THE SELECTNITY OF MICHIGAN MIGRANTS, 1949-1950 THESIS FOR THE DEGREE OF M. A. MICHIGAN STATE UNIVERSITY DONALD LYLE HALSTED 1958 PIACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINE return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE IA! p fifi" 9]_.1 If 4 . l, ' ma mnmmu v.3 SELECTIVITY OF MICHIGAN IIIGPANTS, 1919-1950 By DONALD LYLE HALSTED AN ABSTEACT Submitted to the College of Science and Arts Michiran State University of Agriculture and Applied Science in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Sociology and Anthropology 1958 /7 Approved :;24/LLQUJ4 Dr A (I) 2C 1/ Donald L. Halsted I“ The problem of selectivity Oi migrants has not been studied in tnis country, to any yreat degree, until the last three decades. Vesides the problem of volume there have been taree major difficulties in toe accumulation of the knowledge in these studies, namely, the type of data used, the m thod used and tieoretical orientation. In l9LO and 1950 the problem of comprehensive and reliable data was ret by the publication of migration data by the Bureau of the Sensus. This thesis proposes an approach to the other two difficulties mentioned above. Wh’le due Kettodology in previous studies has many variations, most of the stadies can he said to use the differential method of measuring selectivity of migrants. A new method is proposed on the bases that the differential method is logically unsound, obscure in definition, and is partly a function of the rate of migration. It is believed that tfllS new method meets these criticisms. Studies of migration selectivity have also exhibited a Considerable lack of theoretical orientation. Waile the empirical generalizations found have been insightful, it is hard to evaluate their significance and stability. To this end, the hypothesis prOposed, being derived from tne larger field of ecology and the Specific concept of dominance, allows the placing of migration selectivity in a large perspeciite. Donald L. Halsted The specific, directional hypothesis prOposed is not supported by tie data. However, the results indicate that the relation of-d minance and selectivity of migration is a fruitful area for further research. TIE SELECTTVITY or HISEIGAN ElGhANTS, 19h9-1950 By DONALD LYLE HALSTED A THESIS Submitted to toe College of Science and Arts Michigan State University of Agriculture and Applied Science in partial fulfillment of the requirements for the degree of VASIe‘tu‘AIIs Department of Sociology and Anthropology 1958 CMWmI THE GENEIAL FIELD OF MIGRATION The movement of people from one habitat to another has been a characteristic phenomenon since the beginning of history. From the earliest period when social groups were organized on a mobile basis through the development of towns and settled areas, the rise of nation states, and the discovery and growth of the new world, people have been on the move. Numerous attempts have been made to isolate the essential factors and to characterize and explain such movements of people. Explanations have ranged from environmental conditions, war, political action, popu- lation pressure and desire for freedom, to the call of the wild and direct economic response. Interest in the movement of population probably devel- oped in the Middle Ages. But it was not until much later, along with a growing interest in science and a developing body of scientific knowledge, that any sort of analysis was undertaken. It was perhaps natural enough that certain Western European countries, beset by population and economic problems growing from the Industrial Revolution, began the first systematic count and analysis of populationmovements.l l Taft, Donald R., Human Migration, Ronald Press 00., New York, (1936), p. 56. MIGhATION DIFFERENTIALS IN THIS COUNTRY In this country, internal movement and especially the field of migration differentials, as areas of study, were overshadowed for many years by the problems of immigration. While the general field of migration undoubtedly had several adherents, very few studies of migration differentials were made during this period. 'When immigration to the United States was sharply re- strictly, first by'World War I and then by Federal Legislat- ion in the early 1920's, the population student could still study the problems of assimilation and distribution of the immigrants.2 During this period, industrial growth in the North and the resulting movement of Negro laborers, along with the growing awareness of the effects rural to urban migrations may have, led to a few studies of migration dif- ferentials.3 It was still later, with the crises of the depression of the late twenties and early thirties, and the magnitude ‘ of the concomitant population changes and problems, that the population students turned in earnest to the description and 2 Thomas, Dorothy Swaine, "Research Memorandum on Mi- gration Differentials", Social Science Research Council, Bulletin h3, (1938) pp. 1 and 2. 3 Ibid., pp. 2 and 3 and Appendix A. -3- analysis of migration differentials.h BACKGROUND OF THE PROBLEM Previous to l9hO, population students had to rely on localized studies to obtain data suitable for an intensive analysis of migrants. Unfortunately, both the methods of gathering and analyzing the data varied greatly and in sev- eral cases the data were gathered during a period of crisis for the local area. Beginning in l9h0 and continuing in 1950, the Federal Census has obtained comprehensive, nation— wide data on the movement of people. This thesis is based on the 1950 Federal Census data for Michigan.5 ‘THE PROBLEM This thesis is an examination of the differences be- 6 as exhibited in the 1950 tween migrants and non-migrants data, and an attempt to account for these differences by an explanatory framework. The framework represents an extension h Ibid., Appendix A. Of 111 American studies of migra- tion differentials published up to 1938, only 23 were publish— ed before 1930. And of 23, 15 appeared after l92h. S The l9h0 material has been analyzed by A. H. Hawley, in a report by the University of Michigan Bureau of Govern— ment, called Intrastate Migration in Michigan: l935-l9h0. Because of the discrepancies in collecting and reporting be- tween these two sets of data, no attempt will be made to in- corporate the l9h0 data. 6 Hereafter, the difference dealt with will be referred to as "selectivity". This term will be discussed and defined in the section on method. .b- of ecological principles to migration data. The central problem is the extent to which selectivity is accounted for by this general ecological framework. The significance of this problem is two-fold. On the general scientific level it is a test of one type of hy- pothesis derived from a larger body of theory. Given that the assumptions and the indices used are correct, if the hy- pothesis is borne out then progress has been made in under- standing migration. If the hypothesis is not borne out, this approach to migration may be rejected. On the practical level, an acceptable hypothesis of this type would be of value to population students in pre— dicting the future population characteristics of an area. If this prediction is accurate enough, then the future com— position of areas in terms of age of population, years of school completed, and sex composition, would be available to urban planners and community development programmers. The data, per se, would not be used but rather the implica— tions of the data for type of housing, number and type of jobs, etc., that will be most efficient for the type of popu- lation. Of course, any complete planning program would also require an accurate prediction of types of in—migrants. DERIVATION OF THE HYPOTHESIS Inasmuch as ecology is a general theory concerned with the distribution and movement of population, it should lend itself to an explanation of differential movement. The hypothesis develOped here is derived from the eco— _ -5- logical concept of urban domination. The focus of the con- cept of urban domination is upon the large urban center, which is viewed as extending its characteristics into sur— rounding areas to the degree that these areas are functional— ly interdependent with the urban center. Martin, in a recent review of research concerning urban dominance, finds that dominance is a function of distance to the urban center. He calls this relationship the "gradient principle" and states it as, "the extent of urban influenced changes in rural areas varies inversely with distance to the nearest city and directly with the size of that city".7 This thesis repre- sents an attempt to extend a similar principle to migration selectivity in the State of Michigan. However, two problems are involved: 1) the extension of the principle over a much wider area than in previous studies8 and, 2) the use of a principle developed on static data (characteristics of peo- ple living in an area) to eXplain a dynamic phenomenon (mi- gration). The expectation that the general principle should hold over a wide geographical area is not without basis. R. D. Mckenzie, writing more than two decades ago, says, "One of 7 Martin, Walter T., "Ecological Chants in Satelite Rural Areas", American Sociological Review, Vol. XXII, No. 2, (1957), p- 176. 8 Ibid., The research based on urban domination has, for the most part, dealt with an urban area and its immedi— ate surrounding area. -6... the outstanding trends in modern communal development is that of integration. The entire settlement pattern of the country is becoming knitted together into an ever finer web of functional interrelationships." Further, he notes that, "A national system of 'key' cities, each dominating a more or less definable trade area, is arising..."9. Hence, it would seem reasonable to assume that this principle should cover Michigan and that Detroit would be the dominant urban center for part, if not all of Michigan.10 Inasmuch as the gradient principle is based on research which was not concerned with migration, its application to migration data is postulated on the following considerations. Migration is usually viewed as a result of the "push-pull" factors credited to A. C. Haddon. Hawley interprets these factors as being the relation of population and subsistence. Push is here, the overpopulation of an area, (a larger popu- lation than the number of jobs available will support) and pull, the underpopulation, (a smaller population than the number of jobs available will support).11 'While these fac- tors are set forth as a general principle to explain all 9 McKenzie, R. D., "Integration and Dominance", Read— ings in Human Ecology, George Wahr (Publisher), Ann Arbor, _.C.. (19311;, p 0103 o 10 The effect of Chicago is not considered here. ( 5 ll Hawley, A. H., Human Ecology, Ronald Press, New York, 19 O , p. 329. -7- movement, further specification of factors is necessary to predict the pattern of selectivity. The selectivity pattern is derived from considerations of the types of communities in which migrants originate. Hewley describes two types of communities, the dependent and the independent. He defines the dependent community, here taken to refer to the urban community, in these terms, "The primary orientation of the dependent community is not to the land but to a network of inter-community relations, and that network of relations or market situation, since it consti- tutes a highly flexible and changeable sustenance base, pre- supposes maximum mobility. In consequence, population in general, if not individuals in particular, is prepared for and habituated to readjustment through migration."12 Orien- tation to migration is important, but other factors must be considered. Hawley comments, "Migration is facilitated also by the existence of a highly developed transportation and communication system."13 In contradistinction, the independent community (rural) would exhibit the polar tendencies of a relative lack of transportation and communication facilities, and a lack of orientation to migration. While the concept of a completely independent community can not be used as one end of a continuum, it is possible to 12 Ibid., p. 33h 13 Ibid., p. 335 -8- build a continuum of relative dependence. The continuum is here used to place different areas, with the urban centers located toward the end of relatively complete dependence and the rural areas placed toward the end of relatively lit- tle dependence. It would seem reasonable to assume that the migrants, relative to the populations from which they come, should reflect the variations in orientations to mi- gration and in obstacles to migration, (i.e., lack of trans- portation and communication.) More Specifically, it is ex- pected that the greater the obstacles and the less the orien- tation to migration, the less likely all groups are to move and consequently the greater the difference between migrants and the population from which they come, (i.e., the greater the measure of selectivity). Inasmuch as urban areas are generally more oriented to migration and have relatively ample transportation and communication (relative to the ur- ban areas), it is expected that, in general, migrants from urban areas will exhibit less selectivity than migrants from rural areas. It is not necessary, for this study, to actually find what relation the various areas of Michigan have with Detroit. Bogue has shown that dependence varies directly with dis- tance.lh Consequently, in this study, distance will be used as the measure of dependence and therefore, of the difference 1h Bogue, Donald F., The Structure of the Metropolitan Community, a Study of Dominance and Subdominance, Ann Arbor, (I950). in selectivity expected. On these Considerations the following general hy- pothesis is advanced: The greater the distance an area is from an urban center the greater the difference between the migrants and the population of the area in which the migrants originated. From this is taken the specific hypothesis to be tested in this study, namely, The greater the distance an area is from Detroit the greater the measure of selec- tivity. SO CE OF DATA The data upon which this thesis is based were obtained through the NorthéCentral Regional Project 18 on migration. The special photostat sheets contain information based on a 20% sample of all mobile peOple within, into and out—of State Economic Areas.15 The residence of any particular migrant was obtained for the date one year prior to the date of enumeration in 1950. For practical purposes the dates of April 1, l9h9 and April 1, 1950 are accepted as the dates to 15 State Economic Areas, hereafter abbreviated to SEA, is a general term used to refer to a county or group of coun- ties of similar social and economic characteristics. There are two types of areas, non-metropolitan and metropolitan, referred to as State Economic Areas and State Metropolitan Areas, respectively. The term SEA will refer to both. Dif- ferentiation will be made, where necessary, by the use of the terms Metropolitan and NoneMetrOpolitan. In the Appendix tables and on the map, arabic numbers identify non—metropoli— tan areas and alphabet letters identify metropolitan areas. For a discussion of the construction of these areas see: Bogue, Donald J., State Economic Areas, U. 5. Bureau of the Census, Washington D. C., (1956), p. ll. .10- which the data pertain. A mover is here defined as a person residing in a different house in l9h9 than in 1950. The data are divided into two "counts", the 2—1 count and the 2-14 count . Z—l Count This category includes data on the mobile population that were residing in the State of Michigan on the date of enumeration in 1950. It has four main divisions: Same Coun- ty movers; Same SEA movers; Different SEA movers; and those Abroad and Not Ascertained. Same SEA movers are persons residing in a different county but in the same SEA in 1919 and 1950. Different SEA movers are persons residing in a different SEA, either in Michigan or another) state, in 1919 and 1950. The Abroad and Not Ascertained category refers to all persons whose l9h9 residence was outside the continental boundaries of the United States or whose l9h9 residence could not be obtained from the information gathered. The four classifications of movers are presented for each of the SEA'S in which the movers were residing in 1950. The data are presented for the characteristics of color, residence in 19h9, (farm, non-farm, Not Ascertained), dis- tance moved, age, years of school completed, marital status, employment status, occupation and family income. The char— acteristics are cross—classified by residence in 1950 (urban, rural non—farm, and rural farm) with each residence classi- fication divided into total males, nonéwhite males, total females and nondwhite females. Except for Area F (Detroit -11- MetrOpolitan Area) information is not available for the cross-classification of color of mover. The distance moved characteristic applies only to the different SEA movers, and consists of three categories, Same State, Contiguous State and Non—contiguous State. Z—h Count These data are similar to the different SEA mover classification of the 2-1 count. The data differ in this respect, they represent the characteristics of Different SEA movers by the area in Michigan in which they resided in 19h9. As such they represent the out-migrants of an area between 19h9 and 1950. The format of these data differs from the 2-1 count sheets in this respect. The major divisions of the cross— classification are male and female with each subdivided into total non—farm (19L9 residence), nonawhite non-farm, total farm, nondwhite farm, farm Not Ascertained, nonawhite farm Not Ascertained. The residence characteristics are for urban farm and non-farm in 1950. SAMPLE DESIGN Within each enumeration district five versions of the schedule are used, with each used approximately to the same degree. On each version a line has been preselected as a sample line — a different line for each version. For each individual a separate line has been filled out on the sched- ule. The sample then consists of the peOple found on these -12- preselected lines and represents approximately a 20% sample. To obtain an estimate of the total number of movers by char- acteristics, the sample figures were multiplied by five. The sample is unbiased.16 MIGRANTS For this thesis migrants are defined as those movers who crossed a SEA line. This would be the Different SEA category in the Z—l and Z-h counts. While this definition is to a large degree arbitrary, certain considerations make it seem less so. This category, more than any other, represents the people moving a long distance and for the most part to a new type of area. ‘While distance per se does not make move— ment significant, it is assumed that whatever local ties ex— isted must be broken in large part. Also it is assumed that people moving to a new area are presented with a new situ- ation. This is not likely true to the same degree of the Same SEA movers and eSpecially not true of the Same County movers in general. This combination of a new area and dis— tance are assumed to be the polarity of local movement or no movement and thus should best, within limitations of the data, distinguish nonémovers or local movers and migrants. METHOD OF ANALYSIS Because the approach employed differs considerably from 16 Special Report P-E No. hB, United States Bureau of the Census, Washington, D. C., (1956), p. 11. -13- other methods of analysis, this section will present the method in detail and its rationale. Aside from purely cursory descriptive studies, migra- tion data are usually analyzed by one of two general methods. The first method concentrates entirely on the migrants. The characteristics of the migrants are usually summarized by some measure of central tendency such as the mean or median; or selected parts of a distribution may be utilized, such as the percent of migrants in certain age categories. Usually the migrants are grouped by the residential type of area of origin or destination, (i.e., urban or rural). Comparison is then made between migrants of these residential group- ings using the measures indicated above. The second method is usually referred to as migration differentials. This method compares the characteristics of the migrants with the characteristics of the population af— ter the migrants have left or at the time they were leaving. The important point is that the comparison is between the migrant and the total population as if they were two separ— ate populations. This type of analysis is usually some com- parison of the central tendency of the population such as the mean or median, although a goodness of fit test is sometimes utilized. The method utilized in this thesis attempts to measure more completely and without the tendency for bias of the dif- ferential method, the difference between migrant and non- migrant population. To differentiate the results of this new method, they will be referred to as selectivity. -1h_ Acceptance of the method prOposed is dependent on the answer to the question: what constitutes the valid differ- ence between.migrants and non-migrants? Which is more valid: the difference between two separate and distinct populations as proposed by the differential method or the difference be- tween migrants and some theoretical pOpulation as prOposed by the selectivity method? The answer arrived at by the writer is that the essential difference to establish is the difference between a migrant distribution and a theoretical distribution. The rationale for using the proposed method is statis- tical and theoretical in nature. In order that the argument may be followed more easily the methods will be described and then compared. The discussion is limited to effects the two methods have on measures of differences of distribution. Age will be used as an example of a characteristic under in- vestigation. The differential method would compare the age distri- bution of the migrants with the nonamigrants (the pOpulation in an area after the migrants had left). The nonqmigrant population would represent the original pOpulation minus the migrant population. The selectivity method would compare the age distribu- tion of the migrants with the population as it existed be- fore the migrants had left. The distribution of the origi— nal population is used as a theoretical pOpulation, which would be approximated roughly by the migrant distribution, if only random selection processes are operating. -15... The first point of conflict may be considered theoreti- 021. Of interest to the population student is the effect migration has upon a population and one way of measuring this is to find which groups are affected most by the migra- tion processes. Given this goal it would seem more reason- able to measure how the migrants differ from the population before the migrants left, than to compare migrants with the population after the migrants left. For it is this original pOpulation upon which the migration processes are operating. The population after the migrants have left has already felt the impact of the migration processes. Secondly, the differential method gives one no indica- tion of the extent to which migrants vary from a random se- lection of the base pOpulation. The effect of not allowing for random differences is to arouse concern over the measure obtained. It is not known to what extent this measure of difference is a reflection of random processes. Inasmuch as this is true, the definition of just what is measured by the differential method is at best hazy. It will be noted fur- ther, that any comparison of measures of difference arrive at by the differential method compounds the difficulty in interpretation. The selectivity method does allow for the random selec- tion effect and is interpreted in the following manner. The measure is the effect migration has upon the original popu- lation. The greater the measure the more the migrants vary from the original population and consequently the greater -16- effect the migration has on the original population. The last point of conflict and of great importance to this study is the lack of stability of the differential method. This lack of stability is due to the fact that the measure of difference in the differential method is a func- tion of the rate of migration. That this is not true of the selectivity method can be demonstrated by taking two original populations that have the same proportional distribution but varying in the percent balue they represent of their respec- tive original pepulations. With the above populations, the selectivity method would yield identical values for the dif— ference between the two populations and their respective migrants. The differential method requires that the migrants be subtracted from their original populations; the computat— ion of the resulting populations' (non—migrant) proportional distributions and comparison of these final distributions with the migrants' distributions. The results obtained here will be different. This is the effect of the subtrac- tion process noted above, which has the effect of varying the non-migrants' prOportional distributions to the degree that the rate of migration between the two non—migrant pOpu- lations is different. On the basis of these considerations the selectivity method will be used. The method employed involves the reconstruction of the l9h9 population, i.e., the total populetion before the mi- grants had left. This is referred to as a base population or the theoretical population. This is the distribution on a -17- characteristic that a migrant population would resemble if random selectivity processes were operating. Its compu- tation will be taken up in a separate section. The measure of selectivity is computed by utilizing a chi square test for the differences between the distributions of the migrants and the base population. It is this value we refer to as the degree of selectivity and it is this val- ue that is hypothesized as dependent upon distance from Detroit. Because our unit of analysis is the SEA, we have to ac- complish all the above computations for each SEA. Inasmuch as our final analysis is based on the relative size of these SEA'S, chi square has an immediate disadvantage because it is a function of the total size of the population in an area. To overcome this disadvantage, a method of controlling for different size pOpulations is needed. The control involves giving each pOpulation, both base and migration, in each area, a total size of 1,000. To find the size of a category in a distribution involves finding what percent the category represents of the original total (actual count) and multi- plying it by ten. These proportions multiplied by ten will then equal 1,000.17 It is these final distributions on which 17 This procedure developed by Drs. Joel K. Smith, Assistant Professor of Sociology at Michigan State University and Charles Proctor, Instructor in the Department of Statis- tics at Michigan State University. .18- chi square is computed. To test the hypothesis of the relation of distance and selectivity the Pearsonian coefficient of correlation with the .05 level of significance will be used. 'While the variables do not exactly meet the requirements of normalcy necessary for use of the Pearsonian r, the general effects of this bias are probably negligible. This is especially true of the variables of age and education. The findings in the other variable, percent male (which is subject to the greatest bias), are of such a magnitude that this bias would not cause a change in interpretation. The general character of the hypothesis should be noted here. It deals only with the distribution of the meas— ure of selectivity. An examination of the method will re- veal that the value of selectivity between areas has two possible sources of origin. A difference between areas of the base population and a difference between areas of the migrants. Because of this we cannot say what the exact na- ture of the differences are. ‘While this is not necessary for the hypothesis, some indication of the major source of this difference will be given. BASE POPULATION The base populatiOn is the population as it existed in l9h9. The computation involves taking the 1950 pOpulation of an area, adding the Z-h count Different SEA movers (out—mi- grants), subtracting the Z-l count Different SEA movers (in- migrants), and subtracting the Abroad and Not Ascertained -19- category.18 Inasmuch as the Census pro-rated the Abroad and Not Ascertained category and presents it as part of the total population in 1950, any computation of a base popu- lation from data found in the Census.19 will differ slightly from the base population used here. LllIITAT IONS The method employed here limits our comparison of mi- grants and base population to out-migrants of the areas in Michigan. Lacking stream data-we cannot compute the l9h9 population of the in—migrants of areas in Michigan. It is quite conceivable that control of type of area of destina- tion and origin would be superior to the type of analysis given here and therefore of more value. The method of data collection also limits the characteristics that may be come pared. This is because only 1950 characteristics are col- lected and thus the l9h9 characteristics must be inferred from the 1950 data. Therefore, any characteristidsthat may change other than the same amount and in the same direction in all categories must be dropped. This involves leaving out of analysis: marital status, family income, employment status and occupation. As a result the study is limited to comparisons of age, sex and years of school completed. 'While 18 See the note at the beginning of the Appendix. 19 Data necessary for constructing the base population is available in Special Report P—E, No. hB, o . cit., pp. lhO- 1L5. -20- we reconstruct the l9h9 data on the basis of 1950 charac- teristics, it should be noted that this method is consis- tent for all groups and therefore of no importance here. MEASUREMENT OF DISTANCE The general procedure in determining the distance SEAfs are from Detroit, was to draw on a scale map of Michigan, series of concentric circles from the center of Detroit to the approximate center of the SEA's. Eor SEA's that were outstate no problems developed. For SEA'S near Detroit, because of their number and shape, certain arbitrary decis- ions had to be made. If the SEA was of such a shape that its approximate center could not be readily located and if it was tied or close to another SEA, consideration was given‘ to the area having most of its area closest to Detroit.20 In order to use the Pearsonian coefficient of correlat- ion, measurement of the variables must be in at least an interval scale. To obtain an interval scale, the distance from the center of the Detroit area to the center of the other SEA's was measured on a scale map of Michigan in léths of an inch.21 20 For a map listing the areas and showing the adOpted center of the area, see Appendix Figure I. ' 21 See Appendix Table II. CHAPTER II - FINDINGS AGE Of all the variables dealt'with in migration, age seems to be the most stable.1 Because of this stability, age should provide a better test of the hypothesis than other characteristics might. The hypothesis is tested only for those migrants eigh— teen years and over in 1950. The younger groups, while im~ portant in many studies, have less value here. The decision was made on the basis of keeping the data as close as pos- sible to people capable of making an independent chOice of movement. A second reason for leaving out the young age groups was the probable difference in fertility between rural and urban areas and its close relationship to the ex- pected age of the migrants. If rural migrants have larger families than urban migrants or some other relationship ex—- ists between family size and migration, the-addition of children may add a biasing dimension to the measure of age. This writer feels that this possible bias should be left out. The total correlation of distance and selectivity of 1 Thomas, D. 8., "Migration Differentials", 99. cit., Chapter 1 o (a -2- age is .665h2 (Appendix Table III). This is significant at the .01 level, measured by the Analysis of Variance F test.3 This indicates that the farther an area is from Detroit the less the migrants resemble their base pOpu— lation. It would also be of value to determine what portion of this relationship is due to the differences between areas of the base populations. An indication of this may be obtained from the partial correlation coefficient of age selectivity and distance When median a'e of the base population is taken into account. This relationship has a coefficient of .h995 (Appendix Table III), and is significant at the .05 level. This may be interpreted as indicating that selectivity of age would not be significantly related to distance if all base populations had the same median age. This result is not surprising in view of the high relationships of distance and median age, and of selectivity of age and median age, .5985 and .5562 respectively (rxz and rv in Appendix Table z III). These relationships indicate that median age is non- random with respect to distance and that our selectivity measures vary with median age of the base population. Thus 2 For a discussion of correlation, its computations and interpretations, see Hagood, Margaret, and Price, Daniel, Statistics for Sociologists, Henry Holt & 00., New York, (1952), Chapters 23 and 25. 3 Ibid., p. h30. All tests of significance in this the- sis are by this method. -3- it should be expected that the relationship of selectivity of age and distance would decrease when.median age of the base population is adjusted for. From these correlations, it is concluded that a large part of the selectivity seems to be due to differences in the base population between areas (i.e., the differences in the migrant populations be- tween areas is not enough to account for the measure of selectivity. There is another facet of the data that is important to the hypothesis. ‘While the hypothesis has been set up to cover all areas of Michigan, it might have.been hypothesized that the metropolitan areas will maintain different relat- ionships to Detroit than the non-metropolitan areas exhibit. An indication that this is the case, may be seen by the re- lationships of selectivity and distance for these two types of areas. Here the relationship of the non-metropolitan areas is .6667 and for the metrOpolitan areas .2209 (Appendix Table III). Vhile earlier, the relationship of selectivity of age and distance was spelled out, further data must be presented before the exact relationship of axe and migration can be stated. This is necessary because chi square does not reveal the direction of the differences between migrants and base population. Table IV in the Appendix gives the signed amount of difference between the migrants' and the base populations' distributions. If these signs are in the same direction for all areas then the direction of difference, as well as the 41- amount of difference, may be stated. 'With two exceptions the signs are the same for specific age groups over the areas. If this slight discrepancy is ignored we may conclude that migrants, when compared to the base population, differ in the same way for all ages, i.e., they tend to be over repre- sentative of the younger pepulation and under—representative of the older pepulation with the degree of representative— ness decreasing with the distance from Detroit. In summary several points are of interest. First, the data support the hypothesis, so it must be accepted for the variable of age. Secondly, it will be noted that all of the relationships of age and distance have not been examined for the metrOpolitan - non—metropolitan areas. This has been done because this is an exploratory part of the data and it is not necessary for testing the hypothesis. In connection with this, the total number of cases is so small that any subdivision of the cases such as type of areas, gives a number which is erIrerely small and any relationships based on these smaller numbers is at best only an indication of what might be found with a larger n mber of cases. Third, the control value (median age of the base population) is very weak. To actually test what the effect of the base pOpu— lation has on the difference value, the base pOpulations would have to be standardized to some population distribution. The result would probably be about the same as indicated pre— viously. Finallg the relationships are of such a nature that another test over a larger number of cases selected from the entire United States would be desirable. TEE“ CF SOHO) t—d C3 .2 a ‘i p ”hile age has been the most stable characteristic in migration, many other demographic attributes of less stabili- .y are of at least equal value and interest, especially to urban planners and officials interested in their particu— lar cities. The data for yea‘s of school completed are reported for those persons twenty—five years and older in 1950. The hye potncsized principle is of particularly little value in the area of education. The relationship of distance and selec- tivity of years of school completed, as measured by the coefficient of correlation, is -.0533 (Appendix Table III). This value for practical purposes may be considered zero. When the control for median years of school completed by the base population is added, the relationship becomes .llOl (Appendix Table III). This is in the predicted direction but hardly large enough for serious consideration. This is surprising in view of the relation of distance and median education of base population, —.5960, (Appendix Table III), (i.e., the farther an area is from Detroit, the lower the median education level of the base pepul tion in the area). That selectivity tends to increase with a decrease in median education level of the base population may be inferred from the positive relationship of selectivity of education of mi- grants and median education of the base population, .233? (Appendix Table III). —6— Despite the above relationships of distance and salsa— tivity to the control value, they are only slightly related to each other. The control brings out the positive relat— ionship by affecting the variation in both variables with the result that they are more closely related. The partial Carrelation indicates that educational slectivity would be positively related to distance if all areas had the same median educational level in their base populations. Also of interest is the relationship of distance and educational selectivity for metropolitan and nonémetropoli- tan areas. For non-metropolitan areas this relationship may be considered zero, —.0695 (Appendix Table III). For metropolitan areas the relationship is high, —.6hOl (Appendix Table III). This indicates that the negative of the prin- ciple Operates in metrOpolitan areas, (i.e., the farther a metropolitan area is from Detroit the less the difference between migrants and their base populationsl It will be noted that education tends to be inversely related to the types of areas when compared with the relationship of age to types of areas. Certain considerations may be advanced as reasons for the hypothesis not being supported. The first of these is that the data are not of the same "pureness" as age data and therefore some other factor or factors must be controlled in order for the principle to emerge. One factor that may come to mind is the practice of the Census Bureau, in 1950, to allot the students to the area in which they attend school. -7- However, in view of the fact that the data are only for those persons twenty—five years old and over, the effect of this factor could be small. Another consideration may be that the relationship of pull and of level of education in the population is of such a nature that it does not show in the present data. It is conceivable that the pull for educated people lessens the farther from Detroit an area is. And it is possible that accompanying this lesser pull, is a lower level of educa— tional aspiration by the general population. This would be an especially appropriate proposal for the metropolitan areas. Similar propositions for the non-metrOpolitan areas are not evident, but they might be found upon detailed in— spection of the original data. Finally, it is possible that stream analysis would re— veal patterns not evident in this analysis. Because data for origin and destination of migrants are lacking, this proposal cannot be further eXplored here. The hypothesis must be rejected for the variable of years of school completed. SEX Sex ratios resemble age in having been one of the more stable variables in migration history. But unlike age and more like other variables, sex selectivity seems to bear a more complicated relationship to other variables. Whereas age selectivity seems to hold generally, other types of dem— ographic selectivity seem to hold only in more specific situ— -8- ations. To test the hypothesis, the percent of migrants that are male compared to the percent of the base population that is male, will be used rather than sex ratios. This value is simpler to compute and interpret and the general result is the same. No consideration will be given to Whether the migrant group has more or less percent of males than the base population. This facet of the data will be explored later. The percent difference refers to the people eighteen years and older in 1950. The hypothesis is not supported for the difference in percent of males in the migrant and base populations. The relationship is the reverse of that predicted, -.h065 (Ap— pendix Table III). This indicates that the farther an area is from Detroit, the more the percent of males in the base pOpulation resembles the percent of males in the mi- grant population. No control of variations in the base population will be used here. However, from an inspection of Table I in the Appendix, it will be seen that the preportion of males in the base pOpulation tends to increase with distance from Detroit. Therefore, control of this variable would probably reduce the negative relationship of distance and difference in percent male. With respect to types of area, the relationship is much higher for non-metropolitan areas than for metropolitan areas, (although it is negative in nature. The relationships for -9- non—metrOpolitan and metropolitan areas are, respectively, -.5297 and .0599 (Appendix Table III). It will be noted that the relationship of type of area and percent male is the same as for age and type of area (ignoring signs for tie moment). The relationship of selectivity of sex and distance may be further specified. The final column in Appendix Table II shows the absolute difference in percent males between the migrant and base populations. In only one of the areas, is the migrants' percent of males less than the percent of males in the base pOpulation. If this difference is ignored it can readily be seen that migrants tend to be over—repre- sentative of males and that this over—representativeness decreases with distance from Detroit. It should be noted that this is not saying that migrants tend to be predomin- antly males, but rather that there are more males in the migrant pOpulation than would be expected on the basis of their representation in the base population. While the sizable negative relationship of distance and percent males is interesting, the hypothesis has to be rejected for this variable. CHAPTER III Any decision to accept or reject a hypothesis is depen- dent on the acceptance of four major factors: 1) the statis— tical model used, 2) the measures used in operationalizing the concepts being valid, 3) the number of cases as being large enough to give confidence in the findings, and h) the data being a representative sample of the universe. In this study the statistical model seems appropriate. Chi square, because it makes no assumption about the distri- bution of the population, is utilized on the variables of age and sex. The hypothesis is stated in a manner that is amendable to treatment by correlation analysis. The Operationalizing of the concepts raises the question of using distance as a lone measure of dependence. In a larger study it might be desirable to combine distance with other factors such as the automobiles per capita, public transportation available and, if possible, some measure of communication with other areas. The method of measuring the effect of migration on a population has been discussed in Chapter One. The number of cases in this study is too small to give more than a tentative test of the hypothesis. However, it is large enough to indicate a similar study covering more cases would be worthwhile. The question of the sample being representative of a larger universe refers to the year the data were collected. -2- There is some indication that a slight business recession occured during the period the data covers. However, the effect of this economic condition has been judged to be slight.1 And it should be pointed out that this material is probably better than any other collected to date. ‘While the final decision in this study is to reject the specific hypothesis, a more general relationship appears that might be of value to examine. In the variables of age and y years of school completed, the high relationship (ignoring signs) between distance and the median values of the vari- ables fOr the base pOpulation, will be noted. This indi— cates that people are not distributed at random according to our measure of dominance. The high relationship of cer— tain types of areas and selectivity has already been noted. On the basis of these results it would appear that the eco— logical approach to the problem of distribution and movement of people may be fruitfully explored over a wider number of cases. With more cases, more confidence could be put in the results. Also the larger numbers of cases would permit fur— ther subdivision of the data, which in turn permits a more detailed analysis. It is also recommended that stream anal- ysis be used in this larger study. If the approach utilized in this study proves fruitless in a study similar to that sketched above, then this writer would be in favor of dropping the ecological approach to 1 Op. cit., Special Report, I-E, No. hB. p. 8. understanding migration. DATf NOT 1 ,3 . k; Omitted from the Appendix Tables are the original data tables used in computing the figures found in Table I. The writer felt that the expense and time involved in pre- paring the original tables for use here would not be justi— fied by the negligible value they would add. In comparison to Table I, the original tables would occupy rOUghly eight to ten times as much Space. The tables that would be added are discussed below. The first table required would Te the sunnaiy of county data to obtain 33: data. his operation was necessary be— cause the photostat sheets do not show figures for the total population in 1950. Also, the categories found in the census on county data are not comparable to the categories found in the photostat sheets. Therefore, collapsing of certain cate- gories was necessary, as well as the addition of the specific categories across counties. The second table required, would be composed of the fol— lowing columns: the total 1950 pOpulation; the outmigrants, l9h9—SO; the in-migrants, l9h9~503 the Abroad and Not Ascer— tained population; and theiresulting base population. Besides requiring more columns these new tables require more space (compared to Table I), because the original figures vary in size from figures in the thousands to figures in the millions. Inasmuch as this original data is not used directly, its in— clusion would add little to this study. If researchers are interested in this type of data, similar figures may be —/— ’3 obtained from Special Report P-E, N0. M3. TABLE I. - Populations Used in Computing Chi Square Differences And Difference in Percent Male. AGEa Area 1 ‘ Area 2 Area 3 ' Area h B .P. .B . 0 "72"."? .'" '1? ."13‘ B'."B"." ' ."B'": 13"". “If. 18-19 39 129 AA 126 39 118 LA 126 20-2h 91 206 97 2B9 87 198 90 21L 25-29 98 213 102 157 95 171 98 170 BO-Bh 10h 106 107 109 95 lhh 99 97 35-39 110 78 105 7h 100 72 101 89 uo—hh 99 80 98 61 96 6b 9b 67 us-su 162 77 165 99 16A 93 166 96 55-6b 153 59 1A3 69 155 58 150 77 65 ; 'L 53 139 56 169 82 15 6h Total 1,000 1,001d 1,000 1,000 1,0001L000 1,001 1,000 YEARS OF scu03L comers-.33 Elem. -5 15h A9 122 7b 85 55 90 AB 5-7 167 117 175 106 155 98 163 106 8 213 17h 255 231 27h 199 293 256 H.S. 1-3 167 159 167 193 166 219 171 209 h 193 258 182 225 193 23h 17h 219 0611.1—3 51 100 1 96 y 68 103 6B 80 h uh 130 33 53 AB 79 32 66 N.R. 11 12 15 21 16 13 13 21 Total 1,000 999 1,000 999 1,000 L000 1,000 1,000 ssxc Male 52.h 52.9 51.6 52.5 500.0 51.8 50.h 51.9 Female h7.6 h7.1 h8.h h7.5 SO .0 Oh8.2 h9.6 h8.l Total 100.0 100.0 100.0 100.0 100.0:b0.0 100.0 100.0 - These are the original data' 3 percent distribution multiplied by ten. B. P. stands for Ba se Population and M. P. for Mi grant Population. These are the orig;inal data' 8 percent at distribution. a — These values vary from 999 to 1,001 as a result of rounding. The fig— ures are not adjusted to equal l,OOO because the effect is neglible and because this paper is not prepared for the general public, which might find this discrepancy unacceptable. Cloc‘m I TABLE I. - (Continued) AGE Area 5 __ Area 6.._ Area 7 Area 8 5.2. T.B. B.P. n.P B.P. v55. 8 1. 1.2. 18—19 L5 116 ho 89 to 89 h9 68 20-28 107 281 102 '22 103 205 lhh 280 25—29 111 170 11h 19h 110 179 180 287 30—3u 112 123‘ 110 120 106 121 110 131 35-39 106 81 . 103 75 102 102 97 73 ho-uh 92 71 91 68 93 75 87 85 us—su 129 69 157 88 160 92 186 69 55-68 13 60 139 66 137 65 116 h6 65 7‘ 11.3 70 118 75 150 72 110 81 Total 999 1,001 1,000 999 1,001 1,000 999 1,000 YEARS OF 505001 ammsnm Elem. -5 79 56 65 69 A6 88 57 82 5-7 155 103 133 1h0 109 101 126 78 8 298 199 303 209 257 89 228 127 H.S. 1-3 168 190 178 169 220 227 178 129 u 182 207 200 197 227 219 201 179' Coll. 1-3 65 99 6h 89 7b 117 . 85 llh h an 116 L3 108 no 82 97 315 N.R. 13 29 1h 22 27 17 33 16 Total 1,000 999 1,000 999 1,000 1,000 1,001 1,000 SEX Tale 50.8 51.2 89.6 50.3 50.1 53.3 50.8 52.6 Female h9.6 h8.8 50.8 h9.7 h9.9 h6.7 h9.6 h7.h Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 ,1 TABLE I. - (Continued) AGE Area 9* Area A Area B Area C B”T. v.2 B.P. r.P. B.P. M.P. B.P. 8.}. 18-1 39 92 83 135 81 82 82 95 20-28 106 230 10 221 109 263 106 218 25-29 111 179 120 165 118 180 129 219 30—38 105 126 116 125 107 117 126 138 35-39 101 91 111 76 103 80 116 86 80—88 91 61 99 78 98 66 99 70 85-58 156 88 157 79 166 88 161 77 55-68 138 63 126 62 136 62 122 62 65 g 152 75 118 68 130 65 99 83 Total 999 1,001 999 1,001 1,000 999 1,000 1,000 YEARS OF SCHOOL firmer? Elem. -5 86 53 79 88 89 51 60 58 5-7 105 100 152 127 118 79 158 113 8 253 177 250 187 215 178 283 188 H.S. 1-3 217 192 209 170 219 197 239 288 8 281 288 205 282 228 283 195 225 Coll. 1—3 78 87 51 97 77 137 57 98 8 88 115 81 101 56 100 81 58 N.R. 17 32 13 28 21 18 10 21 Total 1,001 1,000 1,000 1,000 999 999 999 1,001 SEX Male 89.7 50.7 89.8 53.1 88.0 89.2 89.8 52.8 Female 50.3 ’4903 5006 11609 5290 5008 5006 14702 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 TABLE I. - (Continued) AGE Area D Area E Area G BOP. Mop. BOP. Mop. BoPo Mop. 18-19 85 105 58 62 38 86 20-28 118 216 156 272 120 285 25-29 127 186 137 276 118 197 30-38 112 117 106 123 p 108 125 35-39 108 87 95 66 102 88 80-88 108 72 88 88 97 61 85-58 177 93 151 60 163 95 55-68 118 58 113 52 129 55 65 % 91 69 99 80 125 89 Total 1,000 999 999 999 1,000 1,001 YEARS OF SCHOOL COMPLETED Elem. 25 II? 81 30 30 80 19 5-7 119 89 9O 63 103 58 8 227 190 200 132 216 200 H.S. 1—3 283 218 185 130 222 177 8 238 229 280 231 237 233 Coll. 1-3 75 110 130 115 86 128 80 101 113 281 80 166 N.R. 18 23 13 18 17 19 Total 999 1,001 1,001 1,000 1,001 1,000 SEX Male 89.6 87.1 89.1 89.6 88.8 89.8 Female 50.8 52.9 50.9 50.8 51.6 50.2 Total 100.0 100.0 100.0 100.0 100.0 100.0 ['3 ,1 [I TABLE II. - Values Used in Computing Correlation Coefficients X2 Value Absolute Of Years Median Years Difference Distance X2 Value Median of School of School Value in AREA Value-8 Of Age Age BoP. Completed Completed B13. .%.Male_ 1 187.5 873.9 82.95 382.1 8.88 0.5 2 115.5 680.8 82.30 116.8 8.80 0.9 3 76.5 582.5 88.25 116.8 8.95 1.8 8 68.5 507.6 83.60 119.1 8.88 1.5 5 32.5 888.1 81.05 216.8 8.91 0.8 6 57.0 378.2 81.70 131.2 9.00 0.7 7 29.5 317.2 82.10 97.8 10.20 3.2 8 15.5 382.1 37.95 589.5 10.53 2.2 9 39.5 389.2 82.10 136.0 10.33 1.0 A 28.0 82.8 80.05 192.8 9.27 3.7 B 89.0 828.5‘ 81.80 111.1 10.81 1.2 c 63.0 361.9 39.20 78.5 9.58 3.8 D 17.5 282.8 39.55 135.2 10.32 2.5 E 28.5 381.2 37.85 301.3 11.92 0.5 G 87.5 378.7 80.70 158.8 10.89 1.8 * In sixteenths of an inch. TABLE III. — Correlation Coefficients Years of School Difference Agg. Completed in % Males r .6658 -.0533 -.8005 ryz 05562 02337 rXZ 05985 ’05960 ryxoz .8995 .1101 Metropolitan ryz .2209 —.6801 .0599 NonAMetrOpolitan ryz .6667 -.O695 -.5297 y — dependent value x - independent value 2 - value y is adjusted with TABLE IV. — Differences in Distribution by Age Groups. ACE Area Area Area Area Area Area Area Area 1 2 3 8 5 6 7 8 18-19 90 82 79 82 71 89 89 19 20-28 115 152 111 128 138 122 102 96 25-29 115 55 76 72 59 80 69 187 30—38 2 2 89 -2 11 10 15 21 35-39 -32 -31 -28 -12 ~25 ~25 0 —28 80-88 ~19 ~37 ~32 -27 -2 -23 -18 ~82 85-58 ~85 -66 -71 ~70 -80 ~69 ~68 ~77 55-68 ~98 ~78 ~97 ~73 ~78 -73 ~72 ~70 65 ,1 -91 ~83 ~87 ~98 ~73 ~69 ~78 ~69 AGE Area Area Area Area Area Area Area 9 A B C D E 0 18-19 53 92 81 53 60 8 88 20-28 125 112 158 108 98 116 125 25-29 68 85 66 90 59 139 79 30-38 21 9 10 8 5 17 17 35-39 ~10 ~35 ~23 ~30 -21 -29 ~18 80-88 ~30 -25 ~28 ~29 -32 -80 -36 85-58 —72 ~78 -82 ~88 -88 ~91 ~68 55—68 ~75 ~68 ~78 ~60 ~68 ~61 ~78 65 7‘ ~77 ~58 ~65 ~56 -22 ~59 ~76 - indicates the number of migrants is smaller than the number of the base pepulation. ON TON AQON /’ U will ' as “,- . ——-~ —‘-4 Tina-,7 BAtr k ‘ i m. .<.' I --—- -- , 3 r ' ‘ “2 i x \ \ l F‘ 1 \x I \J‘w‘ ,r‘v--- Av) \ v ! .‘auAfiA ‘af-(R VI- "h-I"n Q .‘ 801.1 a. 8* ;, {3119+}: 29 Jun-31-nlgi fig)“ MA‘ “I 11.13301- eg... .