*<_°: c % : Census Tract Papers Series GE-40, No. 9 Social Indicators for Small Areas U.S. DEPARTMENT OF COMMERCE Social and Economic Statistics Administration BUREAU OF THE CENSUS Census Tract Papers Series GE-40, No. 9 Papers Presented at the Conference on Small-Area Statistics American Statistical Association Montreal, Canada August 14, 1972 Issued May 1973 / V \ rr U.S. DEPARTMENT OF COMMERCE Frederick B. Dent, Secretary Social and Economic Statistics Administration Edward D. Failor, Administrator BUREAU OF THE CENSUS, Vincent P. Barabba, Acting Director BUREAU OF THE CENSUS Vincent P. Barabba, Acting Director Robert L. Hagan, Deputy Director Paul R. Squires, Associate Director GEOGRAPHY DIVISION Morton A. Meyer, Chief Library of Congress Card No. A68-7769 SUGGESTED CITATION U.S. Bureau of the Census, Census Tract Papers, Series GE-40, No. 9, Social Indicators for Small Areas, presented at the Conference on Small-Area Statistics, American Statistical Association, Montreal, Canada, August 14, 1972. U.S. Government Printing Office, Washington, D.C., 1973. For sale by the Superintendent of Documents. U.S. Government Printing Office. Washington. DC 20402 Price: $1.00, domestic postpaid; 75 cents, GPO Bookstore. Stock Number 0324-00193 PREFACE The five papers published in this report were presented at the Small-Area Statistics Conference at Montreal, Canada on August 14, 1972 during two sessions of the annual meeting of the American Statistical Association (ASA). This conference was sponsored by the ASA Committee on Small-Area Statistics and the Social Statistics Section (ASA) with the program planned by Albert Mindlin, Chairman of the Committee. All papers deal with Social Indicators for Small Areas, a subject that has received increasing attention in recent years from statisticians and social scientists with the greatly expanded availability of small-area social statistics. Albert Mindlin chaired both morning and afternoon sessions and has provided an introductory statement. This report was organized and prepared under the direction of Robert C. Klove, Geographic Research Advisor, Office of the Associate Director for Statistical Standards and Methodology, Bureau of the Census. CONTENTS Page Introduction 1 Chairman: Albert Mindlin, ASA Committee on Small-Area Statistics Composite Social Indicators for Small Areas— Methodology and Results in Washington, D.C 3 Herbert Bixhorn and Albert Mindlin, Government of the District of Columbia Composite Social Indicators for Small Areas— Census Use Study— Recent Developments in Methodology and Uses 18 Samuel P. Korper, John C. Deshaies, Leo Schuerman, and Ronald Crellin, U.S. Bureau of the Census Application, Limitation, and Methodology of the Scientific Urban Matrix (SUM) (A Case Study of the City of Los Angeles) 25 Alan J. Weiner, President, Mitus Corporation, Los Angeles Indicators and Statistics: Issues in the Generation and the Use of Indicators 37 Harvey A. Garn and Michael J. Flax, The Urban Institute, Washington, D.C. Social Area Analysis: Procedures and Illustrative Applications Based Upon the Mental Health Demographic Profile System 50 Harold F. Goldsmith and Elizabeth L. Unger, National Institute of Mental Health INTRODUCTION Albert Mindlin, Chairman ASA Committee on Small-Area Statistics There is a long history of research efforts to develop typologies of cities and neighborhoods. Most of this history has been academic, although it has interfaced with market research, and to a much lesser extent public administration. More recently, and independently of this history, there has developed a large body of literature on "social indicators." Interest in social indicators has been stimulated, far more than the earlier history of urban typologies, by problems of effective public administra- tion of vast new social programs. It has been to a considerable extent a product of the urban revolution and various social issues, such as poverty and civil rights. These phenomena greatly accelerated in the sixties and are still in full motion. The expression "social indicators" is clearly an echo of "economic indicators." Economic indicators developed over many years as input to an economic accounts system which underpins much of the control and guidance of our national economy. It was hoped during the sixties that a social accounts system could be developed with analogous purpose, and with various "social indicators" as input. Such a system has not emerged, nor does one receiving general agreement appear likely in the foreseeable future. However, the movement to construct "social indicators" has developed a life of its own, and has absorbed much of the earlier research on urban typologies. The earlier literature on urban typologies tended to be dominated by sociologists; and more recent, literature on social indicators has been dominated by political scientists, sociologists and economists. There has been relatively little input from professional statisticians or of a practical, concrete character. However, such input is beginning to appear. The purpose of the 1972 ASA Conference on Small -Area Statistics was to bring together in a single extended presentation some of the best known statistical work now going in the field of social indicators for small areas. All of the work described in these proceedings include both theory and application down to concrete product. Korper, Deshaies, Schuerman, and Crellin, in Composite Social Indicators for Small Areas— Census Use Study— Recent Develop- ments in Methodology and Uses, provide an extended overview of the methodological work that the Census Bureau has been carrying on in this field for several years. The Bureau is drawing on a broad range of statistical tools, and the paper touches, at least briefly, on most of them. It may be noted that in the evolution of Census Bureau research it has moved from analysis of single series indicators to construction of composite indica- tors, and from ranking and related methods of classifying areas to cluster analysis. The Census Bureau paper describes its method- ological efforts to construct a valid trend model, a difficult problem. The afternoon session presented three papers. Weiner, in Application, Limitation, and Methodology of the Scientific Urban Matrix (SUM), describes a rather large scale effort in Los Angeles to create and utilize what he calls a Scientific Urban Matrix. He uses a variety of technics to create composite indicators— factor analysis, subjective grouping, cluster analysis. A unique feature of his methodology is to divide the city into "excellent," "good," "endangered," and "blighted" areas ac- cording to attitudinal perceptions on a variety of characteristics obtained from a sample of residents. He obtains a score for a composite indicator in a neighborhood by a modified summed ranks procedure, and then regresses these scores on the atti tudinal perception rating of the area. His purpose is to rate the various categories of community conditions on a scale of citizen perception of the quality of life in the neighborhood. The improvement of quality of life would focus on those categories that most change the citizens' perceptions. The use of citizen perception as a standard rather than "objective" criteria defined by professionals is a fascinating approach that merits further discussion. U o a. u Q •J The morning session was given over to two papers that describe work based on sophisticated statistical methodology. Bixhorn and Mindlin, in Composite Social Indicators for Small Areas— Methodology and Results in Washington, D.C., describe the effort to develop a practical on-going social indicator system for the District of Columbia. Using census tracts as a geographic unit, the paper focuses on the creation and use of composite indicators, i.e. index numbers based on several individual statistical series. The paper describes the line of reasoning by which individual series are brought together to define a com- posite indicator. It then describes and illustrates the method of principal components by which an index number is calculated from these separate series. It next describes and illustrates a method of cluster analysis which classifies census tracts on the basis of their index numbers. Finally it examines these clusters and illustrates uses in practical public adminiscration. Garn and Flax, in Indicators and Statistics: Issues in the Generation and the Use of Indicators, present a summarized description of the work of the Urban Institute. They review briefly some of the definitional and conceptual issues that have emerged in the literature, and then proceed to a short description of each of several reports prepared by the Urban Institute. The Institute's work for the most part avoids construction of composite indicators. It selects several functional areas of social concern. In each area it selects a single measure. For each measure it attempts comparisons between metropolitan areas, between city and suburbs, and over time in central city and suburbs. The Institute has made inter-city comparisons in greater detail in selected functional areas, in particular racial compari- sons, and comparisons for a variety of statistical series in education. In the latter case an effort has been made to distinguish between social condition variables, educational 1 SOCIAL INDICATORS FOR SMALL AREAS system input variables, and educational condition output vari- ables, and to correlate these, with some interesting results. Having clarified much of its methodology, the Institute is now directing attention to how to make data series immediately useful to program administration. The Goldsmith and Unger paper, Social Area Analysis: Procedures and Illustrative Applications Based Upon the Mental Health Demographic Profile System is essentially taxonomic rather than statistical, and in this sense is a continuation of a substantial body of social indicator literature. The authors examine 1970 census data, create a set of social dimensions, and then select census characteristics which they believe measure each dimension. For each statistical series they establish a set of intervals. They then classify census tracts into these intervals. Thus a census tract is classified on each of a large number of census variables. Since the variables have been grouped into sets that define a social dimension, each census tract can then be described in a profile sense along these social dimensions. The authors do not attempt to create composite indicators. Their taxonomy has been computer programmed and in this form is available to others who might wish to re-create it with their own census data. Thus, the Montreal Conference covered an array of statistical efforts to bring the social indicator movement to small areas, and also to transform it from theory to practical application. The efforts presented ranged from sociological taxonomy, essentially non-statistical, through increasing degrees of statistical method- ology to highly sophisticated technics— but all oriented directly to practical application by program administrators. COMPOSITE SOCIAL INDICATORS FOR SMALL AREAS— METHODOLOGY AND RESULTS IN WASHINGTON, D.C. Herbert Bixhorn and Albert Mindlin Government of the District of Columbia Introduction A large body of literature has developed on "social indica- tors." Interest in the subject is to a considerable extent a product of the urban revolution and various social issues, such as poverty and civil rights. These phenomena greatly accelerated in the sixties and are still in full motion. The expression "social indicators" is clearly an echo of "economic indicators." Eco- nomic indicators developed over many years as input to an eco- nomic accounts system which underpins much of the control and guidance of our national economy. It was hoped during the sixties that a social accounts system could be developed with analogous purpose, and with various social indicators as input. Such a system has not emerged nor does one receiving general agreement appear likely in the foreseeable future. However, the movement to construct social indicators has developed a life of its own. Several features of the present literature are noteworthy. First, it is dominated by political scientists, sociologists, and econo- mists. There has been very little input from professional statisticians. Second, much of it is of a theoretical, conceptual, or subjective taxonomic character with little direct applicability by public administrators. Third, much of it is national or at least macro-oriented. Relatively little is local or micro-oriented, i.e., dealing with cities or neighborhoods. The purpose of this paper is to fill some gaps in all three of these features. It presents the work being done by mathematical and other professional statisticians in the Government of the District of Columbia, oriented toward the city and its neighborhoods, and intended for direct policy and program application by city administrators. So far the work is very much in the "pilot" stage, but enough has been done to warrant public and professional exposure. The context in which we are working imposes an outlook and constraints. We are not seeking to break new theoretical ground, either conceptually or mathematically, nor are we seeking to develop data that do not currently exist on a neighborhood level. Using known mathematical and statistical methodologies and existing data, we are seeking to develop a social indicator model and system whose results are meaningful and usable for planning and management by local public agencies. We define a social indicator as follows: it is a quantitative measure of the quality of community life. Community life of course exists in many dimensions or functional areas— health, welfare, housing, etc. Consequently there must be more than one social indicator. Furthermore, rarely can a single quantitative variable constitute an adequate measure of the quality of life in any functional area. Almost always there are several quantitative variables measuring various aspects of that area. Indeed there are dozens of statistical series measuring aspects of community life. One can be inundated with statistics. One of the objectives of a social indicator system, as we perceive it, is to illuminate the underlying reality by identifying the separate quantitative vari- ables that measure aspects of a functional area, and then to bring these together rationally into a single composite indicator. Thus the thrust of our work is to develop meaningful and useful composite indicators. General Outline of a Local Social Indicator System The system has several objectives. In each dimension or functional area, it seeks to present a comprehensible profile of that dimension over neiahborhoods of the city; to determine the trend of the dimension over time; to illuminate geographic areas of needed governmental attention; to determine for a given neighborhood what social conditions seem more in need of attention than others relative to those conditions in other neighborhoods; and if possible to help in evaluating the impact of governmental programs. In order to accomplish these objectives, the following steps are currently followed: 1. Define specific social dimensions or functional areas of life in broad social fields such as health, welfare, housing, public safety, etc. 2. Determine the available statistical series that constitute quantitative measures of various aspects of the quality of life in each dimension. 3. Construct a methodology to bring these series together into a composite indicator for each dimension. 4. Delineate geo-areas of the city that are relatively homoge- neous with respect to the composite indicator. SOCIAL INDICATORS FOR SMALL AREAS 5. Conduct various analyses of the indicator, comparing it with other indicators in the same neighborhood, comparing neighborhoods, disaggregating it to evaluate its compo- nents, seeking insights heretofore unnoticed, and attempt- ing to draw inferences of policy, programmatic, managerial, and budgetary significance. 6. Develop a presentation package that will most effectively convey the subject to agency administrators. At the present time we have fairly fully developed points 1-4, but have taken only a few steps in points 5 and 6. In this paper we will present the general methodology first, and then illustrate it with actual data from the District of Columbia. Steps 1 and 2 Defining the Dimensions and Determining the Available Statistical Series These steps are inextricably intertwined. There are numerous classification schemes of dimensions of social life. Ideally a taxonomy should be constructed or justified by its relevance to sociological theory, or public administration, or some other intellectual framework quite independent of the availability of data. This has been attempted; the social indicator literature is filled with taxonomies. Regrettably the absence of available data makes almost all of them useless, regardless of other virtues or defects they may have. It simply is not practical to delineate dimensions of social life in order to construct social indicators without having available data play an integral part in the delineation; and indeed, every local social indicator effort that has sought practical development that the authors are aware of (such as efforts of the Census Bureau, Urban Institute, and others) have either taken the body of available data as their point of departure, or have determined data and taxonomy in unison. There are objective mathematical methods for separating a large number of statistical variables into groups of highly correlated variables. The best known of these is factor analysis. We rejected a mathematical approach at this stage for the following reasons: first, factor analysis often produces factors, or variable groupings, either that cannot be understood by adminis- trators ("Just what does this factor measure, as against that factor? What does it mean?") or cannot be used by administra- tors ("Well, your mathematics may tell you that variables a, b, c, d, and e 'hang together,' but that doesn't help me. There is no government program that deals with a and b, and c is not under my control; d and e are the only ones I can deal with"). Second, we want our system to be used. The best way to assure this is to build into it from the beginning the participation of planning and operating agency staff. We are proceeding as follows: starting with a very general, more or less arbitrary and fairly obvious classification (health, welfare, public safety, education, etc.) we conduct a series of meetings with operating officials in each of these fields. It is these officials who hammer out the dimensions of their field and the statistical series that go into each dimension to create a composite indicator. Their deliberations take place under the painful constraint of available data, and with the participation of the statisticians, data processors and other hangers-on. The comments and questions raised are such as: 1. "Is this classification meaningful in terms of administra- tion?" 2. "that's not a measurable variable" 3. "no information exists on that" 4. "yes, we have records on that variable, but they aren't mechanized. Do you have the resources to go through 50,000 records by hand every year?" 5. "that's only available from the census, so you can't make a time series" 6. "that's only available for the whole city, not by census tract, so you can't get a geo-distribution unless you add a coding operation" 7. "the quality of that file is so poor, you could prove black is white from it" 8. "that variable requires a cross-tabulation of two files, and they're incompatible" 9. "this is a community condition variable while that is a program workload variable. It may not be meaningful to put them together in the same composite indicator," etc. These meetings not only engage planning and operating officials in active planning of the system, they also are of considerable educational value to the officials. The officials sometimes lead themselves to recognition of the importance to their operations of statistical series generated elsewhere, and thus themselves create indicators that cross administrative lines. We have no a priori hang-up on taxonomy, but the officials themselves create a practical one. Also they learn some conse- quences of past decisions to mechanize or not to mechanize specific data items, and of the consequences of using geo-codes which are compatible with other files (census tract, city block, etc.); or of unique geo-codes unrentable to anything else (a significant feedback of the sessions we have held has been an increased appreciation of an integrated management information system), and gain insight into some of the types, uses, and possibilities of statistical analysis. Step 3 A Methodology to Construct a Composite Indicator We now have a defined dimension and several available statistical series to measure it. It could be argued that we have an operational definition— it is the series we use that quantitatively defines the dimension. In D.C. the census tracts are our geo-units. We have a value for each of several variables on each census tract for approximately the same time period. The variables are converted to rates, so that for each variable taken separately, the values are comparable from one census tract to another. At this point we cop-out of a theoretically important issue. Should weights be assigned to the variables, reflecting differential importance in the composite? Differential importance is a matter of social values and public policy. It is not a matter to be decided by statisticians. We laid this matter before the program administra- tors, with a result surprising in the event, not so surprising although in hindsight. They refused to pass judgment on the relative social importance of the variables. For example, one functional area hammered out by health officials was "health conditions surrounding birth." The variables they settled on were birth weight under 5Y2 lbs., infant mortality, no or inadequate prenatal care, out of wedlock, and age of mother under 20 years. The administrators refused to rank these in any way in social or health importance. So the variables entered the mathematical maelstrom weightless, all treated equally. However, the mathe- matics we use creates automatic weights. These weights reflect METHODOLOGY AND RESULTS IN D.C. the degree of correlation between the variables. A variable which correlates poorly with the others plays a lesser role in the composite, and may even be rejected. This point will be elaborated later. How to bring the variables together into a single composite? Perhaps the simplest way is to rank each variable over the census tracts and then add the ranks. This "summed ranks" method has decided advantages, such as simplicity of calculation and easy comprehension by non-technicians. But it has a monumental flaw: it throws away too much important information. For a given variable its unique geo-distribution, where it piles up, where it stretches out, is lost; the ability to perceive natural groupings of tracts is lost; poorly correlated variables are not detected and can produce anomalies. Some of these disadvantages can be mitigated by modifying the method, for example converting the values to standard scores within each variable and then summing the standard scores. But modification still leaves too much to be desired. We rejected this method and all modifications of it that we tried It is obvious to think of straightforward multiple regression. This, however, cannot be applied because it requires specification of a dependent variable. There is no dependent variable in this work. However, multiple regression has many desirable prop- erties, and it leads us to the method that we currently employ to create an index number representing a composite indicator— the method of principal components. The easiest way to explain this method nontechnically is to describe it graphically in 2- dimensional space. Consider figure 1. There are two variables, x 1 , x 2 . The points are census tracts. In ordinary regression x 2 is denoted the dependent variable, and the regression line is so chosen as to minimize the sum of squared deviations from the points to the line along the x 2 axis. In the method of principal components the line is so chosen that it minimizes the sum of squared deviations along the perpendicular from the points to the line. Thus variables x, , x 2 are treated equally. There is no dependent variable. SOCIAL INDICATORS FOR SMALL AREAS Figure 1 ordinary regression principal components Figure 2 dispersion along x/ axis dispersion along x/ axis dispersion along x^ axis dispersion along x? axis METHODOLOGY AND RESULTS IN D.C. Consider figure 2. The method first standardizes the variable x-x . This makes all variables directly comparable with each other. °x It also places the origin at x, , * 2 . The method then rotates the original x, , x 2 axes into other axes denoted by x,, x 2 . The formula for each of these new axes is a linear combination of the original variables; that is, x' = ax, + bx.,, x' = ex, + dx.. Each a ' '1 1 2 2 1 2 axis in this new coordinate system is called a "principal component." The regression line which minimizes the perpendic- ular distance from points to line is the x, axis. It is called the "first principal component." The x' 2 axis is called the "second principal component." There is a dispersion of the points along each axis, measured by the variance along that axis. Thus associated with the x, axis is a "first principal component variance"; associated with the x 2 axis is a "second principal component variance." The sum of the variances is called the "total variation of the points." Compare the two graphs in figure 2. It is clear that in the top graph the first principal component variance constitutes by far the largest proportion of the total variation of the points; while in the bottom graph the two variances contribute about equal proportions to the total variation of the points. In the top graph the first principal component "explains" most of the total variation in the points. In the bottom graph the first principal component "explains" much less of the total variation. All of this can be generalized to n dimensions, with n principal components. (See [1] and [2].) We now apply the method of principal components to the construction of a composite social indicator. The formula for the combination of the separate variables into an index number is the equation of the first principal component. If there were only two separate statistical series in the composite, the indicator could be shown graphically as in figure 2. For each point, i.e., each census tract, the index number would be the value of x, : x', = ax, +• bx 2 . It should be noted that this definition of the composite indicator, i.e., this index number, is more meaningful if the first principal component accounts for a high proportion of the total variation of the points (i.e., the points are strung out along the first principal component axis much more than along any other principal component axis), and less meaningful if the first principal component does not account for a high proportion of the total variation. Consider again figure 2. Clearly the desired condition obtains when the variables are highly correlated, and drains away as the correlation between the variables decreases. This conforms to intuition. If several variables are highly correlated, it is reasonable to infer that their combination is measuring some meaningful underlying reality, each variable making a contribution to the index number. But if some variable does not correlate highly with the others then it is reflecting some other underlying functional area. Its presence in this composite distorts the index number and makes it hard to evaluate and less useful in analysis. The actual mathematical expressions of the coefficients, or weights, attached to each variable in the first principal component formula reflect, in an indirect way, the correlation of that variable with the other variables. In practice we adopt the following rule of thumb: we drop a variable from the composite if (a) its weight (i.e., its coefficient) is less than one-half the maximum weight; or (b) the first principal component takes up less than 65 percent of the total variation of the points. Dropping a variable under this rule results in a reduced composite that usually is mathematically satisfactory. This has considerable significance in analysis. Since the subject experts placed that variable with the others, dropping it is equivalent to pointing out something special about it of programmatic significance. This feature is one of the advantages of the principal component method. It will be illustrated later with D.C. data. Step 4 Delineating Geo-areas of the City that are Relatively Homogeneous with Respect to the Composite Indicator We now have a composite indicator with an index number for every census tract. We wish to see how this index number classifies the tracts over the city— i.e., we wish to cluster the tracts so that those that are relatively homogeneous fall into the same cluster. In the first instance we could merely order the tracts by the index number and arbitrarily cut them into, say quintiles or deciles. This has the advantage of simplicity, but it has the disadvantage of utter arbitrariness. What is magical about five groups, or ten— or four or seven or any other number? What is magical about having the same number of tracts in each cluster? Furthermore the range of values in each group may be very dissimilar. For example, in the top group the lowest index number may be only 5 percent less than the highest index number, while in the next group the difference may be, say 30 percent. The means of the two groups might actually be very close together, and there may actually be more within-variability than between-variability. In our judgment ordered cuts into an arbitrary number of groups with an equal number of tracts in each has little intrinsic merit. Alternately, we could plot the index number on a scale and look for "natural gaps." There are two problems here. First, there may not be any obvious "gaps." The plot may look like this: xxxxxxxxxxxxx Second, "natural gaps" can be deceptive. For example, suppose the plot looks like this: -xxx — xxxxxxxxxxx l l A Visually, there would appear to be two "natural" groups— the left three, and the remaining eleven. But in fact tract 4 is closer to tract 1 than it is to tract 14, and is closer to the mean of the left group than it is to the mean of the right group. So "natural gaps" can produce anomalies. We have searched for a more objective way to delineate the city by means of the index number. We believe we have found it in the mathematics of cluster analysis. There are many types of cluster analysis. We employ a technique proposed by Rubin and Friedman [3], pri- marily because it has a simple and precise criterion for clustering the tracts which in addition has several desirable properties. The technique has been reduced to a canned computer routine which prints out on the side a number of secondary results, such as all of the two-way correlations between the variables, and symbolic plots of the tracts along a line according to their index numbers [4] . The technique may be simply explained as follows: suppose there are six census tracts altogether, each having a composite indicator index number. We wish to group the tracts into two clusters such that the tracts in each cluster are relatively homogeneous with respect to their index numbers— that is, they are more like each other than they are like the tracts in the other cluster. N tracts identified by an index number and ordered from low number to high number can be separated into two ordered groups in N-1 ways. For six tracts, let x, be the lowest index 8 SOCIAL INDICATORS FOR SMALL AREAS number and x 6 the highest. Then the tracts can be grouped into < < X 1 X 2 X 3' X 4 X 5 X 6 ) ' (x, x„x,x. , x c x,), (x, x.,x,x„x c , x,). Which is the best cluster- 1234 5o 1 z 3 4 3 o ing? The Rubin-Friedman paper first standardizes the variables, as does principal components. It then sets up the following criterion for selecting the "best" clustering. Denote the total variance of the standardized index numbers over all six tracts by T. For a single clustering, e.g., (x t x 2 , x 3 x 4 x 5 x 6 ), there are two "within" variances, one for each cluster. Denote the sum of these two "within" variances by W. Form the ratio X- This ratio is W called the criterion measure. If six tracts are to be divided into two clusters, there are five such ratios. If heterogeneous tracts are placed in the same cluster, W will be large and, since T is a constant, _J_ will be small. The technique examines all five ratios W and selects the clustering with the largest ratio, i.e., it finds the grouping that maximizes the criterion measure. The best cluster- ing may be e.g., (XjX 2 , x 3 x 4 x 5 x 6 ) or (XjX 2 x 3 , x 4 x 5 x 6 ). The number of tracts in each cluster under the best clustering may vary. This is one advantage of the technique— it does not force an equal number of tracts in each cluster. The technique is not restricted to a single dimension, i.e., a single value for each tract, although we currently employ it this way. It is a general technique. It could be employed even if we had not first used the method of principal components to reduce all of the variables to a single index number. If each tract is represented by several values, then the criterion measure becomes the ratio of two covariance determinants T_. Thus, this method W could be used independently to cluster areas of the city with respect to several variables considered simultaneously. Indeed probably that is its more common use. The user must specify in advance the number of clusters to be created. If he wishes two, the technique finds the "best" two as defined by the criterion measure. It also yields the value of the criterion measure for its "best" selection. If the user wishes three clusters, the technique finds the "best" three and yields the value of the criterion measure. As the number of clusters goes up, the value of the criterion measure also goes up. Hence we use the technique iteratively, calculating each time the difference between successive criterion measure values. We stop the itera- tion when the difference markedly decreases. To cluster beyond that does not seem justified, for further clustering does not appear to increase homogeneity substantially. This ability to indicate an optimum number of clusters is another advantage of the technique. As stated, this technique can be employed to cluster the tracts when each tract has several values, such as the value of the separate variables making up the composite indicator, and there are distinct advantages to doing this. There was a definite line of reasoning that led us to employ principal components first, and then to employ cluster analysis on the single index number so obtained, which is a simplified use of the technique. The cluster analysis technique by itself clusters the tracts but does not yield an actual numerical value of the composite indicator, i.e., an index number. Having an index number to represent the composite indicator is valuable for a variety of purposes, e.g., to see its trend over time, to compare the value for a single tract with the value for the whole city, and other purposes. However, if the cluster analysis is made first and the principal component technique applied independently afterwards, seeming contradic- tions between the results of the techniques can and do appear. In particular, let us assume that low index numbers are possessed by the "best" tracts. The cluster analysis used independently may place a tract in the "best" cluster, while principal components may give that tract an index number that is higher than the index number of a tract that the cluster analysis places in a "poorer" cluster. This is because the two methodologies are different, and an off-beat collection of values in a tract could produce this anomaly. The anomaly cannot occur if the cluster analysis is applied to the single index number produced by the principal component technique. Application of the Methodology to Census Tract Data in the District of Columbia We now show the application of the above methodology to the creation of social indicators in the District of Columbia. We held several meetings with health officials in our Department of Human Resources, during which they hammered out the com- posite indicators that seemed most useful to them as public health administrators. After much discussion, a consensus was reached that indicators based on age seemed the closest related to the general classification of their programs. Hence they arrived at the following five indicators: conditions surrounding birth, childhood, adolescence, maturity, and old age. For each indicator they arrived at a consensus as to the most useful statistical series among those available— i.e., existing in mechanized form on a census tract level and reasonably recurrent. For birth and old age the series consisted entirely of data developed by health officials. For the other ages the officials recognized the importance of some non-health variables. For pilot project purposes we have concentrated on the indicator, "health conditions surrounding birth." As mentioned earlier, this indicator initially contained five variables: a. births to mothers under age 20 b. no or inadequate prenatal care c. birthweight under 5 1 /2lbs. d. births out of wedlock e. infant mortality Discussions with other public officials produced several other composite indicators that became part of the pilot project- dependency, housing, and general socio-economic conditions. The initial variables in each are shown in table 1. Table 1. Composite Indicators 1 . Health Conditions Surrounding Birth a. Births to mothers under age 20 b. No or inadequate prenatal care c. Birthweight under 5V2 lbs. d. Births out of wedlock e. Infant mortality Rates: The denominator for all variables is number of live births 2. Dependency a. Persons under age 18 on Aid to Families with Dependent Children (AFDC) b. Persons on Old Age Assistance (OAA) METHODOLOGY AND RESULTS IN D.C. c. Persons on General Public Assistance (GPA), Aid to the Blind (AB), and Aid to the Permanently and Totally Disabled (APTD) d. Persons receiving food stamps e. Persons registered for medicaid, and persons on D.C. medical care Denominators for: a. total persons under age 18 b. total persons age 65+ c. total persons age 18-64 d. total population e. total population 3. Housing a. Overcrowded households b. Incomplete plumbing c. Low rent d. Value of owner-occupied one-unit structures e. Vacancies Denominators for: a. occupied housing units b. total housing units c. renter-occupied units d. owner-occupied one-unit structures e. all housing units 4. General Socio-economic Conditions a. Overcrowded households b. Incomplete plumbing c. Median family income d. Matriarchy index (female-headed families with children under age 1 8 -s- families with children under age 1 8) Denominators for: a. occupied housing units b. all housing units Indicators 1 and 2 include only locally produced data which can be updated at least annually. Indicators 3 and 4 contain only 1970 census data. Each of the indicators was put through steps 3 and 4. We shall illustrate the results of the procedure with the composite indicator "Health Conditions Surrounding Birth: 1969." Princi- pal components produced the following first principal com- ponent, i.e., the expression that provided the initial index number: x; = .52x x + .47x 2 + .38x 3 + .54x 4 + .28x s Xj = Births to mothers under age 20 x 2 = No or inadequate prenatal care x 3 = Birthweight under 5 1 /2lbs. x 4 = Births out of wedlock x = Infant mortality The proportion of total variation "explained" by this expression was 59 percent. Applying the rule discussed in step 3 for acceptance or rejection of variables, we rejected "infant mortal- ity." The lack of close relationship between infant mortality and the other variables was confirmed by the matrix of 2-way correlations, shown in table 2. Table 2. Health Conditions Surrounding Birth: 1969 Pearsonian Correlation Matrix 67 .40 .87 .27 1 .39 .68 .21 1 .48 .34 1 .32 1 For definitions of variables see text. Eliminating "infant mortality" and rerunning the other variables produced the following first principal component expression: x'j = .54x t + .50x 2 + .38x 3 + .56x 4 The proportion of total variation explained by this expression was 70 percent. We refer to this as the "reduced composite." The census tract index numbers produced by the expression above were not used directly. They were first converted into numbers ranging from 1 to 146, the number of noninstitutional census tracts in 1970. Standardizing the range of values this way expedites understanding and analysis by making each index number of each indicator easily relatable to the range of tracts and allowing comparison of index numbers for a single tract. The index numbers of the "reduced composite" were then put through step 4. The criterion measure values for various numbers of clusters were as follows: 2 clusters 3 clusters 4 clusters Criterion measure 1.16 2.24 2.79 Difference 1.08 .55 The optimum number was taken to be three clusters. The cluster analysis method does not inherently make a judgment that one cluster of tracts is "better" than another. The mathematics simply seeks to put together tracts that are "like each other" when all of the variables are treated collectively. For the purpose of public policy the tracts are judged "better" or "worse" on each variable, e.g., tracts with a low proportion of "birthweight under 5 1 / 2 lbs." are "better" than tracts with a high proportion. All of the variables are structured to run in the same direction, i.e., low values are "better" than high values. Thus, the clusters are designated "B(best)," "M(medium)," "W(worst)." This does not mean that the value of all variables in all tracts in the "best" cluster are lower than those in the "medium" cluster. 10 SOCIAL INDICATORS FOR SMALL AREAS The clustering is based on the index number. It puts together tracts that are most like each other in the composite. The irregularities of the number could easily produce a tract in the "best" cluster with a value on some variable that is worse than some values of that variable in the "medium" cluster. But taking all of the variables in that tract together, it is more like tracts in the "best" cluster than tracts in the "medium" cluster. The mean values of the variables in each cluster in "health conditions surrounding birth: 1969," are shown in table 3. Table 3. Health Conditions Surrounding Birth: 1969 Mean Values in Each Cluster Reduced Composite Xj Births to mothers under age 20 x 2 No or inadequate prenatal care x 3 Birthweight under B'/ilbs. x 4 Births out of wedlock Jest Medium Worst 6.4 23.5 38.6 9.1 24.2 33.3 8.3 12.8 16.6 10.1 32.5 55.3 infant mortality correlates poorly with other health condi- tions surrounding birth. Whatever its role historically, in D.C. today it no longer appears to distinguish good from bad neighborhoods with respect to other health conditions sur- rounding birth. This can be seen clearly in Appendix, figure 2a. This is a cluster map of the birth health composite indicator, 1969. The colors represent the distribution of best, medium, and worst tracts based on the reduced composite. The patterns represent the distribution for infant mortality. It is clear that infant mortality is distributed somewhat haphazardly with respect to the composite indicator. Thus the analysis suggests that judging the health of a community by infant mortality and using infant mortality as a critical variable in directing birth health programs over the city are not for the most part supportable. A side comment: The central mathematical statisticians discovered the low relevance of infant mortality, but it was the research and statistics group in the Health Services Administration that recognized the significance of this discovery. This was a good example of the assertion that effective use of a social indicator system requires the team participation of statisticians and subject experts. Nonsubject oriented statisticians do not have the expertise to draw from the system all that it can yield. The subject experts must be actively involved. Number of tracts On this basis tracts are labeled B, M, W. 30 66 50 The analysis of rejected variables is important. Hence "infant mortality" was separately put through the same cluster analysis to form its own clusters. The "best" and other clusters here are not of course composed of the identical tracts as the reduced composite. The total procedure summarized in the paragraphs above was followed for all four of the composite indicators in the pilot project. Step 5 Using the Indicators We are at an early stage of community analysis by means of composite social indicators. We have not yet developed an integrated analytic package. Our studies so far have been largely restricted to preliminary results of the cluster analysis. We have not yet explored the uses of the index numbers. Consequently this paper does not include the index numbers. We present here some initial findings and paths we are exploring. 1 . Rejected Variables b. Incomplete plumbing— This was rejected from the housing indicator. As with infant mortality, it has a low correlation with other variables in its indicator. However, unlike infant mortality in which high rates are dispersed through the city, high rates of incomplete plumbing are concentrated in a few tracts. This is clearly seen in figure 3a. The vast bulk of census tracts fall into the best cluster on incomplete plumbing. Only three tracts are classified as worst in incomplete plumbing, all in the heart of the inner city. c. Matriarchy index— This was rejected from the general socio-economic indicator. It is similar to infant mortality in being spread through the city more haphazardly than the other variables in its indicator, although less markedly than infant mortality. However, a different phenomenon appears to be at work here— the failure of nonspecialized intuition. The socio-economic indicator is the only one which was put together by the central statisticians alone. It was constructed from selected items from the 1970 census. We assumed, without testing, that the selected items "hang together" geographically as typical city ecology. We were wrong. The two-way Pearsonian correlation mat-'* of these items is shown in table 4. Table 4. Socio-Economic Composite Indicator (1970 Census) a. Infant mortality— This was rejected from the composite health indicator. For many years health administrators, faced with a mountain of health data and a chaotic statistical picture, have accepted infant mortality as a key general measure of community health, using it to judge that one neighborhood is better than another in community health and to point to geo-areas that require special health program emphasis. Results of the composite health indicator challenge this classic role of infant mortality. The analysis shows that Pearsonian Correlation Matrix a b Overcrowded households ... a Incomplete plumbing b Median family income c Matriarchy index d 1 16 -.69 -.11 1 -.31 -.04 1 -.16 1 METHODOLOGY AND RESULTS IN D.C. 11 The poor correlation of matriarchy index with the other variables is clear. However, the other variables also correlate poorly with each other. The rejected variable was simply the one whose correlations were poorest. The only reasonably good correlation is between family income and overcrowding. The remaining correlations involve incomplete plumbing. As indicated earlier, incomplete plumbing is a highly polarized variable which apparently correlates poorly with other com- munity variables. Furthermore, overcrowded households as the single best measure of social deterioration may not be as tenable a generalization as many believe. 2. Some Inferences Suggested by the Cluster Maps We now turn to the reduced composite indicators, whose clusters are shown by colors on the maps. We take up the maps in turn, comparing the census tracts. However, the discussion inevitably will involve comparing maps, i.e., com- paring different indicators for a given census tract. Since this paper is being written for our own administrators as well as for presentation here, certain details of primarily intramural interest will be pointed out. The large numbers on the maps refer to the nine "service areas" (SA) into which the city has been divided by the District Government. SA 6 is the "model cities" area, the heart of the inner city. a. Figure 2a, Health conditions at birth, 1969 First we note certain "islands." In SA 5, ct 66 is a best tract surrounded by poor and worst tracts. This is the heart of the Capitol Hill renovation effort. It would appear that this effort has had some success. In SA 9, ct 54.2 is downtown. It has a small population and few births. In SA 4, ct's 76.2 and 76.3 are of more programmatic signifi- cance. These have always been good residential areas, but the increasing threat to them of deteriorating health-at- birth is clear. Census tract 76.1, which has been a good residential area in the past has deteriorated, and infant mortality, the rejected variable, is already getting a grip on 76.2. If we look at these tracts on figure 5a, dependency, 3a, housing, and 4a, socio-economic, we see that they have relatively few welfare families and other dependency characteristics, and that their general socio-economic condition is still good, but that their housing, while still not bad, is going downhill. These tracts have a good prognosis for improvement because of their seeming basic strengths, provided preventive public action is focused on them. But without public attention they will probably go the way of their neighbors. Let us pursue further the effort to find possible areas of favorable prognosis if public attention is focused. In SA 5, ct's 79.2 and 68.3 are bad in the health-at-birth indicator. However, the general socio-economic condition (fig. 4a) and housing (fig. 3a) are fair, and they have little dependency (fig. 5a). Why is the health indicator so bad? The other indicators show strength. Special attention seems in order. Another such area is SA 9, ct 54.1. This tract also is bad in health-at-birth. Yet it has good housing (3a), fair socio-economic condition (4a), and relatively little dependency (5a). Certainly on the suggestive evi- dence of the maps, ct 54.1 has the strength to benefit substantially from concentrated official attention. The reverse of this type of search can also be seen. Note SA 3, ct's 78.1, 78.4, 78.7, 78.8. In the far northeast section of Washington these are worst in health-at-birth. Prognosis for improvement does not seem good according to the maps, for they are already being overwhelmed in other dimensions, even though their general socio- economic condition is fair. The matriarchy index is very high there (4a); bad housing exists and has already overwhelmed ct 78.3 (3a) and dependency has made inroads (5a). However, this inference is only suggestive and should lead to closer analysis. There is considerable public housing in ct 78.1 and 78.4 which may account for the poor showing of the other indicators in those tracts. If the poor health condition can be identified to public housing occupants to a considerable extent, then ameliorative steps may be more feasible and the prognosis improved. Certain ct's stand out as "peninsulas" that may be the "cutting edge" of serious deterioration, e.g., SA 1, ct 24. This is worst in health at-birth, and seems to be the herald of the expanding deterioration behind it. But it has relatively low infant mortality and is in fair shape on the other indicators. Can concerted public attention to this area hold back the advancing deterioration? Another area in a similar position is SA 7, ct 87. One must beware of possible artifacts. For example, SA 8, ct 4 shows best on the reduced composite (and best on all other indicators) but worst on infant mortality. However, in this tract there were only 24 births through- out 1968 and 1969, and two cases of infant mortality. This is a very high rate of infant mortality on a very small base. It may be statistically unimportant— or it may indicate a serious bad health pocket in an otherwise excellent tract. The same remarks apply to SA 1, ct 26, andSA9, ct57.1. b. Figure 3a, Housing The first thing to bear in mind about this indicator is that the reduced composite has no variable that directly measures the physical condition of housing. All of the variables, including the rejected one, are 1970 census variables. The rejected variable, incomplete plumbing, is the only one dealing directly with physical characteristics. The remaining variables are overcrowding, low rent, vacancy rate, and value of owner-occupied one-unit struc- tures. They are variables usually associated with geographic areas of poor housing, but are not themselves direct measures of that. This caution is appropriate because the thing that strikes the eye of a person familiar with Washington is the concentration of "worst" housing in SA 4, which contrasts with a considerably better showing of that area on the other indicators. Some of the tracts in SA 4 have a good deal of public housing as table 5 indicates. Almost all of the tracts with substantial public housing are worst on the reduced housing indicator. Does this reflect primarily low rent? Or does it also reflect severe overcrowding? Abnor- mal vacancy rate? Comparing tracts with and without public housing in the same general areas, table 5 suggests that there is indeed excessive overcrowding and generally above-average vacancy rate in public housing tracts. The usefulness of comparing indicators is indicated by SA 4, ct 76.2. On the housing map this tract is incon- spicuous. But we saw earlier that, compared to the other indicators, housing seems to be going downhill. Public attention might save this area. In contrast, a tract that appears about to be overwhelmed is SA 7, ct 29. Indeed, 12 SOCIAL INDICATORS FOR SMALL AREAS the housing map is dismaying when compared to the other maps, for it shows the advance of the demographic aspects of bad housing in almost every direction from the core of the inner city, SA 6. It may be noted that almost the entire inner city core, SA 6, is in the worst cluster in housing. c. Figure 5a, Dependency The medium and worst clusters on dependency are clearly not as widespread as on the other indicators. Outside the inner city core, the worst areas tend to be areas with substantial public housing as might be expected, e.g., SA 4, ct's 74.1, 74.4, 97; SA 5, ct's 71, 72; SA 9, ct 60.2. An anomaly seems to be SA 7, ct 36 which is a medium tract almost surrounded by worst tracts. Its social characteristics are generally believed to be not markedly different from its neighbors. This should be investigated. Whatever the explanation, the socially dependent popula- tion is clearly moving up this general corridor. Table 5. Some Aspects of Housing in Selected Census Tracts: 1970 SA CT Percent population in public housing Reduced Percent Vacancy housing over- rate indicator crowded percent 73.4 74.1 74.2 74.4 74.5 75.1 75.2 97 98 78.1 78.3 78.4 78.5 78.7 78.8 70 71 72 60.2 49 8 82 39 46 29 39 46 51 28 33 88 99 W W w w M w M W w w w w M M W M W W w 23 31 13 35 18 22 13 37 29 26 19 20 22 20 27 11 22 28 54 8 9 3 6 3 7 3 13 6 5 3 7 3 2 6 7 7 12 W = worst M = medium These notes are intended to indicate some of the ways that cluster analysis can be used to alert administrators to what is going on in the city in various functional areas. At the present stage of analysis they are hypotheses suggested by the cluster maps. We don't know if the hypotheses are correct, but they illuminate the paths of deeper analysis and possible administra- tive action. Further analysis will involve the index numbers, which provide more precise information, as well as disaggregation of the variables to determine whether certain variables are playing heavier roles than others. There are other uses to which the indicators can be put. For example, the index numbers and the clusters provide rational stratification devices for sample surveys, and similarity groupings for experimental— control experimental designs. In the D.C. Government each service area has its own inter-agency committee, which concentrates on governmental services in the service area of its responsibility. The indicators can be used by these committees seeking program guidance in their own service area. We are only on the threshold of a social indicator system for neighborhoods. We expect to develop more indicators, for example, in public safety and education. We expect to develop time trends (there are method- ological problems here). We expect to improve the methodology. For example, we believe that there is a conceptual difference between program variables, such as those that make up the dependency indicator, and community condition variables such as those that make up the housing indicator; this should be explored. We hope to develop indicators based entirely on locally produced and recurrent data in order to construct useful time trends. From another perspective, the system is impelling a deeper use of the data by operating staff than has occurred in the past. A final note of re-emphasis. Such a system cannot be effectively developed and used by the statisticians alone. They don't know enough. It must be embedded in developmental and analytic teamwork of statisticians, subject experts, and data processors; and it must be used in government planning. Its success as an operation of local government will depend ultimately on how useful it proves to be for concrete govern- mental action. REFERENCES Anderson, T. W., An Introduction to Multivariate Statistical Analysis, John Wiley and Sons, Inc., 1958 Morrison, D. F., Multivariate Statistical Methods, McGraw Hill, Inc., 1967 Rubin, J. and Friedman, H. P., "On Some Invariant Criteria for Grouping Data," Journal of the American Statistical Association, December 1967 Rubin, J. and Friedman, H. P., "A Cluster Analysis and Taxonomy System for Grouping and Classifying Data," IBM Program Library, August 1967 APPENDIX FIGURES 2a. Health Conditions at Birth: 1969 3a. Housing: 1970 4a. Socio-economic Conditions: 1970 5a. Dependency: 1971 2b. Health Conditions at Birth: 1969 3b. Housing Indicator Group Means: 1970 4b. Socio-economic Variables Group Means: 1970 5b. Dependency Indicator Group Means: 1971 Figs. 1a and 1b not referred to and not included in this paper. [1] [2] [3] [4] METHODOLOGY AND RESULTS IN D.C. 13 14 SOCIAL INDICATORS FOR SMALL AREAS ^^T *& r re \ m ^ r CO \ 00 H 3 j?y/\< s -£ • v r~ OS U_ -!^\ * ^ ^_it»T /^>v in 00 "Wr CO \m> r~ O S 8 °\ 00 ^ CT) (D CD ■ — r -J *-> '■5 o u c 0) > O) R) CO > (J c en CU m +- 1 to c 1_ ? (0 r i 1 0) (1) CJ 5 ■■' > O > (0 ' O _l < > METHODOLOGY AND RESULTS IN D.C. 15 CD ■p CD 5 o to E L. c o 11 n h- 75 c 3 > wding ete PI Famil 4-J -1 is t/3 o o. c CO L. Overc Incorr Media D E 16 SOCIAL INDICATORS FOR SMALL AREAS □ SOCIAL INDICATORS FOR SMALL AREAS Figures 17 2b. Health Conditions at Birth: 1969 (Group means in percent ) Best Medium Worst City- wide No or Inadequate Prenatal 6.4 9.1 8.3 10.1 28 1.7 41 23.5 24.2 12.8 32.5 54 3.4 73 38.6 33.3 16.6 55.3 40 6.4 8 26.0 25.0 Birth Weight Under 13.4 36.7 122 NUMBER OF TRACTS 3.0 122 OPM - Statistical Systems Pilot Project 4b. Selected Socio-Economic Variables Group Means: 1970 (Grou 5 means in percent J Best Medium Worst Citywide Overcrowding 3.5 1.0 $16,300 40 13.6 36 14.5 2.1 $8,400 84 30.4 82 24.7 7.1 $5,900 22 45.6 28 12 2 Incomplete Median family NUMBER OF TRACTS Matriarchy index NUMBER OF TRACTS 2.3 $9,600 146 30.1 146 OPM - Statistical Systems Pilot Project 3b. Housing Indicator Group Means: 1970 (Group means in percent) Best Medium Worst Citywide Overcrowding. . . . 3.5 12.7 22.2 12.2 4.2 16.9 39.0 19.9 Average Value of Owner-Occu- pied One-Hous- ing Unit Struc- $37 , 000 2.0 $20,300 3.7 $17,400 7.0 $26,600 4.1 Vacancy Rate . . . , NUMBER OF TRACTS 41 55 50 146 Incomplete Plumbs 1.4 6.2 21.6 2.3 NUMBER OF TRACTS 121 22 3 146 OPM - Statistical Systems Pilot Project 5b. Dependency Indicator Group Means: 1971 (Group means in percent) Aid to families with dependent children Old Age Assistance General Public Assistance; aid to blind; disabled Food stamps Medicaid and D.C. Medical Care NUMBER OF TRACTS OPM - Statistical Systems Pilot Project Best 7.0 1.4 0.8 3.1 8.9 66 Medium 26. 1 7.8 3.0 15.6 34.4 59 Worst 42.8 16.5 7.2 29.4 57.1 21 City- wide 22.0 4.7 2.3 11.3 25.0 146 COMPOSITE SOCIAL INDICATORS FOR SMALL AREAS-CENSUS USE STUDY RECENT DEVELOPMENTS IN METHODOLOGY AND USES Samuel P. Korper, John C. Deshaies, Leo Schuerman, and Ronald Crellin U.S. Bureau of the Census Introduction To lend proper perspective to this paper, let us begin by briefly indicating the general Census Use Study program areas and developments we would like to cover. The earliest of these, the New Haven Health Information System, was begun in 1967 under the dual sponsorship of the Bureau of the Census' Census Use Study and the Maternal and Child Health Division of the Connecticut State Department of Health, with major support from the then Children's Bureau of the Department of Health, Education, and Welfare. Since its early formative years (1967-69), the Health Information System program has been evolving and expanding in a number of ways. First, the subject matter which was originally focused on maternal and child health has been expanded to not only include many other components of health but also other subject matter areas such as welfare, education resource economics and social pathology. Second, there has been a movement from a focus on small-area analysis of a one-time data base to one which could be tracked over time to discern changes in health and social status, the level of delivery of health and social services, the level of community resources, and so forth. The third expansion represents the analysis we have been undertaking in composite social indicators. Our emphasis has always been away from one-dimensional social indicators. Even during the early New Haven days, we were working on composite indicators,- and in those early days, many of the methods we worked with were hit and miss. Later, we shall discuss refinements in the methodology dealing with com- posite indicators. Fourth, is the continuing development of technology by the Census Use Study which has been utilized by the Health Information System program. We believe that new technologi- cal developments can not be overemphasized. Preliminary to active concern about "social indicators" or "composite methods," there is a vast range of processing problems and details that must be dealt with. These include problems of data access, preparation and summarization; but most important, and often most vexing, is the problem of geocoding data to small areas (the issue of geographic base files) and ways and means to disaggregate local data files so that they are useful and meaningful for analyses. The fifth area of expansion is inherent in the fact that our efforts are no longer limited to New Haven. Under a new program the Census Use Study has undertaken for the Office of Economic Opportunity, we are currently transferring our methods and technology to eight new sites— both urban and rural. Social and resource indicators programs are being implemented in Los Angeles, Chicago, Atlanta, Phoenix, Providence, R.I., and Mound Bayou, Miss. Initial forays are also being made into Lowndes County, Ala. and Baldwin, Mich. A basic thrust of the Health Information System program, and one emphasized throughout this report of our activities, is the effort directed toward developing local data files. While we would be denying our heritage by failing to recognize the necessity for utilizing census data, we feel it is unwise to become too dependent on this resource. Census data are only available once every 10 years. Local data are generally to be had on a continuous basis. The problem, which we in no way intend to minimize, of course, is ferreting them out, and making them useful input to our analyses. New Haven Health Information System The major elements, tasks and results of the New Haven Information System program may now be briefly indicated. The fundamental ingredients of the program are: 1. The geographic base file (a computerized source map containing a listing of all possible address ranges and geocodes pertaining to these address ranges. A geographic base file, or the DIME (an acronym for Dual Independent Map Encoding) system, is imperative for supporting the local and census data contained in an information system). 2. Address Matching Computer Programs (a series of computer programs that will break down an unformatted address from a local data file into a number of components, screen a geographic base file to locate these components, and retrieve from the geographic base files the geocodes pertaining to that street address). 3. Data pertaining to public health and related phenomena that could be linked into a system of information through geocoding. 18 CENSUS USE STUDY 19 The original data set was derived from five sources: census information consisting of the results of the dress rehearsal census in New Haven (April 1967), both 100 percent and sample data; the vital records system; a special purpose health survey; and hospital obstetrical records obtained from the Yale-New Haven Hospital. Having these ingredients, a number of subsequent tasks were required in the design of the health information system. These were: days on service, hospital inpatient and outpatient days, admission unit, first treatment unit, interval since last admission, alcohol problems, drug problems, reason for discharge, recidivism, and diagnostic classification. The mental retardation registry includes small-area data on the distribution of retardation by level of retardation. (There are two retardation levels of interest. The first level might better be labeled cultural deprivation since the retardation is not innate. The second level is severe retardation.) 1. to geocode all records in each of the five data files to block groups and census tracts of the City of New Haven. 2. to derive more than 300 data summaries or social indicators for each block group and census tract in the City of New Haven. 3. to delineate neighborhoods (on the basis of scaling and ranking procedures) ascertaining that the characteristics of the population contained therein are relatively homo- geneous. 4. to examine so-called "high risk" health characteristics within and between neighborhoods. 5. to perform multivariate analyses to determine relation- ships between sets of variables and previously formulated research hypotheses. These tasks have been completed with a 1967 data base and are documented in Census Use Study Report No. 12. 1 We have organized a data base in New Haven for 1970 so that comparative time series (temporal trends) analyses may be performed. The 1970 data base which is presently undergoing analysis in New Haven will consist of the following data files: 1. Census Summary Tapes— containing both head count data and sample census data. Social indicators, representing various dimensions of demographic, socio-economic status, migration, and the like, will be derived from these data sets. The emphasis will be on deriving social indicators previously used with the 1967 data base. 2. The Vital Records System— will serve as the source for deriving indicators of fertility, illegitimacy, deaths, and other vital events. The emphasis is again placed on deriving indicators previously used with the 1967 data base. 3. Other sources of information are contained in the 1970 data base which were not available in 1967. These will consist of additional data summaries obtained from agencies such as the Connecticut Mental Health Center, and the mental retardation registry made available through the Connecticut State Office of Mental Retardation. The Mental Health Center data summaries include social and economic information concerning the patient population of the mental health center. Additionally, information will be available for analysis on specific health variables such as 1 U.S. Bureau of the Census, Report No. 12. In addition to these files, the 1970 information base includes further sources which are seen as useful adjuncts to our analysis, and which, it is felt, contribute significantly to a broadly based health planning effort. These are: a. fire department data, including variables such as fire calls and emergency service calls, with further differen- tiations into health problem areas such as seizures, maternity, suicide, drug overdose, etc. b. circuit court statistics, to include such variables as drug offenses, family disruption, and possible relation- ships to mental disorder. c. juvenile court statistics, including drug offenses, cases related to conduct, and several indicators of the need for the appropriate health resources. Several additional files including those of police, transporta- tion, family relocation, and housing code violation authorities are currently being processed for inclusion in the analysis. The processes for analyzing the 1970 data base, with one major revision and these several expansions, is patterned on the 1967 methodology, as documented in Census Use Study Report No. 1 2. The data files containing the obviously sizable number of social indicators will be structured through correlation and factor analysis. However, a major departure from the earlier procedures is that neighborhood delineation will be attempted through cluster analysis rather than the composite scaling method used in the 1967 analysis. The analysis is designed to do two basic things. First, a comparative temporal analysis is being undertaken to identify the changes taking place between 1967 and 1970. The hope is to identify multi-dimensional or configurational changes; the assumption being that one-dimensional social indicators inade- quately measure a dysfunctional situation. From 1967 data, some 20 research hypotheses were formulated and will be reconstructed from 1970 data. The rationale for reconstructing research hypotheses with 1970 data is to test the hypotheses over time, and to determine to what extent, if any, they continue to hold. Another aspect of this comparative analysis will be to examine specific neighborhoods to determine the direction of change, if any, of dysfunctional trends originally discerned in the 1967 analysis. The second thrust of the analysis is to formulate new sets of hypotheses that can be tested on a subsequent data base. Exemplifying the actual and ongoing projects related to this overall work plan is that we have developed in close cooperation with the Connecticut Mental Health Center which serves the South Central Connecticut area. This study is an areal analysis of the utilization of mental health and related services in New Haven, exploring a conceptual area which may best be termed "expressed demand." 20 SOCIAL INDICATORS FOR SMALL AREAS It is obvious that a major goal of the community mental health centers movement is the provision of services that are relevant to the needs of the population served, but the determination of the extent to which this aim is realized is no easy task. While general agreement on the best combination of strategies for the evaluation of community mental health programs has yet to be achieved, there is general agreement that an essential component of evaluation is that of "diagnosis" of the community. To this end, something has to be known about the characteristics of the population at risk (potential consumers of services) and of populations currently being served by existing agencies (current consumers of services) as indicators of demand. Clusters of variables (approximately 300) associated generally with socio- economic class are being analyzed as the best available predictors not only of case outcome, but also as pre-hospitalization determinants of who receives attention, the source of treatment, psychiatric diagnosis, and eventual reentry into treatment. The preliminary analytical components of this study are: (1) the delineation of New Haven neighborhoods (block groups or census tracts) on the basis of factor analytic, scaling, and ranking procedures, and (2) ascertaining that the characteristics of the populations contained therein are relatively homogeneous. The second analytical thrust of the study will be the development and distillation of data from 10 agencies which reflect a wealth of health and social information beyond that available from, but not limited to, census-type data. Two approaches are taken to further explicate the measurement and operationalization of the term "utilization" for this study. First is the recognition and analysis and description of the a priori separation of services which may be termed specifically mental health as opposed to those that are not. The second major approach is the development of "categories" of service utilized, through the identification and designation of common clusters or modules of service. The third emphasis of this study is the assessment of utilization of mental health services and facilities by defined population group from small areas and relating their help-seeking behavior to the utilization of other services which call for or reflect need of psychiatric intervention. The emphasis, then, of this and other studies undertaken in conjunction with the New Haven Health Information System, will be on analyzing non-census data because of the likelihood that the next census will not take place until 1980. What is needed in the interim, therefore, is an information system not completely dependent on census data, hence our conscious emphasis on local records and data sources. An additional important aspect of the analysis will be its capacity for replication in several additional sites throughout the country. Such activities are reviewed below, emphazing research into composite social indicator development and the programmatic role such research plays. Social and Resource Indicators Program As we indicated earlier in this paper, the New Haven Health Information System has long since expanded beyond New Haven and is no longer concerned solely with health or analyses utilizing a one-time data base. The O.E.O. Social and Resource Indicators Program alluded to earlier is now in the data collection stages in Los Angeles; Mound Bayou, Miss.; Atlanta; Providence, R.l.;and Phoenix. Basically, the program is being undertaken in three stages. Stage I consists of identifying and assessing data sources and files, access problems, and estimating necessary initial operations and costs to ready the data to be collected for input to the indicator system to be developed in Stage II. Stage III involves the use of operational data of the O.E.O. Health Program in conjunction with the indicator system to assess certain implications of these programs to the populations they service. Stage III also entails structuring the indicator system to be turned over to local groups to be responsible for operating and using the indicator system on a regular, continually updated basis from 1973 into the future. The study design at the present stage of development entails tapping numerous local data files to obtain health, social, and resource indicators for small geographic areas on an annual basis from 1965 through 1970. Without going into detail, the indicator matrix will include, but not be limited to, data summarized from the vital record system; reportable disease registries; public health clinics; health and allied health manpower; facilities such as hospitals, nursing homes, and extended care facilities; financial data on health department operations; welfare data on the five categorical programs; medicaid and food stamps; mental health; education; employ- ment; public and private housing; taxation; land use and access; and deviant behavior— crime and juvenile delinquency statistics. Basically, the idea is to start with a master matrix that is composed of what would obviously be an unwieldy number of cells dealing with a broad range of subject matter areas and ending up with a structured indicator matrix reduced through multi-variate analytic approaches (such as regression and discrimi- nant analysis) to fewer and more meaningful composite social indicators that are sufficiently viable to be monitored overtime. Through a "before-and-after" research design our intention is to establish this indicator system on an on-going basis; tie it into structured operational data to be provided by O.E.O. health centers or health networks; and finally, to select control areas without health programs and attempt to discern what influence, if any, these health programs have or may exert on the health and social status, delivery of health and social services, and level of resources of the communities they service. The composite indicators, as was the situation with earlier work in New Haven, will consist primarily of logical statements of interrelationships between a number of variables in the system derived and examined through multi-variate analyses. In addition to the hypothetical generalizations we have already mentioned in connection with the early work in New Haven, we will be concerned with a large number of issues which are of some importance to the O.E.O. health programs. We would like to present here two examples of issues we intend to examine in this program. The first concerns health status. We will formulate such questions as: What combination of factors effected by the O.E.O. health programs contribute in what ways to improving the health status of populations served by those programs? Can we define the effects of the health programs themselves? Does health status change over time? If so, what other characteristics associated with health status also change? For example, if an improvement in health status on a certain category or morbidity is discernible, is or was there also a concomitant change in such areas as level of delivery and quality of health services, social pathology, social status, and so forth? Second, what combinations of factors effected by the O.E.O. health programs contribute (and in what manner) to more efficient delivery of health and social services? Example: Medi- caid, which pays for many of the health services received by the poor, is tied inextricably to the welfare system. In some places, a CENSUS USE STUDY 21 prerequisite to the receipt of medicaid benefits is that the person or family be first determined eligible to receive welfare benefits under one of the categorical programs. However, there are many barriers placed in the path of a potentially eligible person who applies for welfare. Unless he is completely disabled, he must usually apply in person to a welfare office which may be difficult to reach. Once he gets to the welfare office, he must fill in complicated application forms which he may have difficulty in completing. Then he must go through an investigation, a series of interviews and a waiting period before the determination is finally made. Once the determination is made, it is subject to review once or twice a year. The point we want to make here is that one must look at a far-reaching complex of issues to begin to investiage a question such as whether medicaid is, or is not, providing benefits which characterize efficient delivery of health services to the poor. O.E.O. health programs can effect this component of the health delivery system in a number of ways, including, as in this case, the assisting of recipients through the cumbersome welfare process. Through the positing of an hypothesis, or, if you will, composite indicators which aid in the modelling of this whole process, we can begin to assess this aspect of delivery of health services. In the analysis of this situation over time, we must attempt to discern the changes in all the components of the configuration. We have indicated two examples of the hypotheses to be formulated with the information base being developed by the Census Use/O. E.O. indicators study. There are, obviously, many more dealing not only with health status and delivery of health services but also with related issues such as improved "levels of living," attrition in the educational system, access to recreational and social services, resource allocation in the community, mental health and deviant behavior, and a plethora of other issues which affect personal and public well-being. At this point in time, we have more questions than we have answers. However, this is what the research and development of composite social indicators is all about. Before any meaningful quantum leaps are forthcoming, it is, we feel, necessary to go through this process of research design and hypothesis formula- tion, attempting to sharpen the definition of intended output as well as underlying theoretical constructs. Obviously crucial in this endeavor is the derivation of a reasonably acceptable definition and interpretation of terms such as "health" and "social indicator," to mention but two. While having reiterated and vowed allegience to such awesome definitions of health as those held by agencies such as the World Health Organization, operationalization dictated a more limited (read realistic) defini- tion. To this end, sights were lowered, and analyses were refocused on "intermediate outputs" and approaches to examina- tion of causal linkages between programmatic inputs and output in terms of measurable health states or outcomes (prenatal visits and birth weight). These quantified states could never have been approached utilizing the original broad definitions of preferred health status at once so moving and elusive. Without reduction to and construction of testable theory, meaningful evaluation becomes impossible. The development of social indicators and the mysticism associated thereto presents a problem certainly as complex. Varying interpretations were assessed and often found wanting. It is obviously too simplistic, for example, to suggest that measurable change in social indicators (given their selection) should result merely from health program impact, without simultaneously controlling for a host of social and economic variables. Reasonable estimation of the effects of the health programs on health states must be normalized for social and economic variables both cross sectionally and prospectively in order to have sound evaluative or analytic purpose. The Los Angeles study provides an example of Census Use Study approaches to these concerns. Composite Indicators of Change— the Los Angeles Case To place the Los Angeles study in perspective, it is worthwhile to briefly recapitulate the framework for the social indicators program research design. In outline, it includes the following broad steps: 1. Obtain and combine numerous local administrative files on an annual basis from 1965 to the present; 2. Generate relevant social and health indicators for small geographic sub-areas; 3. Formulate research questions and hypotheses focused on community health; 4. Develop an analytical perspective that includes a "before-and-after" design around the emergence of O.E.O. health programs, as well as examine the concomitant relationship between program service, social condition and health status; 5. Compare the community service and health trends in the O.E.O. service areas to appropriate control areas without health programs. Through a current arrangement with the Los Angeles County Health Care Department, the plan is to assist in the establishment of a countywide system which will continue the monitoring of these data once the initial stages are complete. In Los Angeles, many of the identified relevant local agency data files, such as those alluded to earlier, are available from 1965 to the present. The files are currently being prepared to cover a common universe of territory corresponding to three designated O.E.O. service areas, with the basic areal unit of analysis being the census tract. For the three Los Angeles centers, there are approximately 200 census tracts identified and data from various governmental, social and voluntary agencies are being aggregated for these sub-areas. In the main, we are interested in the health center service areas as a whole, but realizing the existence of an uneven texture within these areas, it becomes important to indentify the changing trends and "hills and valleys" of health and health related characteristics in the areal cohort. Such changes are typically approached by examining some rate or ratio of difference (e.g. x 2 /x' — 1 x 100 = "percent change") for each derived index, for each year being analyzed. These change measures are often displayed in tables or in figures either as a single line or as a multiple of lines, each line representing a different item over time. The figures can, of course, also be displayed in the form of maps drawn by hand, or by computer graphic techniques. The consequence of the relationships between variables might be analyzed in the form of a descriptive discussion or more elegantly presented in statistical summaries. 22 SOCIAL INDICATORS FOR SMALL AREAS With regard to the former, the diversity in number of variables, cases and temporal observations involved for the three Los Angeles service areas often makes such an approach voluminous, confusing and, at best, a judgmental summary of the relation- ships observed. On the other hand, typically derived rates of change mentioned above are often processed utilizing statistical summary measures. When this is done, many troublesome statistical issues are raised, such as "auto-correlation" and "part-whole correlation" influences. Also, variable universes, when extended over time, may be confounded by problems of measurement which include variable decay, inflation, loss of timeliness, comparative reliability, and so on. The analytical scheme to be implemented in the Los Angeles O.E.O. indicators project is being designed to overcome, insofar as possible, many of the issues that are conceptually, and, ultimately, statistically prevalent in the analysis of multi- dimensional indicators that extend through time. In the main, the procedures described below avoid the creation of ad hoc change variables from the original indexes gathered on a timely basis. Rather, the individual or composite measures are generated out of variant themes using standard regression and correlation techniques. This approach has particular relevance to our study of small areas because it permits the use of various parametric statistics that describe the "fit" and variance of selected indicators. After the formalization of statements that describe the relationship between indicators reflecting the condition of such variables as community health, social structure, and health services, it is important to obtain measures of ratio of change and the concomitant relationship concerning the velocity of a sub-area's rate of change. For example, where specific health services increased (or decreased) rapidly over time, there would be an association with accelerated changes in community health and social well-being. This is significant in understanding the association between dramatic or subtle shifts in the patterns of health indicators and rapidity of change for health services. The development of composite indicators for the O.E.O. health service areas has two facets of multi-dimensionality: (1) each index extends through time and is measured on an annual basis; and (2) a composite of indexes can be a conceptually and statistically related set of independent variables to a target or dependent index. 2 The first type of "composite" is, thus, a single item extended through time; and a composite value for each geographic sub-area is not only useful by itself, but properly constructed, it can be used to determine the degree of relative change over time for each areal cohorts' progression or deteriora- tion on the single indicator. Furthermore, it would seem 2 The statistical arguments presented in the re- mainder of the paper are derived in part from the following sources: Otis D. Duncan and Ray P. Cuzzort, "Regional Differentiation and Socio- economic Change," Papers and Proceedings of the Regional Science Association^ IV (1958); Otis Dudley Duncan, Ray P. Cuzzort and Beverly Duncan, Statistical Geography Glencoe, 111.: The Free Press, 1961; George C. Myers, "Variations in Urban Population Structure," Demography, I (1964), 156- 163; Maurice D. Van Arsdol and Leo A. Schuerman, "Redistribution and Assimilation of Ethnic Popula- tions: The Los Angeles Case," Demography, VIII (November 1971), 459-480. important to know what segments of the distribution change more or less rapidly over the period of observation. These measures can be obtained on an "inter-annual" basis through the extraction of multiple regression residuals, summarized by regression coefficients, and examined for fit by standard error of estimates and multiple coefficient of determination approaches. Inter-annual regressions are calculated for the same variable over two or more points in time. For example, the inter-annual regression for data between 1965 and 1968 is represented as: x 4 = a 4.1 23 + b 41 .23 x 1 + b 42.1 3 X 2 + b 43.1 2 X 3 where x.| = 1965, x 2 = 1966, x 3 = 1967, X4 = 1968 The individual residuals for each observation (residual = x^ — x ntn ) can be interpreted as an inter -annual composite measure of change relative to the universe of cases extended through time for a particular variable. Furthermore, since the same variable for more than one point in time is projected on a hyperplane, a net regression coefficient of unity (i.e. ') is interpreted to mean that any change that took place from time one to time two was, on the average, equal along the entire distribution after all other time measures have been excluded. (A special case is a bivariate distribution along an X and Y axis represented by a linear slope.) It would also follow that coefficients' less than unity indicate that geographic sub-areas low at the earlier time increased, or decreased— whichever is the case— faster between the respective time points than did the higher initial values in the distribution. A similar but reciprocal interpretation is made for coefficients that exceed unity. Thus, there now exists a descriptive summary of the relative average degree of change (in terms of place and degree for an areal cohort for each period of observation). Complimentary to the inter-annual regression coefficient is the inter-annual multiple coefficient of determination. Here we can summarize the degree of stability (or change) for a given indicator. The higher the coefficient of determination, the more relatively stable the particular health or social phenomena are over time. Finally, standard error of estimates gauge the overall variance that might be expected for the specific indicator through time. Thus far, the measures and statistical summaries have centered around a time-related "composite index" for a single variable. Yet, earlier it was suggested that to assess the social or health status of a community, it would be advantageous to develop component measures that isolate the concomitant relationship between the velocity of the rate of change in social or health composition for small areas and the rate of change in, say, neighborhood health services. One procedure that measures the relationship between rates of change for a composite of independent variables to a dependent variable is to calculate coefficients of "multiple deviational" change. This type of correlation provides a statistical description of the degree of association of the velocity of change in each respective measure of health with rates of change for a composite of selected community social or health indicators. Such a statistical procedure also provides a means whereby the residuals can be used as indicators of individual case deviation (away from a relative position) to the universe of cases extended over time. Specifically, the multiple deviational correlation is a multiple correlation wherein both the dependent and independent vari- ables are the residuals of least squares multiple "inter-annual" regressions. Thus the data used in the multiple correlation are CENSUS USE STUDY 23 essentially obtained by first calculating multiple inter-annual regressions based on data with "n" time points of observation for each of the variables to be used in the multiple deviational correlation. To review the inter-annual procedure above, each needed inter-annual regression uses the latest set of nth observa- tions as the dependent variable and each preceding temporal set of cases are treated in turn as independent variables. The deviation of the final temporal observation for each variable from its expected values (X') are the data used in the multiple correlation of deviational change. Therefore, the degree of fit is calculated utilizing these sets of transgenerated forms of the raw data called multiple deviational change measures. The multiple deviational coefficient of determination is particularly suited for the present problem because it reflects the velocity of concomi- tant change between times of observation after "removing" the influence of "natural" relative stability, or change, for each selected variable. Furthermore (and similar to the logic in the case of residuals obtained from calculating inter-annual regres- sions), residuals obtained from deviational regressions can now be viewed as relative measures that are a composite of both multi-dimensional social indicators as well as accounting for these indicators extending through time. Summary real world of local area data files and all that that implies, has stripped much of this work of its glamour, replacing it instead with the basic concerns and compromises necessary to "get on with it." Generally, we are vexed by rather primordial questions such as choice of component indicators and the most appropriate methods of combining them. Working with such large data sets which represents the efforts, programs and services of so many agencies requires vigilance in the avoidance of cross- contamination and almost certain tautological traps. Despite resistance to premature conceptual and theoretical closure, we have found that while it is often quite fruitful and/or popular to use several different indicators to represent a single underlying concept, that in addition to the combined use of multiple indicators, there is still much room for the recording and observation of individual associations between each indicators of the independent and dependent concepts under study. Finally, we feel that it is not an essential characteristic of our indices that they be free from conceptual or operational defects or limita- tions. Rather, we concur with Bickner, who has said "...all that is necessary for such indices is that they be: reasonably interesting, objectively measurable, commonly reported, and reducible to a common number not exceeding three significant digits." 3 We have tried to present an overview of Census Use Study research and programmatic incorporation of the concepts related to composite social indicators. Various levels of grappling with this concept have been evidenced throughout this work, ex- tending from the relatively fundamental work of data acquisition to the more heady reaches of conceptualization and theory testing. Operationalization on a day-to-day basis, working in the 3 Bickner, Robert E., "Measurements and Indices of Health/' Outcomes Conference I - II, Methodology of Ide ntifying, Measuring and Evaluating Outcomes of Health Ser vice Programs, Systems and~Sub- Systems, Carl E. Hopkins, ed., National Center for Health Services Research and Development, Depart- ment of Health, Education, and Welfare, 1969. 24 SOCIAL INDICATORS FOR SMALL AREAS Selected Bibliography Ball, G.H. and Hall, D.J. Applications of Cluster Analysis to Bureau of the Census Data, Menlo Park, California, Stan- ford Research Institute, 1970. Bauer, Raymond A., ed. Social Indicators. Boston: Massachusetts Institute of Technology Press, 1966. Bickner, Robert E. "Measurements and Indices of Health," Outcome Conference I -I I, Methodology of Identifying, Measuring and Evaluating Outcomes of Health Service Programs, Systems and Sub-Systems, Carl E. Hopkins, ed., National Center for Health Services Research and Develop- ment, DHEW, 1969. Bisco, R.L. Data Bases, Computers and the Social Sciences. New York: J. Wiley and Sons, 1970. Bodian, C, Gardner, E.A., Willis, EM. and Bahn, A.K. Socio- economic Indicators from Census Tract Data Related to Rates of Mental Illness. U.S. Department of Commerce, Bureau of the Census, Washington, D.C. Crellin, Ronald. "Los Angeles County Data Files Geocoded with ADMATCH and Their Applications." American Institute of Planning Annual Conference, St. Paul, Minn., October 1970. Deshaies, John C. and Korper, Samuel P. "The Use of Census Data and Local Files for Health Planning." Presented to the Urban Regional Information System Annual Conference, Louisville, Ky. September 5, 1970. Deshaies, John C. and Seidman, David R. "Health Information Systems," Socio- Economic Planning Sciences, Vol. 5, pp. 515-533 (1971) Duncan, Otis D., Cuzzort, R., and Duncan, B. Statistical Geography, the Free Press, Glencoe, III., 1961. Ezekial, Mordecai and Fox, K. Methods of Correlation and Regression Analysis, London: Wiley and Sons, 1967. Greenberg, B.G. "Evaluation of Social Programs," Review of Int. Stat. Institute, Vol. 36, No. 3, 260-278, 1968. Jackson, Alton F. and Curtis, R.F. "Conceptualization and Measurement in the Study of Social Stratification," in Blalock, H.M. and Blalock, A.B., eds., Methodology in Social Research, New York: McGraw-Hill Book Co., 1968. Myers, George C, "Variations in Urban Population Structure" Demography, I, 1964. National Institute of Mental Health (Redick, R.W., Goldsmith, H.F. and Unger, E.L.) 1970 Census Data Used to Indicate Areas with Different Potentials for Mental Health and Related Problems, USPHS, Publ. No. 2171, Washington, D.C. Government Printing Office, 1971 . Reinke, W.A., and Baker, T.D. "Measuring Effects of Demo- graphic Variables on Health Services Utilization" Health Service Report, 2:61-75. Spring, 1967. Tischler, G.L., Henisz, J., Myers, J.K. and Garrison, V.K. The Use of Mental Health Services by Catchmented and Non- Catchmented Populations. Mimeographed report for NIMH, 1971. U.S. Bureau of the Census. Census Use Study. Deshaies, Korper and Siker, Health Information System II, Report No. 12, Washington, D.C. 1971. U.S. Bureau of the Census. Deshaies, Health Information System Report No. 7, Washington, D.C. 1969. U.S. Bureau of the Census. Census Use Study: "ADMATCH: An Address Matching System," Washington, D.C. U.S. Bureau of the Census. Census Use Study: "DIME: A Geographic Base File System," Washington, D.C. U.S. Bureau of the Census. Census Use Study: "GRIDS: A Computer Mapping System," Washington, D.C. U.S. Bureau of the Census, Wallach, Harold and Deshaies, John D. "Development of Social and Health Indicators to Profile and Monitor Populations Served by the Office of Economic Opportunity Health Programs, Atlanta, Ga.— Stage ("Wash- ington, D.C. 1972. U.S. Bureau of the Census, Deshaies, John C. and Wallach, Harold C. "Development of Social and Health Indicators to Profile and Monitor Populations Served by the Office of Economic Opportunity Health Programs, Chicago, III.— Stage I" Washington, D.C. 1972. U.S. Bureau of the Census, Deshaies, John C. and Korper, Samuel P. "Development of Social and Health Indicators to Profile and Monitor Populations Served by the Office of Economic Opportunity Health Programs, Mound Bayou, Miss. -Stage I" Washington, D.C. 1972. U.S. Bureau of the Census, Deshaies, J.C., Humphrey, A.B., Korper, S.P. and Wallach, H.C. "Development of Social and Health Indicators to Profile and Monitor Populations Served by the Office of Economic Opportunity Health Programs, Phoenix, Ariz. -Stage I" Washington, D.C. June 1972. U.S. Bureau of the Census, Korper, Samuel P., Deshaies, John C, and Organic, Harold N. "Development of Social and Health Indicators to Profile and Monitor Populations Served by the Office of Economic Opportunity Health Programs, Provi- dence, R.I .-Stage I" Washington, D.C. June 30, 1972. Van Arsdol, Maurice D. and Schuerman, L. "Redistribution and Assimilation of Ethnic Populations: the Los Angeles Case," Demography, VIII, 1971. APPLICATION, LIMITATION, AND METHODOLOGY OF SCIENTIFIC URBAN MATRIX (SUM) (A Case Study of the City of Los Angeles) Alan J. Weiner, *President Mitus Corporation, Los Angeles Context Due to the vigorous competition for limited resources betwaen the virtually unlimited needs of housing, education, employment, public safety, etc., it is imperative to examine quite critically the entire range of needs that would be affected by given policy and executive decisions. Organization The Los Angeles Community Analysis Program has been funded in order to determine the relative importance and extent of the many problems throughout the City in order to establish priorities for Citywide and community programs to correct existing obsolescence and prevent further inroads of urban deterioration. The Community Analysis Bureau (CAB), an independent City department— in fact, a staff function organized as a line department— was established for this purpose by ordinance of the City Council. Function The CAB study was designed to serve the evaluation of blight, i.e., deficiencies in the quality of living and usually the failure of one or more of the delivery systems that exist for the provision of goods or services necessary to urban living. These include data on morbidity, mortality, births and other vital statistics; probations; crimes and arrests; business taxes; mental health incidents; parcel valuation, use, zoning, and appurtenances; building permits, inspections, and occupancy certificates, employment, net worth, gross sales, and credit ratings of establishments by Standard Industrial Classification (SIC) and situs; public school enrollment, attrition, transiency, and test scores; welfare cases, eligibility, and payments; and various special survey files such as the 1940 through 1970 censuses, transportation surveys, industrial surveys, etc. Scientific Urban Matrix (SUM) A major step forward— from data to information-toward the solution of the problems of evaluation of the urban system based on descriptive statistics was accomplished by the development of a statistical social trade-off matrix named the Scientific Urban Matrix or SUM. In 1970 and again in 1971, this technique of "pre-mathematics" facilitated this operation by not only describing a number of significant elements in the urban system but, even more so, by suggesting common linkages between these elements with the goal of creating a comprehensive theory of urban blight. An important benefit of this approach is that the linkages in the urban system described by the matrix suggest trade-offs in considering various remedial actions. A vehicle is provided for simplistic, static preplan testing of concepts that lie behind the remedial action programs. Resources Critical to an urban system analysis is the flow of pertinent, relevant, and available information from public and private sources. Systematically organized information enables the identi- fication, definition, and analysis of urban problems, the rational recommendation of well-conceived remedial action programs, and the monitoring of the environment to see whether those programs have produced the desired effects. The flow of information, as organized into the Urban Information System, enables the storage and retrieval from a continually increasing number of voluminous and diverse data files that are being acquired from a broad spectrum of public and private sources. Capabilities The Scientific Urban Matrix approach subsumes specific procedures and results which both support and follow from it: 1. It provides a means to coordinate subjective qualitative evaluations with objective quantitative data. 2. It provides a means to identify objectively the areas of the City that were blighted and the criteria by which they are classified as blighted. 3. It provides a means to synthesize identified urban threats and requirements. 'Formerly Manager, Data Analysis Division, Community Analysis Bureau, City of Los Angeles. 25 26 SOCIAL INDICATORS FOR SMALL AREAS 4. It provides, through the linkages seen within the urban system, a means to evaluate the indirect as well as the direct ramifications of remedial actions prior to their implementations. 5. It provides direction for the monitoring of changes in the City that could result from the implementation of remedial action programs. Implementation of the Scientific Urban Matrix requires the application of techniques that can deal with complex relation- ships among many factors. Once these relationships are known, the effects of fluctuations among the characteristics of remedial action programs on the measures of program effectiveness are readily determined through relatively simple linear equations. These, in turn, allow the development of elasticities of change— i.e., measures of the sensitivity of the total urban system to induced changes that derive from remedial actions. Furthermore, through trial and error manipulation of the independent vari- ables, those that represent direct program actions, near optimum solution to urban problems through realistic programs, can be derived. Structure The Scientific Urban Matrix performs its analytical function by logically organizing the universe of available urban data around these assumptions: 1. That the City is a system composed of various delivery subsystems. 2. That blight is the symptom of the failure of these delivery subsystems to adequately provide for the needs of their users. 3. That each delivery system is composed of various dimen- sions that are at least partially measurable by various techniques of social and behavioral science. 4. That the data that are available are sufficient in terms of extensiveness of scope and intensiveness of quality to approximate significant attributes of the various delivery system dimensions after appropriate manipulation for purification and factoring. 5. That the overall urban system performance is best measured by its users rather than by "experts," provided that the criterion of system performance is presented as a user judgment that is simple and that does not require vast quantities of data and techniques of analysis that would normally not be available to the users. SUM is applied by describing the urban system as a series of delivery systems— health care, education, income production, law enforcement and justice, etc. (Planning, Programming, and Budgeting System (PPBS) categories)-with each having a poten- tial for a set of behavioral dimensions defined as follows: 1. The action of the cultural-attitudinal system (of a given society) is to motivate individuals towards the achievement of specific values or goals. Communications are the instrumental vehicles through which culture fulfills its function. (The process of acculturation or acquisition of new values is a learning process wherein an individual adapts to his human environment.) Cultural learning may be viewed as a process of conditioning to group norms. The function of the cultural system, which is a structure of values, is to motivate individuals towards "very long run" behavior. 2. The societal system functions to provide the aggregate behavior or group roles required to reach the image of the "ideal state" embodied in cultural values (in the long run) The structure of the social system is given by the pattern of interrelationships between the members of the group such as the degree of segregation, the number of roles played, etc. 3. The action of the political system is to modify group goals and conflicts and for mobilizing groups for the attainment of intermediate-run goals. The structure of the political system is found in the government and party institutions regulating political life and in the political participation, political pressure, legal constraints and their enforcement, and other governmental responses, particularly in terms of the level of services rendered to a community or group. 4. The economic system consists of the provision of goods and services for the want satisfaction of consumers in the short run (resource allocation behavior). The structure of the economic system is found in production, distribution, and consumption units. Indicators of this dimension include measures of the effectiveness of the market as a resource allocation process, costs of services rendered, and the costs of failures of the delivery systems. 5. The performance dimension includes the elements most symptomatic of the success or failure of the various delivery systems. It is the set of impacts directly felt by changes within the systems. Hence, the variables included in it are those that seemed to best measure the overall effectiveness of each functional system. All available urban data, corresponding to these delivery systems, are retrieved and aggregated to common, reasonably homogenous geographic units, the communities. Residents of the City, the system users, are sampled and polled to rate the communities of which they have knowledge to obtain a subjective evaluation of the overall performance of the system. The condition and services descriptive data or social indicators are statistically evaluated for the degree that they contribute to the overall performance for each delivery system, and they are factored to discern composite indicators that relate to a delivery system. Blight thresholds are found by looking at values of indi- cators that discriminate a blighted or endangered community from safe communities, and the various delivery systems are weighted based upon the relationships that exist between their data and the overall community ambiance scores, the reactions of the polled residents-this last providing the derivation of sys- tem sensitivities or elasticities. SUM is designed to provide a needs analysis; hence, it is essential to translate the threats that are investigated by the above technique into requirements for action. In order to do this post-mathematics, it is essential to examine the threats by community in an integrated fashion, to translate them into units of response, and to examine on-going programs for their relevance and applicability toward fulfilling those needs. The Scientific Urban Matrix serves this purpose by dividing the communities into blighted, endangered, and safe areas and examining them for threats, requirements, and on-going pro- grams. In the context of the SUM, the urban system is defined as the conduct of various activities or delivery systems in a multi- dimensional network where each dimension of the delivery APPLICATION, LIMITATION, AND METHODOLOGY OF (SUT 27 systems has an impact— direct or indirect— upon every other dimension of the delivery systems. These levels of activity and of the ensuing experiences are treated as subsystems of the total system with inputs, outputs, and transfer processes pertaining to a given level, and appropriate linkages between these levels. Each level has an action— denoted by the type of goal that is set— and also an observable structure. Thus, the Scientific Urban Matrix (SUM) is initially based upon the assumptions that the urban phenomenon is composed of a series of delivery systems which, in turn, disaggregate into various common dimensions that are at least partially measurable by various techniques of social and behavioral science utilizing data that are generated by each of those systems in their implementa- tions. Therefore, for a variable to be included in the matrix, it is essential that data measuring it be available, that they be sufficiently extensive to cover all areas of the City, and that they be of a sufficient quality to utilize them as surrogates of the various delivery system dimensions. Variables Selection criteria for the variables included the relevance to each of the 40 cells— i .e., attitudinal, societal, political, economic, and physical performance manifestations arrayed along the horizontal axis; and accessibility, law enforcement, fire protection, recreation, health care, housing and neighborhood, income production, and education functional subsystems arrayed along the vertical; the ability of the variables to be stated operationally in terms of requirements for action where indi- cated; and the availability of data that could readily be developed into statistical area aggregations. Seventy-two candidate variables resulted for which data collection and data development efforts were undertaken. Although most data resided on magnetic tape files of the Urban Information System, in some cases the data did have to be hand crafted. The following table summarizes the variables and their sources. 28 SOCIAL INDICATORS FOR SMALL AREAS Summary of Variables and their Sources ACCESSIBILITY LAW ENFORCEMENT FIRE PROTECTION HEALTH CARE RECREATION HOUSING AND NE IGHBORHOOD INCOME PRODUCTION Attitudinal Modal preference Juvenile proba- Malicious false Inoculable Vandalism losses High school drop Elem. school en- f White collar of transportation tions/ Elem. alarms/capita diseases of ($) per park out rates rollment/Popula- (males 25+ yrs. d = % Bus - school enrollment (FIREINCD, those under 13 ac re ( Rec . & (EDUCATION) tion (EDUCATION, with less than Income - (Los Angeles POPEST) yrs. /Children Parks Dept.) POPEST) 4 yrs. of college) Transit Miles in neighbor- hood (LARTS67, SCRTD) Probation Dept., 1970 Census) AMalicious false alarms 1965-1968 (FIREINCD, age 0-13 yrs. (HEALTH, 1970 Census) Single dwelling unit/Total dwelling units Juvenile delin- quents elem. school enrollment (LAPD, 1970 POPEST) (ASSESSOR, POPEST) $> Owner-occupied single dwelling Census) units (ASSESSOR) Traffic arrests Juvenile depend- Arsons + Suspect- Suicides/100,000 $ of population Largest ethnic Elementary school io Households wi th per street mile (LAPD) °lo Using public transportation i households with 2 or more autos encies/Elemen- tary school en- rollment (LAPD, 1970 Census) ed arsons per 100 people (FIREINCD, POPEST) population (HEALTH, POPEST) age 20-59 °jo of population age 0-19 $ of population age 60+ years (1970 Census) percentage (EDUCATION) ^ Nonwhite en- rollment (EDUCATION) transiency rates (EDUCATION) Elementary school transiency out rates Elementary school working wives (LARTS67) (LARTS 67) transiency in families receiving Aid to Disabled f families receiving Aid to Blind % families receiving old age and survivors' benefits "jo families on medical assistance fo use of public transportation 1971 unemployment rate fo housing lacking facilities % families receiving Aid to Families with Dependent Children 1967 unemployment rate Economic Effects of Immobility Cluster households with "0" autos 4 years old on welfare bypass employment Juvenile dependents ing and Obsolescence Cluster age housing units 20 years and older Land uses by 2-4 families with zoning variance Auto accidents per street mile Alienation Cluster Structural fires fo using public hospitals Malicious false alarms Minority enrollment Felonies % households with 1.51+ persons per room Traffic arrests High school drop-out rate Income foregone fo households in poverty fo housing unit 20 years and older Premature Deaths Cluster Mortality 25-44 Fire losses > $1,000 Neo-natal mortality Neighborhood Instability Cluster Transfers into elementary school Transfers out of elementary schoo Working wives Nonconforming Land Use Cluster Nonconforming residential use Uses by 1-4 families with zoning variance Uses by single families with zoning variance Uses by 2-4 families with zoning variance Housing 20 years and older 34 SOCIAL INDICATORS FOR SMALL AREAS For the discriminant analysis, the communities were divided into four groups according to their ambiance ratings. Since a score of 5 was defined in the survey as an average community, any community with a score of less than 5 was considered below average and, thus, either "endangered" or "blighted." If a community had a score of 5 or above, it was above average and, therefore, "good" or "excellent." The division points of ambiance rating for the four groups of communities were as follows: Group Name Blighted Endangered Good Excellent Ambiance Ratings 0- 3.9 4.0- 4.9 5.0- 6.5 6.6- 10.0 Using all variables, the program chose 10 variables that best discriminated the communities. Using these variables, the discriminant function depicted only three communities as misclassified ("misclassified" meaning the discriminant function classified the communities in a different category than their original classification by their ambiance rating). It was felt that eliminating service-oriented variables from the discriminant function would perhaps give a set of variables whose meaning was more apparent in relating conditions of blight which distinguish communities. This assumed that service variables are more symptomatic than causative. Also eliminated were variables that only described ethnic composition. Using the remaining variables, a new discriminant function was derived which misclassified three communities. None of the three communities misclassified by the first discriminant equation were misclassified by the second equation. In the second equation, however, 10 different communities were misclassified. The two equations have only one variable in common, that being Sixth Grade reading scores. This variable was probably the most important single variable in both equations for classifying the communities because of its explanation of more of the variation between communities than any other variable. In the first equation, the variables after reading scores relate to categories, such as: 1. Segregation (percent largest ethnic group) 2. Income (percent hospital discharges which are from public hospitals and percent who have bypassed employment due to lack of transportation) 3. Family size (percent population aged 0-19 years) 4. Urbanization (percent trips using public transportation) 5. Incompatible land uses (parcels in use 02-04 whose zoning is higher or lower) Other variables are included that are harder to classify, such as, percent of wives who work, value of private recreational investment, and modal preference. Several variables in this equation indicate presence or absence of services or use of services rather than problems or lack of problems which differentiate communities. The equation using the service vari- ables gave much better discrimination than the second equation in which they were deleted. Evidently, then, the service variables give a better description of the communities as far as overall desirability, even though the two equations seem to be approxi- mating much the same functional categories. This is probably partly due to the fact that at least some of the service variables differentiate life in the Central City and the suburbs better than the nonservice variables. It is interesting that income as a variable is in neither equation directly but only indirectly. This points up the fact that the income of the residents of a community is not the decisive factor in the desirability of a community. Several variables, however, in the equations did come from the "affluence" factor determined in the factor analysis. It would seem, then, that it is the amenities commonly, but not necessarily, associated with higher income that are important in a community. It appears that the communities which were misclassified are borderline com- munities in many respects and that a high (or low) ranking on one or two variables can shift the community into a new classification. Since the discriminant analysis program that was used does not give the actual discriminant function in a form that determines the weights of the variables in the equation, it is not altogether clear which variable is most responsible for a community being reclassified. Only through resort to other analytical techniques can the contributing conditions be identified. After interpreta- tion of the results of each discriminant analysis, an assessment is made of each borderline community's problems and problems and probable trends. In order to determine the relative effectiveness of alternative remedial program strategies to upgrade the quality of life in a community, it is vital at least to know the relative effectiveness of the various urban functions if not the specific programs which would be most effective. This determination of the sensitivity of the overall urban system to deficiencies in each functional area is analogous to the economist's measures of elasticity. In other words, it seeks to weigh the primary and secondary impacts that would derive from programmed changes in alternative functional areas so as to indicate in orders of magnitude the relative effectiveness of remedial action. The standard against which a category was measured for its contribution to the vitality of a community was the ambiance ratings. Each category was quantified in a community by its composite score as determined in previous analysis. The composite scores were obtained by summing the ranks of a community by the variables in the category where the ranks were weighted by the correlation of that variable with ambiance. Using the composite score as the observation of each functional category in a community, a regression equation was generated with ambiance as the dependent variable and the categories as the independent variables. From the regression coefficients of each of the categories in the equation can be obtained a standardized regression coefficient. The standardized regression coefficient can be used to rank the categories as to their ability to explain ambiance rating in a community. The rank order of the categories found from this study and their relative weights in the equation are listed below: d i c _,.- i ^ ^ Relative Rank Functional Category ... . , . M y Weight 1 Education 1.000 2 Housing and Neighborhood .442 3 Income Production .438 4 Health .224 5 Law Enforcement .197 6 Accessibility .196 7 Recreation .144 8 Fire Protection .002 APPLICATION, LIMITATION, AND METHODOLOGY OF (SUM) 35 Education has over twice the weight of any other category. This is not surprising since the preceding analyses have shown that the education variables correlate highly with ambiance; that they and environmental quality are good discriminates between communities; and that they are associated with many indicators of affluence and well being. Education, housing, and income production are, in fact, highly associated as can be seen in the factor analysis and again in this study of blight elasticities. In this regression equation, the three above categories explain 88 percent of the variability in ambiance ratings of the communities. The other five categories, together, only explain an additional 1 percent of the variance in ambiance. It is interesting to note that law enforcement is so low on the scale of influences, particularly when the group surveyed to obtain the ambiance ratings indicated that neighborhood crime was a very important aspect in their ratings. It may be that there is a threshold of crime in a neighborhood below which the residents are complacent about the problem and above which they find the situation intolerable. Limitations Technical Four major technical deficiencies characterize the SUM process as applied herein. As expected, they derive from and impair the urban system description. These problems include the lack of discrete classification structures for either delivery systems or the social dimensions that describe them, inadequate attitudinal data, occasionally heterogeneous geographic com- munities, and difficulties with the overall measure of system effectiveness, community ambiance, both as to survey structure and cost and, logically, the scalar nature of the ratings. Unfortunately, aggregation for statistical reporting and analysis can obscure important differences. The definition of the standard statistical areas as communities fails in some cases; for, within each community, characteristics may vary considerably. The concept of community indices demands homogeneity, a relative quality not always available even at the level of the census tract. Naturally, the larger the geographic unit, the more likely it is to have "pocket" conditions contained within it. The Statistical Areas used in this report were tested for the degree of dispersion among the block groups (statistical areas average about 50 actual block groups) comprising them for the following characteristics from the 1970 census: Overcrowded housing (percent units with more than 1.5 persons per room) Housing condition (percent units lacking plumbing facilities) Structural type (percent single-family units) Occupancy type (owner occupied units/single family units) Age of population (percents 0-19, 29-59, 60 plus) Housing value (percent rental units under $100 monthly contract rent and percent owner occupied units under $20,000 estimated market value) Each of these series were examined for their means, standard deviations, and ranges in order to determine if they were relatively homogenous and, if not, if their heterogeneity was due more to "pocket" conditions or to differences across large areas. Only one Statistical Area was considered heterogeneous in four or more of the nine series; and two Areas, in three of the series. It had been expected that the factor analysis would have yielded more large factors that corresponded at least to the functional categories, e.g., housing, law enforcement, recreation, etc.— and, hopefully, factors that more clearly reflected the behavioral dimensions, e.g., cultural-attitudinal, societal, political, etc. However, limitations in the number of observa- tions, i.e., 65 communities— and the likelihood that those observations were not all relatively homogenous aggregations, undoubtedly, contributed to this deficiency. However, the direction of the findings does indicate that these traditional categories are not discrete and throws open to question the use of urban PPBS categories that so characterize the last 5 years work in the field of urban and regional science. With the development of the Digital Character Data Conver- sion System (a computerized system to convert through linear interpolation nonuniform polygon data to point data and, then, reaggregate those points to uniform polygons) being quite imminent, succeeding SUM analyses could be conducted on census tracts which would increase the number of observations more than eleven-fold. Furthermore, the census tracts or aggrega- tions of them to identifiable neighborhoods would be more homogenous units than are the larger Statistical Areas; so, a factor analysis based on the 741 census tracts should yield more details of the interrelationships of the variables. Although the ambiance ratings are continuous or scalar, rather than discrete or categorical, class limits have been assigned. This, of course, raises certain logical objections. However, although the establishment of certain ranges as discrete classes is arbitrary, it is somewhat mitigated by the imprecise nature of the problem. However, it is recognized that the sample selection desperately needs improvement-both as to size and structure. Efforts are currently underway to "piggyback" the Ambiance Survey onto media marketing surveys. Political Although the technical difficulties can be resolved without too much difficulty, there are some far more serious problems that limit the effectiveness of the approach. The generally low level of statistical competence in local government dictates a small audience for full documentation of the Urban Matrix project. The first volume alone contains about 800 pages which, regardless of content, will discourage widespread use in City government. Any substantial reduction in volume would first affect the documentation of the data and their analysis and would degrade the credibility of the findings. The approach to the resolution of this problem that was selected was to produce an analytically sound document for the technical staff of the City Administrative Office and for selected staff of the City Council, Mayor's Office, Planning Department, Redevelopment Agency, etc. However, experience demonstrated that even with more astute management science types, it was not sufficient merely to hand out the document. Seminars on the document findings and methodology have been and will continue to be conducted in order to guide the reader in his exploration and use of the material. We have found this approach particularly effective in communicating the material to the press. Since the seminar approach is available for only a relatively limited audience— they typically consume 3 to 4 hours— a printing of highly abridged summaries of the community threats. 36 SOCIAL INDICATORS FOR SMALL AREAS requirements, and programs— i.e., two of the three volumes of the State of the City Series— and allied work also performed by the Bureau under other projects will be more widely distributed. However, it must be recognized that the summary will almost inevitably oversimplify and under-document. It is also quite clear that the main report, "Conditions of Blight and Obsolescence," is of relatively little value to line operating departments. At best, perhpas it serves as a com- pendium of statistics; but, due to its holistic character, it is probably too broad-brush for its purposes. Furthermore, a principal goal of the report and its methodology is to provide a means to prioritize functions for purposes of allocating resources. Hence, the department managers responsible for those functions which are assigned low priorities will be disinclined to use the findings, and those whose functions are assigned high priorities will tend to abuse the material. Conclusions The Scientific Urban Matrix, as it is applied here, basically simulates the research portion of a community renewal planning program that begins with problem identification, continues through its definition and analysis, and culminates in recommendations for remedial action. This paper details the problem identification, definition, and analysis. In recognition of the difficulty of defining the problem, urban blight, a result of its being a composite of many highly complex and interrelated phenomena, the standards that were explored included qualita- tive composite standards, objective functional standards, and composite statistical standards. It was felt that the integration and synthesis of these into an overall blight standard incorpora- ted the most meaningful and realistic aspects of the parts and produced an effective yardstick by which to identify and measure the extent of urban blight. Intrinsic to this Scientific Urban Matrix approach is the hypothesis that, even within a discrete functional area, blight is probably multifaceted. The statistical analysis confirmed this hypothesis by demonstrating close relationships between various functional measures of performance and socio-attitudinal, socio- political, and socio-economic manifestations of those functions. Consequently, the cause and effect linkages between the under- lying factors and their manifestations in functional performance become far more apparent. This application demonstrates the primacy of social, political, and economic forces, values, attitudes, and goals in the determination of functional per- formance, along with the futility of proposed remedial actions that avoid these dimensions. The Scientific Urban Matrix approach, as performed here, admits to definite gaps: 1. The analysis of the data at gross geographical aggregations that are not sufficiently homogenous. 2. Variation in the time period covered by the data. 3. Subjective forcing of data collected on incompatible geographic limits into statistical area aggregations. 4. Uncontrollable inaccuracies in the data collected by operating departments. Many of these deficiencies will be corrected in succeeding iterations. Yet, in support of the project's ultimate goal of improving the City's environment by optimizing ameliorative programs, this second cycle of the Scientific Urban Matrix Project fulfilled the objectives enumerated in its study design: 1. It synthesized the threats of the functional categories that were studied. 2. By having coordinated qualitatively derived community status indices with conditions and service variables measured by data collected routinely in the City, it expanded the scope of the urban blight phenomenon that can be comprehended and controlled. 3. It provided a firm technical base upon which program recommendations can be founded. 4. It evaluated the relevance of a portion of the Data Base by measuring the sensitivity of the composite phenomenon of urban blight to variations in the state descriptive data contained in and reported from the Data Base. Hence, the second cycle of this project laid the foundation for the Community Analysis Bureau to meet its goals by developing the fundamental capabilities necessary to preplan evaluation of alternative remedial action programs, based on the significant status and trends in areas of the City, and to post- implementation monitoring of the effects of selected action programs for their effectiveness in improving the function of the urban system. INDICATORS AND STATISTICS: ISSUES IN THE GENERATION AND THE USE OF INDICATORS* Harvey A. Garn and Michael J. Flax, The Urban Institute, Washington, D.C. This paper is about a special class of statistics— social indicators— and the experimental program in developing in- dicators, particularly urban indicators, at The Urban Institute. We will discuss: (1) some definitional issues in the indicator field, (2) some critical conceptual problems in developing indicators, and (3) some of our past and current work at the Institute. Defining Indicators This is not the place to engage in an extensive discussion of what indicators are. Some comments, however, are in order. As is characteristic of many relatively new and growing fields, there is not a uniform or completely acceptable definition to all the people doing indicators. 1 Most simply and broadly, indicators are measurements of aspects of life and social conditions related (or believed to be related) to human well-being and satisfaction. The definitional problems revolve around alternative ways of for- mulating appropriate reference points for the indicators— means of attaching significance to the measures. Some have argued, e.g., It is interesting to report, in this connection, some of the history of the definitions of statistics. Yule and Kendall point out in their Introduction to the Theory of Statistics that, "The earliest occurrence of the word 'statistics' yet noted is in The Elements of Universal Erudition, by Baron J. F. Von Biefield. ...One of its chapters is entitled Statistics and contains a definition of the subject as The Science that teaches us what is the political arrangement of all the modern states of the known world.' " E. A. W. Zimmerman said in 1787, "It is about 40 years ago that the branch of political knowledge (statistics) which has for its object the actual and relative power of the several modern states, the power arising from their natural advantages, the industry and civilization of their inhabitants, and the wisdom of their governments has been formed. ..into a separate science." At this time statistics was largely a verbal, rather than numerical, subject. It evolved towards increasing emphasis on numbers and mathematical concerns and away from the emphasis on "what was noteworthy about the state." Indicators return to the earlier conception of statistics in the sense that they are about matters of social concern and the actual and relative status of population groups. *This research was supported by funds from the National Science Foundation under Grant No. GS- 29710X1 to The Urban Institute. Opinions expressed are those of the authors and do not necessarily represent the views of The Urban Institute. We wish to thank Jeremy B. Taylor and Herrington J. Bryce for their assistance and many helpful comments. Mancur Olson, that indicators derive their significance from being direct measures of social welfare. 2 That is, they are measures of variables which appear in a social welfare function. Insofar as one can derive such measures, the argument is that the movement of the variables in the appropriate direction represents an unambiguous improvement in social welfare, other things remaining equal. This definition leaves open the questions of how one is to tell what variables should be in the social welfare function, how one is to aggregate differing (and often con- flicting) individual welfare functions in determining the social welfare functions, and what to say in cases where other things do not remain equal. For these and other reasons, many have found such a definition too restrictive. Some of these, e.g., Sheldon and Freeman, have argued that what is important in attaching significance to indicators is the performance of measures of social conditions over time, both to monitor change and to develop reasonable expectations about the future movements of the measures. 3 This definition leaves open the question of what variables should be measured. Others, e.g., Kenneth Land, see the means of attaching significance to indicators to be their appearance in a social system model in which any individual indicator can be related through the model to other variables in the model. 4 This definition, of course, leaves open the question of how extensive the model must be to qualify, how to choose the appropriate model outputs, and how to assess (in value terms) alternative arrays of model outputs. Each of these definitions calls attention to important (and possibly unresolvable) issues in the generation of indicators. They are mentioned in the context of this paper for two reasons: (1) They explicitly raise some of the problems which permeate indicator work but are frequently left implicit, and (2) they have implications for our work at The Urban Institute and our approach. 2 Therefore,.. ."statistics on the number of doctors or policemen would not be social indicators, whereas figures on health and crime rates could be." U.S. Department of Health, Education, and Welfare, Toward a Social Report, Washington, D.C.: U.S. Government Printing Office, January 11, 1969, p. 97. 3 E.B. Sheldon and H.E. Freeman, "Notes on Social Indicators: Promises and Potential," Policy Sciences, Vol. 1, 1970, pp. 97-111. 4 Kenneth E. Land. "On the Definition of Social Indicators," The American Sociologist, November 1971, p. 323. 37 38 SOCIAL INDICATORS FOR SMALL AREAS Some Conceptual Issues Many of the major problems raised by these alternative definitions and attempts to carry them through in the generation of indicators are conceptual and value-related rather than being primarily statistical or resolvable by statistical techniques, how- ever sophisticated. These problems for those generating indicators include (to summarize some of the implications of the above definitions): (1) Consideration of just how the measurements to be made are related to human well-being and satisfaction, e.g., how can we identify principal components of the "quality of life?" (2) A determination of what measures to take and what units of analysis to use, e.g., our "problems" are often defined by what we measure. What are the best measures in each functional area? How should they be disaggregated? Should they be combined into indices? (3) Since the above are affected by both their intended uses and users, consideration should be given to whether the primary uses of the indicators are likely to be normative, descriptive, predictive, evaluative, or designed to improve understanding of causal relationships. Since the possible uses usually cannot be controlled (or often even anticipated!), what are the probabilities of misuse? Do the anticipated benefits of the particular indicator outweigh the possible costs? 5 (4) Since the indicators, once generated and presented, may be used for any or all of these purposes, some attempt should be made to consider what inferences and assumptions the users are likely to make. Inferences that are supported by adequate research sould be reinforced; where alternate inferences are equally valid they should be encouraged rather than hidden. Given the large number of unresolved conceptual and modeling issues and our sense that progress in the indicator field is likely to be iterative in any case, we have opted for an evolutionary, experimental program. We started with an attempt to describe important social conditions and recent changes in those conditions across metropolitan areas. The absence of a set of generally accepted value-related or objective standards and established system models has led us to emphasize comparisons as an important means of attaching significance to the indicators developed. Clearly there is need for a wide variety of indicator research approaches. We have chosen to confront many of the theoretical and measurement problems just mentioned in a descriptive context by generating and presenting indicators and improving them incrementally. We sense a strong and present demand for indicators that can help certain interested users, e.g., government officials and the informed public, obtain a better perception of the present state of affairs as well as some quantitative estimate of recent changes. While responding to this need (and we feel that some of those working on indicators should respond) we have had the opportunity to confront many theoretical issues. The experimental, iterative, nature of our work permits us to improve both the conceptual adequacy of the indicators and their sharpness as descriptors of important phenomena. In short, we have chosen to develop descriptive indicators as a means of raising conceptual and methodological questions in specific contexts. As stated above we have been concerned with the possible uses and users of indicators. How indicators acquire meaning to the user is central to this concern. This is accomplished in at least two ways: These problems are not likely to be resolved once and for all. We view the development of indicators, therefore, as an inherently iterative process rather than an activity with clear boundaries, fixed purposes, and an accepted theoretical base. General Comments on Our Work In our work, we have attempted to develop and use measures which are related to human well-being and satisfaction in the sense of Olson's definition, i.e., direct measures of social outputs, although we have not found it possible or desirable to do so in all cases. We have attempted to use measures which can be taken over time to facilitate monitoring of social conditions, although we have not attempted to use these time series for predictive purposes in our work to date. Finally, we have relied on knowledge in specialized fields both to guide our selection of indicators and in commenting on possible relationships among our indicators. In most of our work we have not found it possible to follow the prescription implied in the third definition-using only measures which appear in established social system models. This has not been possible because in many of the fields which are interesting (in the sense that they relate to areas of pressing social concern), there are few extant comprehensive models which can be used to isolate appropriate variables and their relationships. 5 The question of who receives the benefits and who pays the costs is usually crucial here. (DSome form of reference or comparison is needed (as mentioned above), and the indicator should be structured so that this comparison process is facilitated for the user. Comparisons can be made with governmental standards, e.g., does pollution exceed accepted safety levels; goals, e.g., is unemployment below 4 percent; comparisons over time, e.g., is the infant mortality rate higher than last year; comparisons across geographical areas, e.g., are the reading scores for New York City higher than those for San Francisco; comparisons of different population groups, e.g., is unemployment higher in the city than in the suburbs. The important point to be made here is that some form of comparison is valuable to the user in attaching meaning to the indicator. (2) Indicators can be said to stand for more than they measure. 6 The inference structure associated with measures varies with the individual and is not subject to precise measurement or prediction. However, the entire process of extracting information from indicators depends on the inferences users make concerning them, and these inferences may not coincide with those used by the indicator generators. Indicators may be designed to be primarily descriptive, evaluative, or causal, but their ultimate use is often decided by their users. However, 6 A frown stands for more than the movements of facial muscles— different observers infer various implications concerning a frowning individual. GENERATION AND USE OF INDICATORS 39 indicators can be designed to facilitate certain inferences, and that choice remains under the control of those who generate them. A Summary of Our Reports Our work so far has resulted in several reports which were, in a sense, "experiments." Each tended to expand our work along a different dimension of indicator problems. Our first study was documented in Paper 136-1, The Quality of Life in Metropolitan Washington (D.C.): Some Statistical Benchmarks, by Martin Jones and Michael Flax. This study had several interesting aspects and has guided the work which followed. In the first place, 136-1 reflected our interest in developing indicators over a broad range of areas of social concern. After extensive research and consultation we selected 14 areas of such concern rather than concentrating on one or two. We presented a single indicator for each of these areas. In each case an attempt was made to present a genuinely representative indicator of the area of social concern with appropriate caveats about the possibility of a single indicator providing a proxy for a whole area of social concern. Next we sought some comparative standards to give the indicators more meaning for our audience. We found that in none of the areas was it possible to adopt either an objective or subjective standard on which most would agree. Therefore, we decided to make our initial comparisons across metropolitan areas. This was done to provide sufficient variation in the measures and because no one had recently attempted to make such inter-city comparisons over such a broad range of social concerns. Further, comparisons were made of the recent changes that had occurred with respect to each indicator. This broad approach enabled the reader to examine both the characteristics of and recent changes in 18 different metropolitan areas over a wide range of social and economic considerations. It highlighted the fact that there are great variations in the comparative conditions that prevail in any one metropolitan area. For example, Washington ranked from second to eighteenth with respect to the different quality areas. This approach underscored the inadequate state of the available, comparative, recent social indicator data. Comparative data for cities (as opposed to metropolitan areas) were even less available. In producing this first document we were faced with many of the issues raised earlier. How were we to relate our variables to human satisfaction and well-being? What kinds of inferences and assumptions did we assume on the part of our readers? Since we chose as our audience a wide spectrum of the population (executives, administrators, and the informed public), we tried to choose variables that were of general interest and enjoyed a wide consensus as to "which way was up." We did not attempt to forecast or guide the inferences drawn from the data by this audience. We did, however, present caveats concerning the data used in as clear a manner as possible. We made no attempt to present an aggregative index of these indicators. We did not feel that such an index would serve any useful purpose, given the state of knowledge, except for the convenience of having a single number. At this stage of our work we feel that the search for a single social indicator or social indicator index would obscure the meaning of the indicators for most users. Eventually it may be possible, as is true of some economic indicators, to formulate composite measures that satisfy both the accuracy requirements of the theoretician and the credibility requirements of most users. Our subsequent work has consisted of further "experiments" representing incremental improvements in some aspects of our initial paper, 136-1. These are listed below: 1. Experimenting with Levels of Aggregation; 2. Experimenting with Multiple Measures; 3. Experimenting with Racial Comparisons; 4. Experimenting with Improved Classification Systems; 5. Experimenting with Aids to Urban Decision Makers. 1. EXPERIMENTING WITH LEVELS OF AGGREGATION It is fairly clear that metropolitan areas are not the only, nor necessarily the most interesting, population aggregates on which to make comparisons. We have, therefore, developed additional work in which the objective was to use different population aggregates. The example of this kind of indicator work which we have completed is "A Study in Comparative Urban Indicators: Conditions in 18 Large Metropolitan Areas," Report 1206-4, by Michael Flax, which is an updated and expanded version of our 136-1 report. As part of this report we displayed data for five (data were available for only five) of the 14 indicators in 136-1, broken down into central city and suburban values for the same 18 metropolitan areas as in 136-1. 7 The point again, in addition to the intrinsic importance of central city/suburban relationships and their recent rates of change, was to display a different kind of geographical disaggregation. 8 We intend to experiment with other forms of aggregation for population units in our future work. These may include aggregations of census tracts within a given city, neighborhoods within a city, admini- strative subdivisions such as school districts or police precincts within the city, or inter-metropolitan comparisons across all of these units. A major reason for the experimental nature of this effort is that serious analytic difficulties exist in establishing criteria for homogeneity of comparison units. We hope that additional insight into these analytic problems will result from experimenting with different forms of aggregation. 2. EXPERIMENTING WITH MULTIPLE MEASURES The next paper, "Selected Education Indicators for 21 Major Cities: Some Statistical Benchmarks," Working Paper 136-4, by the authors, builds on the work which was begun in our initial paper, 136-1, in a different way. As indicated previously, we recognize that a single indicator is an inadequate representation of any given area of social concern. Working Paper 136-4 expands both the number of indicators and the conceptual framework for considering the indicators in the area of education. The paper presents recent data for 12 different educational indicators for the central cities (21) of each of the metropolitan areas (18) examined in 136-1. The development of additional indicators in a given area of social concern requires that attention be paid to the problem of providing a conceptual framework for the indicators. In this case we utilized essentially an input-output framework and then examined the degree to which input indicators are useful in explaining the variation in output measures across the 21 cities. 9 7 In addition to updating the indicators (1968 to 1970), we also expanded our time frame over a 5-year period (instead of the 1-year period in 136-1), and substituted new measurement series for four of the 14 quality areas (education, transportation, narcotics addiction, and pollution). (See Exhibits 1-3 for the variables used, a sample data table and page of caveats, attached.) 8 For an example of this form of disaggregation, see Exhibit 4 attached. 9 For definitions of the indicators used, see Exhibit 5 attached. 40 SOCIAL INDICATORS FOR SMALL AREAS The indicators themselves were divided into four categories. The first of these consisted of measures which tend to be used as output indicators, e.g., standardized reading scores, draft rejection rates, and the like. The second category of indicators contained essentially measures of resources applied in the educational process directly through the educational system, e.g., per-pupil expenditure, percentage of teachers with advanced degrees, pupils per teacher, and so forth. The third category of indicators contained measures thought of as indirect inputs to education, e.g., those related to educational climate and the socioeconomic characteristics of the populations of the cities. The fourth category contained measures of the relative willingness of people in the cities to choose educational expenditures over other possible uses of public funds. 1 ° Since we chose an input-output framework for the indicators, we then tested, using two simple statistical models, the degree to which it is legitimate to use the input measures to explain variations in outputs across the cities. The variation in selected outputs cannot be explained significantly by these input indicators in these city-wide data. For example, it is not true that a lower pupil/teacher ratio tends to be associated across these cities with improved reading scores. Similarly, it is not the case that variations in high school dropout rates can be explained by the proportion of public school students who were either Negro or from low-income families. If nothing more, this work demonstrates the difficulties of using input indicators as proxies for output indicators when making comparisons. 3. EXPERIMENTING WITH RACIAL COMPARISONS Our next paper, "Blacks and Whites— An Experiment in Racial Indicators" 85-136-5, by the authors, represents a different kind of adaptation of the work begun in our initial paper, 136-1. In this case, the relative comparison groups are not, in general, comparisons by geographical aggregates, although we do point out that the nonwhite population is becoming increasingly concentrated in central cities and at a more rapid rate than is true in the white population. In this paper national data for 1960 and 1968, relating to white and nonwhite performance in 16 socioeconomic areas, are presented in a way that highlights the possibility of alternate valid interpretations of the data. In fact, inspection of our charts reveals that for 10 of the 16 functional areas examined, different conclusions as to the relative progress of blacks in our society could be made. 1 ' By discussing implications of both the indicators and ongoing Urban Institute research in assessing the prospects for future income equality between whites and nonwhites, we provide an example of how research results can be used in explaining the implications of relatively simple indicator data. 4. EXPERIMENTING WITH IMPROVED CLASSIFICATION SYSTEMS Our working paper, "Income and General Welfare: An Identification of the Socio-Economic Gaps Between Low and High Income Regions," by Herrington Bryce, represents an experiment in identifying indicator structures which enable policy makers to identify key socioeconomic differences among sets of metropolitan regions. In this paper, metropolitan areas were divided into two classes (high and low income regions) and then four discriminant functions emphasizing demographic, employment, housing, and education variables, respectively, were estimated using 1960 census data. 12 The discriminant functions identify linear combinations of the listed classes of variables which most effectively distinguish between high and low income regions. The variables used reflect reasonable theoretical connections among the variables, are clearly defined, and can possibly be affected by public policy. 1 3 An important advantage of the technique, therefore, is that it calls attention to critical differences in a wide variety of indicators, other than income, which tend to distinguish low and high income metropolitan areas from each other in addition to providing directly a basis for selecting an indicator set for comparative purposes. The technique has the additional advantage of identifying those metropolitan regions which have a typical socioeconomic structures for their income class. 5. EXPERIMENTING WITH AIDS TO URBAN DECISIONMAKERS We are presently preparing an inter- and intra-city (city network) indicator report in cooperation with six cities. We are gathering yearly intra-city data on five functional areas— poverty/welfare, health, education, public safety, and housing— in an effort to answer the following four questions: (1) Given the limitations of available data, can we present a set of relatively few data series which will prove useful to urban decision makers? We expect that the presentation of comparative, understandable, yearly data over a large range of concerns will provide a broader perspective and encourage a higher level of discourse concerning certain urban problems. (2) Given the lack of tested or accepted causal models, can we devise indicators which will provide useful insights that will improve urban decision making? We have constructed a general conceptual schema in order to classify phenomena which are often thought to have some causal relationships. We expect that discussions of this schema will aid us in devising measures which will provide useful insights for urban decision makers. We would also hope to "adjust" some of our indicators so that data from different areas might be more comparable, e.g., age-adjusted death rates would permit comparisons between areas with different age structures. In a sense this will be a form of evaluation falling somewhere between a thorough analytical study and a purely descriptive catalogue of output measures. We would try to structure and present our indicators in a form useful to the user. (3) Given the lack of absolute standards or well-defined goals, can we present national, State, city, and intracity data in a comparable format so as to increase understanding and insights concerning urban problems? Here we hope to locate potentially comparable data from different levels of government and attempt to present them so that users can implement these comparisons to highlight large or unexplained variations wherever they occur. Regression analysis to compute "expected" values controlling for supposedly causal variables such as age, sex, or income may also be used. 1 ° For a summary of these data, see Exhibit 6 attached. 1 ' See Exhibit 7 attached. 12 The study can be replicated using 1970 census data, of course. 1 3 The variables used are listed in Exhibit 8 attached. GENERATION AND USE OF INDICATORS 41 (4) Given the conceptual and methodological difficulties in obtaining attitudinal data, can we combine objective and subjective data so as to be of aid to urban decision makers? We hope to present existing attitudinal and objective data on the same subjects for the same geographic areas in a way that can be useful to decision makers. We believe that unexpected and possibly unexplainable variations will be brought to light by using objective and subjective techniques to describe the same phenomena. We would expect that this descriptive information might provide added insights to decision makers and provide direction, as well as data, for future research efforts. Summary We have discussed several different definitions of indicators. For our own work we have defined them, simply and broadly, as measurements of aspects of life and social conditions related (or believed to be related) to human well-being and satisfaction. Some of the critical conceptual issues that confront generators of indicators are: a) how to relate human well-being and satisfaction to measurable variables, b) how these variables should be defined, aggregated or disaggregated, c) what is the primary use of these indicators, what misuses can be expected, and d) what inferences are likely and which should be facilitated by the data presented. In our work, we are confronting many of these issues in a descriptive context. As a result of our concern about users and uses, two aspects of indicators deserve special mention in this summary: (1) indicators stand for more than they measure; we advocate explicitly considering the inferences likely to be drawn from indicators in their design; and (2) some form of comparison is useful in aiding the user to extract meaning from an indicator. The completed work described, as well as that being planned, are attempts to incrementally address some of the issues raised earlier in this paper. Some Areas for Future Indicator Experimentation 1. A "Use-Users" Study to define in a systematic way indicators that would be most useful, by different classes of users. 2. Incorporating reasonably well-demonstrated, causal effects into evaluative indicators. Adjusting and normalizing data so as to facilitate the use of comparative data in ways which account for known sources of variation. 3. The uses of comparative techniques for combining national, State, and local data in ways that will highlight large variations from expected performance. 4. The design of combined objective/subjective data gathering procedures so as to facilitate comparing objective and subjective data of the same phenomena, and the making of comparisons among similar geographic, demographic or other areas. 5. The application of indicator techniques to additional functional areas, e.g., racial differences, the criminal justice system, so as to improve understanding and aid decision makers. 6. Application of known statistical techniques to better indicator generation, presentation and use, e.g., validating data sources and leading indicators; adjusting for known or assumed causal variables; clustering or segmenting variables for more meaningful comparisons; correlation techniques for establishing, verifying, or testing relationships. 42 SOCIAL INDICATORS FOR SMALL AREAS EXHIBIT 1 Quality Categories and Selected Indicators QUALITY CATEGORIES Latest Year Data INDICATORS USED Unemployment Poverty Income Housing Health Mental Health Public Order Racial Equality Community Concern Citizens Partici- pationb 1970 1970 1969 1969 1969 1969 1970 1970 1970 1968 . a yo of labor force unemployed CO >. a ,r> 3 Sh o w O /-s m 1 cn u CD •H 00 ft ft 3 S 05 CD o CD ai 1 Oi CO T> CD "^. ^ co H CO £h r-i 00 Sh •- ft S3 w S i> 1 ai 3 cd CO H Cfl CO CITY .. oo rH rH +-> CO ai rH -P •H S3 1 03 t£> aj ^-^ •H rH CD T-1 ^-1 -P S3 cn co CD OI +-> -a Sh ■— ' •P al 3 cfl h-i al co U rH C c CD cci ft •o O CD CI "-- CO CD Si CD 3 *— ' -al 3 Sh cfl (D rH o s cn ft ho •a ■a 3 -P Sh ^ CO w +j X ai S3 as 73 cfl •H (D a o •H ft +-> ai S3 rH C O -C Sh ■H Z; rH S3 ^ S3 >> si Cd 3 CD S3 3 T3 ■H Sh CD 00 cd Sh -p •H •a ft Sh CD -P a r-t as a Sh CO •H al •H Sh -P < X CD ft -H CU CD ft Sh 1 •o r-\ ^ CD Cfl i—s W ft x -a PS S3 r-i 3 » CD ai ^-s X! O t- S3 W S3 en U CD s CO cn 00 3 CO •H cfl O CD U) -H (D 01 •— N Sh CO Cfl ■a 01 *"■> P O S3 ft T3 -p CD X! •d th ■O CD •a 0) 01 CD W rH cn t- ■H O O X cfl cfl p o +> 1 +-> O \ S3 rH 1 Cfl rH -p O a) Cfl rH cn oo cn Cd i-H rH cfl cn 3 CO o-w- Cfl r-H ■v Sh X! 3 -H 3 co 3 CD rH •H •H Sh T3 CO y ca si a a bo ■i-j ft •n ci •o E-i cfl ft T3 cfl > ~ei&- s - / ft ft W H -3 mo. 19$ 37 f $865 $8,682 $4,172 34% 21 11.6 1.6% (NA) 35.5f -14 11 34 640 9,141 3,308 26 27 11.4 1.8 $43 38.7 Los Angeles .... -11 9 23 612 9,930 3,793 21 27 12.4 4.5 49 27.7 Philadelphia. . . -13 12 47 670 9,247 3,354 37 24 11.2 3.8 32 23.3 -10 8 35 670 10,543 3,206 34 27 10.9 1.2 31 27.8 -3 11 21 545 7,770 1,784 29 28 12.3 2.1 30 23.0 -10 10 35 633 8,588 4,232 31 24 9.9 1.9 33 30.2 -13 14 30 723 8,581 2,907 24 21 12.0 1.5 (NA) 38.1 (NA) (NA) 9 7 28 31 467 561 7,625 7,559 1,811 2,534 29 18 27 28 12.4 10.3 2.1 3.4 26 44 24.1 24.8 San Francisco.. -16 4 33 639 9,868 3,483 24 22 12.3 4.5 31 37.1 Milwaukee. ..... -15 3 26 533 9,132 3,326 32 28 11.5 1.1 34 27.5 St . Louis ...... -5 6 25 616 8,565 3,896 28 29 9.6 1.0 38 22.2 -10 (NA) 7 4 36 25 618 755 8,711 (NA) 3,695 (NA) 43 26 24 24 12.1 12.0 1.8 1.0 42 25 52.4 30.7 Pittsburgh. .... (NA) 5 35 634 8,357 3,540 27 30 11.8 6.6 45 22.9 Buffalo....... . (NA) 5 31 757 8,003 3,075 23 23 10.5 2.9 44 30.6 Minneapolis .... (NA) 2 21 611 8,623 4,016 35 23 12.2 (NA) (NA) 32.7 Oakland. ....... -15 (NA) 25 632 9,348 3,894 38 24 (NA) 2.2 41 (NA) (NA) (NA) 16 636 10,009 3,369 36 26 (NA) 2.1 42 (NA) (NA) (NA) (NA) 623 9,534 3,503 16 25 (NA) 8.3 35 (NA) WEIGHTED -11 mo. 8°/o 30$, $640 $9,347 $3,345 29% 25 11.5 ■ 2.8$ $37 30.5% NA — Data not available. Source: Selected Education Indicators for 21 Major Cities: Some Statistical Benchmarks, Michael J, Flax, Preliminary Draft Working Paper 136-4, May 1971, 48 SOCIAL INDICATORS FOR SMALL AREAS EXHIBIT 7 An Example of Indicator Presentation Rates of Change and Size of Gaps: 1960 to 1968 W-White; NW-Nonwhite; B-Black. About 92$ of nonwhites are black. Rates tend to: A increase difference Size of gap: A increasing V decrease difference T decreasing Data Anal ysis Evaluation Was the Imputed annual nonwhite Was the Indicator Race rate of change Size of Size of rate of size of the 1960 1968 1960-19680 white/non- white gap, 1960 white/non- white gap, 1968 improve- ment white/non- white gap greater than decreasing? the white? LIVING CONDITIONS AND HEALTH w NW 17.2 26.9 e i4.7 e 23.4 -2.0$ -1.8$ A 9.7$ 8.7$ T no yes f w NW 73.8 yrs. 69.3 yrs. 74.4 yrs. e 70.0 yrs. +0.10$ +0.12$ 4.5 yrs. 4.4 yrs. ▼ ves yes HOUSING W NW 13$ 44$ 6$ 24$ -9.2$ -7.3$ 31$ 18$ T no yes FAMILY W NW 8.7$ 22.4$ 8.9$ 26.4$ +0.3$ +2.1$ A 13.7$ 17.5$ a g no no W NW 92$ 75$ 92$ 69$ 0.0 - -1.1$ A 17$ 23$ A h no no Fertility rates (live birth/1,000 women, age 15-44 e 79.4 6 115.3 W NW 113 154 -4.3$ -3.5$ A 41$ 38.4$ T no yes W NW 2.3$ 21.6$ 6 5.5$ 6 30.7$ +11.4$ +4.5$ 19.3$ 25.2$ * k yes no EDUCATION W B 63$ 36$ 76$ 58$ +2.3$ +6.1$ V 27$ 18$ ▼ yes yes $ Completing at least 4 years of college (25-34 W B 11.7$ 4.3$ 15.7$ 6.3$ +3.8$ +4.9$ V 7.4$ 9.0$ A yes no EMPLOYMENT W NW 4.9$ 10.2$ 3.2$ 8.7$ -5.2$ . -5.1$ 5.3$ 3.5$ T no yes W NW 19.1$ 34.2$ 11$ 25$ -6.7$ -3 . 9$ 15.1$ 14$ T no yes W NW 15.8$ 7.3$ 6 17.5$ 6 12.1$ +1.3$ +6.6$ 8.5$ 5.4$ t yes yes W 12.1$ e 14.2$ e Li +2.0$ 7.3$ 6.4$ t yes yes NW 4.8$ 7.8$ +6.2$ INCOME AND POVERTY W NW 46857 $3794 $8937 $5590 +3.4$ +4.9$ $3063 $3347 A yes no W B 18$ 55$ 10$ 35$ -7.1$ -5.3$ A 37$ 25$ T no yes $ Families with incomes greater than $8000 (in 1968 W NW 39$ 15$ 58$ 32$ +4.8$ +10.0$ V 24$ 26$ A yes no Source of Data: The Social and Economic Status of the Negroes in the United States, 1968. Blacks and Whites - An Experiment in Racial Indicators. Michael J. Flax. Urban Institute Report 85-136-5, 1971. aBiack data are used whenever available. t>This 1S tne annual percentage increase that would give the 1960-1968 change if compounded annually from 1960-1968. Many of these figures are percentage increases in percentages. qviany of these gaps are differences between white and nonwhite percentage figures. d Infants dying at less than 1 month of age. e 1968 values were extrapolated from 1960-1967 or 1960-1969 rates of change. *Data on life expectancy look more favorable as the age group rises. g The number of female-headed families is increasing for both whites and nonwhites, but the number of female-headed nonwhite families is increasing at a faster rate. ^he percent of nonwhite children living with two parents is decreasing, while the percent of white children remains the same. A lowering of the fertility rate is assumed desirable. Where income is limited a decrease would tend to encourage desirable trends in many of the other categories measured. J The white illegitimacy rate is increasing at double the nonwhite rate, therefore the slower nonwhite change may be expressed as a relatively greater "rate of improvement." k The percent of both white and nonwhite illegitimate births is increas- ing, but that of whites is increasing at a faster rate. 1 1960 data are for nonwhites. GENERATION AND USE OF INDICATORS 49 EXHIBIT 8 Variables Included in Discriminant Functions 1 Demography a. Percent nonwhite b. Crude death rate per thousand c. Outmigration rate 1950-60 d. Crude white fertility rate: white population age 0-4 as percent of white females age 15-44 e. Nonwhite dependency ratio: nonwhite population 0-4 and 65+ as percent of nonwhite population 20-64 Economy (Employment) a. Index of the centralization of employment (lack of industrial diversification) bo Index of the concentration of nonwhites in low-paying industries c. Index of the nonfarm occupational mix Housing a. Percent of rental units vacant b. Percent of households with two cars c. Percent of houses with adequate plumbing d. Percent of persons in group quarters Education a. Percent enrolled in private elementary and high school b. Enrollment rate of those aged 15-19 c. Percent white population 25 years and over which completed high school d. Percent nonwhite population 25 and over which completed high school e. Percent of total population 25 and over which completed at least 4 years of college L The variables used in the respective discriminant functions estimated in "identifying Socio-Economic Differences Between High and Low Income Metropolitan Areas," are listed. Other variables were con- sidered. Urban Institute Working Paper 1206-6, Herrington J. Bryce , July 1972. SOCIAL AREA ANALYSIS: PROCEDURES AND ILLUSTRATIVE APPLICATIONS BASED UPON THE MENTAL HEALTH DEMOGRAPHIC PROFILE SYSTEM Harold F. Goldsmith and Elizabeth L. Unger National Institute of Mental Health Introduction The National Institute of Mental Health has designed a small area demographic profile system (MHDP) that provides the requisite data for the delineation of meaningful "Social Areas" within mental health service areas, cities, standard metropolitan statistical areas or other designated census areas. In an earlier paper, we discussed the kinds of data comprising the profile and suggested operational procedures to be used to construct social areas (areas with common demographic characteristics) from the data provided (see Goldsmith and Unger, 1972a). Unfortunately, we found that those procedures presented worked best for racially homogeneous husband-wife/family areas, that is, for suburban-like family areas rather than for inner city residential areas. The inner city tends to be characterized by heavy concentrations of poor people, both white and Negro, large proportions of primary individuals who are household heads (nonfamily households), and/or disproportionate numbers of households headed by females. Thus, what is required are standardized procedures to be used in the construction of social areas that are applicable to a wide range of disparate residential areas. Our primary concern in this paper will be to suggest such procedures. We will use the data items in the MHDP, since we are also concerned with familiarizing the reader with the MHDP. Poor white and Negro census tracts of Baltimore city will be used to exemplify the procedures. Social Area Dimensions Before presenting the operational procedures, we will com- ment briefly on the social area characteristics that need to be measured. "Proponents of social area analysis state that much of residence related behavior can be understood and accounted for in terms of three types of society -wide population characteristics or dimensions: social rank, life style or urbanization, and ethnicity" (Goldsmith and Unger, 1970). In our view, neither the social rank dimension nor the urbanization dimension should be treated as single factors (see Goldsmith and Unger, 1970 and 1972b). The standard components of social rank— economic class (ability to consume goods and services), social class (prestige, position) and, perhaps, educational (information) status— should be considered separately. The life style or urbanization dimen- sion is seen as a composite dimension including the following components: family status (family/nonfamily), family life cycle stage, residential life style, and familism (small family/large family). The ethnicity dimension is maintained without modifica- tion, in addition to the traditional social area variables, residen- tial stability and area homogeneity are viewed as significant dimensions requiring identification. Mental Health Small-Area Demographic Profile System (MHDP) Census measures which serve as indicators of the various social area and related dimensions are presented in appendix tables 1-4 and appendix figure 1. Item selection was based primarily on prior use of specific census measures in demographic and ecological analysis and, secondly, on modifications suggested by subject matter experts. In addition to the data items indexing social area dimensions, we selected (on the basis of examination of epidemiological studies) 14 other census items judged most useful in the identification of subpopulations seen as "high risk" in terms of probable need for mental health services and other related social services (see appendix table 4 and Redick, Goldsmith and Unger, 1971). We feel that the data items in the profile index those demographic and ecological dimensions (1) that best differentiate among residential subareas of American cities, according to the current literature, and (2) that can be measured using census data. Appendix tables 1 and 2 utilize second count summary tapes which contain complete-count population and housing data at the level of census tracts and larger census areas. Appendix tables 3 and 4 utilize not only second count tapes but also fourth count summary tapes which contain such data as income, occupation and education for the 20-, 15-, and 5- percent sample populations. The MHDP provides age-sex population pyramids for Negro, white and total populations and for rural and urban areas (see appendix figure 1). The MHDP has been designed so that appendix tables 1 and 3 contain overall information about an area's population (i.e., the "dominant" characteristics of that population), and appendix tables 2 and 4 contain specific information about the distribution of an area's population (i.e., indicators of area homogeneity). Profiles can be produced for census tracts, minor civil divisions, counties, and aggregations of these units. Further, the MHDP system allows for statistical manipulation of the area units. The profile items are stored on disks for rapid retrieval for description or statistical analysis. Intermediate tapes (tapes that are derived from census tapes and that contain selected second and fourth count tabulations) were used to build the disk data base. 50 IENTAL HEALTH DEMOGRAPHIC PROFILE 51 Procedures For Classifying Areas By Social Area Dimensions In this section, we will discuss procedures for classifying areas by social area dimensions. 1 We will demonstrate procedures that can be used outside of nonintegrated/husband-wife family areas, i.e., that have general applicability to different types of areas. The items selected for use here are from the second count summary tapes since the fourth count data items were not available at the time the paper was being prepared. However, where appropriate, we do refer to the fourth count data items since they are now available. Ethnicity The ethnic populations that are considered most significant in contemporary social area studies are Negroes and persons of Spanish heritage. Since the second count data do not allow identification of persons of Spanish background, our measure of ethnicity in the poor areas of Baltimore city is the proportion of an area population that is Negro. The classification system used is presented below: PERCENT NEGRO (Appendix table 1, Item 9) 0.0% - 4.9% 1 5.0% 9.9% 2 10.0% - 29.9% 3 30.0% - 49.9% 4 50.0% - 74.9% 5 75.0% - 89.9% 6 90.0% - 94.9% 7 95.0% < ar more The code was designed to allow differentiation of areas on the basis of the relative visibility of Negro resident populations. Where other ethnic (non-Negro) populations appear in sufficient number in an area, they may be similarly classified. The second count disk data also identifies the proportion of an area population that is neither white nor Negro (see appendix table 1, item 10). With the fourth count summary tape the percent of the population of Spanish heritage (appendix table 4, item 44), foreign stock (appendix table 3, item 14) and southern or eastern European stock (appendix table 4, item 45) can be identified. Programs that will provide data items in appendix tables 3 and 4 for Negro and Spanish heritage populations are being developed. These programs will use the data stored on our intermediate tapes. Social Rank: Economic Class Only a crude estimate of area economic class can be made using the second count summary tape data (more precise information will be available when the fourth count summary tape data are incorporated in the system). We suggest that median rent and median house value (appendix table 1, items 7 'The authors have omitted from the present article, due to the limitations of space, two appendix tables and an appendix chart. The reader can receive this supplementary appendix by writing directly to the authors, care of the Mental Health Study Center, 2340 University Blvd. E., Adelphi, Md. 20783. The supplementary appendix tables illustrate in detail the application of the various coding procedures discussed in the text to each of the "poor" census tracts in Baltimore city; the supplementary appendix chart details the 4-item and 3-item disk area family life cycle codes mentioned later in the text. and 8) be used to estimate total area economic class (Goldsmith and Unger, 1972a). In order to estimate the economic status of non-Negro and Negro populations of an area separately, the lower quartile values for owner occupied and renter occupied dwelling units by race should be utilized (see appendix table 2, items 1, 2, 3, 4). The decision to use any of the above items as indicators of area economic class entails the implicit assumption that the higher the rent paid or the higher the house value, the higher the economic class or income. Since we are concerned specifically in this paper with identifying poor populations in areas by race, we have used the lower quartile values for Negroes of Baltimore city in classifying and designating a population as poor. First, we looked separately at non-Negro headed and Negro headed owner occupied housing units in a census tract and classified them as "poor" if the lower quartile house value was below $7,000; and we examined non-Negro headed and Negro headed renter occupied housing units and classified them as "poor" if the lower quartile monthly rental was less than $65. Then we looked at the distribution of the above groupings, and designated an entire area (census tract) as "poor" if the subgroups classified "poor" comprised 80 percent or more of the households in the area. Only four types of "poor" areas will be examined in this paper. They are specified by race and by renter/owner status as follows: 1. Predominant population Negro renters: "poor" areas in which 80 percent or more of the housing units are rented by Negroes. 2. Predominant population Negro renters or owners: "poor" areas in which 80 percent or more of the housing units are rented or owned by Negroes (less than 80 percent of the units are rented by Negroes). 3. Predominant population non-Negro renters: "poor" areas in which 80 percent or more of the housing units are rented by non-Negroes. 4. Predominant population non-Negro renters or owners: "poor" areas in which 80 percent or more of the housing units are rented or owned by non-Negroes (less than 80 percent of the units are rented by non-Negroes). In our exploratory analysis of Baltimore city, the first two types are referred to as poor Negro census tracts and the last two as poor white census tracts. When the fourth count summary data is incorporated into the system, data items for areas such as percent of families and persons in poverty by race (see appendix tables 3 and 4), will provide more precise indicators of the extent to which the populations of areas are "poor." Other estimates of economic class can be arrived at by using such items as median income of families or of unrelated individuals by race (appendix table 4, items 1-5), and unemployment and underemployment rates by race and sex (appendix table 4, items 16-20). Social Rank: Social Class and Information Status The system, when it incorporates the fourth count data, will also provide indicators of area social class and area information status. Some examples of appropriate indicators of social status are: percent of males in high status occupations, percent of males in low status occupations (appendix table 3, items 9 and 10), percent of females in low status occupations, etc. (appendix table 4, items 29-32). One indicator of information status would be median school years completed by persons 25 and over (appen- dix table 3, item 11); and appendix table 4 provides other detailed data pertaining to information status by sex and age (items 34-43). 52 SOCIAL INDICATORS FOR SMALL AREAS Family Status In an earlier paper on social area identification procedures, our classification of family status was based solely on the percent of husband-wife households in an area: a tract (or other area) was considered a family area if at least 70 percent of the households in the area were husband-wife households. If an area does contain a high proportion of husband-wife households, there is no question that one is dealing with a family area, representative of what many would describe as "normal family" areas. However, this classification does not tell the whole story. For there remain many households (30.6 percent in the United States) that are not husband-wife households and many households (19.7 percent in the United States) that are not family households, i.e., where the head of household is a primary individual (see U.S. Bureau of the Census, PC(1)C1, June 1972). A classification procedure that will more accurately describe the family status of an area must distinguish family areas and nonfamily areas and, within family areas, husband-wife family areas and other types of family areas. With these requirements in mind, we re-examined the available disk data items and selected the following data items: (1) percent of households with heads who are primary individuals (appendix table 2, item 18); (2) percent of all households husband-wife (appendix table 1, item 11); and (3) percent of households with female heads (appendix table 2, item 23). We were able to formulate at least a preliminary area family status typology using this set of data items, a typology that would meet the requirements of differen- tiation and specification mentioned above. Thus, we are able to distinguish (1) nonfamily areas (high percent of households headed by primary individuals), (2) mixed family/nonfamily areas (moderate percent of households headed by primary individuals), and (3) family areas (low percent of primary individuals). We can further distinguish within family areas husband-wife family areas (high percent of husband-wife house- holds) and female-headed family areas (high percent of female- headed households). It should be recognized that this typology is applicable to largely uni-racial areas since the data items refer to total populations; however, appendix tables 3 and 4 do present two of the data items by race which will enable us to make estimates of area family status for integrated areas. Extensive empirical examination of census tracts in the Washington, D.C., Wilmington, Del., and Baltimore city areas resulted in the following coding system for each of the family status variables: Four-Item Disk Code: A classification system based on four disk items is presented in the paper, Social Areas: Identification Procedures Using 1970 Census Data (Goldsmith and Unger, 1972a). Because the 4-item disk code is most meaningful when applied to predominantly husband-wife family areas, a revised classification procedure that would also be applicable to residen- tial areas that are either not family areas or not husband-wife family areas is needed. In response to this need, we have developed a revised (3-item) code which we will present in the next section, following a brief discussion of the 4-item code. (The 4-item code is detailed in supplementary appendix chart 1.) The development of the 4-item disk code was an outgrowth of a prior analysis of the characteristics of families and areas at different stages in the life cycle (Foote, et al., 1960:96-121; Goldsmith, Stockwell and Unger, 1970). A first step was to assign designations to the various stages in the family life cycle (i.e., pre-family, childbearing, childrearing, childlaunching, post- childlaunching), labels that we hoped would describe families as they age in terms of child-related functions and associated adaptation problems. Then we listed a few related standard and nonstandard demographic characteristics (defining life style characteristics) for area family life cycle stages (see chart 1). This demographic characterization of families proceeds from the general observation that the "average" woman marries by age 20 or 21, has two or three children, and bears her last child before she reaches age 30 (Glick, 1957). We then selected items from the disk data that we felt would reflect and specify the general characteristics outlined in chart 1. The criteria to be applied to these items in classifying areas by family life cycle stage are presented in chart 2. Chart 2 illustrates that a husband-wife family area with a low proportion of children of all ages and young household heads would be classified as a pre-family population; whereas the same sort of area with old household heads would be described as a post-childlaunching population. A husband-wife family area with a high proportion of children in the under 6 only category (other categories low) and young household heads would be classified as a childbearing population; whereas the same sort of area with a high proportion of children in the 6-17 only category (other categories low) would be identified as a late childrearing (or childlaunching) population. PERCENT IN CATEGORY 1 Under 15% (very low) 2 15.0% - 29.9% 3 30.0% - 49.9% 4 50.0% - 69.9% 5 70.0% - 84.9% 6 85% and above (very high) We will give examples of some of the area family status types that we found in Baltimore city later on in the paper. Area Family Life Cycle In this section, we will present classification procedures of areas by family life cycle utilizing (a) data stored on the disk and (b) the age-sex population pyramids. The procedures using disk data will be discussed first. Three-Item Disk Code: The theoretical basis for classifying areas by family life cycle stage included consideration of parental age; however, we were obliged to infer age of parents from the disk item "age of household head." Obviously, we are justified in making such an inference only for husband-wife areas (70 percent or more of the households husband-wife) or for family areas (less than 30 percent of household heads primary indivi- duals). Because of this limitation, we decided to delete the "age of household head" item. The remaining 3-item code would still provide information about the families in an area even if that area had a sizeable nonfamily population. We then compared these 2-disk codes to determine the consequences of the revision on the adequacy of the classification system (see supplementary appendix tables 1 and 2). In our judgment, the 3-item code, although less sensitive than the 4-item scale in identifying specific area family life cycle stages, does produce a more reliable index for the families in nonfamily areas. (Both codes are presented in supplementary appendix chart 1.) MENTAL HEALTH DEMOGRAPHIC PROFILE 53 Chart 1.-DEFIIMING LIFE CYCLE CHARACTERISTICS FOR AREA FAMILY LIFE CYCLE STAGES Pre-family Childbearing Early childrearing Middle childrearing Late childrearing Childlaunching Post-childlaunching Families with no children and single persons Families with children under 5 only Families with children under 5 and 5-9 Families with children 5-9 and 10-14 Families with children 10-14 and 15-19 Families with children 15 and over Families with no children under 18 Young household heads Relatively young household heads Young to middle-aged household heads Middle-aged household heads Middle-aged and slightly older household heads Middle-aged and slightly older household heads Old household heads Chart 2.-DISK DATA CRITERIA FOR AREA FAMILY LIFE CYCLE STAGES Family life cycle stage of area Pre-family Childbearing Early childrearing Middle childrearing Late childrearing (or childlaunching) Post-childlaunching Families: Husband-wife households Age of household head With children With children under 6 With children under 6 only and 6-17 6-17 only High Young Low Low Low High Young High Low Low High Young Low High Low High Middle-aged Low Low High High Old Low Low High High Old Low Low Low Age-Sex Population Pyramid Codes: Our exploration of two central cities, Washington, D.C. and Baltimore city, and of suburban Prince George's County made us aware that it was necessary to provide estimates of family life cycle of area populations by race, primarily because the Negro and white populations of a given area may be at different life cycle stages. Appendix figure 1 gives a dramatic illustration of very different age-sex distributions of the white population and the Negro population of a single census tract; we assume that such a discrepancy between the two population pyramids is a reflection of different life cycle stages characterizing the populations. The data items comprising the disk code are not presently available by race (although we do contemplate adding these items by race to the intermediate tape); in addition, the disk codes do not provide a sensitive index for nonfamily areas. Since the age-sex population pyramids are provided by race (see appendix figure 1), and since a classification system of area family life cycle based on the age-sex pyramids would not be contingent upon either area family status or median age of household head; we decided to develop a classification system using the age-sex population pyramids to indicate the family life cycle stages of an area. In an earlier paper (Goldsmith, Stockwell and Unger, 1970), we developed a procedure using the age-sex pyramids to estimate area family life cycle. Our present procedure, though related to the earlier one, is not concerned with the dominant adult age grouping which was an integral part of the 1970 procedure (see Goldsmith, Stockwell and Unger, 1970). The present procedure, like the disk code, follows logically from charts 1 and 2, and, like the revised disk code, ignores parental age and family status. The 1972 age code primarily relies on the presence (or absence) of children in specified 5-year age categories to estimate area family life cycle, as detailed in chart 3. Attention is paid to adult age grouping only for two life cycle stage designations: pre-family and post childlaunching. These exceptions will be discussed later. An area often can be classified in more than one stage of the family life cycle. This is because designations are made generally on the basis of Highs always taking into account the adjacent categories. If one or both of adjacent categories are High, then the area receives appropriate designations as indicated in chart 3. Thus, if an area's pattern was H H L H, it would be classified as early childrearing (on the basis of the two Highs in the "under 5" and "5-9" categories) and childlaunching (on the basis of the Low and High in the "10-14" and "15-19" categories). Determi- nation as to the primary (first mentioned) designation is based on which High age category is highest. Area family life cycle designations can be preceded by "with some." This wording indicates Mediums rather than Highs in the defining categories (see supplementary appendix tables 1 and 2). 54 SOCIAL INDICATORS FOR SMALL AREAS Chart 3.-REVISED AGE CATEGORY CRITERIA FOR DETERMINING FAMILY LIFE CYCLE STAGES OF SUBAREA POPULATION Percentage of population in specified age category Family life cycle stage Type Under 5 5-9 10-14 15-19 Pre-family and/or post-child- launching 1 Low Low Low Low Childbearing 2 High Low Low Low Childbearing with early child- rearing 3 High Medium Low Low Early childrearing 4 High High Low Low 5 Medium High Low Low 6 Low High Low Low Early childrearing with middle childrearing 7 Low High Medium Low Middle childrearing 8 Low High High Low 9 Low Medium High Low 10 Low Low High Low Middle childrearing with late childrearing 11 Low Low High Medium Late childrearing 12 Low Low High High 13 Low Low Medium High Childlaunching 14 Low L Low Low High 1 Specific criteria for high, medium and low classifications: High 12.5% or more of total population Medium 7.5% to 12.4% of total population Low below 7.5% of total population An area population is designated as including a post- childlaunching sub-population if 25 percent or more of the population is age 50 or over; the designation "with some post-childlaunching" is used to indicate that from 20 to 24 percent of the population is 50 or over. Designation of a population as prefamily depends on the distribution of population in two adult age groupings, "20-24" and "25-29" in combination with the distribution in two child age groupings, "under 5" and "5-9." Thus, an area is designated as "prefamily" if either of the adult age groupings accounts for 12.5 percent or more of the population and neither of the child age groupings accounts for more than 7.5 percent of the population. The above designation is qualified and referred to as "with some prefamily" under the following conditions: when the proportion of the population in either of the adult age groupings is high (12.5 percent or more), but the proportion in one or both of the child age groupings is medium (7.5-12.5 percent); or when the proportion of the population in either of the adult age groupings is medium (7.5-12.5 percent), provided the proportions in both child categories are low (less than 7.5 percent). In addition to the 1972 age code, we have included a visual classification of the age-sex pyramids (in the supplementary appendix tables) because we think that it serves as an additional measure against which the other codes can be checked for reliability. Our previous experience has provided us with consid- erable skill in coding the age-sex pyramids according to family life cycle (Goldsmith, Stockwell and Unger, 1970). Though admittedly subjective, the visual code incorporates the logic in chart 1 and also takes into account how selective a population is in terms of child and adult age categories. The designations assigned to each census tract according to the different codes (detailed in the supplementary appendix tables) illustrate the similarities and dissimilarities among the systems. These criteria are based on earlier work and recent exploration of the Washington and Baltimore Standard Metropolitan Statis- tical Areas. Further, we have summarized the comparison of the 1972 age codes with the visual codes (table 1) and with the revised disk codes (table 2). One can see that there is considerable agreement among the designations resulting from the different systems, and that the similarity is most marked with respect to classification of family areas. Therefore, we suggest that the disk codes (since the items can be retrieved rapidly) be used to identify family life cycle in family areas, and that the 1972 age code be used in mixed family/nonfamily areas and nonfamily areas. Residential Life Style Residential life style indicates the day-to-day activities con- ducted in the immediate vicinity of the residence. It appears that residential life style of an area is related to the type and condition of available housing units (Goldsmith and Unger, 1970:13-17). We use the item "percent of occupied and vacant year-round housing units that are one unit structures (one unit at an address)" to index type of housing; and the item "percent of persons in households with 1.01 or more persons per room" to index condition of housing (overcrowding). Both of these items are from the second count census tapes. Our codes for these two items are as follows: TYPE OF HOUSING: PERCENT ONE UNIT STRUCTURES (Appendix table 1, item 15) 1 80% or more (single dwelling units) 2 50%-79% (mixed: single dwelling units with apartments) 3 Less than 50% (predominantly apartments) 1ENTAL HEALTH DEMOGRAPHIC PROFILE 55 CONDITION OF HOUSING: PERCENT NON-NEGRO OR NEGRO POPULATIONS IN OVERCROWDED HOUSING (Appendix table 2, items 32 and 33) 1 Less than 5% (not overcrowded) 2 5%-29% (slightly overcrowded) 3 30% or more (overcrowded) When the data from the fourth count summary tapes are incorporated into the MHDP System, we will be able to identify type of housing by race (percent single dwelling detached units) and will have access to additional useful indicators of residential life style such as percent of housing units that are mobile homes, in large apartment complexes or in highrise apartments (see appendix tables 3 and 4). Other Relevant Area Characteristics Area residential instability and area subpopulations with high potential need for health and related services have not been considered in this paper because indicators are derived from the fourth count summary tapes; however, the relevant items are contained in appendix tables 3 and 4. The following census measures are used as indicators of an area's residential instability: percent living in different house 5 years earlier, percent living in different county 5 years earlier, and percent moved into present residence within the last year (see appendix tables 3 and 4). Items identifying "high risk populations" (shown in appendix table 4) encompass data on school dropouts, working mothers, aged persons living alone, disabled populations, and specific poverty groups. Social Area Characteristics of "Poor" Areas of Baltimore City: 1970 In this section, we describe the kinds of "poor social areas" that we discovered when the procedures described above were applied to Baltimore city (see supplementary appendix tables). These social area types provide a preliminary typology of "poor" residential areas. Poor Negro Census Tracts The 23 poor Negro tracts are grouped into 6 basic types according to shared social area characteristics. The social area characteristics of each of the 23 census tracts (presented in supplementary appendix table 1) are summarized by type below: Set A— Family areas; Negro female-headed families with many school-aged children; overcrowded housing conditions. The family status patterns (224, 225, 124 and 134) for these tracts give credence to their being described as female-headed family areas: in each tract at least 50 percent of the households are headed by females—female headed— and less than 30 percent of the households are headed by primary individuals— family areas. The predominant family life cycle classification for these area families is middle childrearing and early childrearing, with some late childrearing. The age-sex pyramids for the tracts comprising Set A are quite distinctive (e.g., see figure 1), indicating the presence of adult women with many children and the absence of adult males. All census tracts in this set are classified as overcrowded. There is no uniformity in set A in respect to type of housing. Set B— Mixed family/nonfamily areas; Negro female-headed families in the middle childrearing stage of the family life cycle and some post-childlaunching households; mixed (single dwelling units with apartments) housing areas. The main difference between set B and set A is that set B contains more primary individuals. The family status pattern for this set of 324 describes mixed family/nonfamily areas with many (50 to 70 percent of all households) female-headed households and families. In each tract 30 to 50 percent of all households are headed by primary individuals (i.e., nonfamilies); on the other hand, at least 50 percent are not headed by primary individuals (i.e., families). The predominant family life cycle classification for these areas is middle childrearing or post-childlaunching: specifically, tracts 1002 and 1501 were classified as middle childrearing with some early and late childrearing and some post-childlaunching, and tracts 1702 and 1801 as post-childlaunching with some middle childrearing, late childrearing and early childrearing. The propor- tion of children for this set of tracts is not as large as that for set A (see figure 2 below). The age-sex pyramid also shows more adult females than males between the ages of 20 and 50, indicating the presence of many female-headed families. Set B tracts are classified as mixed housing areas (single dwelling units with apartments); moreover, three of the four tracts in this set are classified as overcrowded. Set C— Family areas; Negro families husband-wife and female- headed in the middle childrearing family life cycle stage, also some post- childlaunching, late childrearing and early childrearing households. The family status pattern of 233 for set C tracts describes predominantly family areas (no more than 30 percent of all households are headed by primary individuals). Husband- wife families account for from 30 to 50 percent of all households; the same proportion obtains for female-headed households. In terms of family life cycle classification, all but one of the set C tracts are predominantly middle childrearing with some post-childlaunching, early childrearing and late child- rearing. For the exception (tract 1502), the post-childlaunching stage was dominant with some middle, late and early childrearing also present. None of the five tracts in set C are classified as predominantly apartment areas, and only one of the tracts is a single dwelling unit area. Four tracts are classified as uncrowded. (See figure 3.) Set D— Family/nonfamily areas; mixture of Negro nonfamily households, husband-wife families and female-headed families or households; area households in post-childlaunching family life cycle stage, some middle, early and late childrearing households also present. The social area characteristics of this set of tracts are similar to those of set C. The sets differ primarily in the presence of more households headed by primary individuals and in the predominance of the post-childlaunching family life cycle stage. The family status pattern of 333 indicates a mixture of family and nonfamily households, and among families many husband-wife families with a moderate proportion of female- headed families. The predominant area family life cycle stage for these tracts is post-childlaunching with some middle, early and late childrearing families also present. The sex ratios appear to be fairly balanced. None of the set D tracts are classified as single dwelling unit areas. Six of the seven tracts in this set are not overcrowded. (See figure 4.) Set E— Predominantly nonfamily areas; mixture of Negro and white nonfamily households, and Negro husband-wife and female-headed households; area households in post-child- launching family life cycle stage, some middle, early and late childrearing households also present; uncrowded, predominantly apartment areas. This set (tract 1402 only) has a family status 56 SOCIAL INDICATORS FOR SMALL AREAS Fig. 1 AN EXAMPLE OF A SET A AGE-SEX PYRAMID NEGRO POPULATION OF TRACT 402, COUNTY 510, MARYLAND, INNER CITY CMHC TOTAL POPULATION 3509 MALES 1503 FEMALES 2006 NUMBER % AGE 11 0.31 85+ 7 0.20 80-4 20 0.57 75-9 33 0.94 70-4 35 1.00 65-9 48 1.37 60-4 50 1.42 55-9 45 1.28 50-4 39 1.11 45-9 47 1.34 40-4 46 1.31 35-9 32 0.91 30-4 45 1.28 25-9 70 1.99 20-4 160 4.56 15-9 298 8.49 10-4 295 8.41 5-9 222 6.33 0-4 5 4 3 2 MM MM MMMM MMMM MMM MMM MMMMMMMMMf MMMMMMMMMP IMMMMMMMMMMMMMMM IMMMMMMMMMMMMMMM IMMMMMMMMMMMMMMM 1 F F F FF FFF FFF FFF FFF FFF FFF FFF FFF FFF FFF FFF FFF FFF FFF FF FFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FF FFFF FFF FFFFF FFFFFFFFF FFFFFFFFFFFFFFFFFF FFFFFFFFFFFFFFFFFFF FFFFFFFFFFFF % 0.48 0.37 0.63 0.85 1.17 1.11 1.85 2.17 2.48 2.34 3.08 3.70 3.48 4.02 5.56 8.52 8.75 6.61 NUMBER 17 13 22 30 41 39 65 76 87 82 108 130 122 141 195 299 307 232 Fig. 2. AN EXAMPLE OF A SET B AGE-SEX PYRAMID NEGRO POPULATION OF TRACT 1501, COUNTY 510, MARYLAND, PROVIDENT HOSPITAL CMHC TOTAL POPULATION 5174 MALES 2291 FEMALES 2883 NUMBER % AGE 987654321 1 23456789 % NUMBER 12 0.23 85+ F 0.46 24 19 0.37 80-4 M FF 0.77 40 58 1.12 75-9 MMM FFF 1.06 55 64 1.24 70-4 MMM FFFFF 1.89 98 94 1.82 65-9 MMMMM FFFFFFFF 2.94 152 105 2.03 60-4 MMMMMM FFFFF 1.78 92 85 1.64 55-9 MMMM FFFFFFF 2.42 125 104 2.01 50-4 MMMMMM FFFFFFFF 2.80 145 134 2.59 45-9 MMMMMMM FFFFFFFF 2.86 148 109 2.11 40-4 MMMMMM FFFFFFFFF 3.02 156 81 1.57 35-9 MMMM FFFFFFFFF 3.02 156 70 1.35 30-4 MMMM FFFFFFF 2.47 128 77 1.49 25-9 MMMM FFFFFFFF 3.00 155 116 2.24 20-4 MMMMMM FFFFFFFFF 3.27 169 260 5.03 15-9 MMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFF 5.76 298 330 6.38 10-4 MMMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFFFFF 7.02 363 320 6.18 5-9 MMMMMMMMMMMM MMMMMM FFFFFFFFFFFFFFFFFFF 6.44 333 253 4.89 04 MMMMMMMMMMMMMM FFFFFFFFFFFFFF 4.75 246 Fig. 3. AN EXAMPLE OF A SET C AGE-SEX PYRAMID NEGRO POPULATION OF TRACT 1001, COUNTY 510, MARYLAND, JOHNS HOPKINS HOSPITAL TOTAL POPULATION 3725 MALES 1800 FEMALES 1925 NUMBER % AGE 987654321 1 23456789 % NUMBER 10 0.27 85+ 0.13 5 10 0.27 80-4 0.19 7 17 0.46 75-9 M FF 0.75 28 26 0.70 70-4 MM FF 0.97 36 57 1.53 65-9 MMMM FFFFF 1.77 66 78 2.09 60-4 MMMMMM FFFFFF 2.12 79 95 2.55 55-9 MMMMMMM FFFFFFFF 2.68 100 97 2.60 50-4 MMMMMMM FFFFFFFF 2.74 102 92 2.47 45-9 MMMMMMM FFFFFFFF 2.95 110 93 2.50 40-4 MMMMMMM FFFFFFFF 2.90 108 78 2.09 35-9 MMMMMM FFFFFFF 2.39 89 86 2.31 30-4 MMMMMM FFFFFFF 2.66 99 68 1.83 25-9 MMMMM FFFFFFFF 2.74 102 103 2.77 20-4 MMMMMMMM FFFFFFFFFF 3.46 129 210 5.64 15-9 MMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFF 5.91 220 252 6.77 10-4 MMMMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFFFF 6.71 250 251 6.74 5-9 MMMMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFF 6.31 235 177 4.75 0-4 MMMMMMMMMMMMMM FFFFFFFFFFFF 4.30 160 MENTAL HEALTH DEMOGRAPHIC PROFILE 57 Fig. 4. AN EXAMPLE OF A SET D AGE-SEX PYRAMID NEGRO POPULATION OF TRACT 1601, COUNTY 510, MARYLAND, PROVIDENT HOSPITAL CMHC TOTAL POPULATION 6103 MALES 2817 FEMALES 3286 NUMBER % AGE 987654321 1 23456789 % NUMBER 17 0.28 85+ FF 0.69 42 31 0.51 804 M FF 0.88 54 77 1.26 75-9 MMM FFFF 1.64 100 119 1.95 70-4 MMMMM FFFFFFF 2.56 156 171 2.80 65-9 MMMMMMMM FFFFFFFFF 3.26 199 178 2.92 60-4 MMMMMMMM FFFFFFFFFF 3.54 216 194 3.18 55-9 MMMMMMMMM FFFFFFFFFFF 3.83 234 168 2.75 50-4 MMMMMMMM FFFFFFFFFF 3.65 223 152 2.49 45-9 MMMMMMM FFFFFFF 2.64 161 150 2.46 40-4 MMMMMMM FFFFFFF 2.59 158 126 2.06 35-9 MMMMMM FFFFFF 2.33 142 90 1.47 30-4 MMMM FFFFFF 2.31 141 141 2.31 25-9 MMMMMM FFFFFFF 2.65 162 188 3.08 20-4 MMMMMMMMM FFFFFFFFFFF 3.80 232 221 3.62 15-9 MMMMMMMMMM FFFFFFFFFFFFF 4.42 270 254 4.16 10-4 MMMMMMMMMMMM FFFFFFFFFFFF 4.31 263 279 4.57 59 MMMMMMMMMMMMM FFFFFFFFFFFFFF 4.83 295 261 4.28 0-4 MMMMMMMMMMMM FFFFFFFFFFF 3.90 238 Fig. 5 EXAMPLES OF A SET E AGE-SEX PYRAMID WHITE POPULATION OF TRACT 1402, COUNTY 510, MARYLAND, PROVIDENT HOSPITAL CMHC TOTAL POPULATION 349 MALES 121 FEMALES 228 NUMBER % AGE 987654321 123456789 % NUMBER 9 2.58 85+ MMMMMMM FFFFFF 2.01 7 5 1.43 80-4 MMMM FFFFFFFFFFFFFFFFFFFF 6.88 24 10 2.87 75-9 MMMMMMMM FFFFFFFFFFFFFFFFFFFFFFFFFFF* 10.60 37 14 4.01 70-4 MMMMMMMMMMMM FFFFFFFFFFFFFFFFFFFFFFFFFFF* 16.91 59 24 6.88 65-9 MMMMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFFFFFFFFFFF* 13.47 47 6 1.72 60-4 MMMMM FFFFFFFFFFFF 4.30 15 3 0.86 55-9 MM FFFF 1.43 5 6 1.72 50-4 MMMMM FFF 1.15 4 5 1.43 45-9 MMMM 0.29 1 5 1.43 40-4 MMMM FF 0.86 3 2 0.57 359 M 0.29 1 2 0.57 30-4 M FFFF 1.43 5 2 0.57 25-9 M FFF 1.15 4 13 3.72 20-4 MMMMMMMMMMM FFFFFFF 2.58 9 2 0.57 15-9 M FFF 1.15 4 4 1.15 10-4 MMM 0.29 1 3 0.86 59 MM 0.0 6 1.72 0-4 MMMMM F 0.57 2 NEGRO POPULATION OF TRACT TOTAL POPULATION 5223 1402, COUNTY 510, MARYLAND, MALES 2471 PROVIDENT HOSPITAL CMHC FEMALES 2752 NUMBER % AGE 987654321 123456789 % NUMBER 12 0.23 85+ F 0.48 25 17 0.33 80-4 FF 0.80 42 44 0.84 75-9 MM FF 1.00 52 94 1.80 70-4 MMMMM FFFFF 1.91 100 125 2.39 65-9 MMMMMMM FFFFFFF 2.51 131 108 2.07 60-4 MMMMMM FFFFFFFF 2.80 146 155 2.97 55-9 MMMMMMMM FFFFFFFFF 3.18 166 156 2.99 50-4 MMMMMMMM FFFFFFFFF 3.24 169 161 3.08 45-9 MMMMMMMMM FFFFFFFFF 3.08 161 167 3.20 40-4 MMMMMMMMM FFFFFFFF 2.89 151 120 2.30 35-9 MMMMMM FFFFFFF 2.37 124 116 2.22 30-4 MMMMMM FFFFFFFF 2.78 145 154 2.95 25-9 MMMMMMMM FFFFFFFF 2.89 151 164 3.14 20-4 MMMMMMMMM FFFFFFFFFFFFF 4.35 227 183 3.50 15-9 MMMMMMMMMM FFFFFFFFFFFF 4.23 221 220 4.21 10-4 MMMMMMMMMMMM FFFFFFFFFFFFFF 4.75 248 248 4.75 5-9 MMMMMMMMMMMMMM FFFFFFFFFFFFFF 5.00 261 227 4.35 0-4 MMMMMMMMMMMMM FFFFFFFFFFFFF 4.44 232 58 SOCIAL INDICATORS FOR SMALL AREAS pattern of 423 which indicates that most of the households of this tract are headed by primary individuals. This characteristic is unique to tract 1402; for that reason we have designated the tract as a separate type rather than placing it in set D. In terms of the Negro age-sex pyramid, the tract could have been a subtype of set D (see figure 5). Part of the difficulty in classification may relate to the fact that from 5 to 10 percent of the total population in the tract is white, predominantly female and over 65 (see figure 5). Even though the relative size of the white population is not large, it is likely that most of these individuals are female household heads living alone. The disproportionate number of such white households might distort the total figures so that they do not represent the characteristics of the dominant Negro population. Accordingly, in order to make more accurate estimates of family status for such tracts we need data by race. Such data will, of course, be available from the fourth count census tapes. The area family life cycle and residential life style characteristics are like those of set D. Set F— Family/nonfamily areas; predominantly Negro non- family households, husband-wife families and female-headed families or households; area family life cycle stage is middle childrearing with some early and late childrearing households; uncrowded, mixed (single dwelling units with apartments) housing areas. This set is represented by tract 1204. The family status pattern of the total population is 334. Quite likely, the family status pattern of the Negro population is 333 or 233. As in Set E, the difference between the total population pattern and the Negro pattern may be due to the small aged predominantly white population. The family life cycle of this tract is middle childrearing with some early and late childrearing households. It is an uncrowded, mixed (single dwelling units with apartments) housing area. (See Figure 6.) Poor White (Non-Negro) Census Tracts The 27 census tracts classified as "poor" and white are grouped into 6 sets (G through L). Together, Sets G and H contain 78 percent of the white tracts. The social area charac- teristics of each tract (presented in supplementary Appendix table 2) are summarized by type below. Sets G and H— Family areas; white husband-wife families with some female-headed households in the post-childlaunching family life cycle stage, other life cycle stages also present; uncrowded housing conditions. The 21 census tracts comprising sets G and H are described by one of two family status patterns: 242 or, less often, 243. These patterns describe family areas, specifically characterized by a preponderance of husband-wife families with some female-headed households. The post-childlaunching stage of the family life cycle is the primary classification for all tracts in sets G and H. In each instance, post childlaunching is accom- panied by at least one (often several) of the following stages: Middle childrearing, late childrearing, early childrearing, pre- family, childlaunching, childbearing. None of the tracts in sets G or H are overcrowded. Further, none of the tracts are predomi- nantly apartment areas; i.e., at least 50 percent, and often 80 percent or more, of the housing units are single dwelling units. Set H tracts differ from set G tracts only in terms of the proportion of the population Negro: 10 to 30 percent in set H and less than 5 percent in set G. (See figures 7 and 8.) Set I— Family/nonfamily areas; marked predominance of white husband-wife families with some female-headed households; many households in post-childlaunching family life cycle stage, some middle and early childrearing households also present; uncrowded, mixed (single dwelling units with apartments) housing areas. The family status pattern for set I of 342 describes family/nonfamily areas and, among families, mostly husband- wife families with some female-headed families. Set I (containing only tract 105) resembles set G in all but one respect, i.e., the larger number of primary individuals who are household heads. The predominant family life cycle stage is post-childlaunching with some middle and early childrearing. This is an uncrowded area, with single dwelling units accounting for from 50 to 80 percent of all housing units. (See figure 9.) Set J— Family/nonfamily areas; white husband-wife families and female-headed households in post-childlaunching family life cycle stage, some late and middle childrearing households also present; uncrowded, mixed (single dwelling units with apart- ments) housing areas. The family status pattern for set J of 333 describes family/nonfamily areas with families predominating (50 to 70 percent of households are families). Among families, husband-wife families predominate over female-headed families, but to a lesser extent than in set I areas. Post-childlaunching is the predominant family life cycle stage with some middle and iate childrearing (also some early childrearing in one tract). Set J has the same residential life style characteristics as set I. (See figure 10.) Set K— Family areas; predominantly white husband-wife families and female-headed households in the late and middle childrearing family life cycle stages; overcrowded, single dwelling units areas. The family status pattern for predominantly white Set K (tract 2603.03 only) of 233 delineates a family area (70 to 84 percent families) with many husband-wife families and female-headed families. The family life cycle classification for the white population of this tract is iate and middle childrearing with some early childrearing and post-childlaunching. With respect to residential life style, over 80 percent of the housing units are single dwelling units and the area is overcrowded. (See Figure 11.) Set L— Nonfamily areas; predominantly white prefamily and post-childlaunching households; uncrowded, predominantly apartment areas. The family status pattern for predominantly white set L (tract 1206 only) of 423 describes a nonfamily area (50 to 70 percent of the households are headed by primary individuals) with many female-headed households (from 30 to 50 percent of all households). Consistent with this nonfamily designation, set L receives a family life cycle classification of prefamily and post-childlaunching. The area is uncrowded, and over 50 percent of the housing units are apartments. (See figure 12.) Summary And Conclusions We have presented the Mental Health Small Area Demographic Profile System (MHDP) and described procedures (using that system) for identifying the major social area axes— economic class, social class, information status, family status, family life cycle and residential life style. Specific items to be used in indexing these dimensions (and other social area dimensions such as familism and community stability) are presented in appendix tables 1 through 4. The social area procedures as presented have general applicability to central city populations as well as to populations of suburban or rural areas. We have illustrated the use of these procedures with selected poor areas of Baltimore city. This descriptive analysis can be seen as yielding a preliminary typology (made possible by the MHDP system) of "poor" areas that merits refinement and further work. The system allows identification not only of types of poor MENTAL HEALTH DEMOGRAPHIC PROFILES Fig. 6 EXAMPLES OF A SET F AGE-SEX PYRAMID WHITE POPULATION OF TRACT 1204, COUNTY 510, MARYLAND, NORTH CENTRAL CMHC TOTAL POPULATION 235 MALES 106 FEMALES 129 59 NUMBER % AGE 987654321 1 23456789 % NUMBER 0.0 85+ F 0.43 1 2 0.85 80-4 MM FFF 1.28 3 3 1.28 75-9 MMM FFFFFFFFFFF 3.83 9 9 3.83 70-4 MMMMMMMMMMM FFFFFFFFFFFFFFFF 5.53 13 6 2.55 659 MMMMMMM FFFFFFFFFFFF 4.26 10 11 4.68 60-4 MMMMMMMMMMMMMM FFFFFFFFFFFFFFF 5.11 12 10 4.26 55-9 MMMMMMMMMMMM FFFFFFFFFF FFFFFFFFFFFF 7.66 18 9 3.83 50-4 MMMMMMMMMMM FFFFFFFFFFF 3.83 9 9 3.83 45-9 MMMMMMMMMMM FFFFFFFFFFFFFFF 5.11 12 8 3.40 40-4 MMMMMMMMMM FFFFFFF 2.55 6 7 2.98 35-9 MMMMMMMM 0.0 4 1.70 30-4 MMMMM FF 0.85 2 3 1.28 25-9 MMM FFFFF 1.70 4 13 5.53 20-4 MMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFF 5.96 14 3 1.28 15-9 MMM FFFFF 1.70 4 2 0.85 10-4 MM F 0.43 1 0.0 5-9 FFF 1.28 3 7 2.98 0-4 MMMMMMMM FFFFFFFFFF 3.40 8 NEGRO POPULATION OF TRACT 1204, TOTAL POPULATION 3222 COUNTY 510, MARYLAND, MALES 1533 NORTH CENTRAL CMHC FEMALES 1689 NUMBER % AGE 987654321 1 23456789 % NUMBER 3 0.09 85+ 0.22 7 7 0.22 80-4 F 0.50 16 17 0.53 75-9 M F 0.50 16 23 0.71 70-4 MM FFF 1.06 34 35 1.09 65-9 MMM FFFF 1.40 45 46 1.43 60-4 MMMM FFFF 1.58 51 64 1.99 559 MMMMM FFFFF 1.92 62 84 2.61 50-4 MMMMMMM FFFFFFF 2.64 85 96 2.98 45-9 MMMMMMMM FFFFFFFF 2.82 91 83 2.58 40-4 MMMMMMM FFFFFFFF 2.95 95 70 2.17 35-9 MMMMMM FFFFFFFF 2.89 93 71 2.20 304 MMMMMM FFFFFFFFF 3.26 105 67 2.08 25-9 MMMMMM FFFFFFFF 2.86 92 99 3.07 204 MMMMMMMMM FFFFFFFFFF 3.63 117 177 5.49 15-9 MMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFFF 6.58 212 203 6.30 10-4 MMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFFFF 6.70 216 202 6.27 5-9 MMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFF 5.71 184 186 5.77 0-4 MMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFF 5.21 168 Fig 7. AN EXAMPLE OF A SET G AGE-SEX PYRAMID WHITE POPULATION OF TRACT 101, COUNTY 510, MARYLAND, BALTIMORE CITY HOSPITALS TOTAL POPULATION 3944 MALES 1867 FEMALES 2077 NUMBER % AGE 987654321 1 23456789 % NUMBER 7 0.18 85+ F 0.41 16 12 0.30 80-4 FF 0.76 30 32 0.81 75-9 MM FFFF 1.50 59 51 1.29 70-4 MMM FFFFFFF 2.38 94 84 2.13 65-9 MMMMMM FFFFFFFF 2.84 112 98 2.48 60-4 MMMMMMM FFFFFFFFF 3.04 120 154 3.90 559 MMMMMMMMMMM FFFFFFFFFFF 3.68 145 130 3.30 50-4 MMMMMMMMM FFFFFFFFFFFF 4.01 158 135 3.42 45-9 MMMMMMMMMM FFFFFFFFFFF 3.73 147 102 2.59 40-4 MMMMMMM FFFFFFFFF 3.32 131 99 2.51 35-9 MMMMMMM FFFFF 1.95 77 73 1.85 30-4 MMMMM FFFFFFF 2.54 100 108 2.74 25-9 MMMMMMMM FFFFFFF 2.56 101 138 3.50 20-4 MMMMMMMMMM FFFFFFFFFFFFF 4.41 174 158 4.01 15-9 MMMMMMMMMMMM FFFFFFFFFFFFF 4.46 176 181 4.59 10-4 MMMMMMMMMMMMM FFFFFFFFFFFF 4.13 163 165 4.18 5-9 MMMMMMMMMMMM FFFFFFFFFFFF 4.11 162 140 3.55 0-4 MMMMMMMMMM FFFFFFFF 2.84 112 60 SOCIAL INDICATORS FOR SMALL AREAS Fig. 8. AN EXAMPLE OF A SET H AGE-SEX PYRAMID WHITE POPULATION OF TRACT 1903. COUNTY 510, MARYLAND, INNER CITY CMHC TOTAL POPULATION 3735 MALES 1796 FEMALES 1939 NUMBER % AGE 987654321 123456789 % NUMBER 11 0.29 85+ 0.29 11 11 0.29 80-4 FF 0.70 26 9 0.24 75-9 FF 0.99 37 36 0.96 70-4 MM FFFF 1.39 52 72 1.93 65-9 MMMMM FFFFFF 2.09 78 106 2.84 60-4 MMMMMMMM FFFFFF 2.01 75 105 2.81 55-9 MMMMMMMM FFFFFFFF 2.92 109 106 2.84 50-4 MMMMMMMM FFFFFFFF 3.00 112 99 2.65 45-9 MMMMMMM FFFFFFFFF 3.24 121 105 2.81 40-4 MMMMMMMM FFFFFFFF 2.70 101 84 2.25 35-9 MMMMMM FFFFFF 2.17 81 85 2.28 30-4 MMMMMM FFFFFFF 2.36 88 119 3.19 25-9 MMMMMMMMM FFFFFFFFFF 3.37 126 117 3.13 20-4 MMMMMMMMM FFFFFFFFFFFF 4.31 161 151 4.04 15-9 MMMMMMMMMMMM FFFFFFFFFFFF 4.07 152 179 4.79 10-4 MMMMMMMMMMMMMM FFFFFFFFFFFFFFF 5.19 194 186 4.98 5-9 MMMMMMMMMMMMMM FFFFFFFFFFFFFFF 5.01 187 215 5.76 0-4 MMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFF 6.10 228 Fig. 9 AN EXAMPLE OF A SET I AGE-SEX PYRAMID WHITE POPULATION OF TRACT 105, COUNTY 510, MARYLAND, BALTIMORE CITY HOSPITALS TOTAL POPULATION 2754 MALES 1362 FEMALES 1392 NUMBER % AGE 987654321 123456789 % NUMBER 5 0.18 85+ 0.25 7 20 0.73 80-4 MM FF 0.80 22 23 0.84 75-9 MM FFFF 1.63 45 33 1.20 70-4 MMM FFFFF 1.85 51 50 1.82 65-9 MMMMM FFFFFFF 2.54 70 81 2.94 60-4 MMMMMMMM FFFFFFFFF 3.12 86 86 3.12 55-9 MMMMMMMMM FFFFFFFFF 3.30 91 92 3.34 50-4 MMMMMMMMMM FFFFFFFFF 3.01 83 87 3.16 45-9 MMMMMMMMM FFFFFFFFFF 3.45 95 85 3.09 40-4 MMMMMMMMM FFFFFFFFF 3.23 89 81 2.94 35-9 MMMMMMMM FFFFFF 2.29 63 73 2.65 30-4 MMMMMMM FFFFFF 2.32 64 77 2.80 259 MMMMMMMM FFFFFFF 2.58 71 131 4.76 20-4 MMMMMMMMMMMMMM FFFFFFFFFFFFFF 4.90 135 97 3.52 15-9 MMMMMMMMMM FFFFFFFFFF 3.52 97 115 4.18 10-4 MMMMMMMMMMMM FFFFFFFFFFF 3.89 107 116 4.21 5-9 MMMMMMMMMMMM FFFFFFFFFFF 3.78 104 110 3.99 0-4 MMMMMMMMMMM FFFFFFFFFFFF 4.07 112 Fig. 10. AN EXAMPLE OF A SET J AGE-SEX PYRAMID WHITE POPULATION OF TRACT 202, COUNTY 510, MARYLAND, BALTIMORE CITY HOSPITALS TOTAL POPULATION 2763 MALES 1399 FEMALES 1364 NUMBER % AGE 987654321 123456789 % NUMBER 7 0.25 85+ 0.22 6 11 0.40 80-4 M FF 0.80 22 28 1.01 75-9 MMM FFF 1.05 29 55 1.99 70-4 MMMMM FFFFFF 2.06 57 57 2.06 65-9 MMMMMM FFFFFF 2.14 59 107 3.87 60-4 MMMMMMMMMMM FFFFFFF 2.50 69 97 3.51 55-9 MMMMMMMMMM FFFFFFFFFFF 3.73 103 137 4.96 50-4 MMMMMMMMMMMMMM FFFFFFFFF 3.08 85 110 3.98 45-9 MMMMMMMMMMM FFFFFFFFFFF 3.87 107 92 3.33 40-4 MMMMMMMMM FFFFFFFFF 3.11 86 64 2.32 35-9 MMMMMM FFFFFFFF 2.68 74 44 1.59 30-4 MMMM FFFFFF 2.14 59 59 2.14 25-9 MMMMMM FFFFFFFF 2.97 82 121 4.38 20-4 MMMMMMMMMMMMM FFFFFFFFF 3.08 85 96 3.47 15-9 MMMMMMMMMM FFFFFFFFFFFFFF 4.74 131 103 3.73 10-4 MMMMMMMMMMM FFFFFFFFFFFF 4.05 112 106 3.84 5-9 MMMMMMMMMMM FFFFFFFFFFF 3.76 104 105 3.80 0-4 MMMMMMMMMMM FFFFFFFFFF 3.40 94 MENTAL HEALTH DEMOGRAPHIC PROFILES Fig. 1 1 . AIM EXAMPLE OF A SET K AGE-SEX PYRAMID WHITE POPULATION OF TRACT 2603.03, COUNTY 510, MARYLAND, AREA VII TOTAL POPULATION 1368 MALES 619 FEMALES 749 61 NUMBER % AGE 987654321 123456789 % NUMBER 2 15 85+ F 0.37 5 6 0.44 80-4 M FF 0.73 10 3 0.22 75-9 FF 0.88 12 12 0.88 70-4 MM FFFF 1.39 19 15 1.10 65-9 MMM FFF 1.17 16 22 1.61 60-4 MMMM FFFFF 1.75 24 19 1.39 55-9 MMMM FFFFFFFFFF 3.36 46 26 1.90 50-4 MMMMM FFFFFFFF 2.78 38 37 2.70 45-9 MMMMMMMM FFFFFFFF 2.70 37 23 1.68 40-4 MMMMM FFFFFFFF 3.00 41 25 1.83 35-9 MMMMM FFFFFFFFF 3.22 44 21 1.54 30-4 MMMM FFFFFFFFF 3.14 43 30 2.19 25-9 MMMMMM FFFFFFFF 3.00 41 39 2.85 20-4 MMMMMMMM FFFFFFFFFF 3.44 47 92 6.73 15-9 MMMMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFF 6.14 84 96 7.02 10-4 MMMMMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFFFFF 7.24 99 90 6.58 59 MMMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFF 5.85 80 61 4.46 0-4 MMMMMMMMMMMMM FFFFFFFFFFFFF 4.61 63 Fig. 1 2. AN EXAMPLE OF A SET L AGE-SEX PYRAMID WHITE POPULATION OF TRACT 1206, COUNTY 510, MARYLAND, NORTH CENTRAL CMHC TOTAL POPULATION 2415 MALES 1226 FEMALES 1189 NUMBER % AGE 987654321 1 23456789 % NUMBER 7 0.29 85+ FF 0.99 24 22 0.91 80-4 MM FFFF 1.49 36 37 1.53 75-9 MMMM FFFFFF 2.11 51 40 1.66 70-4 MMMM FFFFFFFF 2.90 70 65 2.69 65-9 MMMMMMMM FFFFFFFFF 3.23 78 95 3.93 60-4 MMMMMMMMMMM FFFFFFFFFF 3.64 88 79 3.27 55-9 MMMMMMMMM FFFFFFFFFFF 3.89 94 80 3.31 50-4 MMMMMMMMM FFFFFFFFFFFF 4.02 97 75 3.11 45-9 MMMMMMMMM FFFFFFFF 2.77 67 65 2.69 40-4 MMMMMMMM FFFFFFF 2.44 59 57 2.36 35-9 MMMMMMM FFFFF 1.95 47 67 2.77 30-4 MMMMMMMM FFF 1.20 29 118 4.89 25-9 MMMMMMMMMMMMMM FFFFFFFFFFF 3.69 89 197 8.16 20-4 MMMMMMMMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFF 6.00 145 78 3.23 15-9 MMMMMMMMM FFFFFFFFF 3.31 80 46 1.90 10-4 MMMMM FFFF 1.49 36 46 1.90 5-9 MMMMM FFFF 1.57 38 52 2.15 0-4 MMMMMM FFFFFFF 2.53 61 62 SOCIAL INDICATORS FOR SMALL AREAS populations but also of other subpopulations (such as the aged, female-headed households with children, and populations living in overcrowded quarters) that may run a high risk of health and related problems. In conclusion, we see the identification of distinctive residen- tial subareas, which in turn have distinctive problems, as a first step in and an integral part of dealing preventively with social problems. In our view, the application of the social area procedures that we have outlined here will result in the identification of distinctive subareas. Thus, with access to the programs that provide the NIMH small area profiles, one can obtain relevant census data rapidly, at a low cost, and in a useful format for effective planning. Given the profiles, one can begin to analyze local census data by applying the procedures suggested in this paper. References Foote, Nelson N., Janet L. Abu-Lughod, Mary M. Foley, and Louis Winnick. 1960 Housing Choices and Housing Constraints. New York- McGraw-Hill. Glick, Paul C. 1957 American Families. New York: John Wiley. Goldsmith, Harold F., Edward G. Stockwell, and Elizabeth Unger. 1970 A Technique for Classifying Population Age Profiles. Laboratory Paper No. 33, Adelphi, Maryland: Men- tal Health Study Center, National Institute of Mental Health. Goldsmith, Harold F., and Elizabeth L. Unger. 1970 Differentiation of Urban Subareas: A Re-Examina- tion of Social Area Dimensions. Laboratory Paper No. 35, Adelphi, Maryland: Mental Health Study Center. National Institute of Mental Health. 1972a Social Areas: Identification Procedures Using 1970 Census Data. Laboratory Paper No. 37, Adelphi, Maryland: Mental Health Study Center, National Institute of Mental Health. 1972b Social Rank and Family Life Cycle: An Ecological Analysis. Laboratory Paper No. 43, Adelphi, Mary- land: Mental Health Study Center, National Insti- tute of Mental Health. Redick, Richard W., Harold F. Goldsmith, and Elizabeth L. Unger. 1971 1970 Census Data Used to Indicate Areas with Different Potentials for Mental Health and Related Problems. National Institute of Mental Health Statis- tics Methodology Reports, Series C, No. 3. Wash- ington, D.C.: U.S. Government Printing Office. U.S. Bureau of the Census. 1972 Census of Population: 1970, General Social and Economic Characteristics, Final Report PC(1)-C1, United States Summary. Washington, D.C.: U.S. Government Printing Office. MENTAL HEALTH DEMOGRAPHIC PROFILE 63 Table 1. Similarity and Disparity Between Family Life Cycle Classifications of Census Tracts Produced by Visual Code and 1972 Age Code (Specific Measure of Comparison: First Two Visual Code Designations) by Ethnic Status and Family Status: Poor Census Tracts-Baltimore City, 1970 Number of tracts First 2 visual code designations occur as first two 1972 age code designations One of first 2 visual code designations occurs in first two 1972 age designations First two visual Selected poor census tracts 1 by ethnic status and family status 2 Remaining visual code designation occurs after first two age code designations Remaining visual code designation does not occur among 1972 age code designations code designations do not occur among 197 2 age code designations 50 23 10 12 1 27 22 4 1 42 20 10 9 1 22 19 2 1 6 3 3 3 2 1 2 2 1 1 _ - _ _ 'See text for designations. 2 Family status classification: Family areas — less than 30% of households headed by primary individuals. Family/Nonf amily areas — 30-4 9% of households headed by primary individuals. Nonfamily areas — 50% or more of households headed by primary individuals. SOURCE: Supplementary appendix tables 1 and 2. Table 2. Similarity and Disparity Between Family Life Cycle Classifications of Census Tracts Produced by 4-ltem Disk Code and 1972 Age Code (Specific Measure of Comparison: First Two Four-Item Disk Code Designations) by Ethnic Status and Family Status: Poor Census Tracts-Baltimore City, 1970 Selected poor census tracts 1 by ethnic status and family status 2 Number of tracts First 2-disk code designations occur as first two 1972 age code designations One of first two 4-item disk code designations occurs in first two 1972 age designations Remaining disk code designation occurs after first two age code designations Remaining disk code designation does not occur among 1972 age code designations First 2-disk code designations do not occur among 197 2 age code designations Selected poor tracts. Poor Negro tracts Family areas Family/nonf amily areas. Nonfamily areas Poor white tracts Family areas Family/nonfamily areas. Nonfamily areas 23 10 12 1 27 22 4 1 'See text for designations. 2 Family status classification: Family areas — less than 30% of households headed by primary individuals. Family/Nonf amily areas — 30-49% of households headed by primary individuals. Nonfamily areas — 50% or more of households headed by primary individuals. SOURCE: Supplementary appendix tables 1 and 2. 64 SOCIAL INDICATORS FOR SMALL AREAS APPENDIX TABLE 1 PRINCE GEORGE'S COUNTY, MARYLAND SELFCTED STATISTICS FROM 1970 CENSUS OF POPULATION AND HOUSING. SECOND COUNT (100? SAMPLF ) DATA FILES COMPARISONS WITH OTHER AREAS STATISTIC DESCRIPTION NUMPEP, DENOMINATOR PERCENT, FOR PFRCFNTS SMSA STATE U.S. PERCENT MEDIAN RATIO OR AND PATIOS OP — OF VALUES LESS THAN VALUE VALUE FOR THE POPULATION RATIO RATIO RATIO THE ALL COUNTIES TOTAL FOR MEDIANS OR OR OR IN AREA AND AVFRAGFS VALUE (A) VALUE VALUE COUNTY (C) MARYLAND (C) 660567 2861123 3922399 203211905 92.0 53764 316987 1331512 1851019 95456663 92.0 24804 32706" 1443391 1966563 101943250 92.0 26225 16511 86220 104817 5812013 88.0 1589 561476 2124903 3194888 177748975 92.0 46423 91808 703745 699479 22580289 92.0 7424 $237*3 91656 128542 $18776 $17450 88.0 $15849 $143 97629 $140 $110 $ 90 (S) 92.0 $ 74 GENERAL POPULATION DATA (1) TOTAL POPULATION (2) NUMPFR OF MALES ( I'M HOUSEHOLDS) (3) NUMBER OF FEMALFS (IN HOUSEHOLDS) (4) POPULATION IN GROUP QUARTERS (5) POPULATION WHITE (6) POPULATION NEGRO SOCIOECONOMIC STATUS (7) HOUSE VALUE: MEDIAN VALUE OF OWNER- OCCUPIFD AND VACANT FOR SALE HOUSING UNITS (8) RENT: MEDIAN RENT HP P FMTER-OCCUPIED AND VACANT FOR RFNT HOUSING UNITS ETHNIC COMPOSITION (9) NEGRO: PERCENT OF POPULATION NFGRO (IN HOUSEHOLDS) (10) OTHEP NCNWHITE PACES: PERCENT OF POPULATION NEITHER WHITE NOP NEGRO (IN HOUSEHOLDS) HOUSEHOLD COMPOSITION AND FAMILY STRUCTURE (11) HUSRANP-WIFE HOUSEHOLDS: PERCENT OF ALL HOUSEHOLDS HUSBAND-WIFE (12) AGE OF HOUSFHOLD HEADS: MFDIAN AGE OF HOUSEHOLD HEAD (13) YOUTH DEPENDENCY RATIO: PERSONS UNDER 18 PEP 100 PERSONS 18-64 (IN HOUSEHOLDS) (14) AGED DEPENDENCY RATIO: PERSONS 65 AND OVFR PER 100 PERSONS 18-64 (IN HOUSEHOLDS) 13.9 64^.056 24.6 17.7 11.1 46.0 15.9 1. 1 6440 56 1.1 0.7 1.4 92.0 0.2 75.1 192962 64.8 70.9 69.4 54.0 74.3 39.2 192962 45.0 45.7 48.2 4.0 48.7 65.3 374295 60.1 63.4 63.4 63.0 64.1 6.8 374295 9.8 13.1 17.4 0.0 17.2 TYPE OF HOUSING (URBANIZATION) (15) ONE UNIT STRUCTURES: PERCENT OF OCCUPIED AND VACANT YEAR-ROUND HOUSING UNITS WITH ONE UNIT AT ADDRESS 55.6 200211 53.9 0.0 CONDITION OF HOUSING (16) OVFPCROWDING: PERCENT OF PERSONS IN HOUSE- HOLDS WITH 1.01 OR MORE PERSONS PER ROOM (17) STANDARD HOUSING: PERCENT OF OCCUPIED HOUSING UNITS WITH DIRECT ACCESS AND WITH COMPLETE PLUMBING AND KITCHEN FACILITIES FOR EXCLUSIVF USE 12.5 644056 13.3 13.6 16.9 25.0 14.5 98.2 192962 97.8 95.4 93.4 92.0 88.9 Symbols apply to this and all subsequent tables. ... Category not applicable. NA Not available . (A) Medians estimated from SMSA county medians and county populations. (B) Occupied housing units. (C) Based on 23 counties and Baltimore city. MENTAL HEALTH DEMOGRAPHIC PROFILE 6b APPENDIX TABLE 2 PRINCE GEORGE'S COUNTY, MARYLAND ADDITIONAL STATISTICS FROM 1970 CENSUS OF POPULATION AND HOUSING, SfcCONC COUNT (100X SAMPLF) DATA FILES COMPARISONS WITH OTHER AREAS STATISTIC QESCRIPTION NUMBER, PFRCFNT, RATIO OP DENOMINATOR FOR PERCFNTS AND RATIOS OR POPULATION FOP MEDIANS AND AVERAGES SMSA STATF U.S. RATIO OR VALUE PERCENT MEDIAN OF VALUES VALUE VALUF FOP THE TOTAL AREA RATIO OR VALUE (A) RATIO OR VALUE LFSS THAN THE ALL COUNTIES IN COUNTY (B) MARYLAND (B) $19455 R- $15«91 70790 10816 $20015 $15590 $12581 $ 7433 $11265 $6330 88.0 $11004 92.0 $ 5651 SOCIOECONOMIC STATUS (1) LOW HOUSF VALUE, NON-NEGRO: LOWER OUAR- TILE HOUSE VALUE OF NON-NEGRO HEAOFr, OWNER OCCUPIED HOUSING UNITS (2) LOW HOUSE VALUE, NEGRO: LOWER OUARTILF HOUSE VALUE OF NFGRO HFAOFD, OWNER-OCCUPIED HOUSING UNITS (3) LOW RFNT, NON-NEGPO: LEWER OUARTILF MONTHLY RFNTAL OF NON-NEGRO HEADED, RENTER- OCCUPIED HOUSING UNITS (4) LOW RFNT, NEGRO: LPWFR OUARTILE MONTHLY RENTAL OF NFGRO HEADFD, RENTER-OCCUPIED HOUSING UNITS $103 P?107 $ 85 $65 $49 92.0 92.0 $ 58 HOUSFHOLD COMPOSITION AMD FAMILY STPUCTUPF GENERAL CHARACTERISTICS »5) MEOIAN HOUSEHOLD SIZF (6) SMALL HOUSEHOLDS: PERCENT PF HOUSEHOLDS WITH ONE PERSON (7) LA»GE HPUSFHOLPS: PERCENT PF HOUSEHOLDS WITH 6 OR MORF PERSONS <8) SEX RATIO: MALES PER 1U0 FEMALES (POPULA- TION IN HOUSEHOLDS) (O) FERTILITY RATIO: CHILCRtN UNPFF. 5 PER 100C FEMALES 15-44 YFARS (HOUSEHOLD POPULATION) (10) CHILDREN LIVING WITH °PTH PARENTS: PERCENT OF PERSONS UNDER IP IN HOUSEHOLDS WHO APE OWN CHILDREN IN HUSBAND-WIFf HOUSEHOLDS 3.1 1°?062 2.G 2.9 2.7 58.0 2.9 12.0 l°? c '6? 1°.4 14.9 17.6 29.0 13.4 11.1 19296;: 9.C 11.3 10.4 50.0 11.0 c-b .0 327069 9T.3 94.1 93.6 58.0 95.5 443.4 l? c 706 :- 1 7 . 7 412.6 415 .7 71.0 420.9 £6.0 24444o ei.2 81.0 02. 6 75.0 83.8 FAMILY LIFE CYCLE (12) FAMILIES WITH CHILDREN: PERCENT PF FAM- ILIES WITH OWN CHILDREN UNCER 18 (13) CHILDBFAPIMG ONLY FAMILIES: PFRCFNT OF FAMILIES WITH OWN CHILDREN UNDER 6, NO CHILDREN 6-17 (14) CHILOBEARING AND REARING FAMILIES: PERCENT OF FAMILIFS WITH OWN CHILDREN BOTH UNDER 6 AND 6-17 (15) CHILDRFARING ONLY FAMILIFS: PFRCENT (16) CHILDRFARING COMPLETED: PERCENT OF HUSBAND -WIFE FAMILIES WITH HUSBAND 4= OR OVER AND WITH NO CHILDREN PRESENT PEPSONS NOT IN FAMILIES (17) GROUP QUARTERS: PFRCENT OF TOTAL POPULA- TION LIVING IN GROUP QUARTERS (18) HOUSEHOLDS WITH PRIMARY INDIVIDUALS: PERCENT OF HOUSEHOLDS WITH HEADS WHO ARF PRIMARY INDIVIDUALS (19) NPN-PFLATI VES: PERCENT PF PFPSONS IN HOUSEHOLDS WHO APE NDN-P EL AT I VES OF THF HOUSEHOLD HEAD 63.5 163379 59.7 57.4 54.9 88.0 55.3 18.4 163379 15.3 13.5 12.7 92.0 11.9 15.7 163370 2.2 644056 14.1 14.1 2.0 13.4 1 .6 71.0 13.6 29.3 163370 30.3 29.8 28.7 67.0 28.3 20.7 144880 26.0 30.7 34.7 4.0 35.5 660567 3.0 2.7 2.9 54.0 2.3 15.3 192962 23.3 17.3 19.7 46.0 15.5 1.7 DISRUPTED FAMILIES (20) DIVORCFD OR SEPARATED MALES: PERCENT OF MALFS 14 AND OVFP WHO ARE DIVORCED OR SEPARATFO (21) DIVORCED OP SEPAPATFD FEMALES: PERCENT OF FEMALES 14 AND OVER WHP ARE DIVORCED OR SEPARATED (22) WIDOWED FEMALES: PFRCENT OF FEMALES 14 AND OVFP WHO ARE WIDOWS (23) FEMALE HEADED HOUSFHOLPS: PFPCFNT OF HOUSFHOLOS WITH FE^ALF HEAD (?4) FEMALE HEADFD HOUSEHOLDS WITH OWN CHILDREN: PFRCFNT OF HOUSEHOLDS WITH OWN CHILDREN THAT ARE HFADFD BY FFMALFS 4.0 224616 6.3 237020 5.5 P. 2 5.0 7.0 4.3 6.2 38.0 79.0 4.1 5.2 7.5 237020 9.6 11.1 12.4 0.0 12.0 15.9 192962 23.4 20.0 21.0 33.0 17.1 9.0 103756 12.4 11.2 10.7 50.0 8.7 SINGLE ADULTS (25) SINGLF MALES: PERCENT OF MALES 25 AND OVER WHO HAVE NEVER MARRIED (26) SINGLE FEMALES: PERCENT OF FEMALES 25 AND OVER WHO HAVE NEVER MARRIED 7.8 156744 11.3 9.1 8.9 25.0 8.8 5.6 163183 10.2 7.1 7.0 21.0 6.7 TYPE OF HOUSING (URBANIZATION) (27) LOW OWNER-OCCUPANCY: PERCENT OF OCCUPIED HOUSING UNITS WHICH ARE RENTER-OCCUPIED (28) TRAILERS: PERCENT OF ALL HOUSING UNITS THAT ARE MOBILE HOMES OR TRAILERS 49.9 192962 54.0 41.2 37.1 92.0 32.1 0.7 200211 0.7 1.5 2.7 8.0 3.5 SEE FOOTNOTES AT END OF TABLE. 66 SOCIAL INDICATORS FOR SMALL AREAS APPENDIX TABLE 2 — CONTINUED PRINCE GEORGE'S COUNTY, MARYLAND ADDITIONAL STATISTICS FROM 1970 CENSUS OF POPULATION AND HOUSING, SECOND COUNT UOOX SAMPLE) DATA FILES STATISTIC DESCRIPTION COMPARISONS WITH OTHER AREAS NUMBER, DENOMINATOR PFPCENT, FOP PFRCFNTS SMSA RATIO OP AND PATIOS VALUE OR FOP THE POPULATION RATIO TOTAL FOP MEDIANS OR AREA AND VALUE (A) &VEP,AL-r.s CONDITION OF HOUSING STATF U.S. PERCENT MEDIAN OF VALUES LFSS THAN VALUE RATIO RATIO THE ALL COUNTIES OR OR IN VALUE VALUE COUNTY (B) MARYLAND (B) HPUSING (29) VACANCY INDEX: PEPCFNT OF TOTAL HOUSING UNITS VACANT YF AP-RPUNC (30) STANDAPD HOUSING, NPN-N C GRP: PERCFNT OF NON-NFGRO HFAPEP, OCCUPIFP HOUSING UNITS WITH BIPECT ACCESS AND WITH COMPLETE PLUMB- ING AND KITCHEN FACILITIFS FOP FXCLUSIVF USE (31) STANDARD HOUSING, NEGRO: PFPCENT OF NF&PP HEACFD, CCCUP1FD HOUSING UNITS WITH DIRECT ACCFSS AND WITH COMPLETE PLUMBING AND KIT- CHEN FACILITIES FC'P EXCLUSIVF USE 200211 4.2 4.8 6.2 13.0 6.2 169792 9P.3 96.6 94.6 92.0 92.4 23170 88.0 58.5 DENSITY (32) NON-NFGPO POPULATION IN OVERCROWDED HOUS- ING: PERCENT OF T>-F NON-NF&RO HEADED HOUSEHOLD POPULATION IN HOUSING WITH 1.01 OP MORF PFRSONS P r P ROOM (33) NFGRC POPULATION IN CVFRCRPWDEO HOUSING: PERCENT PF THE NEGPP HFACLO HOUSEHOLD POPULATION TN HOUSING WTTH 1.01 OP MOPE PER- SONS PFR ROOM (34) OVFRCRPWDFD HOUSING: PFPCFNT OF CCCUPIFT UNITS WITH 1.01 OP ^OPF PFRSONS PEP ROOM (35) PERSONS IN HIGHLY OVERCROWDED HOUSING: PFPCENT OF THE HOUSEHOLD POPULATION IN HOUS- ING WITH 1.51 OR MC3F PERSPNS PER ROOM 10. 554364 27. 9 itQfup 7.5 31.1 9.7 14.3 38.1 17.0 10.3 35.6 6.4 1>??962 6.9 6.6 8.2 33.0 7.0 2.7 644056 3.9 3.0 5.0 25.0 3.8 GROUP CUAPTFRS 136) INMATES PF INSTITUTIONS: L«TIPN IN GPCUP QUARTERS OF INSTITUTIONS PFPCENT OF PPPU- WHO ARE INMATES 14.5 16511 37.9 42.0 (A) Medians estimated from SMSA county medians and county populations. (B) Based on 23 counties and Baltimore city. MENTAL HEALTH DEMOGRAPHIC PROFILE 67 APPENDIX TABLE J PRINCE GEORGE'S COUNTY . MARYLAND SFLFCTFD STATISTICS FROM 1970 CENSUS OF POPULATION ANU HOUSING, SFCOND TOLL'S FOURTH COUNT (ALL FILES) COMPARISONS WITH OTHER AREAS STATISTIC DFSCPIPTION GFNFRAL POPULATION ^ATA (1) TOTAL POPULATION (2) NUMPER OF MALES (IN HOUSEHOLDS) (?) NUMBF" OF FEMALFS (IN HOUSFHOLDS) (4) POPULATION IN GROUP QUARTERS (5) POPULATION WHITE (6) POPULATION NFGRO NUMBFR, DENOMINATOR PERCFNT, FOR PfRCENTS SMSA STATF U.S. PERCENT MEDIAN RATIO OR AND RATIOS OR OF VALUES LESS THAN VALUE VALUE OF FOP THF POPULATION RATIO RATIO RATIO THE ALL COUNTIES TOTAL FOR MEDIANS OR OR OR IN AREA AND . AVERAGFS VALUF (A) VALUF VALUE COUNTY B) MARYLAND (B) 660567 2861123 3922399 203211905 92 53764 316987 13315 12 1851019 95456663 92 24804 32706° 1443391 1966563 101943250 92 26225 16511 P6220 104017 5812013 88 1589 561476 2124903 3194886 177748975 92 46423 V1808 703745 6Q9479 22580289 92 7424 SOCIOECONOMIC STATUS ECONOMIC STATUS (7) INCOME OF FAMILIES AND UNRELATED INDIVIDUALS: MEDIAN INCOMF OF FAMILIES AND UNRELATED INDIVIDUALS (P) FAMILIES IN POVERTY: PFRCENT OF ALL FAMILIES "FLOW POVERTY LEVEL J10467 218761 4.3 163400 $ 9130 $ 7,699 7.7 10.7 13 0.0 11.9 SOCIAL STATUS (o) LOW OCCUPATIONAL STATUS, MALFS: PFPCENT OF 22.6 160697 EMPLOYED MALES 16 AND OVER WHO ARE OPERATIVES, SERVICE WORKERS, AND LABORERS INCLUDING FARM LABORERS (10) HIGH OCCUPATIONAL STATUS, MALES: PERCENT OF 37.2 160697 EMPLOYED MALFS 16 AND OVER WHO ARF PROFESSIONALS, TFCHNICAL AND KINDRfcD WORKFRS, AND MANAGERS EXCEPT FARM 23.5 42.2 30.3 31.4 36.0 25.4 88 37.2 20.6 FDUCATIONAL STATUS (11) SCHOOL YEARS COMPLETED: MEDIAN SCHOOL YEARS COMPLETED EY PERSONS 25 AND OVER 12.5 12.1 11.0 ETHNIC COMPOSITION (12) NEGRO: PERCENT OF HOUSEHOLD POPULATION NECRO (13) OTHER NONWHITF: PERCENT OF HOUSEHOLD POPULATION NONWHITF AND NON-NEGRO (14) FOREIGN STOCK: PFPCFNT OF POPULATION WHO /SRF FOREIGN eOPN OR NATIVE BORN OF FOPEIGN OP MIXED PAPFNTAGF 13.9 644056 24.6 17.7 11.1 46 15.9 1.1 644056 1.1 0.7 1.4 92 0.2 12.4 660564 13.5 11.6 16.5 88 5.0 HOUSEHOLD COMPOSITION AND FAMILY STRUCTURE (15) HUSB«ND-WIFE HOUSFHOLDS: PFRCENT OF ALL HOUSFHOLDS WITH HUSBAND-WIFE FAMILIES (16) AGF OF HOUSEHOLD HFADS: MFDIAN AGF OF HOUSEHOLD HEADS (17) YOUTH DEPFNDFNCY RATIO: PERSONS UNDER 18 PER 100 PERSONS 18-64 IN HOUSEHOLD POPULATION (IE) AGFD DEPENDENCY RATIO: PERSONS 65 AND OVER PEP 100 PERSONS 18-64 IN HOUSEHOLD POPULATION 75.1 l Q ?9t2 64.8 70.9 69.4 54 74.3 39.2 192962 45.0 45.7 48.2 4 48.7 65.3 374295 60.1 63.4 63.4 63 64.1 6.8 374295 9.E 13.1 17.4 17.2 TYPE OF HOUSING (URBANIZATION) (19) SINGLE DWELLING UNITS: PERCENT OF ALL YEAR-ROUND HOUSING UNITS THAT ARE SINGLE DETACHED (EXCLUDING MOBILE HOMES AND TRAILERS) (20) HIGH RISE APARTMENTS: PERCENT OF ALL YFAR-ROUND HOUSING UNITS THAT ARE IN STPUCTURES OF 7 OR MORF STORIES 2.7 20017° 200185 43.1 10.0 51.1 2.8 (NA) (NA) 75.4 CONDITION OF HOUSING (21) OVERCROWDING: PERCFNT OF PERSONS IN HOUSEHOLDS IN HOUSING UNITS WITH 1.01 OR MORF PERSONS PER ROOM (22) STANDARD HOUSING: PERCFNT OF OCCUPIED HOUSING UNITS WITH DIRECT ACCESS/COMPLETE PLUMBING AND KITCHEN FACILITIES FOR EXCLUSIVE USE 98.2 192962 13.3 97.8 13.6 95.4 93.4 25 92 14.5 88.9 COMMUNITY INSTABILITY (23) RFCFNT MOVERS: PERCENT OF POPULATION WHO MOVFD INTO PRESENT RESIDENCE 1969-1970 22.6 23.5 20.5 (A) Medians estimated from SMSA county medians and county populations. (B) Based on 23 counties and Baltimore city. 68 SOCIAL INDICATORS FOR SMALL AREAS APPENDIX TABLE 4 PRINCE GEORGE'S COUNTY, MARYLAND ADDITIONAL STATISTICS FROM 1970 CENSUS OF POPULATION AND HOUSINGr SECOND PLUS FOURTH COUNT (ALL FILES' COMPARISONS WITH OTHER AREAS STATISTIC DESCRIPTION NUMBFR, DENOMINATOR PERCENT, FOR PERCENTS SMSA STATE U.S. PERCENT MEDIAN RATIO OR AND RATIOS OR OF VALUES LESS THAN VALUE VALUE OF FOR THF POPULATION RATIO RATIO RATIO THE ALL COUNTIES TOTAL FOR MEDIANS OR OR OR IN AREA AND AVERAGES VALUE (A) VALUE VALUE COUNTY (B) MARYLAND (B) GENERAL POPULATION DATA SOCIOECONOMIC STATUS ECONOMIC STATUS INCOME (1) INCOME OF FAMILIES, WH OF WHITE FAMILIES (2) INCOME OF FAMILIES OF NEGRO FAMILIES (3) INCOME OF UNRELATED IN INCOME OF UNRELATED IN 14) INCOME OF UNRELATED IN MEDIAN INCOME OF WHITE INDIVIDUALS 14 AND OVF (?) INCOMF OF UNRELATED IN MEDIAN INCOME OF NFGRO INDIVIDUALS 14 AND n VE (6) FAMILIFS IN POVERTY, W FAMILIFS BELOW POVERTY (7) FAMILIES IN POVERTY, N FAMILIES BELOW POVERTY (8) POPULATION IN POVERTY: POPULATION BELOW POVER (9) POPULATION IN POVERTY, OF WHITE POPULATION BE (10) POPULATION IN POVERTY, OF NEGRO POPULATION BE (111 HIGH INCOME FAMILIES: FAMILY INCOME ITE: MEDIAN INCOME NEGRO: MEDIAN INCOMF DIVIDUALS: MEDIAN DIVIDUALS 14 AND OVER DIVIDUALS, WHITE: UNRELATED R DIVIDUALS, NEGPO: UNPFLATFD P HITE: PEPCENT OF WHITE LEVEL tGRO: PFPCtNT OF NEGRO LEVEL PERCENT OF TY LEVEL WHITE: PEPCENT LOW POVERTY LEVEL NEGRO: PERCENT LOW POVERTY LEVEL UPPER OUARTILE 112748 1412^8 $14841 $11635 $ 9,961 88 $ 9333. 110624 20578 $ 8513 $ 7701 $ 6,067 96 $ 6431. J 3 809 55361 J 4539 $ 3099 $ 2,489 83 $ 2207. $ 3859 49813 $ 5274 $ 3349 $ 2,568 83 $ 2352. * 3436 4789 $ 3839 $ 2325 $ 1,936 96 $ 1834. 3.8 141258 3.5 5.3 8.6 13 8.9 8.0 20578 14.6 20.9 29.8 4 24.5 5.8 644486 8.3 10.1 13.7 13 15.4 5.0 54P636 5.0 6.9 10.9 13 11.3 10.4 89320 18.4 24.7 35.0 4 29.1 $18311 163400 $22103 $16674 $14176 88 $12805. VALUE OF HOUSING (12) HOUSE VALUE, NON-NECRO: MEDIAN VALUF OF NON-NEGRO OCCUPIED HOUSING UNITS (13) HOUSE VALUE, NEGRO: MEDIAN VALUE OF NEGRO OWNER-OCCUPIED HOUSING UNITS (14) RENT, NON-NEGRO: MEDIAN MONTHLY RENTAL OF NON-NEGRO RENTER-OCCUPIED HOUSING UNITS (15) RENT, NEGRO: MEDIAN MONTHLY RENTAL OF NEGRO RENTER-OCCUPIED HOUSING UNITS EMPLOYMENT LEVEL AND LABOR (16) UNEMPLOYMENT: PERCENT AND OVFR UNEMPLOYED (17) UNEMPLOYMENT, WHITE: P FORCE 16 AND OVER UNEM (18) UNEMPLOYMENT, NEGRO: P FORCE 16 AND OVER UNEM (19) UNDER EMPLOYMENT, MALE MALE LABOR FORCE 25-64 THAN 40 WEEKS IN 196« (20) UNDFP EMPLOYMENT, WHITE MAL OF WHITE MALE LAPOP FORCE 2 LESS THAN 40 WEEKS IN 1°69 (21) UNDER EMPLOYMENT, NFGRO MAL OF NFGRO MALE LAE'OR FORCE 2 LESS THAN 40 WEEKS IN 1969 (2?) LABOR FORCE PARTICIPATION, OF FEMALES 16 AND OVER IN T FORCE (INCLUDING ARMFE EORC (23) LABOP FORCE PARTICIPATION, PERCFNT OF WHITE =EMALFS 16 LABOR FORCE (INCLUDING ARME (24) LABOR FORCF PARTICIPATION PERCENT OF NEGRO FEMALES 16 LAPOR FORCE (INCLUDING ARME FORCE PARTICIPATION OF LABOR FORCE 16 FRCENT OF WHITE LABOR PLOYED ERCENT OF NEGRO LABOR PLOYED S 25-64: PERCFNT OF WHO WORKED LESS ES 25-64: PEPCENT 5-64 WHO WORKED ES 25-64: PERCFNT 5-64 WHO WOPKFD FEMALES: PERCENT HE LABOR ES) WHITE FEMALES: AND OVFR IN THE D FORCfcS) NEGRO FEMALE*: AND OVER IN THF D FORCES) SOCIAL ST (25) LOW PERC WHO LARO (26) HIGH °ERC WHO WORK (27) LOW PFRC WHO LABO (28) HIGH PEPC WHO WORK (29) LOW PERC WHO AND ATUS OCCUPA ENT OF ARE OP RERS I OCCUP ENT OF APE PR FR C , A. OCCUPA FNT OF APE OP REPS I OCCUP ENT OF ARF PR ERS, A OCCUPA ENT OF ARF OP LAPORE TIONAL ST EMPLOYEb ERATIVES, MCLUDING AT1HN4L S EMPLOYED OFESSIONA NO MANAGE TIONtL ST EMPLOYED EP4TIVFS, NCLUDING ATIONAL S EMPLOYEC OFESSIONA NO MANAGE TIONAL ST EMPLOYED ERATIVES, RS INCLUD ATUS, WEI WHITE MA SERVICE FARM LABO TATUS, WH WHITE MS LS, TFCHN P c tXCEPT ATUS, NLG NEGRO MA SEPVICF FARM LABO TATUS, Nfc NFGRO MA LS, TECHN RS EXCEPT ATUS, FFM FEMALES SFRVICE ING FARM $24170 79799 $32394 $19637 (NA) 88 $16484. $1<>783 10816 $18732 $11107 (NA) 88 $ 9560. $145 82107 $145 $121 (NA) 92 $ 76. $130 10740 $ 98 $ 85 (NA) 88 $ 49. 2.2 293776 2.5 3.1 4.4 8 2.9 2.1 251506 2.0 2.6 4.1 21 2.5 3.2 39351 3.9 5.5 7.0 21 4.7 3.7 139022 3.9 4.1 (NA) 29 4.6 3.3 120425 3.2 3.5 (NA) 33 4.0 6.0 17139 6.1 7.2 (NA) 38 7.4 50.2 224804 50.3 44.4 41.4 92 42.0 TE MALFS: Lf ? 16 AND OVER WORKERS, AND RERS ITE MALES: LES 16 AND OVER ICAL AND KINDRED FARM RO MALES: LFS 16 AND OVER WORKERS, AND RFRS GRO MALES: LES 16 AND OVER ICAL AND KINDRED FARM ALES: 16 AND OVER WORKERS, LABORFRS 19.9 13«<.33 130433 108708 50.1 15.6 53.1 25.0 34.7 61.4 11.3 33.2 64.9 36.2 96 88 51.1 30.1 25.35 68.0 42.3 SEE FOOTNOTES AT END OF TABLE. MENTAL HEALTH DEMOGRAPHIC PROFILE 69 APPENDIX TABLE 4 — CONTINUED PRINCE GEORGE'S COUNTY, MARYLAND ADDITIONAL STATISTICS FROM 1970 CENSUS OF POPULATION AND HOUSING* SECOND PLUS FOURTH COUNT (ALL FILES* COMPARISONS WITH OTHER AREAS SMSA STATISTIC DESCRIPTION NUMBER. DENOMINATOR PERCENT, FOR PC^CENTS RATK OR AND PATIOS VALUE OR FOP THF POPULATION PATIO TOTAL FOR MEDIANS OR AREA ANC VALUE (A) AVf PAC-fcS STATE U.S. PERCENT MEDIAN OF VALUES LESS THAN VALUE OF PATIO RATIO THE ALL COUNTIES OR OR IN VALUF VALUE COUNTY (B) MARYLAND (B) SOCIAL STATUS — CONTINUED (30) LOW OCCUPATIONAL STATUS. WHITE FEMALES: 13.1 90810 PERCENT OF EMPLOYED WHITE FEMALES 16 AND OVER WHO ARE OPERATIVES. SERVICE WORKERS, AND LABORERS INCLUDING FARM LABORERS (31) MID OCCUPATIONAL STATUS, WHITE FEMALES: 62.5 90810 PERCENT OF EMPLOYED WHITE FFMALES 16 AND OVER WHO ARF SALES, CLERICAL AND KINDRED, WORKERS, ANO CRAFTSMEN, FOREMFN AND KINDRED WORKERS (32) LOW OCCUPATIONAL STATUS, NEGRO FEMALES: 30. A 16830 PFRCENT OF EMPLOYED NFGRO FEMALES 16 AND OVER WHO ARE OPERATIVES, SERVICF WORKERS, AND LABORERS INCLUDING FARM LABORERS (33) MID OCCUPATIONAL STATUS, NEGRO FEMALES: 46.3 16830 PERCFNT OF EMPLOYED NEGRO FEMALES 16 AND OVER WHO ARE SALES, CLEPTCAL AND KINDRED WORKERS, AND CRAFTSMEN, FORFMFN AND KINDRFD WORKERS 41.7 40.1 43.2 22.5 54.5 55.1 29.7 32.7 46. 8 62.4 24.7 96 96 32.5 4B.5 74.6 17.3 FDUCATIONAL STATUS (34) SCHOOL YEARS COMPL YEARS COMPLFTED PY (35) SCHOOL YEARS COMPL YEARS COMPLETFC PY (36) LOW EDUCATIONAL ST 25 AND OVER WITH 8 (37) LOW EDUCATIONAL 5 T WHITE PFRSONS ?5 OP LESS EDUCATION OBJ LOW FDUCATIONAL ST NFGPO PERSONS 25 A OR LESS EDUCATION (39) HIGH SCHOOL COMPLE WHITE PERSONS 18 AN COMPLETED AT LEAST 140) HIGH SCHOOL COMPLE NEGRO PERSONS 18 COMPLETED AT LEAST (41) HIGH SCHOOL COMPLE PERCENT OF WHITE P COMPLETED AT LEAST (42) HIGH SCHOOL COMPLE PERCENT OF NFGRO P COMPLETED AT LEAST (43) HIGH EDUCATIONAL S POPULATION 25 AND YEARS OF COLLEGE FTFC, WHITF: MEDIAN SCHOOL WHITE PERSONS 25 AND OVER FTED, NEGRO: MEDIAN SCHOOL NECRC PERSONS 25 AND OVER A1US: PEPCFNT OF PERSONS YEARS OR LESS EDUCATION ATUS, WHITF: PERCENT OF ND OVER WITH 8 YEARS ATU*. NEGRO: PFPCENT OF NO OVER WITH 8 YEARS TION, D 0\/t YE TION, ND OV 4 YE TION. EPSON 4 YF TION, ERSON 4 YE TATUS OVER WPITF R WHO ARS OF NFGRO ER WHO ARS OF YOUNG S 18-2 ARS OF YOUNG S 18-2 ARS OF PERC WITH 4 : PFRCFNT OF HAVE HIGH SCHOOL PERCENT OF HAVE HIGH SCHOOL WHITF ADULTS: 4 WHO HAVE HIGH SCHOOL NEGRO ADULTS: 4 WHO HAVE HIGH SCHOOL ENT OF OP MORE 12.5 14.2 276886 12.9 12.2 12.1 92 11.4 39413 11.4 9.9 9.8 96 9.0 319843 15.9 27.4 2b. 3 4 32.6 276886 11. ° 24.7 26.6 4 31.3 39413 29.3 42.1 43.8 4 49.1 359682 74.4 58.2 (NA) 92 48.5 50476 48.0 36.3 (NA) 96 29.4 72.7 82796 65.4 11063 17.1 319843 23.4 52.6 13.9 55.2 10.7 88 83 86 68.7 50.5 7.0 ETHNIC COMPOSITION (44) SOUTHERN OP EASTERN EUROPEAN STOCK: PERCENT 3.3 OF POPULATION WHO ARE FOPFIGN STOCK OF SOUTHERN OR EASTERN EUROPEAN BACKGROUND (POLAND, CZECHOSLOVAKIA, AUSTRIA, HUNGARY, U.S.S.R., AND ITALY) (45) SPANISH AMERICANS: PFRCENT OF POPULATION 2.2 WHO ARE OF SPANISH HERITAGE 660567 3.4 2.5 1.4 4.6 83 92 1.1 0.6 HOUSEHOLD COMPOSITION AND FAMILY STRUCTURE GENFRAL CHARAC (46) MFDIAN HOU (47) SMALL HOUS WITH ONLY (48) LARGE HOUS WITH 6 OR (49) CHILDREN L PERCENT OF WITH BOTH (50) SEX RATIO: IN HOUSFHO (51) FERTILITY FEMALES 15 (5?) HUSBAND-WI WHITE HOUS (53) HUSBAND-WI NEGRO HOUS T ERISTICS SFHOLD SIZE FHOLDS: PERCENT OF HOUSEHOLDS ONE PERSON EHOLDS: PERCENT OF HOUSEHOLDS MORF PFRSONS IVING WITH THEIR PAPFNTS; PERSONS UNDER IP LIVING PARENTS MALES PEc 100 FEMALES LD POPULATION RATIO: CHILDRFN UNDER 5 PER 1000 -44 IN HOUSEHOLD POPULATION FE HOUSEHOLDS, WHITE: PEPCFNT OF EHOLDS WITH HUSBAND-WIFE FAMILIES FE HOUSEHOLDS. NEGRO: PERCENT OF EHOLDS WITH HUSBAND-WIFE FAMILIFS 3.1 192962 2.9 2.9 2.7 58 2.9 12.0 192962 19.4 14.9 17.6 29 13.4 11.1 192962 9.9 11.3 10.4 50 11.0 U6.0 244446 El. 2 81.0 82.7 75 83.8 96.9 327069 92.3 94.1 94.8 58 95.5 443.4 155706 387.7 412.6 404.2 71 420.9 75.3 167889 68.9 73. e 71.3 42 75.6 73.3 23170 *2.4 54.9 52.6 96 55.3 FAMILY LIFE CYCLE (54) AGE OF HOUSEHOLD HFADS, WHITE: AGE OF WHITE HOUSEHOLD HFADS (55) AGE OF HOUSEHOLD HEADS, NEGRO: OF NEGRO HOUSEHOLD HEADS (56) YOUTH DEPENDENCY PATIO, WHITE: PERSONS UNDER 18 PFR 100 WHITE 18-64 IN HOUSEHOLD POPULATION SEE FOOTNOTES AT END OF TABLE. MEDIAN 39.7 167889 42.9 46.2 48.5 4 49.0 MEDIAN AGE 36.2 23170 42.0 43.5 45.5 4 48.0 WHITE 62.3 323170 57.5 59.8 60.8 50 61.1 PERSONS 70 SOCIAL INDICATORS FOR SMALL AREAS APPENDIX TABLE 4 — CONTINUED PRINCE GEORGE'S COUNTY, MARYLAND ADDITIONAL STATISTICS FROM 1970 CENSUS OF POPULATION AND HOUSING* SECOND PLUS FOURTH COUNT (ALL FILES' COMPARISONS WITH OTHER AREAS STATISTIC DESCRIPTION NUMBFR, PERCCNT, RATIO OR VALUE FOR THF TOTAL AREA DENOMINATOR FOR PERCENTS AND RATIOS OR POPULATION FOR MEDIANS AND A V ER A GES U.S. PATIO OR VALUE (A) RATIO OR VALUE RATIO OR VALUE PERCENT MEDIAN OF VALUES VALUE LESS THAN OF THE ALL COUNTIES IN COUNTY (B) MARYLAND (B) FAMILY LIFE CYCLE — CONTINUED (57) YOUTH DEPENDENCY RATIOt NEGRO: NEGRO PERSONS UNDER 16 PER 100 NEGRO PERSONS 18-64 IN HOUSEHOLD POPULATION (58) AGED DEPENDENCY RATIOt WHITE: WHITE PERSONS 65 AND OVER PER 100 WHITE PERSONS 18-64 IN HOUSEHOLD POPULATION (59) AGED DEPENDENCY RATIO, NEGRO: NEGRO PERSONS 65 AND OVER PER 100 NFGRO PERSONS 18-64 IH HOUSEHOLD POPULATION (60) FAMILIES WITH CHILDRFN: PFRCENT C* FAMILIES WITH OWN CHILDREN UNDER 18 (61) CHILDBEARING ONLY FAMILIES: PERCENT OF FAMILIES WITH OWN CHILDREN UNDER 6, NO CHILDRFN 6-17 (62) CHILDBEARING AND CHILDREARING FAMILIFS: PERCENT OF FAMILIES WITH OWN CHILDREN BOTH UNDER 6 AND 6-17 (63) CHILDREARING ONLY FAMILIES: PEPCENT OF FAMILIES WITH OWN CHILDREN 6-17, NO CHILDREN UNDER 6 (64) CHILDRFARING COMPLETED: PERCENT OF HUSBAND-WIFE FAMILIES WITH HUSBAND 45 OR OVER AND WITH NO CHILDREN PRESENT 87.1 7.1 46737 323170 4.8 46737 63.5 163379 18.4 163379 15.7 16337° 29.3 20.7 163379 68.7 10.2 8.7 59.7 15.3 14.1 30.3 26.0 82.3 13.6 10.7 57.4 13.5 14.1 29.8 30.7 85. S 17.9 13.6 54.9 12.7 13.4 28.7 (NA) 63 88 92 71 67 83.3 18.1 19.1 55.3 11.9 13.6 28.3 35.5 PERSONS NOT IN FAMIL (65) GROUP CUARTERS: POPULATION WHO L (66) INMATES OF INSTI POPULATION IN CR INMATFS OF INSTI (67) INMATES OF MENTA POPULATION IN GR INMATFS OF MENTA (6P) GROUP QUARTERS E PERCENT OF TOTAL IN ROOMING HOUSE NONINSTITUTIONAL (69) HOUSEHOLD HEADS OF HOUSFHOLDS WI INDIVIDUALS (70) NON-RELATIVES: P HOUSEHOLDS WHO A HOUSEHOLD HEAD IES PFRCFNT OF TOTAL IVF IN GROUP OUARTERS TUTIONS: PERCENT OF OHP OUARTERS WHO ARE TUTIONS L HOSPITALS: PERCENT OF OUP OUARTFRS WHO ARE L HOSPITALS XCLUDINC INSTITUTIONS: POPULATION WHO LIVE S AND OTHER GROUP OUAPTEPS PRIMARY INDIVIDUALS: PERCENT TH HEADS WHO ARE PRIMARY ERCENT OF PERSONS IN RE NOT RELATED TO THE POPULATION NOT CURRENTLY MARRIED (71) SINGLE MALES: PERCENT OF MALFS 25 AND OVER WHO HAVE NEVER MARRIED (72) SINGLE FEMALES: PERCENT OF FFMALES 25 ANO OVER WHO HAVE NEVER MARRIED (73) DIVORCED OR SEPARATED MALES: PERCENT OF MALES 14 AND OVER WHO ARE DIVORCED OR SEPARATED (74) DIVORCED OR SEPARATED FEMALES: PEPCENT OF FEMALES 14 AND OVER WHO ARE DIVORCED OR SEPARATED (75) WIDOWED FEMALES: PERCENT OF FFMALES 14 AND OVER WHO ARE WIDOWS NON HUSBAND-WIFE HOUSEHOLDS (76) FEMALE HEADED HOUSEHOLDS: PERCENT OF HOUSEHOLDS WITH FEMALE HEAD (77) FEMALE HEADED HOUSEHOLDS WITH OWN CHILDREN: PERCENT OF HOUSFHOLDS WITH OWN CHILDREN UNDER 18 THAT ARE HFADFD BY FEMALES CONDITION CF HOUSING 2.5 660567 3.0 2.7 2.9 54 2.3 14.5 16511 23.2 37.9 36.5 13 42.0 0.2 16286 2.1 660f-67 15.3 192*62 2.2 644056 6.3 2.1 2.9 8.6 1.7 17.3 2.0 7.3 1.8 19.7 1.6 46 63 79 0.3 1.1 15.5 1.7 7.8 156744 11.3 9.1 (NA) 25 8.8 5.6 163183 10.2 7.1 (NA) 21 6.7 4.0 224616 5.5 5.0 4.3 38 4.1 6.3 237020 8.2 7.0 6.2 79 5.2 7.5 237020 9.6 11.1 12.4 12.0 15.9 192962 23.4 20.0 21.0 33 17.1 9.0 103756 12.4 11.2 10.7 50 8.7 HOUSING (78) VACANCY INDFX: PERCENT OF ALL YEAR-POUND HOUSING UNITS THAT APF VACANT (79) STANDARD HOUSING, NON-NFGRO: PFRCFNT OF NON-NFGRO HOUSING UNITS WITH DIRECT ACCESS AND WITH COMPLETE PLUMPING AND KITCHFN FACILITIES FOR EXCLUSIVE USE (80) STANDARD HOUSING, N1GR0: PEkCFNT OF NEGRO HOUSING UNITS WITH DIPFCI ACFSS AND WITH COMPLETE PLUMPING AND KITCHEN FACILITIES FOR EXCLUSIVE USE DENSITY (81) OVERCROWDED HOUSING: PERCENT OF OCCUPIED HOUSING UNITS WITH 1.01 OR MORE PFRSONS PER ROOM (82) NON-NEGRO POPULATION IN OVERCROWDED HOUSING: PERCENT OF NON-NEGRO HOUSEHOLD POPULATION IN HOUSING UNITS WITH 1.01 OR MORE MORE PERSONS PER ROOM SEE FOOTNOTES AT END OF TABLE. 3.6 200211 4.2 4.8 6.2 13 6.2 98.8 lc>v792 98.3 96.6 (NA) 92 92.4 93.5 23170 6.4 19296? 10.0 554364 6.9 6.6 9.7 (NA) 14.3 88 33 42 58.5 7.0 10.3 MENTAL HEALTH DEMOGRAPHIC PROFILE /l APPENDIX TABLE 4 — CONTINUED PRINCE GEORGE'S COUNTY, MARYLAND ADDITIONAL STATISTICS FROM 1970 CENSUS OF POPULATION AND HOUSING, SECOND PLUS FOURTH COUNT (ALL FILES* COMPARISONS WITH OTHER AREAS STATISTIC DESCRIPTION NL'MBFR , OF NOMINATOR PERCENT, FOP PEPCENTS SMSA STATE U.S. PERCENT MEDIAN RATIO OR AND P.ATI05 OR OF VALUES LESS THAN VALUE VALUE OF FOP THF POPULATION PATIO PATIO RATIO THE ALL COUNTIES TOTAL FOR MEDIANS OR OR OR IN AREA ANC AVERAGES VALUE (A) VALUF VALUE COUNTY (B) MARYLAND (B) DENSITY — CONTINUED (83) NEGRO POPULATION IN OVERCPOWOFD HOUSING: PERCENT OF NEGRO HOUSEHOLD POPULATION IN HOUSING WITH 1.51 OR MORE PERSONS PFR ROOM (84) PERSONS IN HIGHLY OVERCROWDED HOUSING: PERCFNT OF HOUSFHOLO POPULATION IN HOUSING UNITS WITH 1.51 OR MORF PFRSONS PFR ROOM TYPE OF HOUSING(URBANIZATION) (85) RENTER OCCUPANCY: PFRCfcNT PF OCCUPIED HOUSING UNITS THAT ARF RFNTFR OCCUPIED (86) TRAILFRS: PERCENT OF ALL YFAP-ROUNP HOUSINC- UNITS THAT ARE MOBILE HOMES OP TRAILERS (87) LARGF APARTMFNT STRUCTUR E c : PERCENT OF ALL YEAR POUND HOUSING UNITS THAT ARE IN STRUCTURES WITH 20 OP MORF UNITS (88) SINGLF OWFLLING UNITS, NON-NEGRO: PFRCENT OF NON-NEGRO HOUSING UNITS THST ARE SINGLE DETACHED (89) SINGLF OWFLLING UNITS, NEGRO: PFRCENT OF NEGRO HOUSING UNITS THAT /RE SINGLE DETACHED (90) RURAL POPULATION: PFRCENT OF POPULATION COMMUNITY INSTABILITY (91) RECFNT MOVFRS, WHITE: PFRCENT OF WHITE POPULATION WHO MCVED INTU PPESENT PESIDFNCF 1969-1Q70 (92) RECFNT MOVEPS, NEC-RP: PFRCENT OF NFGRC POPULATION WHO MOVED INTO PRFSFN1 RFSIDENCF 1?«62 2*492 192962 351414 243269 4.3 0.0 21. e 6.8 11.2 42.0 32.2 5.3 21.9 0.3 9.9 11.5 7.3 (na) 40.8 (NA) 19.2 (NA) (NA) 5.1 11.2 4.3 15.1 4 17 13 21 13 8.7 14.5 43.2 6.5 27.4 0.6 13.6 30.9 4.8 9.2 3.5 (A) Medians estimated from SMSA county medians and county populations. (B) Based on 23 counties and Baltimore city. 72 SOCIAL INDICATORS FOR SMALL AREAS App. Fig. 1A TOTAL POPULATION OF TRACT 8052.01, COUNTY 33, MARYLAND, COMPREHENSIVE CMHC I TOTAL POPULATION 4674 MALES 2209 FEMALES 2465 NUMBER % AGE 987654321 1 23456789 % NUMBER 14 0.30 85+ 0.26 12 15 0.32 80-4 F 0.47 22 25 0.53 759 M FF 0.73 34 43 0.92 70-4 MM FFFFF 1.71 80 71 1.52 65-9 MMMM FFFFF 1.71 80 89 1.90 60-4 MMMMM FFFFFFF 2.40 112 96 2.05 55-9 MMMMMM FFFFFFFF 2.70 126 126 2.70 50-4 MMMMMMMM FFFFFFFFF 3.06 143 113 2.42 45-9 MMMMMMM FFFFFFFFF 3.19 149 117 2.50 40-4 MMMMMMM FFFFFFFF 2.89 135 120 2.57 35-9 MMMMMMM FFFFFFF 2.55 119 164 3.51 30-4 MMMMMMMMMM FFFFFFFFF 3.27 153 196 4.19 25-9 MMMMMMMMMMMM FFFFFFFFFFFFFFFF 5.54 259 198 4.24 20-4 MMMMMMMMMMMM FFFFFFFFFFFFFFFF 5.50 257 169 3.62 15-9 MMMMMMMMMM FFFFFFFFFFFF 4.04 189 165 3.53 10-4 MMMMMMMMMM FFFFFFFFFFF 3.94 184 217 4.64 5-9 MMMMMMMMMMMMM FFFFFFFFFFFF 4.09 191 271 5.80 0-4 MMMMMMMMMMMMMMMMM FFFFFFFFFFFFFF 4.71 220 App. Fig. 1B WHITE POPULATION OF TRACT 8052.01, TOTAL POPULATION 2249 COUNTY 33, MARYLAND, COMPREHENSIVE CMHC I MALES 1056 FEMALES 1193 NUMBER % AGE 987654321 123456789 % NUMBER 13 0.58 85+ M M F 0.44 10 15 0.67 80-4 MM FF 0.93 21 23 1.02 75-9 MMM FFFF 1.33 30 40 1.78 70-4 MMMMM FFFFFFFFF 3.07 69 68 3.02 65-9 MMMMMMMMM FFFFFFFF 2.93 66 76 3.38 60-4 MMMMMMMMMM FFFFFFFFFFFFF 4.45 100 81 3.60 55-9 MMMMMMMMMM FFFFFFFFFFFFFFF 5.02 113 97 4.31 50-4 MMMMMMMMMMMM FFFFFFFFFFFFFFF 5.11 115 71 3.16 45-9 MMMMMMMMM FFFFFFFFFFFFF 4.45 100 46 2.05 40-4 MMMMMM FFFFFFF 2.45 55 39 1.73 35-9 MMMMM FFFFF 1.78 40 45 2.00 30-4 MMMMMM FFFFF 1.69 38 58 2.58 25-9 MMMMMMM FFFFFFFF 2.98 67 86 3.82 20-4 MMMMMMMMMMM FFFFFFFFF 3.11 70 100 4.45 15-9 MMMMMMMMMMMMM FFFFFFFFFFFFFF 4.71 106 80 3.56 10-4 MMMMMMMMMM FFFFFFFFFF 3.56 80 62 2.76 5-9 MMMMMMMM FFFFFFFF 2.85 64 56 2.49 0-4 MMMMMMM FFFFFF 2.18 49 App. Fig. 1C NEGRO POPULATION OF TRACT 8052.01, COUNTY 33, MARYLAND, COMPREHENSIVE CMHC I TOTAL POPULATION 2372 MALES 1124 FEMALES 1248 NUMBER % AGE 987654321 123456789 % NUMBER 1 0.04 85+ 0.08 2 0.0 80-4 0.04 1 2 0.08 75-9 0.17 4 3 0.13 70-4 F 0.46 11 2 0.08 65-9 F 0.59 14 12 0.51 60-4 M F 0.51 12 12 0.51 55-9 M F 0.55 13 28 1.18 50-4 MMM FFF 1.18 28 41 1.73 45-9 MMMMM FFFFF 1.94 46 70 2.95 40-4 MMMMMMMM FFFFFFFFF 3.29 78 80 3.37 359 MMMMMMMMMM FFFFFFFFF 3.29 78 118 4.97 30-4 MMMMMMMMMMMMMM FFFFFFFFFFFFFF 4.81 114 132 5.56 25-9 MMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFFFFFFF 7.93 188 112 4.72 204 MMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFFFFFFF 7.80 185 69 2.91 15-9 MMMMMMMM FFFFFFFFFF 3.41 81 84 3.54 10-4 MMMMMMMMMM FFFFFFFFFFFF 4.30 102 148 6.24 5-9 MMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFF 5.31 126 210 8.85 0-4 MMMMMMMMMMMMMMMMMMMMMMMMMM FFFFFFFFFFFFFFFFFFFF 6.96 165 4 U. S. GOVERNMENT PRINTING OFFICE : 1973 O - 505-780 CENSUS TRACT PAPERS Series GE-40 No. 1. Papers Presented at the Census Tract Conference, Septem- ber 10, 1965, Philadelphia, Pa. No. 2. Needs and Plans for the 1970 Censuses (Census Tract Conference of August 15, 1966, at Los Angeles, Calif.) No. 3. Some Uses of Census Tracts in Private Business, by Wilbur McCann (1967) No. 4. Papers Presented at the Conference on Small-Area Sta- tistics, American Statistical Association, Washington, D.C., December 27, 1967, and Related Papers No. 5. Papers Presented at the Conference on Small-Area Sta- tistics, American Statistical Association, Pittsburgh, Pa., August 23, 1968, and Related Papers No. 6. Final 1970 Census Plans and Four Programing Systems for Computerized Data Retrieval and Manipulation (Confer- ence on Small-Area Statistics, American Statistical Association,August 21, 1969, New York, N.Y.) No. 7. New Uses of Census Resources, The Southern California Census Use Study and a Related Paper (Conference on Small-Area Statistics, December 29, 1970, Detroit, Mich.) American Statistical Association. No. 8. Small-Area Statistics: Strengthening Their Role in Federal Government and Their Use in Criminal Justice Programs (Papers Presented at the Conference on Small-Area Statistics, American Statistical Association, August 23, 1971, Ft. Collins, Colo.) No. 9. Social Indicators for Small Areas, Papers Presented at the Conference on Small-Area Statistics,American Statistical Association, August 14, 1972, Montreal, Canada. Copies of early editions may no longer be in print but are usually available in Federal Depository Libraries. Papers pre- sented at conferences held from 1958 through 1964 were originally published by the Bureau of the Census in its Working Paper Series. For information on all census tract reports, write to: Publications Distribution Section Social and Economic Statistics Administration Washington, D.C. 20233 UNITED STATES GOVERNMENT PRINTING OFFICE DIVISION OF PUBLIC DOCUMENTS Washington, D.C. 20402 OFFICIAL BUSINESS FIRST CLASS MAIL PENN STATE UNIVERSITY LIBRARIES DDDD7E PRINTING n flbOlbfl