Do reservation Native Americans vote with their feet? A Friendly Comment - article clarifies distinctions regarding 2001 research on the influence of welfare on Native Americans

American Journal of Economics and Sociology, The,  Oct, 2002  by Richard J. Cebula

• E-mail
• Print
• Link

I

IN A RECENT ISSUE of this journal, Geib (2001) provides an interesting and well-executed analysis of the potential impact of "welfare" on migration of Native Americans over the 1985-1990 period. The study by Geib (2001) can be interpreted as paralleling a co-authored paper by Cebula and Belton (1994). This brief, friendly Comment seeks to clarify certain distinctions between the two studies and to raise a few concerns regarding the Geib (2001) study that can presumably be addressed by the author herself.

II

THE STUDY BY CELBULA AND BELTON (1994) deals with aggregate-level data of an interstate nature over the 1985-1990 study period. The study by Geib (2001, p. 819) deals with micro-level data that involves, among other things, whether" ... a household head moved from a reservation area to a non-reservation area in another state . . ." during the same study period. It is important to stress that the "dependent" variables in the two studies are different not only because in one case aggregate migration at the state level is considered whereas in the other case migration at the micro level from reservations is considered, but also because in Geib (2001), the migration variable is gross out-migration, unscaled by the size of the Native American or total state population, whereas in Cebula and Belton (1994) the migration variable is the net in-migration rate, expressed as a percent of the state's 1985 total state population. Thus, the differences between what is being measured and investigated in the two studies are not trivial.

There are also potential issues regarding the right-hand side (explanatory) variables in the Geib (2001) study. For example, consider the variable rent, defined as the "average rent by state of residence in 1990" (Geib 2001, p. 820). Presumably, this variable is adopted as a control variable to correspond to the variable COL, which in Cebula and Belton (1994) was a geographically comparable index of the average cost of living in each state. It appears that both the variables averent and COL are state-level variables. If not, then the variable averent should have been defined expressly as the average rent paid at the reservation level. There are two potential problems here. First, the year chosen is 1990; however, since 1990 is the end of the study period, it would seem that use of a 1985 (or average rent over the 1985-1990 period) rent variable would have been logically more relevant (and less likely as well to reflect the impact of general migration patterns over the 1985-1990 period). Moreover, the variable COL as constructed is a geographically comparable series, whereas there appears to be no such comparability for the averent variable. Indeed, it would seem that the latter should itself have been scaled by a variable such as COL in order that the average rent levels between states could have been compared in terms that were in fact comparable.

In the Geib (2001) study, there may be a question as to why the variables assoc, to reflect whether a party has earned an associates degree, and bach, to reflect whether a party has earned a BS or BA degree, are both included in the same estimate. It would seem not improbable that many parties who have earned a BS or BA degree may also have availed themselves of cost-effective local community colleges before proceeding with bacheloriate-level study. Thus there would seem to be the potential for a multicollinearity problem. Indeed, this is suggested in Geib (2001, Table A), where neither of these variables exhibits a statistically significant coefficient. One would certainly expect the more educated parties to have greater mobility and job opportunities.

Finally, if indeed the averent series are not geographically comparable, one might then also question the expwaged variable. For the results for this variable to be truly insightful, either the wage comparison must involve geographically comparable reservation and nonreservation area real wages or geographic living-cost differences must have somehow been controlled for by the inclusion of a geographically comparable living-cost index. Thus, if the averent data prove not to have been scaled by a geographically comparable living-cost index, then the wage data used in generating variable expwaged should have been so scaled. Related to this observation, witness footnote 3 in Geib (2001, p. 824), where "money illusion" is implied in the migration-investment formulation. This formualtion can be contrasted with mew (1973), where, e.g., the expected earnings in the areas being compared have been divided by area living-cost differentials to generate geographically comparable expected real earnings.

III

THE REMARKS IN THE COMMENT are not a criticism. They are intended: (1) to make the point that the two studies are parallel but that they are different enough that comparisons/contrasts between them are quite difficult (and made all the more difficult by the introduction of new factors such as age, gender, and so forth); (2) to raise certain questions that Professor Geib might wish to clarify and/or otherwise address; and (3) to make the point that these two studies complement each other in certain ways that may make further inquiry into the topic at hand seem worthwhile. Frankly, in certain ways, I regard the study by Geib (2001) a welcome and well-done extension of the earlier paper by Cebula and Belton (1994).