are reservation-focused support systems in economic downturns, which may be stretched further than support systems in urban or non-reserve areas. If the return to homeland areas is a stop-gap that reduces individual s ability to resume whatever path they were formerly on and reduces their long-term welfare, then an alternative response would be to increase employment assistance to Native Americans residing outside of traditional homelands and lessen the pressures on tribal governments. In the absence of additional government supports, it is clear that the Great Recession will have placed substantial strains on existing homeland-based safety nets that could be persistent given the slow pace of recovery. It is also important from a policy perspective to understand whether or not the population increase in homelands should be seen as the same phenomenon as young adults either moving back home in response to economic shocks or not leaving the home because of lack of economic opportunity (Johnson 2011; Bitler and Hoynes 2015). While the findings in this paper have commonalities with this phenomenon in the general population, especially given the changes in in-migration are most significant among those between the ages of 25 and 35, it is not clear that the same phenomenon is at play given the fundamentally different institutional nature of the Native American experience. We encourage future work to continue to consider the Native American economic experience in order to support Native Americans in achieving their long-term goals. Acknowledgments We would like to thank Maggie Jones, Dan Giedeman, David Scoones, and Rebecca Wortzman for their insights, suggestions, and support. Funding No grants or other external funding was used for this project. Compliance with Ethical Standards Conflict of interest The authors declare that they have no conflict of interest.