The Survival of Sheltered Care Homes: Facility and Neighborhood Contributions

Abstract

We use data from a representative probability sample of all 1973 California Sheltered Care facilities for ex-mental patients. Facilities were recontacted between 1983 and 1985. Based on several literatures, we hypothesized variables that might contribute to facility survival over this period including: neighborhood factors such as community reaction and gentrification and organizational and institutional characteristics such as profit motivation and legitimacy. Contrary to expectations, age of facility, appreciation in housing values, vacancy rates, neighborhood antagonism, gentrification and conservatism were not related to closure. Instead, the facilities were more likely to stay open when they possessed a steady income stream and when they were more businesslike and licensed. They were also likely to stay open when they were located in very poor and mixed use neighborhoods.

The transformation of urban space has gripped popular attention. In particular, attention has focused on the declining availability of moderate and low-income housing. The ability of sheltered care facilities -- halfway houses, psycho-social rehabilitation facilities, board and care facilities and family homes -- to survive within a changing urban ecology has not commanded the same public awareness. Yet, these facilities are a major resource for people with mental disabilities. They provide a low cost “bed of last resort” for a very needy population at risk of becoming homeless. Ideally, they also provide supportive care and access to supplemental services (Segal & Kotler 1989). Indeed, in 1981 they housed almost as many people with mental disabilities as did state and county hospitals at the outset of deinstitutionalization (Goldman, Gattozzi, & Taube 1981).

If sheltered care facilities are to continue to provide housing for ex-mental patients, they must be able to hold their own in the changing urban housing market. We know the supply of other forms of low cost housing has decreased (Gilderbloom & Applebaum 1988; Bratt, Hartman, & Meyerson 1986). Advocates are concerned about the possible loss of sheltered care beds (Blaustein & Viek 1987). Here, we look at facility survival over a ten year period. Using data gathered in 1973 and again in 1983–1985, we asked what organizational, institutional and neighborhood characteristics help foster sheltered care survival. We focus on the predictors of survival rather than the more typically investigated formation of new facilities. The study then helps ascertain what kinds of facilities in what kinds of neighborhoods are able to survive over time. This study is part of a larger longitudinal study of sheltered care and its inhabitants. It should be noted that while the data were gathered some 7 years ago, this quality of data are not available elsewhere. Further, there is no reason why the results should not be relevant today.

Background

The mortality of sheltered care facilities is an understudied topic. Our work is guided by two literatures. The sheltered care literature looks directly at facilities but lacks a strong empirical base. The organizational survival literature is theoretically and empirically grounded, but is not specific to sheltered care.

The sheltered care literature typically looks at where facilities cluster and assumes this is where they survive (Berdiansky & Parker, 1977; Davidson, 1981; Dear & Wolch, 1987). Authors discuss the importance of neighborhood factors in predicting facility survival (Robert Wood Johnson Foundation, 1990; Dear 1991). In particular, facilities are thought to do less well when the local social climate is inhospitable and when local real estate values support more profitable land uses. Thus, gentrification, rising housing values and “NIMBYism” (not in my backyard), are thought to inhibit new facility formation and contribute to the failure of existing ones (Baustein & Viek, 1987; Currie, Trute, Tefft, & Segall, 1989; Wenocur & Belcher, 1990). The Mental Health Law Project has estimated that fifty percent of the sites originally chosen for group homes are never developed because of community opposition (Rizzo, Zipple, Pisciotta, & Bycoff, 1992). Certainly, the impression is that facilities concentrate in areas that have not gentrified, have a mixture of land uses and thus less opposition to non-standard residences, and a poor and transient population less capable of mounting organized opposition (Dear, 1991). However, do the factors that favor formation of new facilities necessarily favor their survival? A facility strong enough to form in a hostile environment might better survive pressures that cause other facilities to close. Thus, this literature needs empirical support.

Work on general issues of organizational survival is empirically grounded. Researchers have tested the importance of organizational characteristics, institutional processes and processes linked to competition and cooperation among populations of organizations. We do not place ourselves within the ongoing debate about the relative importance of ecological and institutional factors; we lack the data to do justice to an analysis of the organizational ecology of sheltered care. We can, however, identify factors shown to be important in the literature and applicable to sheltered care.

One of the most consistently identified predictors of organizational mortality is age. Stinchcombe (1965) first argued for the liability of newness. More recent work has shown the organization’s age predicts mortality, albeit in differing ways. Some find risk of failure declines monotonically with age (Freeman, Carroll, & Harmon, 1983; Carroll, 1987), others that it declines monotonically after an initial adolescence when the risk is low (Bruderl & Schüssler, 1990). Yet others find the association is curvilinear with high risks among new and old organizations (Aldrich, Staber, Zimmer, & Beggs, 1989; Staber, 1989). Allied work hypothesizes, with some empirical support (Freeman, Carroll & Hannon, 1983), that larger organizations have a greatest chance of survival. They are thought to have more resources to enable them to withstand temporary downturns.

Yet other work hypothesizes that organizations demonstrating greater conformity to institutional norms are better able to survive (Meyer & Rowan. 1977; Scott & Meyer, 1983). Such organizations are held to be more legitimate. more accountable and to inspire greater trust. Legitimacy buffers the organization from questioning and sanctions, regardless of actual performance (Meyer & Rowan, 1977; Scott, 1982). Further, they are more likely to be linked to other legitimating organizations and, as such, to have more access to buffering resources. The importance of conformity is held to be particularly key for institutionalized organizations, ones held accountable to standards unrelated to productivity concerning appropriate behavior, role allocation and the like (Meyer & Rowan, 1977). As discussed below, these expectations should be relevant to sheltered care organizations since they are dependent on professionals for referrals to keep their beds full.

The Nature of Sheltered Care Facilities

In order to adapt the existing literature to sheltered care, we must consider the nature of a for-profit enterprise with a specific history and nature. With deinstitutionalization, employees in state mental hospitals also left for the community, sometimes to open shelter care facilities for ex-patients. Since sheltered care facilities were first homes for their tenants, they needed to locate in areas where property was designed for residential purposes or convertible to this use. This was most likely to be in residential areas. For example, in the 1973 survey of California sheltered care homes analyzed in this paper 70% of the homes were in totally residential areas, 28% in areas of mixed commercial and residential uses and only 2% in primarily commercial areas. In some cases, people converted their homes to this purpose; in others they bought or leased inexpensive marginal housing (Segal & Aviram, 1978). Concomitantly, the nature of the organization varied (Segal & Moyles, 1988). Some facilities were extensions of the owner’s family, involving a long-term relationship with a small and sometimes dependent population. These were influenced by the social and moral constraints arising from the needs of their residents. In such cases, the profit from the facility could well be of lesser importance. Indeed, over half of the owners who were operators in this sample were only partially or not at all dependent upon the income from the facilities. In other cases, the facility was a small business, perhaps with hired staff, a separate plant, and the potential for expansion through the leasing of other housing.

Although clients or their families could find a space in the facility on their own, this was not the typical pattern. Seventy-eight percent of the California facilities in 1973 relied primarily on professional referrals. Organizations dependent on professionals fit the definition of institutionalized environments; they secure legitimacy and resources by virtue of their proper connections to their environment (Meyer & Rowan, 1977; Meyer, Scott, & Deal, 1981). The most obvious such connection for sheltered care was the state certification provided by licensure. Initially, facilities were not regulated, except for the need to gain use and other business permits. States initiated licensure laws fairly early on. In 1974, California began licensing facilities.

Licensure placed certain demands upon facilities including (a) the need for an initial inspection, (b) certain quality of the physical plant, (c) programming demands, and (d) the need for annual licensing fees. A licensed facility would be listed in the licensing book and presumably better able to attract placements from professionals, would be able to define itself as a supervised living arrangement and eligible for the Supplemental Security Income (SSI) fee, and would generally have greater legitimacy.

Our hypothesized reasons for closure then must reflect the nature of an entity that may more resemble an extended family or a business. Since it “sells” a residence, and residences are located in neighborhoods with the power to make the life of the owner uncomfortable or even untenable, the ecology of the local neighborhood may be important. Finally, because it shares important elements of “institutionalized organizations,” it is dependent on institutional linkages and legitimacy.

Facility Characteristics and Survival

Our first hypothesis is that sheltered care facilities run as a for-profit business will more often remain open. More businesslike facilities will more likely be those where the owner has had to get a lease for the property because it is not run out of the owner’s home, where the owner must get a use permit, and where the owner is clearly interested in the profit to be gained from the facility’s operation. In such cases, the costs of opening and maintaining the facility are higher, the business itself is seen as having worth separate from the value of the property and more is lost if the facility is simply closed rather than sold to a new operator. Whether or not the facility owner has a business orientation, the ability to at least break even. if not make a profit, is important (Shadish, Lurigio, & Lewis. 1989). The financial security of the sheltered care facility depends less on its ability to sell a product than its capacity to ensure beds are filled by a paying population. Facilities housing a more permanently and seriously disabled population are more likely to possess these guarantees because of the residents’ lack of alternatives and the steady income provided by government transfer payments.

We should note that we are not able to consider the age of the operator. In the data analyzed here, ages are only available for owner/operators and not for owners who do not manage their own facilities (38 percent of this sample) and therefore cannot be included in this analysis. We should note that a separate analysis of whether operators (rather than facilities) remained in the sheltered care profession (Segal, Hazan, & Kotler, 1990) found that age was not a significant predictor.

We are not able to make any firm prediction about the effects of the facility’s age, but follow the literature in expecting it to be important and include it as a control. Given the conflicting predictions about the shape of the association, we will examine the plot to determine the best fit.

We anticipate that facilities that have greater legitimacy will be more likely to stay open. Legitimacy, for sheltered care, is best represented by whether or not the facility is licensed.

Hypothesis 1: Business-oriented facilities will be more likely to stay open.

Hypothesis 2: Facilities guaranteed a more steady stream of income are more likely to stay in business.

Hypothesis 3: Age affects the survival of the facility (direction unspecified).

Hypothesis 4: Licensed facilities are more likely to remain open.

Neighborhood Characteristics and Survival

Based upon the general trends in urban neighborhoods, the sheltered care literature and popular perceptions we hypothesize several possible sources of closure:

Neighborhood Reaction

The popular press discovered the notion of NIMBY -- community resistance to unwanted land uses -- during the 80’s. While the topic of this analysis is the survival of existing facilities rather than reactions to proposed new ones, there is evidence that existing facilities are the targets of complaints, if not always organized resistance (Segal & Aviram, 1978; see also Blaustein & Vick, 1987 who mention community resistance as one reason given for closure).

We might expect neighborhoods to be least likely to organize against or otherwise display hostility to facility owners when their population is poor, historically the group with the least capacity to create and sustain a community organization without outside help (Taylor, Hall, Hughes, & Dear, 1984; Logan & Molotch 1987; Wenocur & Belcher, 1990). We also hypothesize that neighborhoods with a mixture of land uses would have no single constituency capable of mobilizing the entire neighborhood. They also have less sense of what an appropriate land use might be (Davidson, 1981; Dear & Wolsh 1987). We further hypothesize that neighborhoods changing toward more affluent populations should show greater hostility since the facilities would be seen as standing in the way of the full gentrification of the area. Finally, we expect that more conservative neighborhoods show more resistance to facilities located in their midst.

Hypothesis 5: Facilities in neighborhoods with the least capacity to organize against them should be less likely to close.

Hypothesis 6: Facilities in neighborhoods that show direct hostility should be more likely to close.

Hypothesis 7: Facilities in gentrifying and conservative neighborhoods should be more likely to close.

Changes in House Values and Increased Competition for Urban Land

Over the past few decades, real estate prices have appreciated greatly nationwide. This is particularly true in California where our data are gathered. Proposition 13, the property tax revolt, insulated sheltered care operators from any direct increase in taxes. However, the amount of income they would hope to realize upon sale might convince owners of financially unrewarding facilities to sell. Furthermore, in 1973 about 24% of the residential care facilities in California were located in urban working class or poor neighborhoods. These neighborhoods are the most suitable for gentrification and thus under pressure to sell either for residential or commercial uses.

Hypothesis 8: Facilities in neighborhoods with the greatest increase in house prices and the greatest demand should be the most likely to close.

Methods

Our analysis is based on data gathered in California in 1973 using a probability cluster sample of 214 sheltered care facilities. The initial sampling frame included all family care, board and care, and half-way houses in California serving at least one resident with a mental disability.

In order to secure the frame, the state was divided into three master strata (a) Los Angeles county, (b) the nine county Bay Area, and (c) all other counties. Facilities were stratified by size in both Los Angeles and the Bay Area and samples of facilities were drawn with probabilities proportionate to bed capacities. In the stratum composed of all other counties, two counties were selected from the north of the state and two from the south with probabilities proportionate to capacity. From each pair, a sample of facilities was selected, again with probabilities proportionate to size. The sample of facilities then is a self-weighting probability sample representative of all 1,155 facilities in the state during the summer of 1973 (Segal & Aviram, 1978, contains further details of the sampling procedures).

Facilities were recontacted between 1983 and 1985 and current managers interviewed. Of the original 214 facilities, 156 were still open at the later date. Interviews were completed with 151 (97%) of the managers. Of those facilities which had closed, over half of the buildings had been converted to simple residential use (63%) and the majority of the remainder were either converted to business use or parking lots (26%).

Because interviews were done over an extended period, there is the problem that facilities closing between 1983 and 1985 might be coded as open if contacted earlier in the interview period and closed if contacted later. All analyses presented in this paper were replicated, treating any facility closed after 1983 as open. This did not alter the predictive models.

Data were gathered in 1973 from the operators on all aspects of the facility operation including size, population served and neighbor reactions to the facility. Census tract information was compiled for all facilities, both open and closed, from the 1970 and 1980 censuses. Data on voter registration was gathered at the level of the census tract from listings compiled in 1973 by the California State Data Repository System and in 1983 by the Rose Institute.

Variable Measurement

(All variables measured in 1973, unless otherwise specified.) Factor analysis is an analytic technique that permits several variables to be combined into a single underlying dimension. Each variable “loads” onto the resulting factor, with the weight showing how much each variable contributes to defining the factor. We used this technique to create single measures for some of our hypothesized variables. (All factor analyses are in Table 1).

Table 1.

Factor Analyses Used to Construct Facility - Neighborhood Variables

I. Business Orientation	Factor Loading	% Varance Explained
Own and live in facility	−.86	.56
Over 6 bed	.78
Profit orientation	.57
II. Degree of Disability among Residents	Factor Loading	% Varance Explained
Need help with life functions	.43	.39
In state hospital more than 5 years	.81
Unable to hold regular employment	.56
III. Poverty (’70)	Factor Loading	% Varance Explained
% males in labor force (LF) employed	−.76	.63
% females in LF employed	−.78
% below poverty line	.84
Poverty (’80)		% Varance Explained
% males in LF employed	−.78	.56
% females in LF employed	−.78
% below poverty line	.67
IV. Class (’70)	Factor Loading	% Varance Explained
Median income	.88	.70
Median education	.82
% professional/managerial	.82
V. Class (’80)	Factor Loading	% Varance Explained
Median income	.75	.48
Median education	.75
% professional/managerial	.55

Hypothesis 1: Business Orientation

Three variables were combined into a single factor score:

1. Whether the facility was located in the home of the owner.

2. Whether the facility included six or fewer beds. In California, special use permits are required for larger facilities.

3. Whether the income gained from the facility was important. Operator/owners were asked how important the income from the facility was to them. Those reporting very or somewhat important were coded as 1 and those saying not too, or not at all important as O. Since this question was not asked of the non-operator owners, we had to make the assumption that income was important to them.

A principle components factor analysis was computed and the resulting “business orientation” factor score used.

Hypothesis 2: Steady income stream

Since a steady income can depend on housing a more disabled population, level of disability of the residents was measured by combining three items into a single factor score:

1. Operator evaluation of the proportion of residents with mental disabilities in the facility who were unable to hold a job.

2. Operator evaluation of the proportion of residents with mental disabilities needing help with basic life functions.

3. Operator evaluation of proportion of those with a mental disability who had been in a state hospital for more than 5 years.

Income also depended on the amount charged per resident: The amount charged per resident reflects two things. As a continuous measure, it is first a measure of the amount of income the operator receives (we have no measures of actual cost per resident and thus of profit in 1973). However, it also shows the type of resident sought by the facility. Aid to the Totally Disabled (ATD to become SSI) in 1973 in California, paid residents $231 a month, at the high rate. Assuming that the resident was given 31 dollars spending money (as required by state law) and the operator the remainder, operators who charged $200 or more were targeting the ATD population and thus the more seriously disabled. Because of the multicolinearity of income as a continuous and collapsed measure and the absence of data on operator costs, we use it in the collapsed form.

Hypothesis 3: Age is measured by months the facility had been open before the first interview

(Since the literature makes competing predictions about the effects of age on survival, we considered several measures of the age variable to test this hypothesis). First we included the continuous measure of age in months to test the linear effect in all the analyses. Next, we explored the bivariate scatterplot of age and survival to define appropriate nonlinear measures (i.e., new and old). Examining the scatterplot of the association, we noted that the distribution was truncated among new facilities. We were, therefore, unable to include a dummy variable to measure the effects of newness. The scatterplot did show evidence of greater closure rates among the older facilities. To model this, we constructed a dummy variable that separated facilities that had been opened ten years or more from the remainder. Since, however, the correlation between survival and this dummy variable approached zero, we abandoned the nonlinear hypothesis.

Hypothesis 4: Legitimacy

We measured legitimacy by whether the facility had become licensed in the period since the initial interview. See Segal and Hwang (in press) for further details about the construction of this measure.

Ideally, we would be able to model the cumulative effects of neighborhood characteristics up to the time of closure. However, census data are only available for 1970 and 1980. Even splitting the sample and looking separately at those that closed closer in time to 1980 would yield too skewed a distribution to permit multivariate analysis, given the relatively small number of closures. We first consider 1970 census variables, applicable to all facilities. We then compute the difference in scores between 1970 and 1980, under the expectation that such represent a summary of a process that took place across the decade and are therefore relevant to all save, perhaps, the very earliest closures.

Hypothesis 5: Absence of organized constituency was assessed by two indicators

1. Poverty: measured by constructing factor scores derived from two principle components factor analyses, one for 1970 and one for 1980. To measure changes in poverty we took the difference between the two scores. The scores were each based on the percent below the poverty level and the percent of males and females in the labor force and unemployed.

2. Mixed land use: The census does not code non-residential land uses. Therefore, we use the operator’s description of land use in the neighborhood. This variable was collapsed into mixed use (commercial and residential or apartments and houses) and single use.

Hypothesis 6: Direct measures of opposition

Actual neighborhood negative reaction to facilities (only available for 1973): coded present if the operator reported any of the following as occurring: neighbor complaints about the facility, operator or family ever threatened or harassed because of the facility, and neighbors complained to local authorities. We should note that there is no way to standardize this measure for a constant time referent -- that is, some facilities had been open and non-owner operators employed for longer periods of time.

Hypothesis 7: Presence of gentrification and conservatism

1. Gentrification: We first modeled socio-economic status. Two principle components factor analyses were constructed: one for 1970 and one for 1980. Variables included in the separate factor analyses were: median income, education and percent of those employed who hold professional or managerial occupations. Following a modification of Covington & Taylor (1989), we rank ordered all census tracks on this factor for 1970 and then again for 1980. The one with the lowest SES had a rank order score of 1, the next lowest of 2 and so forth. We then took the difference between the two rank-order scores as our measure of gentrification.

2. Conservatism: measured by percent registered Republican in 1970 and change in percent Republican between 1970 and 1980.

Hypothesis 8. Inducement to sell

1. median house value in 1970 and change in median house value between 1970 and 1980 (both measured in 1965 dollars).

2. percent vacant units and change in percent vacant units between 1970 and 1980.

Results

The mean number of clients served in a facility was 28 (s.d. 35.78); seventy-one percent of the facilities were targeted entirely or predominately to clients with a history of having a mental disability. Sixty-two percent of the operators were also owners and 74 percent employed additional staff. The median amount charged per client was $200 a month (x=$206; s.d. =43.01). Finally, the average amount of time the facility had been open, serving as an out of home placement, was 92 months (s.d.=72,61 months) or 7.7 years.

The analysis will precede in several stages. Since events post 1973 are sometimes correlated with the 1973 variables, we first develop a model for the characteristics of the facility and its neighborhood in 1973. We then include the post 1973 variables.

Since we have a large number of possible explanations, we look at the bivariate correlations to test the plausibility of each of our possible types of explanations for closure and derive our final model from those which yield the most fruitful results. In this case, the absence of results is as interesting as the positive findings.

Table 2 shows the bivariate correlation coefficients for survival and each of the hypothesized independent variables (see vertical column). It further shows the correlations among the variables included in the final regression model. All facility variables were significant with the exception of the linear age variable, as well as the dummy variable which looked at facilities opened 10 years or more. (Note: we will keep age in the multivariate model because of its importance in the literature.)

Table 2.

Bivariate Correlations

Correlations all independent variables with Remain Open O Closed; 1 Open			Correlations Among Independent Variables in Multivariate Model
Hyp.		Open	Business Orientation	Disability	At/Above ATD	Age	Licensed	Poverty	Increase in Poverty	Mixed Use
1.	Business Orientation	.23***	1.0	−.07	.23*	−.04	.13*	−.01	.10	.31***
2.	Disability	.24***		1.0	.21**	.02	.20**	.02	.13	.08
	At/Above				1.0	.08	.27***	−.08	.04	.02
	ATD	.28***
3.	Age	.00				1.0	−.02	−.06	−.00	−.07
4.	Licensed	.48***					1.0	.05	.08	−.08
5.	Poverty ’70	.18**						1.0	.52***	.22***
	Change in Poverty	.18**							1.0	.14*
	Mixed Use	.25***								1.0
6.	Neighbor Opposition	.05
7.	Change in Class	.15*
	% Republ. ’70	−.12
	Change in % Republ.	.00
8.	House Value ’70	−.14*
	Change in House Value	−.02
	% Vacant 70	.10
	Change in % Vacant	−.09

^*p<.05

^**p<.01.

^***p<.001

We test four related sets of neighborhood factors (a) direct evidence of neighbor opposition, (b) housing demand, (c) absence of possible predictors of organized resistance (poverty and mixed use), and (d) presence of indicators of possible antagonism (conservatism and gentrification).

The results contradict much of what the literature would have us expect. The direct measure of neighborhood opposition in 1973 (hypothesis 6) and the measure of conservativism (hypothesis 7) were not significantly associated with closure. The housing and demand indicators likewise reveal no significant results (hypothesis 8). (Note all non-significant variables were added, one at a time, to the final multivariate model without changing either their importance or the interpretation of the model. Nonsignificant multivariate models available from the authors on request.) Only house value in 1970 (hypothesis 8) and gentrification (hypothesis 7) are weakly, albeit significantly, associated with closure. Appreciation in house value, percent vacant and change in percent vacant (hypothesis 8) do not predict closure. Even the correlation between 1970 house price, gentrification and closure is probably due to their joint association with poverty. When a partial correlation between house value and closure or gentrification and closure is computed, controlling for the poverty of the area, the significant association disappears (r=−.09/house value; r=.1/gentrification). Further none of the housing and demand items retain significance when included in the multivariate model.

The first column in Table 3 shows the logistic model predicting closure by the 1973 variables. Most facility variables maintained their significance in the multivariate model (hypotheses 1 and 2). Facilities were more likely to stay open when they possessed a steady income stream (as revealed by the disability and income measures) and when they were more likely to be run as a business. Age (hypothesis 3) as a continuous variable did not predict the probability of a facility remaining open in the multivariate model. Looking at the neighborhood variables in Table 3, closure is best predicted by poverty and mixed land use in 1970 (hypothesis 5).

Table 3.

Logistic Regressions Predicting Staying Open (O = closed; 1 = open) (N = 214)

	Facility and Neighborhood Characteristics up to ’73		Facility and Neighborhood Characteristics ’73–’83
	B	p	B	p
Intercept	.01	.966	2.94	.000
Business Orientation	.39	.05	.38	.076
At or Above ATD rate	1.11	.004	.83	.05
Disability Residents	.66	.002	.53	.027
Licensure			2.16	.000
Age Facility	.000	.927	−.000	.736
Poverty ’70	.44	.051	.46	.064
Increase in Poverty ’70–’80			.22	.450
Mixed Land Use	.98	.010	1.05	.042
p for model	.0000		.0000
% correctly classified	78.		80.
Somer Dyx	.61		.77

The SAS algorithm for computing logistic co-efficients predicts the probability of the non-event (here staying closed). To maintain consistency with the bivariate correlation, signs were switched to predict probability of predicting staying open.

The second model in Table 3 considers the post-1973 variables: change in poverty (hypothesis 5) and licensure (hypothesis 4). Change in poverty is not significant but licensure is strongly so -- sufficiently that it reduces the significance of the “business orientation” factor. This reflects the fact that the business oriented facilities were more likely to become licensed.

The all-over model is significant (p=.0000) and correctly classifies 80 percent of the cases (Sommer’s Dx = .77). (Sommer’s Dx is a PRE measure varying between 0 and 1 and showing the strength of the association. In this case, it shows how well the logistic regression model is able to correctly classify facilities as open or closed.)

Discussion

Our results both support and refute the existing literature. The hypothesized facility variables are important, with the exception of age. The neighborhood effects are not as anticipated. Rising house values, changes in population composition, price appreciation and decreasing vacancies do not predict facility closure in this period. Instead, our final models show that locating in any SES neighborhood, save for the very poorest, or in one with a single land use lead to a greater likelihood that the facility will not survive.

The effects of neighborhood opposition do not support a straightforward interpretation. Mixed use and poverty were included as indicators of opposition. Yet, the direct measure of neighborhood opposition is neither a significant predictor of closure nor are the remainder of our assumed indicators of neighborhood opposition. One possible explanation might be that opposition occurred during the decade of the 70s in all except the very poor and mixed use neighborhoods and that this opposition differed in character from that found in the earlier period. However, analysis looking at the relationship between poverty, mixed land use and neighborhood opposition in 1983 (for those facilities that remained open) found no association, leading us to abandon this hypothesis. Perhaps mixed use and poverty are important for other reasons. The strong bivariate correlation between business orientation, poverty and mixed use explain a part of this pattern. Facilities run as a business were more likely to locate in mixed use and poor areas, presumably because of the greater ease in securing available housing at lower costs and necessary permits. This explains part of the correlation between poverty, mixed use and closure. Yet each maintains significance in the final model, showing that something else is also at work. Perhaps we are measuring a pull rather than a push factor, (i.e., owners in single use and more affluent neighborhoods, could more easily and profitably convert their facilities to other uses). This hypothesis awaits additional data suited to testing it. Thus our analysis, at the neighborhood level, permits us to identify where facilities are most likely close but can not show why this is the case.

NIMBYism may then be an important factor in preventing facilities from opening (Dear & Wolsh 1987), or in affecting the social functioning of residents of existing facilities (Segal, Baumohl, & Moyles 1980); but it has negligible importance for the continuing existence of a facility. Further, our analysis shows importance of facility rather than neighborhood factors in keeping facilities open. This suggests that policy efforts should be directed towards more direct facility investments in order to maintain the existing sheltered care housing stock. Since facility survival depends on keeping beds full, underwriting of these beds would be a fruitful target of planning efforts. Direct subsidization of these facilities through tax advantages, supplemental payments beyond the resident’s own income and other strategies would not only help guarantee the future viability of these facilities, it would also provide a mechanism to encourage higher quality living environments.

This research shows that small non-business oriented and unlicensed facilities do not survive. Although these data permit us to say nothing about the characteristics of facilities that have opened since 1973, we do know that the cohort of operators from the closed state hospitals was a one-time event. It seems reasonable that the newly formed facilities are also more business oriented. Certainly, the demise of the small, family oriented facility means a truncation in the diversity of sheltered care placements available, which we hypothesize weakens the system (Segal, Silveiman, & Baumohl, 1989.)

Our conclusions that sheltered care facilities in the poorest and mixed use neighborhoods are the most likely to survive should not be taken as a proscription for locating new facilities. Simply because a neighborhood is good for continued existence of a facility does not mean that it is good for its residents. Poor and mixed use neighborhoods, for example, also have high crime rates -- an undesirable living situation for any person and for these sometimes very vulnerable people in particular. Instead, our findings should be taken as informative in that sheltered care facilities outside of poor and mixed use neighborhoods may be vulnerable to closure and needing of special assistance.