AND CENSUS MIGRATION ESTIMATES 233 A COMPARISON OF THE ESTIMATES OF NET MIGRATION, 1950-60 AND THE CENSUS ESTIMATES, 1955-60 FOR THE UNITED STATES* K. E. VAIDYANATHAN University of Pennsylvania ABSTRACT This paper compares two types of estimates of net migration for the United States, one by the forward survival ratio method derived at the Population Studies Center of the University of Pennsylvania ("CRS Estimates") and the other based on the place of residence question in the 1960 framed by the U.S. Bureau of (" Estimates"). Restricting the study to the white male population, and specifically to the two groups aged 20-24 and 25-29 in 1960, the author has attempted to explain the relationship between the two types of estimate in terms of several factors, namely, the impact of multiple moves, mortality among migrants, changing probability of migration with age, military migration and economic variables such as median income, percent urban and percent unemployed. As anticipated by the author, the net migration for the 20-24 cohort was closely associated with military concentration whereas for the 25-29 cohort the association of net migration with economic factors was stronger. In the 1960 of the United States, a question on the place of residence five years prior to the was asked of a 25 percent sample of the population. On the basis of this information estimates of in- and out-migration were framed by the U.S. Bureau of (U.S. Bureau of, 1962 :Table 24). We have used this information to obtain estimates of net internal migration by states. More recently, the Population Studies Center of the University of Pennsylvania has worked out another series of estimates of net migration by states for 1950-60 by the forward survival ratio method (Miller, 1964; Eldridge, 1965). For simplicity, we designate the former as estimates and the latter as estimates. The object of this paper is to compare the two types of estimate and to attempt an explanation of the relationship between them. Our attention will be confined to the white male population, and specifically to the two groups aged 20-24 and 25-29 in 1960. 1 *The author is indebted to Dr. Hope T. Eldridge for suggesting the investigation and for her guidance throughout the study. Thanks are also due to Dr. Dorothy S. Thomas for her comments and criticisms. The charts in the paper were prepared by Mrs. Lydia Christaldi of the University of Pennsylvania Population Studies Center. 1 Dr. Eldridge has carried out adj ustments of These are the age groups that experience maximum rates of migration, and an understanding of the factors affecting the estimates for these groups may perhaps be helpful in understanding the differences for other ages or for all ages combined. The and estimates for the two age groups, by states, are presented in columns 2, 3, 5 and 6 of Table 1. We have also plotted the estimates against the estimates in Chart I-A and Chart I-B. These charts show that there is a positive relationship between the two estimates. This is what we would normally expect in view of the fact that the period covered by the estimates includes that of the estimates. The coefficients of correlation (r) are 0.97 for the younger cohort and 0.90 for the older. Both are significant at the.05 level. We have drawn a rectangle around the major cluster of points in order to bring out the states which have high positive or negative estimates. The states falling outside this arbitrary region are: the estimates for native whites to include the effects of internal migration of foreign-born whites and to exclude the effects of the external migration of native whites. Hence the estimates used in this study are comparable with the estimates.
234 SOCIAL FORCES TABLE 1. AND CENSUS ESTIMATES OF NET MIGRATION, WHITE MALES, UNITED STATES 20-24 25-29 CSB 1950-60 1955-60 1950-60 1955-60 State ('00) ('00) Ratio ('00) ('00) Ratio Maine... -36-5 0.14-66 -23 0.35 New Hampshire... 1-1 -2.13-15 6-0.37 Vermont... -28-10 0.37-35 -4 0.12 Massachusetts... 12-8 -0.64-65 -43 0.67 Rhode Island... 65 56 0.86-46 -37 0.80 Connecticut... 29-12 -0.41 81 39 0.49 New york... -507-370 0.73-133 -34 0.25 New Jersey... 64 20 0.31 114 83 0.73 Pennyslvania... -613-420 0.69-423 -112 0.26 Ohio... 45-88 -1.94 340 119 0.35 Indiana... 3-39 -1.40 51 37 0.74 Illinois... -67-159 2.37 77 60 0.77 M~chiga?... -156-187 1.20 103 14 0.14 WIsconsIn... -157-107 0.69-65 28-0.43 Minnesota... -135-89 0.65-56 39-0.69 Iowa... -187-116 0.62-154 -13 0.08 Missouri... '" -65 5-0.08-109 -25 0.23 North Dakota... -59-24 0.41-69 -11 0.16 South Dakota... -45-26 0.58-59 -23 o 39 Nebraska... -58-35 0.61-69 -19 0.28 Kansas... 18-13 -0.76-31 -56 1.77 Delaware... 24 14 0.58 41 12 0.29 Maryland... 173 107 0.62 168 41 0.24 D.C... 44 21 0.48-13 -59 4.66 Virginia... 337 265 0.79-2 -70 3.79 West Virginia... -357-168 0.47-293 -51 0.17 North Carolina... 25 20 0.79-167 -77 0.46 South Carolina... 93 95 1.03-25 -42 1.67 Georgia... 84 98 1.17-37 -63 1.71 Florida... 444 266 0.60 459 180 0.39 Kentucky... -243-78 0.32-359 -102 0.28 Tennessee... '" -179-86 0.48-209 -35 0.17 Alabama... -152-43 0.29-126 2-0.01 Mississippi... -79-45 0.57-136 -18 0.13 Arkansas... -237-93 0.39-226 -29 0.13 Louisiana... -12-1 0.06 51 18 0.34 Oklahoma... -104-37 0.36-186 -76 0.41 Texas... 259 130 0.76 66-150 -2.25 Montana... -19-13 0.69-6 2-0.36 Idaho... -51-25 0.49-36 5-0.14 Wyoming... -11-18 1.63-14 4-0.25 Colorado... 74 30 0.40 60-2 -0.03 New Mexico... 35 49 1.39 49 10 0.20 Arizona... 96 56 0.59 124 29 0.23 Utah... -9-5 0.50 7 12 1.73 Nevada... 37 19 0.53 43 1 0.03 Washington... 123 105 0.85-16 -16 0.95 Oregon... -47-36 0.75-7 20-2.71 California....... 1528 831 0.54 1421 494 o 35 Alaska... 116 74 0.64 44-40 -0.91 Hawaii... 155 110 0.71 42-25 -0.59 20-24-Pennsylvania, New York, West Virginia, and Michigan with low net migration (i.e., high net out-migration) and California, Virginia, and Florida with high net migration. 25-29-Pennsylvania and Kentucky with low net migration and California, Florida, and Ohio with high net migration. These states contribute heavily to the high positive correlation between the two sets of estimates. California especially had large amounts of migration according to both sets of estimates. (California is not shown in the charts because of limitations of space.) To test the possibility that the figures for California had given rise
AND CENSUS MIGRATION ESTIMATES 235 CHART I. SCATTER DIAGRAM SHOWING THE RELATIONSHIP BETWEEN THE ESTIMATES AND THE CENSUS ESTIMATES OF NET MIGRATION FOR THE UNITED STATE S 1950-1960 A: Age 20-24 estimate +30 -: Flo..Pa. estimate estimate -45 B: Age 25-29 -30 r------------ +15 +---...-----,"--...,.'-..., -15-30 -45 L---,L---...'---- - 154----------.J I estimate +15 +15 +30 Scale I" = 15,000-45 -30-15.30 +45 +45-15 Texas -30
236 SOCIAL FORCES to a spuriously high correlation between the two series, we computed the coefficients excluding California. The resulting coefficients were only slightly lower than the ones cited above. We did not, however compute an r excluding all the states falling outside our arbitrary rectangle. In view of the fact that these states together account for a large proportion of redistribution due to migration in the United States, such a procedure would be unrealistic. All four series are highly variable, but uniformly so. The coefficient of variation (a/x) is 1.6 for each set. To facilitate the comparison of the two series, we have defined the correspondence ratio for a state as the ratio of the estimate to the estimate for any specified age group.p The correspondence ratios are shown in columns 3 and 6 of Table 1. In the following sections, we examine the factors contributing to the levels and variability of the correspondence ratios and attempt to take account of some of them in predicting the estimates from the estimates. FACTORS AFFECTING THE LEVELS AND VARIATIONS IN THE CORRESPONDENCE RATIOS Although one might suppose offhand that net migration for the second half of a ten-year period would be roughly one-half of that for the ten-year period as a whole, and that this ratio would be reasonably stable from one area to another, there are a number of factors that lead one to expect ratios that differ from 0.5 and that vary from state to state. We shall consider some of the more important of these. Impact of Multiple Moves N either the nor the estimates take multiple moves into account. The defines a migrant as a person whose state of residence on April 1, 1960 differs from his state of residence on April 1, 1950. The method defines a migrant as a person whose state of residence on April 1, 1960 differs from his state of residence on April 1, 1955. Both series are affected by multiple moves, but the impact is different on the two series. The period of exposure to the cumulation of multiple 2 In a similar context, Everett S. Lee (1957) has used the term Index of Conformity. moves is greater for estimates than for estimates and therefore the ratio of the estimates to the estimates for both cohorts would tend to be above 0.5. Mortality Factor The and estimates measure net changes due to the migration of persons who survived to April 1, 1960. Since the period of exposure to the force of dying is longer for the estimates than for the estimates, the estimates are more affected by the failure to count as migrants those who died after migration but before the date. On this score, also, the ratio of the estimates to the estimates for both cohorts would tend to be over one-half. Impact of Changing Probability of Migration With Age A third factor that has a differential impact on the two estimates is the changing probability of migration with age. Unlike the first two factors, this one involves changes in the actual amounts of current migration and its impact is different for the two cohorts of ages 20-24 and 25-29 in 1960. An understanding of the age groups that are exposed to migration in the two estimates is facilitated by the Lexis diagram in Chart II. The estimates for the cohort aged 20-24 in 1960 relate to persons who would have been 10-14 at the beginning of the decade, whereas the estimates for the same cohort relate to persons who would have been 15-19 at the beginning of the fiveyear period for which the estimates apply. Similarly, the estimates for the cohort aged 25-29 in 1960 relate to persons who would have been 15-19 at the beginning of the decade; whereas the estimates for the same cohort relate to persons who would have been 15-19 at the beginning of the five-year period for which estimates apply. Taking the central age of the cohort at the middle of the interval as an approximation of its average age of exposure to migration, we find that the cohort aged 20-24 in 1960 had an average age of 15 during the first five years and an average age of 20 during the second five years. For the cohort 25-29 in 1960, the corresponding central ages were 20 and 25 for
AND CENSUS MIGRATION ESTIMATES 237 CHART II. A LEXIS DIAGRAM OF THE AND CENSUS ESTIMATES OF NET MIGRATION Age 20-24 Age estimate 25 25-----------}20_24 estimate 201------ 15 10IL-----I ------I 1950 1955 1960 Period of observation Age Age 25-29 estimate 30-----...------ 251 ---- 20 20 ~------'l 15 JL----_4----~ 10 I'-----f-----~ 1950 1955 1960 Period of observation estimate 25-29 30r------~---:.--}25-29 251------ 20 I'- ---,,~--- l 15~----I --- 4 10 I------~--- f 1950 1955 1960 Period of observation 10 I------I-----~ 1950 1955 1960 Period of observe tion
238 SOCIAL FORCES the respective time periods. The probability of migration reaches its peak in the early twenties. For the 20-24 cohort the risk of migration is greater in the second half of the ten-year period, and therefore the impact of the age factor for this cohort would lead to an expectation of a mean correspondence ratio of over O.S. For the 25-29 cohort the peak probability of migration would have occurred about half way through the decade, or perhaps near the end of the first half of the decade. The impact of the age factor for this cohort would therefore lead to an expected ratio of approximately O.S, or perhaps less. The Impact of Military Migration The above relationships between the and estimates for the two cohorts are likely to be altered by the differential impact of migration incident to military service, namely, "induction" migration, the movement of persons entering the armed forces, and "separation" migration, the movement of persons returning to civilian life, especially the former. Eldridge (1965 :21) found that the estimates for native white males 20-24 years old in 1960 and, to a lesser extent, the estimates for native white males 15-19 years old in 1960 are affected by induction migration, whereas the estimates for 30-34 and, to a lesser extent, the estimates for native white males 25-29 years old in 1960 are affected by separation migration. This fact is calculated to have an impact not only upon the level of correspondence ratios but, to the extent the distribution of military installations differs from the distribution of the general population, upon variations in that ratio as well. The cohort of white males 20-24 years old in 1960 would have the maximum number in the armed forces in 1960, whereas most of them were civilians in 1950. In the first five years of the decade (i.e., as the cohort grew from the 10-14 age group to the 15-19 age group), there would have been some induction migration, but this would be small compared to the heavy induction migration during the latter half of the decade (i.e., as the cohort passed from the 15-19 age group to the 20-24 age group). In the latter period there would be some separation migration as well, but the magnitude of this would be rather small. The impact of induction migration is thus concentrated in the period between 1955 and 1960. All the four factors examined so far-the impact of multiple moves, the mortality factor, the impact of changing probability of migration with age, and the impact of military migration-tend to reinforce each other with regard to the cohort of ages 20-24 in 1960. We would, therefore, expect the ratios of the five-year estimates to the ten-year estimates to be in general in excess of one-half, Indeed, this is what we found in a majority of the ratios for this cohort. With regard to the variability of the ratios, we would expect the following pattern as a result of the differential impact of military concentrations: 1. High ratios for states of net in-migration (both estimates positive) that are states of high military concentration. 2. High ratios for states of net out-migration (both estimates negative) that are also states of low military concentration. 3. Low ratios or mixed signs for states of net out-migration () that are states of high military concentration. 4. Low ratios or mixed signs for states of net in-migration () and low military concentration. With these several possible patterns of ratios, one should expect the high variability in the ratios for this cohort that we noted earlier. On the other hand, the relationships between the and estimates for the 25-29 cohort are somewhat complicated by the impact of military migration which is different from what we observed for the 20-24 cohort. Most of the cohorts aged 25-29 in 1960 were too young for military service in 1950, and most of them would have begun their military service after 1950 and completed it before 1960. In the ten-year period, therefore, this factor leads to no particular expectation regarding the general level of the correspondence ratios. estimates reflect the impact of separation migration, while the estimates probably feel very little of the impact of either induction or separation migration. With regard to the variability of the ratios, the
AND CENSUS MIGRATION ESTIMATES 239 differential impact of military concentration may be expected to give rise to the following pattern of ratios: 1. High ratios for states of net out-migration (both estimates negative) that are also states of high military concentration; 2. Low ratios (or conflicting signs) for states of net in-migration () that are also states of high military concentration; 3. High ratios for states of net in-migration (both estimates positive) that are states of low military concentration; 4. Low ratios (or conflicting signs) for states of net out-migration () that are also states of low military concentration. Thus, of the four factors examined, the age selectivity of migration, other than for military service, would tend to make the ratios of the estimates to the estimates onehalf or less; whereas, the greater impact of multiple moves and the mortality factor in a ten-year period than in a five-year period would tend to make the ratios be more than one-half. Military migration would have the opposite effects on the two estimates and contribute to the variability of the ratios. Thus, for the 25-29 cohort, the relationship between the two estimates is more complex than for the 20-24 cohort and would depend upon the predominant influence for each state. Economic Factors Thomas (1964) has shown that economic factors play an important role in determining patterns of migration. Because of geographic variations in the availability of economic opportunity, we would expect economic factors to contribute to the variability of the correspondence ratios rather than having any systematic effect on the levels of these ratios. The dispersion of the correspondence ratios can be understood in terms of economic factors in conjunction with the military. We have used the median income of white males in 1959 as a measure of economic status of different states, and the percentage of employed males in the armed forces as an index of military concentration. It is of interest to divide the states into gaining and losing categories and examine if there is any association between the factors discussed above and the status with respect to gain or loss. Excluding the states that have and estimates of different signs, we have 19 gaining states ( and estimates both positive) and 25 losing states for the 20 24 cohort, and 15 gaining states and 24 losing states for the 25-29 cohort. On the basis of the percentage of employed males in the armed forces, the states were grouped into two groups: states having more than their proportionate share of the military population; and states that have less than their proportionate share of the military population (denoted by high and low respectively). The states were further grouped according to whether they have median income above or below the U.S. average (denoted by high and low respectively). The distribution of the states in these groups is as shown below.* For the cohort aged 20-24 in 1960 all the gaining states except one (New Jersey) are states having more than their proportionate share of the military population, and all but four of the losing states have less than their proportionate share of military population. Thus, status with respect to gain or loss is clearly associated with high and low military concentration. Looking at median income, we find that only 8 out of the 19 gaining states had median incomes above the U.S. average and of the 25 losing states, 17 had median incomes * Military 20-24 25-29 Concen- Median Income Median Income tration High Low All High Low All Gaining states High 7 11 18 5 3 8 Low 1 1 6 1 7 All 8 11 19 11 4 15 Losing states High 1 4 5 2 10 12 Low 7 13 20 3 9 12 All 8 17 25 5 19 24
240 SOCIAL FORCES below the U.S. average. Thus, among gaining states high military concentration appears to be the important factor in accounting for gaining status; but among the losing states, both low military concentration and low income appear to be contributing factors to losing status. For the cohort aged 25-29 in 1960 both the gaining and the losing states are more or less evenly distributed between the two levels of military concentration, but the gaining states are predominantly high-income states and the losing states are predominantly low-income states. For this cohort the association appears to be stronger with median income than with military concentration. Had there been no other factors operating, one would have expected the states with less than their prorata share of military population to have been the losing states and states with more than their pro-rata share of military population to have been the gaining states. It is apparent that such an association has been lost for this cohort because of the welter of factors operating in conflicting directions, as we have noted earlier. Also, the military migration that affects the 25-29 cohort is predominantly separation migration which has an economic character, since many of the persons separated from the armed forces tend to move to areas of economic opportunity. Therefore, the stronger association of net migration for this cohort with median income is what one should expect. These considerations apply to states with consistent directions for the two estimates. We shall now examine briefly the states that have and estimates of different signs. positive and negative 20-24 1. New Hampshire 2. Massachusetts 3. Connecticut 4. Ohio S. Indiana 6. Kansas 25-29 1. Texas 2. Colorado 3. Alaska 4. Hawaii negative and positiue 1. Missouri 1. New Hampshire 2. Wisconsin 3. Minnesota 4. Alabama S. Montana 6. Idaho 7. Wyoming 8. Oregon For the cohort aged 20-24 in 1960 only 7 out out of the 51 states have different signs for the two estimates. In all cases the estimates yield a small net in-migration in one and a small net out-migration in the other or vice versa. Some of these estimates may not differ significantly from O. Nevertheless, we may examine the degree to which they meet expectations. For states having low military concentration the impact of military migration would be outward and would be concentrated in the second half of the decade, leading to the expectation of negative estimates. On the other hand, those states with relatively high median income would be less likely to have negative estimates than negative estimates. Four out of the 7 states-viz., Massachusetts, Connecticut, Ohio, and Indiana-meet these expectations. The other 3 states do not fit into either this pattern or its reverse, viz., low median income and high military concentration. For the cohort aged 25-29 in 1960, 12 states, or nearly one-fourth of the total, have and estimates of different signs. The largest differences are those for Alabama, which has a net out-migration of 12 J600 according to the estimate and a net in-migration of 200 according to the estimate and Texas which has a net in-migration of 6 J600 according to the and a net outmigration of 15 JOOO, according to the estimate. For the remainder the amounts are small and the differences may not be significant. However, the 4 states with negative estimates are states having more than their prorata share in the military population, and all but 1 of the states having positive estimates are states with less than their pro-rata share of the military population. This is in conformity with our observations concerning the effects of separation migration discussed above. Further, 2 of the states with positive estimates had above average income and 5 of the states with negative estimates had below average income. We have intuitively established the existence of association of the two estimates to the economic and military factors; and now we will attempt to measure this association through the use of correlation techniques, using the
AND CENSUS MIGRATION ESTIMATES 241 TABLE 2. VARIABLES CORRELATED WITH NET MIGRATION IN THIS STUDY Armed Median Percent Unemployed Forces Income Percent State (000) ($00) Urban 20-24 25-29 ----- Maine... 16 33 51.3 7.62 5.21 New Hampshire... 6 38 58.3 5.59 2.74 Vermont... 1 33 38.5 7.37 3.52 Massachusetts... 37 44 83.6 5.57 3.77 Rhode Island... 23 38 86.4 4.37 4.27 Connecticut... 12 50 78.3 4.96 3.36 New York... 36 48 85.4 7.32 4.77 New Jersey... 46 52 88.6 5.09 2.90 Pennsylvania... 16 43 71.6 10.71 6.48 Ohio... 15 47 73.4 8.46 4.88 Indiana... 8 49 62.4 9.92 3.10 Illinois... 38 45 80.7 5.22 3.09 Michigan... 12 51 73.4 9.64 5.67 Wisconsin... 5 50 63.8 6.05 3.26 Minnesota... 5 40 62.2 8.67 4.82 Iowa... 2 37 53.0 6.16 3.08 Missouri... 31 39 66.6 5.46 3.45 North Dakcta... 4 31 35.2 9.78 5.12 South Dakota... 5 30 39.3 6.44 2.56 Nebraska... 13 35 54.3 3.41 2.17 Kansas... 30 40 61.0 4:.42 2.78 Delaware... 6 49 65.6 6.48 3.40 Maryland... 46 49 72.7 4.39 2.90 D.C... 9 47 100.0 3.17 1.99 Virginia... 122 37 55.6 3.55 2.74 West Virginia... 1 35 38.2 17.05 9.91 North Carolina... 67 30 39.5 3.59 2.38 South Carolina... 53 32 41.2 3.04 2.12 Georgia... 60 34 55.3 3.60 2.56 Florida... 71 37 73.9 4.11 2.82 Kentucky... 34 29 44.5 7.94 6.60 Tennessee... 22 29 52.3 7.52 4.95 Alabama... 20 34 54.8 6.70 4.72 Mississippi... 20 28 37.7 4.77 2.79 Arkansas... 8 25 42.8 6.73 4.26 Louisiana... 18 40 63.3 6.99 4.26 Oklahoma... 29 34 62.9 5.26 3.49 Texas... 146 37 75.0 4.83 2.98 Montana... 6 40 50.2 9.00 5.85 Idaho... 4 39 47.5 6.53 4.35 Wyoming... 2 44 56.8 6.53 3.66 Colorado... 25 '.1:2 73.7 5.45 2.96 New Mexico... 20 41 65.9 6.46 4.25 Arizona... 16 43 74.5 5.51 3.10 Utah... 3 46 74.9 5.62 2.98 Nevada... 7 51 70.4 4.44 3 55 Washington... 47 47 68.1 7.08 4.17 Oregon... 5 45 62.2 8.32 4.91 California... 83 51 86.4 6.30 4.46 Ala aka... 30 47 37.9 2.44 4.04 Hawaii.... 40 36 76.5 0.94 0.89 numbers in the armed forces in 1960 as a measure of military concentration, the median income of white males in 1959 as a measure of economic status, and two other variables of an economic character, viz., percent urban and percent unemployed in 1960 (see Table 2). We have also worked out the coefficients of correlation with rates in the place of absolute numbers where the rates were obtained by dividing the and estimates of net migration by the number of white males in the 20-24 and 25-29 age groups as of 1960. 3 The coefficients of correlation (r) obtained are as follows: 3 For correlating with rates, we have used the percentage males in the armed forces as our measure of military concentration.
242 SOCIAL FORCES Military concentration Medianincome Percent urban Percent.97*.52*.22.26 20-24 Rate.96*.74*.41*.42* unemployed -.37* -.69* * Significant at the.05 level. Military concentration Median income f>trcenturban Percent.90*.33*.44*.41*.45*.66*.53* unemployed -.19 -.40* * Significant at the.05 level..53* 0.08 0.15-0.40* 25-29 Rate.19 Rate.83*.24.30* -.68* Rate.04 -.59*.41*.26.32*.14 -.02.08 At the.05 level the numbers in the armed forces and the percent unemployed are significant factors in explaining the variation in both the and estimates for the 20-24 cohort, whereas median income and percent urban have a significant influence in explaining the variation in the and estimates for the 25-29 cohort, especially the former. The high positive correlation of the rates for the 20-24 cohort with the percentage of employed males in the armed forces reflects the impact of induction migration that we had anticipated for this cohort. However, the results are somewhat surprising for the 25-29 cohort for which the rates are poorly correlated with each other (r =.19) in contrast to the strong correlation that we found between the absolute numbers (r =.90). The correlations of the rates with the percentage of employed males in the armed forces, median income, percent urban and percent unemployed are all significant at the.05 level, whereas the rates are significantly correlated only with the percentage of employed males in the armed forces. The latter is positively correlated with the rates and negatively correlated with the rates, reflecting the opposite ef- fects of the separation migration. This is again in accordance with our expectations about the impact of separation migration on the two estimates for this cohort. The strong negative correlations that we find between the rates and the percent unemployed in contrast to the positive correlation between the rates and the percentage of employed males in the armed forces may be due to the concentration of the armed forces in the low income areas. CONCLUSIONS A high proportion of the variations in the estimates is explained by the variations in the estimates (RS = 94 percent for the 20-24 cohort and 81 percent for the 25-29 cohort). This clearly validates the estimates as indirect measures of net migration. For both cohorts, the correlation of the absolute numbers (for the two estimates) is higher than the correlation of the two rates, and for the older cohort the correlation of the rates is not significant at the.05 level. We have speculated that for the 20-24 cohort, the impact of multiple moves, the impact of changing probability of migration with age, and the impact of military migration tend to reinforce each other. For this cohort we expected the ratios of the five-year estimates to the ten-year estimates to be in general in excess of one-half. This was found to be the case. On the other hand, for the 25-29 cohort the age selectivity of migration, other than for military service, would tend to make the ratios of the estimates to the estimates one-half or less; whereas the greater impact of multiple moves and the mortality factor in a ten-year period than in a five-year period would tend to make the ratios be more than one-half. Military migration would have the opposite effects on the two estimates and contribute to the variability of the ratios. For the 20-24 cohort it was speculated that net migration would be closely associated with military concentration, whereas for the 25-29 cohort, the association of net migration with economic factors would be stronger. These speculations are borne out by our correlation analysis using median income in 1959 as a mea-
sure of the economic factor and number (and percentage of employed males) in armed forces as a measure of military concentration. REFERENCES Eldridge, Hope T. 1965 Net Intercensal Migration for States and Geographic Divisions of the United States, 1950-60: Methodological and Substantive Aspects, Analytical and Technical Report No.5. Philadelphia: Population Studies Center, University of Pennsylvania. Lee, Everett S., et al. 1957 Population Redistribution and Economic Growth, United States, 1870-1950. Vol. 1. Methodological Considerations and Reference Tables. Philadelphia: American Philosophical Society. Miller, Ann Ratner AMERICAN INDIAN MIGRATION 243 1964 Net Intercensal Migration to Large Urban Areas of the United States, 1930-40, 1940-50, 1950-60. Analytical and Technical Report No.4. Philadelphia: Population Studies Center, University of Pennsylvania. Thomas, Dorothy S. 1964 "Temporal and Spatial Interrelations between Migration and Economic Opportunities." In Hope T. Eldridge and Dorothy S. Thomas, Population Redistribution and Economic Growth, United States, 1870-1950. Vol 3: Demographic Analysis and Interrelations. Philadelphia: American Philosophical Society. U.S. Bureau of 1962 of Population, 1960, Mobility for States and State Economic Areas. Final Report PC (2)-2B. Washington, D.C.: Government Printing Office. SOME ASPECTS OF AMERICAN INDIAN MIGRATION* ALAN L. SORKIN The Johns Hopkins University ABSTRACT This paper is a study of federally assisted American Indian migration from the reservations to urban areas. The education of the migrants, their earnings before and after relocation, and the change in the degree of antisocial behavior after leaving the reservation are analyzed. It is found that while relocation can enhance the standard of living of those participating in federal programs, budget limitations prevent these programs from assisting enough applicants, to markedly reduce the level of surplus labor on the reservations. T he purpose of this paper is to present information on the magnitude and character of migration of American Indians from the reservations to urban areas, and its effect on the reservation economy. There are approximately 380,000 American Indians residing on or adjacent to reservations (U.S. Public Health Service, 1966:10). These individuals comprise the most povertystricken minority group in the United States. The median family income for reservation Indians is $1,800 per annum, with 76 percent of all reservation families earning incomes below *This study was in part financed by funds provided by the William H. Donner Foundation, Inc., to the Brookings Institution where the author was a Research Associate. the poverty threshold (U.S. Bureau of Indian Affairs, 1967). Unemployment of reservation males in 1967 was 37.3 percent of the labor force, or SO percent higher than in the United States as a whole during the worst part of the Great Depression (U.S. Department of Labor, 1968:68). In 1966, according to a task force on Indian housing, over 75 percent of all reservation homes were substandard, with over SO percent needing to be replaced (U.S. Bureau of Indian Affairs, 1966:5). In order to ameliorate the problems of poverty and surplus labor on the reservations, the Bureau of Indian Affairs operates two separate relocation or employment assistance programs for reservation Indians. The first is a