ELIZABETH ZAHRT GEIB (*)
ABSTRACT. Native Americans suffer some of the highest rates of poverty and unemployment and the lowest rates of human capital attainment among racial minority groups in the United States, but economists understand very little about the impact these conditions have on the migration patterns of Native Americans. In 1994, a seminal article on this topic appeared in this journal (Cebula and Belton 1994). In their article, the authors suggest that the low levels of human capital and poor conditions in Native American reservations should make Native American migration sensitive to interstate differences in AFDC spending levels. Their hypothesis is confirmed for Native American migration over the 1985-1990 period. This paper refines their analysis by using micro-level rather than aggregate data, and by controlling for reservation residence and the impact of informal social safety nets in the source region. It is found that human capital factors and expected wage differences overwhelm interstate differences in public welfare spending and that informal safety nets in the source region dampen off-reservation migration. These findings suggest that state AFDC spending levels cannot explain contemporary Native American migration patterns.
I
Introduction
NATIVE AMERICANS, or American Indians, suffer some of the highest rates of poverty and unemployment among racial minority groups in the United States, and conditions are even worse on Native American reservations. In 1989, 27.2 percent of Native American families lived below the poverty level while 10 percent of all American families fell into this category (U.S Bureau of the Census 1990a, Table 112). The 1989 Native American family median income was $21,619, only 67 percent of the average family median income for the total U.S. population (ibid). Census Bureau estimates of Native American unemployment rates across selected reservations in 1990 vary from 14 percent to 44 percent (U.S. Bureau of the Census 1990b, Summary Tape File 3c). The Bureau of Indian Affairs reports even higher unemployment rates for these areas, estimating rates as high as 70 percent for some reservations (Stuart 1987). Both series place reservation unemployment rates far above average rates for other races or regions.
Despite our long-standing knowledge of Native American poverty, few economists have investigated the impact of poor economic conditions on Native American migration. In a seminal article published in this journal in 1994, Cebula and Belton were the first to do so (Cebula and Belton 1994). In their article, Cebula and Belton examine the role of state differences in welfare payments on interstate Native American migration patterns over the 1985-1990 period. They hypothesize that the low human capital attainment of Native Americans, and the resulting reliance on public assistance, will draw Native Americans to areas with higher than average AFDC payments. Using 1990 census data, they find that Native American migration is most strongly influenced by, among other factors, state differences in AFDC cash benefit levels.
The current analysis continues the inquiry into the impact of economic conditions in the source region on Native American migration. While more favorable economic conditions and/or higher public welfare spending in other areas may provide incentives to migration, it is hypothesized that the presence of informal social safety nets in, and cultural ties to, a reservation will dampen these incentives. Using the 1990 Integrated Public Use Microdata Series (Ruggles and Sobek 1995) this analysis finds that the effects of human capital and expected wage differences by state overwhelm state differences in public welfare spending as factors inducing Native American migration over the 1985-1990 period. In addition, the data reveal that cultural institutions in reservation areas do influence the migration decision. So, unlike the earlier study, this analysis finds interstate differences in public welfare spending to be a weak indicator of the migration decision for reservation Native Americans.
II
Differences in the Models
TO TEST whether AFDC payments affect Native American interstate migration, Cebula and Belton (1994:275) estimate two equations:
Mj = a + b [AFDC.sub.j] + c [PCI.sub.j] + d [COL.sub.j] + e [Un.sub.j] + f [DDAY.sub.j] + u [1]
NMj = g + h [AFDC.sub.j] + i [PCI.sub.j] + k [COL.sub.j] + l [Un.sub.j] + m [DDAY.sub.j] + n [POP.sub.j] + u* [2]
where [M.sub.j] is the net in-migration rate of the Native American population to state j over the 1985-1990 period, expressed as a percent of state j's 1985 population; [AFDC.sub.j] is the average per recipient family monthly Aid to Families with Dependent Children in state j in 1985; [PCI.sub.j] measures per capita income in state j in 1985; [COL.sub.j] is an index of the cost of living in state j in 1985; [UN.sub.j] is the unemployment rate in state j, averaged over 1985; and [DDAY.sub.j] measures the annual degree days in state j. Equation 2 represents the level of net in-migration ([NM.sub.j]) rather than representing migration as a ratio to the total population in the state in 1985 (1994:277).
Controlling for differences in average real incomes in a state (PCI and COL), for employment opportunities (UN), and for the effect of amenities (DDAY), Cebula and Belton test whether net in-migration by state is associated with AFDC payments. Heteroskedastic corrected estimations of both forms of the model indicate that the average per recipient AFDC aid is positively correlated with in-migration (1994:277). Additionally, they find that Native American migrants prefer areas with lower living costs and warmer climates, and are only slightly affected by per capita incomes and unemployment rates (1994:278).
The current analysis differs from Cebula and Belton's in several ways. First, and most importantly, unlike their analysis, which measures aggregate net migration flows, this paper considers the migration decisions of individual household heads. The use of micro-level data allows for control of human capital and demographic characteristics, which have been theorized or shown in previous migration studies to affect the migration decision (Sjaastad 1962; Mincer 1978; Borjas, Bronars, and Trejo 1992; and Borjas 1994). Control for these personal characteristics generates a more stringent test of the role of public spending on migration.
Second, the analysis considers the migration decisions of individuals living in reservation areas. (1) The analysis is restricted to this geographic region for two reasons. First, economic conditions in reservations are poor and individuals there rely to a greater extent on public assistance than do Native Americans living off reservations. By restricting the analysis to this set of individuals, I intend to capture the behavior most likely to be influenced by higher levels of aid in other states. Second, restriction to reservation areas allows for an analysis of the role of informal safety nets in these regions and for cultural ties to a reservation.
Third, the measurement of public welfare spending used in this analysis is broader than that used by Cebula and Belton. As noted above, Cebula and Belton include in their analysis AFDC cash payments as the measure of public welfare spending. However, individuals deciding to move consider not only AFDC cash benefits but also the level of food stamps, general assistance payments (GA), and other benefits available to them at various locations. (2) In addition to these social expenditures, this analysis accounts for public health programs available to Native Americans who live reasonably close to Indian Health Service facilities located in reservation areas.
III
The Empirical Model
THE HUMAN CAPITAL migration model offers the basis for the empirical model employed for this analysis (Sjaastad 1962). (3) Equation 3 expresses the expected net benefit of migration, which is computed as the discounted expected wage difference in the two regions, net of moving costs. Following Nakosteen and Zimmer (1980), costs are represented as a linear combination of a subset of personal (P) and locational (L) characteristics. Unlike previous studies, however, a third component to costs is added. This component represents traditional/cultural variables (7) and acts as a proxy for informal institutions in the host region.
E[[PV.sub.ij]] = [b.sub.0] + [b.sub.1](log [wage.sub.j]-log [wage.sub.i]) + [b.sub.2]P + [b.sub.3]L + [b.sub.4]T + [micro] [3]
Equation 4 translates the empirical model into a logistic econometric model, where I = 1 indicates that a household head moved from a reservation area to a non-reservation area in another state during the 1985 to 1990 period. The migration decision is determined by personal characteristics, which include demographic (age, age2, sex, marist, nchild) and human capital (english, hishc, assoc, bach) variables, the difference in the expected wage in the two regions (expwaged), (4) and regional characteristics, which include the general economic conditions in the area (empgrow5, averent), and other benefits available in both regions (pubwelf5, ihsper5). In addition to these conventional variables, two cultural institutional variables (numres5, perlang5) are constructed and included.
Prob(Y = 1\ X) = [e.sup.z]/(1 + [e.sup.z]) [4]
where z = [B.sub.0] + [B.sub.1]*age + [B.sub.2]*[age.sup.2] + [B.sub.3]*sex + [B.sub.4]*marist + [B.sub.5]*nchild + [B.sub.6]*english + [B.sub.7]*hisch + [B.sub.8]*assoc + [B.sub.9]*bach + [B.sub.10]*expwaged + [B.sub.11]*empgrow5 + [B.sub.12]*averent + [B.sub.13]*pubwelf5 + [B.sub.14*]ihsper5 + [B.sub.15]*numres5 + [B.sub.16]*perlang5 + v
and: age = age of household head in 1985
age2 = age squared of household head in 1985
sex = 0 if male, 1 if female
marist = 0 if single, separated, or divorced, 1 if married
nchild = number of own children in household
english = 0 if doesn't speak English, 1 if speaks English
hisch = 0 if did not graduate from high school, 1 if obtained a high school diploma or equivalent
assoc = 0 if does not have an associates degree, 1 if obtained an associates degree
bach = 0 if did not graduate from college, 1 if obtained a BS or BA degree
expwaged = selection corrected difference in the log expected wages in the two regions
empgrow5 = employment growth rate by state of residence in 1985
averent = average rent by state of residence in 1990
pubwelf5 = per capita log monthly public welfare payments by state of residence in 1985
ihsper5 = per capita log monthly expenditures on Indian Health Service programs by IHS service area, 1985-1990 average
numres5 = number of reservations in PUMA of residence in 1985
perlang5 = percentage of Native Americans in PUMA of residence in 1985 that speak a native language
A few of the variables merit examination. First are the regional variables that measure the impact of formal social safety nets in the sending region. These include public welfare programs at the state level and government-sponsored health services in reservation areas. A measure of log monthly per capita public welfare spending by state for 1986-1987 (pubwelf5) was collected from the U.S. Department of Commerce (1990). It is expected that as the per capita amount of public assistance in a state increases, a person will be less likely to move from that state.
Another component to the social wage measurement for Native Americans located in reservation areas is the assistance they receive through federally funded health programs in reservation areas. The Indian Health Service (IHS) operates hospitals, clinics, and health centers in reservation areas. Health services are provided to individuals who are enrolled members of a tribe and who can access the services in the areas where they are located.(5) It is expected that as the per capita IHS spending in an area increases, individuals will be less likely to move from that area.
Because much of the work that can be found on and near reservations is seasonal and because income derived from lease contracts on trust resources is erratic, Native Americans develop informal ways to share resources and smooth consumption (Lamphere 1979; Ashworth 1986; Moore 1993). The movement of resources through these traditional safety nets, which include the extended family network, powwows, and giveaways, is more difficult to measure directly.(6) To the extent that these institutions exist and are active they may decrease the migration sensitivity of Native Americans to more standard economic measurements. Two variables that act as proxies for the extent of traditional institutions are constructed and included.
The first is the number of reservations contained in a reservation area (numres5). The larger the number of reservations in a reservation area the more likely it is that there is a general awareness of Native American culture and active cultural institutions. When this is the case, individuals living in these areas should be less likely to adopt a market philosophy, be less sensitive to economic variables, and be less likely to move out of the area, all else equal. (7)
The second is a variable that measures the percent of the Native American population in a reservation area that speak a native language (perlang5). This variable is also intended to represent cultural integrity, or the extent to which informal safety nets have survived. Controlling for English ability of the individual, it is expected that the probability of migrating to a non-reservation area in another state will decline as the number of individuals in the reservation area who speak a native language increases.
The data used for this analysis are the Integrated Public Use Microdata Series or IPUMS (Ruggles and Sobek 1995). (8) Logistic migration equations are estimated over Native American household heads aged 16 to 65 in 1985. This age group is chosen to match those most likely to be in the labor force and therefore those most likely to be affected by migration variables. Restriction to household heads eliminates members of the household who move strictly because the household head is moving and maintains those most truly sensitive to migration variables. Additionally, as discussed above, restriction is made to those Native American individuals who resided in a reservation area in 1985 but were living in a non-reservation area in a different state by 1990. This restriction on interstate migration makes this analysis comparable to Cebula and Belton's.
IV
Regression Results
LOGISTIC REGRESSION results are presented in Table A. Column 2 of Table A shows which of the estimated coefficients is statistically significant. The last column of this table indicates the magnitude of each variable's influence on migration.
The data reveal that an individual's age, whether or not she or he has a high school diploma, the expected wage difference between regions, and the percent of Native Americans in a reservation area who speak a native language all contribute to the migration decision. Older household heads are less likely to move from a reservation area than are younger household heads. Individuals with a high school diploma are more likely to move from a reservation than those without. Individuals who face a positive expected wage difference are more likely to move to areas where their wage will be greater. And, as expected, individuals who currently live in a reservation where the traditional language is still spoken are less likely to move from that reservation, all else equal.
These results show also that differences in the levels of social spending by state have no real impact on off-reservation interstate migration. While social welfare spending is statistically significant, higher levels of spending in certain states do not significantly affect the probability that a reservation area Native American household head will migrate from the reservation to that state. It appears that human capital and expected wage differences overwhelm the impact of all other variables in the model.
V
Conclusion
IN AN EFFORT to learn more about Native American migration, this article extends the analysis previously undertaken by Cebula and Belton (1994). While Cebula and Belton find that AFDC cash payments are a significant factor in the aggregate migration of Native Americans over the 1985-1990 period, the current research indicates that public welfare spending as a whole does not substantially affect the reservation area migration decision. Instead, what seems to matter most is the human capital an individual possesses as well as expected wage differences between regions. This outcome is consistent with other micro-level empirical studies of migration.
What are we to make of the different conclusions? While both studies ask the same question, the different conclusions may arise from the different methodologies utilized. Cebula and Belton's analysis of aggregate migration flows cannot control for differences across individuals. Omission of variables that capture individual characteristics may influence the significance of their AFDC variable. And, while we expect that individuals who rely on public assistance will be influenced by state differences in public welfare spending, it is important to measure regional factors that could influence the migration decision. For Native Americans living in reservation areas, one such factor is the support they may receive from informal social safety nets that act to redistribute income to those with need. Because Cebula and Belton do not limit their analysis to off-reservation migration, they are not able to control for these forms of social spending that take place in reservation areas. This could bias their results in a direction toward states with a higher measured level of public welfare spending.
It is clear that Native American migration, and in particular, off-reservation migration, is complex. Indeed, Native Americans are likely to be more sensitive to both formal and informal social welfare spending than other individuals. The two studies discussed here support this notion, and move us a small step closer to understanding the weight these factors receive in the migration decisions of Native Americans.
(*.) Elizabeth Zahrt Geib is Assistant Professor of Economics, Lewis and Clark College, 0615 SW Palatine Hill Road, Portland, OR, 97219, zahrt@lclark.edu. Professor Zahrt Geib's interests include the economics of Native American reservations, tribal TANF participation, and the economic role of social institutions. Other published papers include "Expansionary Policy for Full Employment in the United States: Retrospective on the 1960s and Current Period Prospects," with Robert Pollin (1997), and "A New Look at U.S. Agricultural Productivity Growth, 1800-1910," with Lisa Geib-Gundersen (1996).
Notes
(1.) A reservation area is defined as a public use microdata area (PUMA) that contains at least one federally recognized reservation. The Census Bureau identifies public use microdata areas (PUMAs) as areas that contain between 100,000 and 200,000 residents. In urban areas a PUMA may be part of a, but not an entire, city. In less populated areas a PUMA may contain several contiguous counties.
(2.) The public welfare-spending variable includes categorical assistance programs (Aid to Families with Dependent Children and Supplemental Security Income), other cash assistance payments (General Assistance), vendor payments (Medicaid), and other public welfare (U.S. Department of Commerce 1990, Table 36).
(3.) The behavioral model is expressed as follows. Suppose an individual's expected earnings in region i are [E.sub.i] while expected earnings in region j are [E.sub.j]. Then the expected present value of the investment in migrating from region i to region j, E[[PV.sub.ij]], is represented below, where the term [C.sub.ijt] represents the costs associated with moving. Migration to region j occurs when the expected present value is positive.
E[PV.sub.ij] = [[[summation over].sup.n].sub.t=1] [(([E.sub.jt] - [E.sub.it]) - [C.sub.ijt])/[(1 + r).sup.t]]
(4.) For each individual, an actual wage is observed only in his or her place of residence in 1990. It is expected that the choice of residence is not strictly random. The entire migration analysis is a three-step procedure that takes into account possible selectivity bias in wages. In the first two steps, unbiased log wage equations are estimated and predictions for reservation and non-reservation area wages are made. The difference between an individual's actual wage in 1990 and his or her expected wage in the other region is computed. This difference (expwaged) is included in the logistic migration equation, the estimation of which constitutes the third step of the procedure.
(5.) A monthly per capita measurement of the Indian Health Service expenditures by IHS area (ihsper5) is computed by dividing a measurement of average IHS fiscal year 1985-1990 allocations by IHS region, provided by the Indian Health Service (Long 1997), by the number of IHS users by IHS region in 1985 as reported by the U.S. Department of Health and Human Resources (1994).
(6.) Moore defines giveaways as get-togethers for an occasion--such as a death, a graduation, or the return of an individual from military service--at which gifts are given to attendees other than the honored guest. These giveaways are marked by free food and extensive rules involving gift giving. When the giveaway includes a "drum," it is characterized as a pow-wow. There is no admission charge at rural functions and food is free. Especially needy families can bring bags and pots to be filled with food, and can survive for weeks by moving from one function to another. In addition to food, there may be requests for gas money, telephone use, and airfare. Both Ashworth (1986) and Moore (1993) find a positive relationship between the frequency of pow-wows and poverty in the local area and suggest that these institutions serve as safety nets in those communities.
(7.) This variable may also proxy economic conditions in an area. If a reservation area contains many reservations and these reservations are not economically self-sufficient, then this area is less able to support employment and individuals will be more likely to move from this PUMA, all else equal.
(8.) Currently, there is no micro-level data set that identifies Native American individuals by reservation status. The 1990 IPUMS was chosen for this analysis because it yields a large number of observations on Native Americans as well as migration information and information useful for approximating reservation status. Other data sets either yield too few observations on Native Americans and/or do not allow for a determination of reservation area status. The IPUMS data record whether an individual lived at the same residence five years earlier as well as his or her state of residence and PUMA five years ago. The 1990 IPUMS does not record, however, whether a person's residence was on an identified Native American reservation five years ago and whether he or she currently lives on a reservation because this information might identify some individuals in the sample. Since the primary interest of this research is to analyze off-reservation migration and the influence of certain reservation institutions on the decision to migrate, reservation area status is first estimated before proceeding with the analysis. Note that it is reservation area status and not reservation status that is estimated. For a description of the estimation of reservation area status, please contact the author.
References
Ashworth, K. (1986). The Contemporary Oklahoma Pow-wow. Ph.D. diss., University of Oklahoma.
Borjas, G. (1994). "The Economics of Immigration, "Journal of Economic Literature 32.
Borjas, G. J., S. G. Bronars, and S. J. Trejo. (1992). "Self-selection and Internal Migration in the United States," Journal of Urban Economics 32.
Cebula, R. J. and W. J. Belton. (1994). "Voting with One's Feet: An Empirical Analysis of Public Welfare and Migration of the American Indian, 1985-1990," American Journal of Economics and Sociology 54:3.
Cornell S. and J. P. Knlt, (1990). "Pathways from Poverty: Economic Development and Institution-Building on American Indian Reservations," American Indian Culture and Research Journal 14:1.
Cushing, B. J. (1993). "The Effect of the Social Welfare System on Metropolitan Migration in the U.S., by Income Group, Gender and Family Structure," Urban Studies 30:2.
Graves, P. (1979). "A Life-Cycle Empirical Analysis of Migration and Climate, by Race," Journal of Urban Economics 6:2.
Lamphere, L. (1979). "Traditional Pastoral Economy," Economic Development in American Indian Reservations, edited by Roxanne Dunbar Ortiz. University of New Mexico: Native American Studies.
Long, K. (1997). Personal communication to the author from Karen Long, Budget Analyst, Indian Health Service Division of Resources Management. Jan. 27.
Mincer, J. (1978). "Family Migration Decisions," Journal of Political Economy 86:5.
Moore, J. H. (1993). "How Giveaways and Pow-wows Redistribute the Means of Subsistence," The Political Economy of North American Indians, edited by John Moore. Norman: University of Oklahoma Press.
Nakosteen, R. A. and M. Zimmer. (1980). "Migration and Income: The Question of Self-selection," Southern Economic Journal 46:3.
Ruggles, S. and M. Sobek. (1995). Integrated Public Use Microdata Series: Version 1.0. Minneapolis: University of Minnesota, Social History Research Laboratory.
Sjaastad, L. A. (1962). "The Costs and Returns of Human Migration," Journal of Political Economy, Supplement 70.
Stuart, P. (1987). Nations Within a Nation: Historical Statistics of American Indians. Connecticut: Greenwood Press.
Tiebout, C. M. (1956). "A Pure Theory of Local Expenditures," Journal of Political Economy 64:5.
U.S. Bureau of the Census. (1990a). 1990 Census of Population, Social and Economic Characteristics, CP-2.
-----. (1990b). 1990 Census of Population and Housing, Summary Tape File 3c.
U.S. Department of Commerce. (1990). 1987 Census of Governments, Government Finances, Compendium of Government Finances 4:5.
U.S. Department of Health and Human Resources. (1994). Regional Differences in Indian Health, Washington: Public Health Service, Indian Health Service.
Zahrt, E. C. (1997). There's No Place Like Home?: Reservation Institutions and American Indian Migration, 1985-1990. Ph.D. diss., University of California, Riverside.
Table A
Determinants of Interstate Out-of-Reservation Area Migration: Native
American Houseold Heads Aged 16 to 65 in 1985
Variable Coefficient Z value Mean Marginal Effect
Age -0.382 (***) -3.403 41.73 -0.50
age2 0.003 (**) 2.420 1864.09 0.00
Sex -0.506 -0.984 0.26 -0.70
Marist 0.102 0.255 0.40 0.10
Nchild -0.009 -0.071 1.54 0.00
english 0.525 0.997 0.63 0.70
Hisch 1.464 (**) 2.097 0.74 2.10
Assoc -0.745 -0.850 0.17 -1.10
Bach 1.196 1.181 0.09 1.70
expwaged 1.314 (***) 2.663 -0.40 1.90
empgrow5 -0.006 -0.557 4.57 0.00
averent 0.007 (***) 4.728 222.08 0.00
pubwelf5 -0.009 (***) -3.194 310.03 0.00
Ihsper5 0.002 (**) 2.338 547.98 0.00
numres5 0.035 0.337 4.14 0.00
perlang5 -0.070 (***) -3.091 27.02 -0.10
Constant 7.281 (***) 2.934 NA 10.30
N=737 chi2(18)=91.03 Psedo [R.sup.2]
= 0.4494
(***)=Significant at the 1% level.
(**)=significant at the 5% level.
(*)=significant at the 10% level.