1 THE STRUCTURAL ECOLOGY AND POPULATION HEALTH OF UPSTATE NEW YORK: ANALYSIS AND APPLICATIONS Frank W. Young Cornell University 2009 Acknowledgements: My thanks to Robin Blakely at CARDI and to Florio Arguillas, Mike Kunz and Jiyoun An at CISER for indispensable assistance in producing this report.
2 2 TABLE OF CONTENTS PAGE ABSTRACT 3 INTRODUCTION 4 Sociology, economics and biomedical theory 4 The basic mortality figures 6 I. THREATS, RESPONSE CAPACITY AND MORTALITY 11 The theory in brief 11 Macro threats to the county communities 14 The impact on mortality 23 The challenge of macro threats 25 II. THE ANALYSIS OF ETHNIC SITUATIONS 27 Distribution of mortality by race and region 28 The structural context of black and white mortality 31 III. APPLICATIONS OF STRUCTURAL ECOLOGY 43 Programmatic steps 43 Interventions 47 Institutionalizing a new branch of public health 49
3 3 ABSTRACT This report explains how a structural ecology theory of population health applies to the 50 counties of upstate New York (UNY). It begins with a presentation of the trends of age-adjusted black and white mortality over , which are the criteria of improvement. The supporting theory claims that communities, like the counties of New York State, can face serious threats and the research identifies three such threats: resource/location poverty, deindustrialization and the influx of poor minorities. On the theoretical side, the study conceptualizes three master problem-solving strategies that all communities use in dealing with threats. These are the application of specialized knowledge, the weighing of alternative policy directions, and mobilization behind a leader who offers a new perspective on the problem. Specialized knowledge is examined in its upstate urbanization form, but the other two dimensions have yet to be measured for this level of community. Poverty and new arrivals predict higher mortality, and analysis suggests that interventions are feasible, if at all, only with respect to the immigration of minorities. Accordingly, a second factor analysis probes these ethnic patterns, suggesting the conclusion that of the several minority situations, the simple increase of minorities is the most tractable problem. The report concludes with an analysis of interventions designed to reduce the impact of threats. It also discusses how social problems public health could be institutionalized.
4 INTRODUCTION Sociology, economics and biomedical theory This report summarizes a new explanation of population health, defined as the rate of an appropriate measure of biological functioning across comparable communities. The specific criterion used in this study is age-adjusted mortality per 100,000 population. Community is defined as a multifunctional group that is concerned with the well-being of the members (Selznick, 1996). It is also the basic unit of adaptation. Individuals manage personal threats but this theory claims that communities, from households to nationstates, are more important in predicting population health. The proposed theory is sociological and neo-darwinian because it interprets the major dimensions of the structure of communities as master problem-solving strategies that determine the adaptation to the environment. All communities have some capacity for applying specialized knowledge to problems and they are usually able to argue over policy options. If these fail, they can mobilize behind a leader or a platform with the aim of looking at the problem from a new perspective. These processes can be combined into a formula: Ph =C/t, where Ph is an appropriate measure of population health, C is institutionalized problem-solving capacity as defined by the three strategies, and t is a serious threat to the community. Thus, communities with strong problem-solving capacity relative to the threat tend to have better population health. One impediment to any attempt to introduce a sociological theory of development that is centered on population health is that it necessarily challenges both economics and
5 5 the biomedical theory of disease. That sets up a conflict with two strong disciplines and puts a burden on community leaders as they try to evaluate possible programs. Structural ecology attempts to sidestep these dominant paradigms by starting from different premises and accepting their accomplishments. But conflict is ultimately unavoidable because structural ecology claims that it can make better sense out of development (or lack of it) in places like Upstate New York and that it can come up with initiatives that, like the community-wide preventions of public health, protect all the residents. The alternative spelled out in this report is to create an addition to conventional public health organization that focuses on the social problems that communities must deal with. Solutions to these result in improved health, at lower cost, which then becomes a regional comparative advantage for attracting people who may be able to create new enterprises. It should also decrease the cost of curative medicine by preventing the onset of diseases and related maladies. This can happen because the structural theory that guides this program locks onto the most powerful engine of development, the human community. When communities are organized in certain ways, people can often solve problems that seem impossible at first. Admittedly, structural changes at any community level are infrequent and we have yet to learn how to bring about such shifts. But we do have a rich history of such changes social security, welfare programs, unemployment insurance, affirmative action, and the like. There is one other reason for pursuing a sociological path. Communities deserve and need their own research backup. Industry, agriculture, the military and the biomedical institutions, to name only the most prominent, are all supported by specialists
6 6 in state and federal bureaucracies and by private companies. But communities do not have a designated research backup. True, institutes of local government exist and universities conduct scattered research on communities as adaptive units. But until recently it has been difficult for sociology to analyze communities rigorously. In fact, we still rely on household surveys for much of our community information or, even worse, a single community study that is always atypical in some respects. Demographers routinely compare counties and similar units, and much can be done with the composite scales that are available but census data is still mostly individual-level. Fortunately there are other sources of information and new techniques for combining the information. The Basic Mortality Figures The analysis begins with the contrasting mortality rates for the four geo-historical regions of Upstate New York as shown in Map 1. The Canal Corridor and the Hudson Valley are self-explanatory. The other well-known region is the Southern Tier, which is the northern edge of Appalachia. What is left in the north and south are interstitial counties.
8 Table 1: Mortality statistics for upstate New York regions Statistic Upstate Southern tier Canal corridor Hudson Valley Interstitial Aam Aam Aam Aamb Aamw Aam7998gap N Aam00=age-adjusted mortality per 100,000, Source for this and the other mortality figures: CDC Wonder site. Aam79gap=difference between Aamb7998 and Aamw7998. (mean=240; SD=152; min-max= ) Table 1 begins with the upstate averages and it is apparent that with the exception of aam00 the interstitial counties account for most of the higher averages over the three decades. The remarkable fact is that the regional means are approximately equal and all of them show significant decline. In the lower part of the table, the second pair of mortality rates, for blacks and whites from , throws light on the long standing gap between these rates. The gaps are wide in the Southern Tier and the Canal Corridor, in contrast to the Hudson and Interstitial counties. Understanding this geographical pattern of the black-white gap is an obvious task for analysis.
9 Diagram 1 Mortality Trends for Upstate Regions S. TIER CANAL HUDSON INTERSTIAL
10 Diagram 1 shows the downward trend for the four regions. The Canal Corridor, especially the cluster of counties around Rochester, leads the trend toward improved population health. The straight catch up path of the interstitial region contrasts with the bent line of the other three regions. But the 2000 figures are higher than would be expected from the slope and a later section will propose a surprising explanation for this contrast. The larger question is: what accounts for the basic similarity of all four trend lines? The conventional answer is improved medical care, both preventive and curative. That is certainly part of the answer even though the reduction of mortality rates is not directly addressed by curative medicine. A second explanation that is usually mentioned for other regions is an improved economy but UNY has lost much of its manufacturing economy and the service economy has not replaced it. There may be other specific explanations, such as the strength of the educational institutions (both high school and college) but such sectoral emphases require a theoretical context. Structural theory invokes the principle of institutionalized problem-solving capacity and points to the increase in a wide range of differentiated institutions, more pluralism as reflected in the changing voting patterns and, to a lesser degree, regional solidarity as Upstate competes with Downstate.
11 11 I. THREATS, RESPONSE CAPACITY AND MORTALITY The theory in brief The theory that guides this report begins with the community unit which may be as small as the household and as large as the nation-state. As a fact of common observation, they are organized all over the world in the form of nested hierarchies. This structure is the backdrop for many community studies, including this one, but the analysis will focus on one level of community, the county. The model (Young, 2004; 2006b; 2009) begins with threats and ends with three major outcomes, productivity, exchange and population health. (See diagram) As already suggested, we will concentrate on the last. In between, are structure and transaction organization. The latter consists of all the familiar organizations and agencies that do the daily work of the community, which can be summarized as production, commerce and maintaining the health of the residents. Because they are constantly adjusting to a changing environment, it is difficult to measure these organizations. Their denotation in lower case letters (o) emphasizes this fact. By contrast, the three dimensions of structure, differentiation, pluralism and solidarity, are denoted in upper case letters (S). They derive from theory and pose the problem of finding measures that are congruent with the concepts. They also raise the question of how structure is related to organization. The answer to that question can be summarized in a formula: ph=(s*o), where ph is an appropriate measure of population health, S is one or more of the structural dimensions, and o is an appropriate community organization. The multiplication sign signifies their hypothesized mutual reinforcement.
12 Diagram 2: Model showing the elements of the structural ecology model of community adaptation THREATS CAPACITY OUTCOME Structure Transaction organization Productivity Mining Factories Example: Specialized skills Farming Resource depletion Corporate exploitation Technical obsolescence Commerce Exchange Minority influx Policy Contests Tourism Infrastructure loss Urban-rural Epidemics Reforms Public health Medical facilities Population health
13 The three principal structural dimensions in this formula may be defined briefly as follows: differentiation is the number of occupational specialties that the community draws on for problem-solving; pluralism is the degree to which a community engages in open debate, not confined to elected officials, over policy directions; and solidarity is the capacity to mobilize behind a leader or a program when the other problem-solving strategies fall short. The hypothesis that explains the adaptation that insures superior population health levels is summarized by another formula: ph = C/t where ph is an appropriate measure of population health as before, C is institutionalized problem-solving capacity and t is one or more threats. Capacity is the (S*o) reinforcement mentioned above. In other words, population health improves when community problem-solving capacity is superior to the threat. It declines when the community s capacity for managing threats is not up to the challenge. Although it was not designed to be analogous to the germ theory of disease, this theory takes the same general form because a strong immune system, and host resistance generally, defends the body from pathogen attacks, protecting it from disease and death. The difference, of course, is that the germ theory is a physiological explanation whereas the capacity/threat hypothesis is sociological. It is true that a measure like age-adjusted mortality involves physiology but even as measured by death statistics, population health is different. It is an overall rate across comparable counties in contrast to the disease rate criterion of the germ theory. The C/t formula omits the intervening processes of the capacity/threat hypothesis but they can be filled in as follows. When the C/t ratio is below 1, the community is
14 14 vulnerable to threats and a gamut of trial and error responses appear. Community leaders hold meetings and negotiations, citizens influence their representatives in different directions, and subordinate communities like neighborhoods and households may react on their own. Typical responses are deviant behaviors such as strikes, family instability or youth problems. Although these occasionally succeed in blunting or blocking the threat, they typically compound it, making matters worse. If the trial and error process fails, the rate of disruption of body maintenance habits increases: eating, sleeping, exercise, socializing and the like. These disruptions, in turn, affect the body s equilibrium and hasten death, regardless of disease. (See the detailed discussion and documentation in Young, 2009) Macro Threats to the County Communities Almost all communities face serious problems from time to time. But how does a community know when it is seriously threatened? If it is overwhelmed by a natural disaster, there is little doubt in anybody s mind, and the community leaders usually call for outside help. But many threats are insidious and may be invisible. Usually, however, there are indications, such as empty churches, abandoned factory buildings or an influx of people who are different from the natives. All these indicators require interpretation by community leaders, aided by experts in the bureaucracies of superordinate units. In the U.S., the demographers, public health officials, state police and the like work with counties and villages to define problems. In some cases they may assume the responsibility for managing them by prevention and/or control.
15 15 The nested hierarchy of communities and their specialists from superordinate communities is crucial in defining threats and allocating resources to manage them. Even so, many threats defy accurate diagnosis and it may take years before the community understands them in a way that fosters successful problem-solving. Alternatively, local officials may respond to a perceived threat such as the arrival of minorities with the primordial responses that all human groups make: hostility, control or exclusion. As the C/t ratio indicates, the impact of a threat depends on the problem-solving capacity of the community, as summarized in the (S*o) relationship. Unfortunately, measurement of this interaction is impossible with this dataset, although urbanization is controlled in all equations as an all-purpose measure of community capacity. The emphasis here is on the threats to community and Table 2 presents a factor analysis (descriptive statistics at the foot of the table) that defines the three threats that have impacted UNY over the years. The urbanization factor in the first column is a reliable finding for all community levels (where it may be called by the generic differentiation ). It is especially valuable because urbanization and the differentiation that it reflects is quite comprehensive and may indirectly measure the enhanced pluralism that is frequently found in urban centers. Urbanization is indicated positively by the percent of the county population that lives in places of 2500 or more, the number of cities, the number of small manufacturing firms, physicians per 100,000 population and the percent of female-headed households. The negative pole of this factor is indicated by the number of churches per 10,000 population and mobile homes per thousand. Few churches per capita reflect the larger congregations in cities, while mobile homes are infrequent in the urbanized counties.
16 16 Table 2: Factor analysis of threats to counties in UNY N=50 Indicator Urbanization Poverty Migration Decline Purban90.90 Churptt -.80 Mobilpt -.79 Cities.72 Mfgsmal90.76 Phypht.69 Pfemhh00.59 Povfam90.87 Mdnval Pnowork.84 Pservf00.75 Punemp89.66 Pblk Phisp Pservm00.69 Pctkids Page65plus.85 Pman Score range R Definitions and descriptive statistics (mean, std. deviation, minimum-maximum) Purban= Percent of population living in places of 2,500 or more (39.7; 23.4; ) Churptt=Churches per 100,000 population (124.8; 47.8; ) Mobilpt=Mobile homes per 1,000 (50.7; 37.1; ) Mfgsmal=Number of small manufacturing firms (101.7; 146.1; ) Cities=Number of officially designated cities (.9;.9; 0-3) Phypht=Physicians per 100,000 (158.6; 92.0; ) Pfemhh00=percent female headed households, (17.5; 1.4; 14 20) Povfam90=percent of families in poverty, (7.9; 1.8; ). Mdnval Median value of homes, ( ; ; ) Pnowork=percent of working age people who reported on the census form that they had a disability that kept them from working, (5.6; 1.2; ). Pservf00=percent of labor force in service jobs, female, (20.5; 2.8; ). Punemp89=percent unemployed, (6.1; 1.8; ) Pblk8090=percent increase of black population, (.9;.9; ). Phisp8090= percent increase of Hispanic population, (90.1;.62.3; ). Pservm00=percent of labor force in service jobs, (19.2; 6.4; ) Pctkids=percent of families with four or more children, (.5;.0;.4.6) Page65p=percent of population 65 years or older, (12.8; 2.1; 9 19). Pman8090=percent change in manufacturing workforce, (-27.0; 7.8; )
17 17 The second factor measures poverty. It is reflected in the high loadings for family poverty, low median values for housing, a high percent of the population that reported a disability that kept them from working, a high percent of the adult female population in service jobs, and high unemployment. This factor may also be interpreted as stagnation because most of the indicators do not change much over a decade or more. The Eberts and Merschrod study of trends in New York counties (2004, p. 152) provides independent corroboration. The third factor, Immigration, is indicated by positive loadings on the percent increase in the black population over , percent increase in the Hispanic population, and the percent of the male labor force in service jobs. The factor is called Immigration even though in this case it is mostly minority relocation. The general term is used, despite it vagueness, because majority (white) immigration is probably just as frequent as the minority influx. Even vacation amenities can attract whites. The fourth factor is labeled Decline because, although some of the indicators are cross sectional, they reflect negative change. The negative loading on the percent of the labor force in manufacturing is the core indicator. The percent of the population that is 65 years or over reflects the depopulation of the declining counties, while a negative loading on percent of families with children reinforces the picture. This factor analysis has generated measures of three threats that have been named, with only minor interpretation, Stagnation, Decline and Minority Immigration. These are probably viewed as serious by the residents of the affected counties but we have no independent validation on this point. In the present era, Minority Immigration is
18 18 probably perceived as an ominous development that may generate ethnic conflict. The maps display the factors. The first map locates the eight metro counties which are high on the Urbanization factor. These form a belt across the state that counter-balances the heavily urbanized downstate region. Because of their size and importance, they will be excluded from some statistical analyses. The second map shows the long term poverty that is associated with the Southern Tier counties. But as the map shows, persistent poverty is also marked in the northern and central counties (one of which is the largest state park in the U.S.). These 16 counties constitute a major drag on the UNY economy. The third map shows that the 16 high migration counties form a corridor along the Hudson River and points of concentration near the metropolitan centers. These counties have a lower cost of living but still permit access to jobs in the metro counties. The fourth map shows the 15 counties that the factor analysis identified as Declining. These counties include the aging metro center of Buffalo as well as the contiguous counties north of the New York City region. This region extends north to the two Adirondack Park counties because the census definition of manufacturing includes the processing of logs and minerals that once thrived in this region. Of course, decline does not necessarily mean collapse; as we shall see, there is evidence of a still strong middle class way of life in these counties.
19 Map 1 Metro Counties in Upstate New York Ranges for fac1d Means 1 to 1 (8) 0 to 1 (42)
20 Map 2 Poverty in Upstate NewYork Ranges for fac2d Means 1 to 1 (16) 0 to 1 (34)
21 Map 3 Migration to Upstate New York Ranges for Fac3d Means 1 to 1 (16) 0 to 1 (34)
22 Map 4 Economic Decline in Upstate New York Ranges for fac4d Means 1 to 1 (14) 0 to 1 (36)
23 The Impact on Mortality The next step in the analysis is to show the statistical relationship of black and white mortality when the four factors are the predictors. Inasmuch as the varimax rotation (a standard default) of the factor analysis sets to zero the correlation of each of the factors to the others, problems of collinearity are eliminated. But, of course, the small sample, which is further reduced by case loss and the elimination of outliers, severely handicaps the analysis. Table 3: Regression analyses of black and white mortality using factor scores as predictors (N=42**) Predictors Aamb7998 (black) Aamw7998 (white) F1Urbanization -.33* -.01 F2 Poverty.08.39* F3 Immigration -.41*.56* F4 Decline R * Significant at the.05 level or better. **The four largest metro centers were omitted and four cases were lost because of missing data for Amb7998. The rationale for using the average as the criterion variable is discussed below. Aamb7998=age-adjusted mortality, black, per 100,000, average. Aamw7998=age-adjusted mortality, white, per 100,000, average. F1 Urbanization factor score from Table 1. F2 Poverty/stagnation factor score from Table 1 F3 Immigration factor score, minorities, from Table 1. F4 Decline factor score from Table 1.
24 24 Urbanization shows a negative association with black mortality but the correlation with white mortality is zero. African-Americans seem to thrive in the large urban centers. The more interesting coefficient is the association of mortality with the Immigration factor. It is significantly negative for blacks, which means that the greater the increase in black residents during the decade, the lower the mortality rate of all the people in the county. We do not know from this correlation whether it is the minority migrants who have the lower mortality rates, although that is probable, but the county-specific association always reflects a community-level dynamic. That means that the networks of communication, neighborhoods, ethnic clubs, etc. of the migrants work to improve their health. The structural hypothesis rejects a composition effect which would claim that the migrants were healthy when they arrived and their increased number affected the rate. The coefficients for white mortality in Table 3 are equally surprising. The Poverty factor predicts positively as expected, but the Minority Immigration factor is also significantly positive, which means that a higher proportion of migrants in a county is associated with higher white mortality. Are the whites threatened so much by the minority arrivals that they die at higher rates? Or are the job opportunities attractive to blacks but not so for the whites and the dominant group suffers from unemployment and general economic stagnation? Note that the Decline factor is not a significant predictor for either mortality rate.
25 25 As noted, this table used a 20-year average, , for the mortality criterion. The CDC website calculated these estimates and they provide a single figure for what amounts to a two decade cross-section. But using such averages is quite different from using estimates for the year 2000 and treating the factor scores as antecedent conditions. The cross-section estimates produced significant findings and they captured the impact of the 1980s minority influx. The regressions that used the 2000 criterion were weaker. But, of course, the data point referred to a later and more stable period. The Challenge of Macro Threats The correlations linking macro threats to black and white mortality raise the question of how they might be causally connected. Supporters of the dominant biomedical theory would point to a physiological sequence, starting with the stressors in the environment that weaken the immune system and leave the bodies vulnerable to diseases. Some of the diseases shorten lives and affect the mortality rates. This study was not designed to test that pathway and it would be difficult at best. It would require indicator(s) of stress and of a complicated chain of physiological processes that included the onset of diseases. The focus of this study is the parallel causal sequence that begins with serious threats to communities, such as a factory closing or a rush of minority arrivals. When community problem-solving capacity is weak, these threats disrupt health habits. Sleep is disturbed, consumption habits fluctuate, physical activity is
26 26 reduced and social interactions patterns become troubled. Such disruption impacts on the overall biological functioning of the body, not just the immune system. A complete test of this sequence would require individual-level data and multilevel statistics, but the threat-mortality test shown here is valid even without evidence for the intervening processes because the rates of habit disruption are considered emergent properties at the community level along with mortality. The whole causal sequence works at the community level (see Sampson and Wilson, 1995 and Menzel, 1950 for a similar argument). The basic hypothesis claims that it is the ratio of structure to threats, not the direct impact of one or the other, that is causal, and this ratio is difficult to measure. The all-purpose control on urbanization used in this study does not measure the capacity/threat ratio. Another kind of problem is that the Decline factor did not predict. One explanation for this unexpected finding is that the older people in the declining counties are the remnants of the manufacturing elite and have maintained good health over the years. Poverty/stagnation predicted higher mortality for whites as expected. That leaves the Minority Immigration factor that is associated with lower black mortality and higher white mortality. The first correlation can be explained by hypothesizing a solidarity effect of increasing size. The literature seems not to contain supporting studies (but see Cooper, Steinhauer, Schatzkin and Miller, 1981) but it is plausible and fits the structural model. The literature does contain theory and findings (Fossett and Kiecolt, 1989; Frisbie and Neidert, 1977; Quillian, 1995) on the minority threat hypothesis and these are relevant to the cross-group correlation (i.e. blacks threatening whites). The new idea here is that serious threats of this type can increase death rates of the native residents.
27 27 II. THE ANALYSIS OF ETHNIC SITUATIONS The classical format for applying a theory is to identify problems that appear to be open to solution and then extrapolate feasible interventions that promise to solve the problem. Thus, in the well-known story, Edward Jenner and others around 1800 applied an embryonic immune system theory to smallpox. Guided in part by observations of the apparent immunity of milk maids to the dreaded disease, they invented vaccination, the procedure of infecting a person with a small dose of the smallpox germ to stimulate the production of antibodies in the human body, thereby protecting people from the epidemics of smallpox. Both the theory and the diagnosis were crude, and the vaccination procedure was primitive, but the basic strategy is clear. The fact that they began without complete understanding is typical of many applications. The version of social ecology outlined here points to social problems that threaten a given level of community, depending on its problem-solving capacity, as the causes of mortality differences among the counties under study. Of the three serious threats, long term location/ resource poverty and deindustrialization reflect long-term nation-wide trends, involving outsourcing of traditional jobs and hiring highly educated specialists for new jobs. There is not much that regional officials can do to ameliorate this process. That leaves the influx of poor minorities as the only threat that is open to improvement by means of interventions that the region can implement. But minority influx takes many
28 28 different forms. How can we find the situations that offer the best chance of success? This section of the report introduces the focused factor analysis ( see Young and Rodriguez, 2005, for another example) as a technique for examining ethnic situations under the statistical microscope. It identifies four such situations and analyzes their potential for lowering mortality. It also looks at the possibility of interventions. It appears that only one of these situations, voluntary migration of new people, offers the best bet. Arriving at that judgment required some initial explorations. These preliminaries may be called successive approximation because they take the form of locating the high and low values across a range of well-known criteria such as regions within a state, race/ethnicity and decade, the 1980s in particular. These rough distributions set the stage for the more refined factor analysis. But these descriptive tables are tricky because some of them, such as race, reflect the individual and household level of community and it is not possible with this dataset to separate these levels. The Distribution of Mortality by Race and Region The map of the geo-historical regions of Upstate New York presented earlier offered an initial picture of the contrasts in mortality. Such geographical subregions reflect different histories and forms of social organization, contrasts that may help to interpret the different ethnic problems that communities face. Accordingly, Table 4 uses the regions for comparing black and white mortality in
29 29 the geo-historical regions of New York. The average is used and the regions are listed clockwise beginning with the southwest corner. The black-white gap is redefined as the level of mortality for whites (including Hispanics) subtracted from that of blacks for each county. Table 4: Means for black and white mortality by region, upstate New York. N=50 Region Aamb9901 Aamw9901 Aambwgap00 S. Tier Canal corridor Hudson valley Interstitial Total The mortality estimate is the average for 1999, 2000 and In these tables, the measure of mortality gap aggregates the separate county differences. By this method, the gap for the Canal counties is 212, not the difference between 1071 and 848. The fact that 204 in the first row is the difference between 1067 and 863 is coincidental. For blacks in New York, the rates for the Southern Tier and the Canal corridor are, surprisingly, almost the same despite the contrast between these regions. The Canal Corridor, along with the Hudson Valley, is the principal urban axis of New York State while the Southern Tier is marginal to these. This contrast is also reflected in the Hudson Valley s low mortality and the higher mortality of the adjacent interstitial region. With respect to the black-white gap, the Hudson Valley shows a smaller gap of 148. Have these counties discovered the secret for reducing racial
30 30 differences in mortality? An analysis of the means for several key variables across the regions turned up only one relevant fact: the median value of housing in the Hudson Valley is $88,589, much higher than the $68,000 figure in the interstitial region. This figure may reflect a stronger social structure that protects both races. Moving to the three decade comparison in Table 5, the first fact is that mortality for both races has been declining since 1980 by approximately 100 deaths per 100,000 per decade. The two races have declined at approximately the same rates, as reflected in the comparable black-white gaps. These estimates indicate that the factors causing the change impacted on the two races in ways that maintains the gap. Table 5: Comparison of age-adjusted mortality per 100,000 population, for blacks and whites, in Upstate New York, Variable UpstateNY counties N=50* 1980** Aamblack Aamwhite BW gap * The New York sample excludes the 12 downstate counties. **These marker years actually refer to the five year averages starting The figures for black mortality are based on reduced Ns due to deletion of counties with too few minorities.
31 31 Do the patterns in Table 5 suggest serious problems confronting the counties? On the one hand, the mortality rates are declining. That, according to structural theory, implies either reduced state-wide threats or improved capacity for dealing with problems or both. Biomedical specialists will claim that the mortality rates are declining because they have been successful in solving the many heath problems of these residents. But this claim is weakened by two facts. First, it implies that the aggregate of medical improvements accounts for the changes. But the composition of improvements almost certainly varies across counties and it is hard to see how these different bundles could produce the same trends. It is possible that a single major improvement, like reducing blood pressure, accounts for the decline. If so, that factor should be identified. The second problem is reflected in the black-white gap. Are biomedical improvements less effective for African-Americans? Or do the blacks have less access to the treatments? To a sociologist, the black-white contrast suggests structural causes that impact differentially on the two races, a topic to which we may now turn. The Structural Contexts of Black and White Mortality The starting assumption is that ethnic threats are quite varied and require closer analysis. Accordingly, the list of indicators to be factored included all the available ethnic indicators for Hispanics and blacks, with whites the unmeasured reference group. As shown in Table 6, the New York analysis generated five factors, starting with the familiar Urbanization.
32 32 Table 6: Factor analysis of structural, especially ethnic, indicators for upstate New York counties. N=50 Indicator F1 Urban F2 Prison F3 Ethnic Crime F4 Old Housing estadmin97.93 esthith97.93 estprof97t.92 church90.90 large89t.90 purban90.78 phyphtt.74 pblack crmvioht phisp8090r.76 plfm pblk8090r.74 pcprrinst00.69 pservm00.69 pservf crmprpht97.76 phispan00.76 pfemhh00.63 pser page65p00.81 phos9000r -.73 phos8090r -.65 phisp9000r.83 pblk9000r.79 F5 Migration Range R
33 33 Definitions of the factor analysis indicators, and sources (US Census, unless otherwise indicated): Estadmin97= administrative establishments, County and City Data Book, (CCDB, 2000) Esthlth97 = health establishments, (CCDB, 2000) Estprof97t = professional establishments, truncated, (CCDB, 2000) Church90 = churches, Large89t = large manufacturing establishments, truncated, Purban90 = percent of the population in centers with 2,500 or more, Phyphtt = physicians per 10,000 population, Pblack00 = percent African American, Crmvioht97 = violent crimes per 100,000 population, (CCDB, 2000). Phisp8090r = percent increase (relative change) Hispanic, Plfm00 = percent of the male population in the labor force, Pblk8090r = percent increase in the black population, Pcorrinst00 = percent of institutionalized population in correctional institutions, Pservm00 = percent of males in service jobs, Pservf00 = percent of females in service jobs, Crmprpht97 = property crimes per 100,000, (CCDB, 2000) Phispan00 = percent Hispanic, Pfemhh00 = percent female headed households, Pser9000 = percent change in service jobs, Page65p00 = percent of population aged 65 or over, Phos9000r = percent increase in housing, Phos8090r = percent increase in housing, Phisp9000r = percent increase in Hispanic population, Pblk9000r = percent increase in the black population. The Urbanization factor lists nine variables that load.50 or more, the standard threshold. There are five counts of establishments that deal with administration, health, professional services, churches, and large firms. They are not standardized by population because the raw counts define the variety of specialties, which is the definition of differentiation. The percent of the population in centers of 2,500 or more (the census definition of urban places) is included as a validator and because population concentration typically implies
34 34 specialized occupations). In addition to these, physicians per 10,000, the percent black population and violent crimes per 100,000 load on this factor. The physician indicator reflects specialization and, in an abstract way, so do violent crimes, if we assume that variety correlates with number of violent crimes. That leaves the higher percent of black people in 2000 which, although a typical attribute of large cities, reflects specialized knowledge only in the loose sense of ethnic diversity. The second factor, Prisons, is reflected in the percent increase in Hispanics, , a lower percent of the population in the labor force due to the uncounted prisoners, a high percentage increase in the black population, , a high percent of the institutionalized population in correctional facilities (rather than in nursing homes or college quarters) and high proportions of both male and female service workers. The event that allows us to make sense of these indicators is the 1980 crack cocaine epidemic in New York and the harsh incarceration laws that were passed in an attempt to contain it. The prisoners were mostly from the two minorities and the crowded prisons required many non-prisoner service workers. The Ethnic Crime factor 3 is reflected in the high loadings for percent black for the marker year 2000 (loading on two factors), violent crimes and property crimes per 100,000, percent Hispanic, percent female households and lower (as reflected in the negative sign) percent change in the service labor force in the 1990 decade. Thus, ethnicity, crime and poverty are combined in this factor that accounts for 14 percent of the total variance.
35 35 The fourth factor, Older Housing, is defined by the high loading on the percent of residents 65 and older in 2000 and the low percent increase in houses in the 1990s and in the 1980s. In other words, housing in these two decades was stagnant, leaving the old people in their older homes. The fifth factor, Migration, is defined by two measures of Hispanic and black population increase during the 1990 decade. A search for other indicators that would throw light on this pair of indicators was unsuccessful, but the Migration label accurately summarizes the two indicators. Table 7 shows the regressions for black and white mortality for each factor score (a summary measure that ranges from -3 to about +3), holding the other four constant (by virtue of the varimax rotation, which is a standard default). For the 46 counties with a black population that was large enough for calculating reliable estimates for , only two factors, Prisons and Migration, predict, and the signs for both are negative. Taking the correlations at face value, they indicate that the counties with expanding prison populations have lower black mortality. Likewise, the counties with large percentage increases of black and Hispanic newcomers have lower mortality. For the whites in all 50 counties, the Prison factor predicts higher white mortality. The Ethnic Crime factor predicts higher white mortality only when the eight metro counties are excluded.
36 36 Table 7a: Regression analysis of black mortality, , in NYS counties, N=46 Predictors Coefficients FAC1urban -.12 FAC2prisons -.32* FAC3crime -.04 FAC4old housing -.08 FAC5migration -.49* R2.28 Table 7b: Regression analysis of white mortality, , in NYS counties, N=50 Predictors Coefficients FAC1urban -.12 FAC2prisons.62* FAC3crime.15 FAC4old housing.05 FAC5migration.03 R2.35 Note: the factor scores for the above predictors were derived from the factor matrix in Table 6. By definition, each of them has a mean of 0 and a standard deviation of 1. Their ranges are shown in a lower row of Table 6.
37 37 How can we explain the negative (low mortality) coefficients of the blacks and the high mortality of the whites? Are prisons really beneficial with respect to black mortality? And, assuming that the number of white prisoners is large enough to influence the county rate, do prisons cause an increase in white mortality? That, in fact, is the most plausible interpretation. For young black males, a safe environment, a stable routine, good food and, we might add, black ethnic dominance in the prisons, actually fosters population health. Despite the loss of freedom, mortality is lowered as compared to black populations in nonprison counties. This association of incarceration and lower mortality has been independently noted in a recent study by the U. S. Department of Justice (Mumola, 2007) that reports that the mortality rate of black inmates was 57 percent lower than the adult U.S. (of all races) resident population. The literature on recent developments in New York prisons throws light on the positive impact on the white mortality rate. While the data do not tell us whether it is the white prisoners that are boosting the mortality of the whole county, the evidence strongly suggests that it is. By 2000, the black prison population comprised 51 percent of the inmates as compared to 16 percent (non- Hispanic) white (Table H-21 New York Dept. of Correctional Services, online) and the numerical strength was reinforced by the spread of Islam among blacks (Jacobs, 1983:66; Zoll, 2005). Still a third aspect, according to Jacobs, was racial polarization in the form of increased black hostility toward white guards and black dominance over white inmates. Even when blacks were a minority, they were able
38 38 to intimidate the other inmates by means of rape, control of space and the prison culture, everything from the music played over the loud speakers to the type of food served. The accounts of life in prisons (Jacobs, 1983; Canover, 2000) are unanimous in finding it extremely stressful for the guards. They were constantly anxious, fearful of sudden attacks, and nervous about arguments and confrontations with prisoners. Their frequent depressions impacted their home life. These authors note their absenteeism, heart attack rates, alcoholism, and frequent resignations. Although there are no official figures, some of these writers reported that guard death rates were also high. Insofar as this condition holds for most of the prison counties in New York, the white mortality rate should be significantly higher as indeed it is. The high scoring counties on each of the factors can be mapped as shown in the following series. Map 1 displays the counties that score high on the Prison factor and it is apparent that they are located in the poorer northern counties where they provide employment for the population. Other prisons are near the metropolitan centered counties.
39 39 Map 1 FPrisons Upst Ranges for F2p Means 1 to 1 (16) 0 to 1 (34)
40 40 Map 2 FCrime Upsta Ranges for F3c Means 1 to 1 (16) 0 to 1 (34)
41 41 F4 Oldhousing Upst Ranges for F4old Means 1 to 1 (16) 0 to 1 (34)
42 42 Upstate NY Mig Ranges for F5m Means 1 to 1 (16) 0 to 1 (34) Map 2 shows the high scoring counties on the Ethnic Crime factor. As is apparent, crime is high in the counties that lie just outside the downstate region and also adjacent to the metro counties. The third map lack of increase of the housing stock as implied by the negative signs for the 1980 and Many of these counties have experienced decline in manufacturing such as in around the city of Buffalo, which at one time was the leading city in western New York. The fourth map shows the counties that serve as corridors or targets for minority migration. Many of these counties have experienced deindustrialization while others are centered on a large city. Both kinds of counties offer work and services for the migrants.
43 43 III. APPLICATIONS OF STRUCTURAL ECOLOGY Programmatic Steps The broad perspective that derives from structural ecology has been called social problems public health (Young, 2006; 2008) because it claims that, in addition to the many invisible pathogens that conventional public health departments work to prevent, social problems can cause lower population health. These problems fall into two groups, those like deindustrialization that are impersonal, and those, like an influx of poor minorities, that seem to be caused by a particular group of people. From a sociological perspective, these newcomers are part of an impersonal process that is no different in kind from deindustrialization, but that is not the usual perception of the older residents. Another difference is that the problems that immigrants bring can be handled more adequately with local resources. So the first guideline is to select an external problem that appears to be tractable. At the present time, higher migration seems to be the best choice for Upstate New York. The second implication of the theory is to recognize the levels of communities even if, as is true of this report, the researcher must focus on one level. In New York State, the subdivision of the county is the town, which is different from a small city although one or more of these may exist within the town. Below the incorporated cities are the incorporated villages and the
44 44 unincorporated hamlets which the census classifies as Census Designated Places (CDPs). The availability of data is limited at the lower community levels, but analysis of these smaller units is often possible in the context of county structure. A third implication is the necessity to work with communities as wholes, as defined by official boundaries and their socially reinforced limits. The many county-level institutions that function within these boundaries are the basis of measures of the three master dimensions. Structural theory does not assume that communities tend toward integration or equilibrium. On the contrary, the structural dimensions of differentiation, pluralism and solidarity measure the unit character of communities. Beyond these broad guidelines are a number of operational rules that merit discussion. The first is that structural initiatives must be kept separate from the biomedical based public health and private medicine. Otherwise, structural practicioners will have difficulty claiming that the social changes they have identified are causal. Of course, evaluations must show that one or a cluster of interventions account for a significant deflection of the mortality slopes. Analyzing the causes of such trends is a central task of structural research. Even when development programs are low key, it is risky to confine them to one or two communities except as pilot projects. If they succeed, they are likely to attract an overflow of immigrants. So a given program must cover all relevant communities, a rule that could require efforts in many upstate counties. That is the public health ideal, of course, but it is also a financial impossibility in most cases.