Child Abuse & Neglect 41 (2015) 136–145
Contents lists available at ScienceDirect
Child Abuse & Neglect
Community characteristics, social service allocation, and child maltreatment reporting Asher Ben-Arieh The Hebrew University of Jerusalem, Jerusalem, 91905 Israel
a r t i c l e
i n f o
Article history: Received 14 May 2014 Received in revised form 21 July 2014 Accepted 26 August 2014 Available online 15 September 2014 Keywords: Neighborhood and child maltreatment Communities Context Reported cases of child maltreatment
a b s t r a c t This study expands research on the relationship between community (defined here as a locality) characteristics and child maltreatment. Research in this field is not new, but it is scarce. Our study is unique by examining changes between two periods rather than focusing on one point in time. Furthermore, our study examines structural conditions in small and medium size localities in Israel, a non-Western and non-Christian society. We compare our results with those from studies on inner-city and suburban neighborhoods in Western countries and earlier studies in Israel. We collected data on 169 Israeli localities, ranging from small ones (with as few as 1,500 residents) to medium size localities (i.e., towns) (with as many as 50,000 residents) in which approximately 34% of the Israeli child population resides. Our study tested four hypotheses: (1) Socioeconomic characteristics of the locality will be negatively correlated with the availability of social services; (2) Reported child maltreatment rates will be negatively correlated with the socioeconomic characteristics of the locality; (3) The availability of social services will be positively correlated with reported child maltreatment rates; and (4) Overall reported child maltreatment rates will be negatively correlated with the overall status of the localities. We have supported our second and third hypothesis in full, and partially supported our first and fourth hypothesis. In particular we have demonstrated that while demographics play a different role in Israel than in other countries in regard to child maltreatment, social, economic and cultural context are crucial for understating reported rates of child maltreatment. © 2014 Elsevier Ltd. All rights reserved.
Introduction More than 20 years ago, the American National Research Council (1993) proposed an ecological-developmental approach to understand child maltreatment using a framework in which risk and protective factors interact across ecological levels of the individual, family, community, and sociocultural context (Cicchetti & Lynch, 1993). Others have more recently expanded this framework to include the effects of relationships within social networks, community connectedness, neighborhood, and child maltreatment (Attar-Schwartz, Ben-Arieh, & Khoury-Kassabri, 2011; Coulton, Crampton, Irwin, Spilsbury, & Korbin, 2007). However, several issues complicate research on communities and/or the effects of neighborhoods on child maltreatment. First, defining neighborhoods or communities is complicated, and relying on census blocks (i.e., administrative demographic units) is problematic. Recent efforts tried to overcome the census block issue by examining small towns (Ben-Arieh, 2010a) or by defining communities more accurately with alternative methods or mapping than the administrative census blocks (Spilsbury, Korbin, & Coulton, 2012). Second, early research focused on communities in Western and Christian societies, and studies have only recently begun looking beyond those societies (Ben-Arieh, 2010c). Third, most studies have focused on a specific moment in time, neglecting possible changes in child maltreatment that may occur as time progresses. Our http://dx.doi.org/10.1016/j.chiabu.2014.08.018 0145-2134/© 2014 Elsevier Ltd. All rights reserved.
A. Ben-Arieh / Child Abuse & Neglect 41 (2015) 136–145
137
study attempts to bridge these shortcomings by focusing on the relation between community characteristics and child maltreatment at two points in time. Furthermore, we examine these changes in small to medium-sized localities in Israel, a non-Christian society outside traditional definitions of the West. Although contextual factors (i.e., community characteristics) and their relation to child maltreatment have been established in research, less extensive knowledge exists on the relationship among these contextual factors, and the availability of social services (Ben-Arieh, 2010c). In many countries, national or state governments determine social service expenditures and the number of social worker positions in a locality, regardless of the actual local characteristics or need, even though the local authorities are responsible for providing and administering the services. Thus, children and families receive services on the basis of where they live but not necessarily their needs. Unequal distribution of social services is a major issue (Saunders, 1994), yet our knowledge of these inequalities, particularly in different localities within a country, is lacking. In Israel, inequalities between localities are evident in varying levels of human and financial resources available to local social services and in the range, level, and quality of these services. Inequality is pronounced in Arab localities compared with Jewish localities, but inequality also exists within each sector (Kop, 2007). Furthermore, like studies examining child maltreatment rates, studies on unequal distribution of social services have not considered changes that occur between two periods. Finally, few studies adequately examine the relation between the availability of social services, particularly child protective services, and rates of reported child maltreatment (Ben-Arieh, 2010c). However, some research has suggested that the availability and quality of child protective services might be directly correlated with rates of child maltreatment reporting (Ben-Arieh, 2010b; Ben-Arieh & Haj-Yahia, 2006). Our study explores the relation between three variables during two periods. The independent variables include a 10-item set of locality characteristics and two measures of social service availability. The dependent variable is the rate of reported cases of child maltreatment. Using locality as the unit of analysis, we examine if and how socioeconomic characteristics of the locality relate to the availability of social services and child protective services, and how these two independent variables are related to a dependent measure of reported child maltreatment. The Theoretical Framework Community social organization theory and the ecological perspective support the notion that the characteristics of communities or localities are connected to rates of child maltreatment (Coulton et al., 2007; Swanson, 2001). Community social organization theory addresses patterns and functions of formal and informal networks, institutions, or organizations in a community (Kasarda & Janowitz, 1974). It includes the protection and provision of resources for families (Leventhal & Brooks-Gunn, 2000; Warren, 1971). The concept of social organization is useful for explaining the relation between child maltreatment and locality characteristics, such as poverty, economic decline, and residential mobility (Hirschi, 1969; Sampson, 1992; Sampson, Raudenbush, & Earls, 1997). In the ecological perspective, a growing number of studies have found child maltreatment rates in neighborhoods to be highly correlated with socioeconomic measures (Zuravin, 1989). In addition, researchers have found that lack of community organization and social coherence were most apparent in areas where child maltreatment was more prevalent (Freisthler, 2004; Garbarino & Crouter, 1978; Garbarino & Kostelny, 1992; Garbarino & Sherman, 1980) and, specifically, that low-income communities had fewer social services. Additional studies have embraced both theoretical frameworks and have shown that communities and neighborhoods are etiologically significant to child maltreatment (Coulton et al., 2007; Korbin & Coulton, 1997; Stanley, Tomison, & Pocock, 2003; Swanson, 2001; Tomison & Wise, 1999). Literature is consistent in suggesting that maltreatment is associated with economic characteristics of communities, including unemployment rates (Deccio, Horner, & Wilson, 1994; Freisthler, 2004; Gillham et al., 1998), poverty rates (Drake & Pandey, 1996; Freisthler, Midanik, & Gruenewald, 2004; Swanson, 2000), income levels (Garbarino & Crouter, 1978), median residential housing/property values (Drake & Pandey, 1996; Swanson, 2000), and low economic status (Zuravin & Taylor, 1987). Other community characteristics that were found to correlate with child maltreatment rates include child care burdens (Coulton, Korbin, & Su, 1999; Korbin, Coulton, Chard, Platt-Houston, & Su, 1998), residential instability (Deccio et al., 1994; Hyde, 1999; Swanson, 2001), vacant and abandoned housing (Deccio et al., 1994; Zuravin, 1989), lower participation in the labor force among females (Swanson, 2000), overcrowdedness (Garbarino & Kostelny, 1992), and per capita density of liquor stores (Freisthler, 2004; Freisthler et al., 2004). These studies and others raise a number of issues. First, when using both an ecological approach and a social organization approach, it is necessary to look beyond the socioeconomic structure of a community. Crucial elements of both theories are the availability, adequacy, and quality of the social services in a community (i.e., do deprived communities receive a larger allocation and, therefore, enjoy greater availability of social services?). Furthermore, as Coulton et al. (2007) argued, the availability of social workers and other human services professionals partly explains neighborhood variations in rates of reported child abuse and neglect. In addition to the availability of social services and child protection officers, the quality, training, and skills of social workers are critical. Scholars have argued that the availability and practices of social services, particularly child protective services, are major contributors to rates of reported child maltreatment (Melton, 2005). Second, the majority of studies were conducted in urban settings; only a few ventured to suburban neighborhoods (Swanson, 2001), and even fewer have looked at rural, small, or medium-sized localities (Ben-Arieh, 2010a). Defining urban and suburban communities is more difficult because neighborhoods are close to one another and many times borders are
138
A. Ben-Arieh / Child Abuse & Neglect 41 (2015) 136–145
ambiguous. Therefore, it is surprising that localities—which tend to be more geographically homogenous and easier to define as a community—have not been studied more. Third, most studies were conducted in the United States. Only a few were conducted in other Western countries, and few were conducted in other (less Western) and non-Christian communities (Ben-Arieh, Gal, Nepomnyaschy, & Garfinkel, 2007; Ben-Arieh & Haj-Yahia, 2006). Fourth, most studies examined social and economic factors while neglecting cultural or religious factors. Yet, communities are not defined by their social or economic status alone. Nationality (particularly among minorities) and other cultural characteristics are crucial to understanding and studying communities. Finally, most were cross-sectional studies, focusing on one point in time. To better understand the relation between locality characteristics and child maltreatment, a longitudinal perspective is needed. Methodology This study assesses the correlation between community characteristics, social service availability and rates of reported child maltreatment in Israel during two periods. The unit of analysis used to define a community was locality (i.e., the equivalent unit of a township or city or any other administrative geographical unit). Data were collected on 169 localities in Israel, ranging from small localities with as few as 1,500 residents to medium-sized localities with as many as 50,000 residents. Our study tested four hypotheses: (a) Socioeconomic characteristics of the locality will be negatively correlated with the availability of social services (i.e., less advantaged localities will have more social services available to the population); (b) Reported child maltreatment rates will be negatively correlated with the socioeconomic characteristics of the locality (i.e., more advantaged localities will have lower rates of reported child maltreatment); (c) The availability of social services will be positively correlated with reported child maltreatment rates (i.e., localities with more social workers will have higher rates of reported child maltreatment); and (d) Overall reported child maltreatment rates will be negatively correlated with the overall status of the localities (i.e., if the status of a locality worsens, the rates of reported child maltreatment will increase). Using locality as our unit of analysis enabled us to consider a range of variables and data in the social, economic, religious, and national realms. These localities were also relatively homogenous, unlike larger, diverse cities in Israel, and thus served as a better alternative definition for communities. Dependent Variable: Rate of Reported Child Maltreatment Our study used official investigation reports on the maltreatment of children ages 0–14. In Israel, all of these investigations are carried out by special child investigators (see Sternberg, Lamb, & Hershkowitz, 1996, for details). We used official data, published annually (as an official report) by the child investigation unit of the Ministry of Welfare on the number of investigations (either substantiated or not). Because annual outliers in the number of reports can occur, we calculated the average number of investigations per locality from 2004 to 2006 and from 2009 to 2011. In addition, we calculated child maltreatment rates based on the average child population size (ages 0–14) for the same periods. Official reports of child maltreatment are often considered an imperfect source of data because of problems in recordkeeping, variability in definitions, and misidentification of cases. In that regard, investigation reports are considered more accurate than reports to child protection officers. Moreover, child maltreatment investigation reports are indicators of the distribution of, recognition of, and response to child maltreatment and have been usefully applied in past research (Coulton, Korbin, Su, & Chow, 1995) including earlier studies in Israel (Ben-Arieh, 2010c; Ben-Arieh & Haj-Yahia, 2006). Independent Variable: Availability of Social Services Based on official administrative data from the Israeli Ministry of Social Affairs, we calculated the number of full-time posts of social workers and child protection officers per locality. In Israel, the ministry allocates social workers and child protection officers (see Kop, 2007). We calculated the rates of social workers and child protection officers as follows: (1) Rate of social workers per 1,000 residents: The average number of social workers in each locality was divided by the average number of residents in each locality (for 2004–2006 and 2009–2011). (2) Rate of child protection officers per 1,000 children: The average number of child protection officers in each locality was divided by the average number of children ages 0–17 in each locality (for 2004–2006 and 2009–2011). Independent Variable: Socioeconomic Characteristics of Localities Literature on the etiology of child maltreatment has suggested that economic status, residential mobility, family structure, race or ethnicity, age distribution, locality size, and nationality are related macro structural factors. In our study, as in earlier studies (Attar-Schwartz et al., 2011; Ben-Arieh, 2010c; Swanson, 2001), we chose a set of 10 indicators that we believed were consistent with the aforementioned concepts and most of them enabled a dynamic measure during the two periods we studied. For each measure, we calculated an average for 2004–2006 and 2009–2011.
A. Ben-Arieh / Child Abuse & Neglect 41 (2015) 136–145
139
Economic Status. (1) We considered the number of adults receiving unemployment benefits relative to the number of working-aged adults (ages 20–64). We divided the average number of people receiving unemployment benefits in each month in 2004–2006 and 2009–2011 by the average adult population in the same periods for the overall unemployment rate. Residential Mobility. (2) We considered population loss or gain by calculating the number of new residents in a locality minus the number of residents who left the locality in a given calendar year. (3) We considered the rate of new immigrants in each locality by calculating the rate of new immigrants who arrived to Israel after 1990. Family Structure. (4) We considered the rate of children living in single-parent households by dividing the number of children in single-parent households by the overall number of children in the locality. (5) We considered the rate of families with five or more children by dividing the number of families with five or more children by the number of families with children in the locality. Age Distribution. (6) We considered the percentage of children ages 0–17 among the population in the locality. (7) We considered the dependency ratio by dividing the number of children ages 0–19 and adults older than age 65 by the number of working-aged adults (20–64 years old). (8) We considered the percentage of adults aged 75 and older among the population in the locality. Locality Size. (9) We considered the locale’s population. Nationality. (10) We considered the nationality of each locality using the following categories: Arab, Jewish, or mixed Jewish and Arab nationalities.
Data Collection Data were collected from the Israel Central Bureau of Statistics and from the service of child investigators at the Ministry of Social Welfare and the annual State of the Child publication in Israel (Tzionit & Berman, 2013). All data were merged into one database, which we used for analysis.
Data Analysis We conducted our data analysis in three stages. First, we performed a frequency analysis to characterize the localities included in our study. Second, we looked at the availability of social services in our localities at two points. We did so by calculating correlations (using the Pearson correlation test for continuous factors and analysis of variance [ANOVA] for categorical factors) between the independent variables. In the third stage we used a two-points-of time fixed effects regression analysis to explore changes in reported child maltreatment rates and the various independent variables.
Findings Reported child maltreatment rates increased from an average of 0.57% (SD = 0.43) from 2004–2006 to 0.69% (SD = 0.95) from 2009–2011. These figures represent approximately 2,735 investigations from 2004 to 2006 and 2,782 investigations from 2009 to 2011.
Localities Table 1 provides a summary of the localities, including community characteristics during the two periods.
Table 1 Overview of localities. Characteristics
Rate of unemployed adults Population loss or gain Rate of new immigrants Rate of children in single-parent households Rate of families with 5 or more children Nationality of town % of children Dependency ratio % 75 or older Size (in thousands)
M (SD) 2004–2006
2009–2011
1.48 (0.57) 40.93 (282.16) 14.54 (12.15) 6.94 (5.07) 10.57 (10.44) 1.82 (0.72) 38.26 (9.14) 95.49 (27.70) 2.45 (2.02) 14.12 (11.52)
1.50 (0.51) 38.24 (254.66) 13.29 (10.70) 7.31 (5.22) 8.79 (9.67) 1.49 (0.53) 36.80 (8.74) 92.54 (27.46) 2.78 (2.16) 15.6 (12.62)
140
A. Ben-Arieh / Child Abuse & Neglect 41 (2015) 136–145
Availability of Social Services The average rate of social workers was stable between periods at 0.51 (SD = 0.27) from 2004 to 2006 and 0.52 (SD = 0.22) from 2009 to 2011. But the average rate of child protection officers per 1,000 children grew substantially from 0.39 (SD = 0.29) in 2004–2006 to 0.54 (SD = 0.44) in 2009–2011. Finally, the rate of social workers was significantly correlated with the rate of child protection officers (r = .147, p < .000 from 2004 to 2006; r = .237, p < .000 from 2009 to 2011). Bivariate Analysis Results of a comparison of the independent variables are presented in Table 2. Only 3 of the 10 independent variables were correlated with both social service variables, and only one was correlated in both periods. The rate of children in single-parent households was positively correlated with both social service variables during both periods. In other words, a higher rate of children in single-parent households was linked with more social workers and child protection officers, both in 2004–2006 and 2009–2011. The percentage of children ages 0 to 17 in 2004–2006 was negatively related to both social service variables, and in 2009–2011, it was negatively related only to the child protection officers. The nationality variable was related to the social workers only in 2004–2006, and a post hoc analysis showed that in these years, the towns with mixed nationality had more social workers than Jewish and Arab towns. The nationality variable was related to the child protection officers only in 2009–2011, and a post hoc analysis showed that in this period, the Jewish towns had more child protection officers than the Arab towns. Four independent variables were correlated only with the rate of social workers in both periods. The variables of unemployment, new immigrants, and rate of residents aged 75 and older were positively correlated with the rate of social workers, so increases in these variables were linked to more social workers in the locality. The population loss or gain variable was negatively correlated with social workers (i.e., more social workers was correlated with a decrease in the population size). Two independent variables were correlated only with the rate of child protection officers during both periods. The dependency ratio and size variables were both negatively correlated with child protection officers (i.e., smaller localities and those with smaller dependency ratios had more child protection officers). Examining Both Periods Table 3 presents the changes in the independent and dependent variables between periods. Table 3 clearly illustrates that the rate of reported cases of child maltreatment increased between the periods. Yet the direction of change in the independent variables was mixed. Three variables remained stable (rate of social workers, rate of unemployed adults, and dependency ratio). We witnessed a decrease in three variables (rate of large families, rate of children, and population gain or loss). Finally, four variables increased (rate of child protection officers, rate of new immigrants, rate of children in single-parents households, and rate of persons aged 75 and older). Analyzing Two Periods of Time with Fixed Effects Regressions. We conducted this analysis between the independent and dependent variables to analyze differences across time resulting in differences in reported cases of child maltreatment. The results are presented in Table 4. Changes in six of the independent variables resulted in changes in the reported rates of child maltreatment across the two points of time, while the remaining four independent variables were not related. The economic status variable—the rate of unemployment—was not correlated over time with rates of child maltreatment. This is interesting given that when using a Pearson correlation in each period independently, we found a significant correlation (r = .60, p < .001 in 2004–2006; r = .46, p < .001 in 2009–2011). Both variables of residential mobility—population loss or gain and rate of new immigrants—were related across time but in opposite directions. Although population loss across time correlated with a rise in child maltreatment rates, a rise in the rate of new immigrants also correlated with a rise in reported rates of child maltreatment. Our study used two variables on family structure. First, a rise in rates of children in single-parent families was strongly correlated with a rise in rates of child maltreatment (for each unit of change in child maltreatment rates, there was a 0.06 change in rates of children in single-parent families). The second variable, children in families with more than 5 children, was not correlated across time with child maltreatment rates, even though when using a Pearson correlation in each period independently, we found a significant correlation (r = −.40, p < .001 in 2004–2006; and r = −.27, p < .001 in 2009–2011). We used three variables to analyze the age distribution relation to child maltreatment. None of these variables correlated with child maltreatment across time. Yet when looking at each period independently, all three variables were related and in the same direction. The percentage of children in the population was negatively related to child maltreatment rates (r = −.57, p < .001 in 2004–2006; and r = −.58, p < .001 in 2009–2011). The same was true for the dependency ratio (r = −.43, p < .001 in 2004–2006; and r = −.38, p < .001 in 2009–2011). In the same vein, the rate of elders (75+) in the locality was positively related with child maltreatment rates (r = .54, p < .001 in 2004–2006; and r = .70, p < .001 in 2009–2011). One possible explanation may be the slow changes in demographic trends and the short period between the two points of time. Locality size was positively correlated with changes in child maltreatment rates. Localities that grew between the two points witnessed an increase in child maltreatment rate, if only a moderate one. Each unit increase in the dependent variable
Table 2 Correlations between independent variables. Test
Rate of social workers 2004–2006
Rate of child protection officers 2009–2011
2004–2006
2009–2011
Rate of unemployed adults
Pearson
r = .47** ; p < .001; n = 167
r = .36** ; p < .001; n = 168
r = .04; p = .66; n = 141
r = .06; p = .45; n = 169
Population loss or gain
Pearson
r = −.32** ; p < .001; n = 167
r = −.39** ; p < .001; n = 168
r = −.13; p = .43; n = 141
r = −.09; p = .23; n = 169
Rate of new immigrants
Pearson
r = .37 ; p < .001; n = 94
r = .51 ; p < .001; n = 94
r = −.13; p = .27; n = 79
r = .00; p = .97; n = 95
Rate of children in single-parent households
Pearson
r = .46** ; p < .001; n = 134
r = .36** ; p < .001; n = 139
r = .33** ; p < .001; n = 118
r = .33** ; p < .001; n = 139
Rate of families with 5 or more children
Pearson
r = .−07; p = .41; n = 164
r = .05; p = .51; n = 165
r = −.15; p = .07; n = 141
r = −.13; p = .09; n = 166
**
Nationality of town
ANOVA
f = 16.17 ; df = 2.00; p < .001; n = 167
f = 2.76; df = 2.00; p > .05; n = 168
f = .19; df = 2.00; p < .001; n = 141
f = 4.16* ; df = 2.00; p = .02; n = 169
% of children
Pearson
r = −.17* ; p = .03; n = 167
r = −.13; p = .09; n = 168
r = −.21* ; p = .01; n = 141
r = −.26** ; p = .001; n = 169
Pearson
r = −.13; p < .11; n = 167
r = −.09; p = .26; n = 168
Dependency ratio
**
**
*
*
r = .−21 ; p = .02; n = 141
r = −.25** ; p = .001; n = 169
*
% 75 or older
Pearson
r = .20 ; p = .01; n = 167
r = .16 ; p = .03; n = 168
r = .07; p = .40; n = 141
r = .08; p = .32; n = 169
Size (in thousands)
Pearson
r = .02; p = .81; n = 167
r = −.01; p = .92; n = 168
r = −.49** ; p < .001; n = 141
r = −.30** ; p < .001; n = 169
* **
A. Ben-Arieh / Child Abuse & Neglect 41 (2015) 136–145
Independent variable
Contents lists available at ScienceDirect
Child Abuse & Neglect
Community characteristics, social service allocation, and child maltreatment reporting Asher Ben-Arieh The Hebrew University of Jerusalem, Jerusalem, 91905 Israel
a r t i c l e
i n f o
Article history: Received 14 May 2014 Received in revised form 21 July 2014 Accepted 26 August 2014 Available online 15 September 2014 Keywords: Neighborhood and child maltreatment Communities Context Reported cases of child maltreatment
a b s t r a c t This study expands research on the relationship between community (defined here as a locality) characteristics and child maltreatment. Research in this field is not new, but it is scarce. Our study is unique by examining changes between two periods rather than focusing on one point in time. Furthermore, our study examines structural conditions in small and medium size localities in Israel, a non-Western and non-Christian society. We compare our results with those from studies on inner-city and suburban neighborhoods in Western countries and earlier studies in Israel. We collected data on 169 Israeli localities, ranging from small ones (with as few as 1,500 residents) to medium size localities (i.e., towns) (with as many as 50,000 residents) in which approximately 34% of the Israeli child population resides. Our study tested four hypotheses: (1) Socioeconomic characteristics of the locality will be negatively correlated with the availability of social services; (2) Reported child maltreatment rates will be negatively correlated with the socioeconomic characteristics of the locality; (3) The availability of social services will be positively correlated with reported child maltreatment rates; and (4) Overall reported child maltreatment rates will be negatively correlated with the overall status of the localities. We have supported our second and third hypothesis in full, and partially supported our first and fourth hypothesis. In particular we have demonstrated that while demographics play a different role in Israel than in other countries in regard to child maltreatment, social, economic and cultural context are crucial for understating reported rates of child maltreatment. © 2014 Elsevier Ltd. All rights reserved.
Introduction More than 20 years ago, the American National Research Council (1993) proposed an ecological-developmental approach to understand child maltreatment using a framework in which risk and protective factors interact across ecological levels of the individual, family, community, and sociocultural context (Cicchetti & Lynch, 1993). Others have more recently expanded this framework to include the effects of relationships within social networks, community connectedness, neighborhood, and child maltreatment (Attar-Schwartz, Ben-Arieh, & Khoury-Kassabri, 2011; Coulton, Crampton, Irwin, Spilsbury, & Korbin, 2007). However, several issues complicate research on communities and/or the effects of neighborhoods on child maltreatment. First, defining neighborhoods or communities is complicated, and relying on census blocks (i.e., administrative demographic units) is problematic. Recent efforts tried to overcome the census block issue by examining small towns (Ben-Arieh, 2010a) or by defining communities more accurately with alternative methods or mapping than the administrative census blocks (Spilsbury, Korbin, & Coulton, 2012). Second, early research focused on communities in Western and Christian societies, and studies have only recently begun looking beyond those societies (Ben-Arieh, 2010c). Third, most studies have focused on a specific moment in time, neglecting possible changes in child maltreatment that may occur as time progresses. Our http://dx.doi.org/10.1016/j.chiabu.2014.08.018 0145-2134/© 2014 Elsevier Ltd. All rights reserved.
A. Ben-Arieh / Child Abuse & Neglect 41 (2015) 136–145
137
study attempts to bridge these shortcomings by focusing on the relation between community characteristics and child maltreatment at two points in time. Furthermore, we examine these changes in small to medium-sized localities in Israel, a non-Christian society outside traditional definitions of the West. Although contextual factors (i.e., community characteristics) and their relation to child maltreatment have been established in research, less extensive knowledge exists on the relationship among these contextual factors, and the availability of social services (Ben-Arieh, 2010c). In many countries, national or state governments determine social service expenditures and the number of social worker positions in a locality, regardless of the actual local characteristics or need, even though the local authorities are responsible for providing and administering the services. Thus, children and families receive services on the basis of where they live but not necessarily their needs. Unequal distribution of social services is a major issue (Saunders, 1994), yet our knowledge of these inequalities, particularly in different localities within a country, is lacking. In Israel, inequalities between localities are evident in varying levels of human and financial resources available to local social services and in the range, level, and quality of these services. Inequality is pronounced in Arab localities compared with Jewish localities, but inequality also exists within each sector (Kop, 2007). Furthermore, like studies examining child maltreatment rates, studies on unequal distribution of social services have not considered changes that occur between two periods. Finally, few studies adequately examine the relation between the availability of social services, particularly child protective services, and rates of reported child maltreatment (Ben-Arieh, 2010c). However, some research has suggested that the availability and quality of child protective services might be directly correlated with rates of child maltreatment reporting (Ben-Arieh, 2010b; Ben-Arieh & Haj-Yahia, 2006). Our study explores the relation between three variables during two periods. The independent variables include a 10-item set of locality characteristics and two measures of social service availability. The dependent variable is the rate of reported cases of child maltreatment. Using locality as the unit of analysis, we examine if and how socioeconomic characteristics of the locality relate to the availability of social services and child protective services, and how these two independent variables are related to a dependent measure of reported child maltreatment. The Theoretical Framework Community social organization theory and the ecological perspective support the notion that the characteristics of communities or localities are connected to rates of child maltreatment (Coulton et al., 2007; Swanson, 2001). Community social organization theory addresses patterns and functions of formal and informal networks, institutions, or organizations in a community (Kasarda & Janowitz, 1974). It includes the protection and provision of resources for families (Leventhal & Brooks-Gunn, 2000; Warren, 1971). The concept of social organization is useful for explaining the relation between child maltreatment and locality characteristics, such as poverty, economic decline, and residential mobility (Hirschi, 1969; Sampson, 1992; Sampson, Raudenbush, & Earls, 1997). In the ecological perspective, a growing number of studies have found child maltreatment rates in neighborhoods to be highly correlated with socioeconomic measures (Zuravin, 1989). In addition, researchers have found that lack of community organization and social coherence were most apparent in areas where child maltreatment was more prevalent (Freisthler, 2004; Garbarino & Crouter, 1978; Garbarino & Kostelny, 1992; Garbarino & Sherman, 1980) and, specifically, that low-income communities had fewer social services. Additional studies have embraced both theoretical frameworks and have shown that communities and neighborhoods are etiologically significant to child maltreatment (Coulton et al., 2007; Korbin & Coulton, 1997; Stanley, Tomison, & Pocock, 2003; Swanson, 2001; Tomison & Wise, 1999). Literature is consistent in suggesting that maltreatment is associated with economic characteristics of communities, including unemployment rates (Deccio, Horner, & Wilson, 1994; Freisthler, 2004; Gillham et al., 1998), poverty rates (Drake & Pandey, 1996; Freisthler, Midanik, & Gruenewald, 2004; Swanson, 2000), income levels (Garbarino & Crouter, 1978), median residential housing/property values (Drake & Pandey, 1996; Swanson, 2000), and low economic status (Zuravin & Taylor, 1987). Other community characteristics that were found to correlate with child maltreatment rates include child care burdens (Coulton, Korbin, & Su, 1999; Korbin, Coulton, Chard, Platt-Houston, & Su, 1998), residential instability (Deccio et al., 1994; Hyde, 1999; Swanson, 2001), vacant and abandoned housing (Deccio et al., 1994; Zuravin, 1989), lower participation in the labor force among females (Swanson, 2000), overcrowdedness (Garbarino & Kostelny, 1992), and per capita density of liquor stores (Freisthler, 2004; Freisthler et al., 2004). These studies and others raise a number of issues. First, when using both an ecological approach and a social organization approach, it is necessary to look beyond the socioeconomic structure of a community. Crucial elements of both theories are the availability, adequacy, and quality of the social services in a community (i.e., do deprived communities receive a larger allocation and, therefore, enjoy greater availability of social services?). Furthermore, as Coulton et al. (2007) argued, the availability of social workers and other human services professionals partly explains neighborhood variations in rates of reported child abuse and neglect. In addition to the availability of social services and child protection officers, the quality, training, and skills of social workers are critical. Scholars have argued that the availability and practices of social services, particularly child protective services, are major contributors to rates of reported child maltreatment (Melton, 2005). Second, the majority of studies were conducted in urban settings; only a few ventured to suburban neighborhoods (Swanson, 2001), and even fewer have looked at rural, small, or medium-sized localities (Ben-Arieh, 2010a). Defining urban and suburban communities is more difficult because neighborhoods are close to one another and many times borders are
138
A. Ben-Arieh / Child Abuse & Neglect 41 (2015) 136–145
ambiguous. Therefore, it is surprising that localities—which tend to be more geographically homogenous and easier to define as a community—have not been studied more. Third, most studies were conducted in the United States. Only a few were conducted in other Western countries, and few were conducted in other (less Western) and non-Christian communities (Ben-Arieh, Gal, Nepomnyaschy, & Garfinkel, 2007; Ben-Arieh & Haj-Yahia, 2006). Fourth, most studies examined social and economic factors while neglecting cultural or religious factors. Yet, communities are not defined by their social or economic status alone. Nationality (particularly among minorities) and other cultural characteristics are crucial to understanding and studying communities. Finally, most were cross-sectional studies, focusing on one point in time. To better understand the relation between locality characteristics and child maltreatment, a longitudinal perspective is needed. Methodology This study assesses the correlation between community characteristics, social service availability and rates of reported child maltreatment in Israel during two periods. The unit of analysis used to define a community was locality (i.e., the equivalent unit of a township or city or any other administrative geographical unit). Data were collected on 169 localities in Israel, ranging from small localities with as few as 1,500 residents to medium-sized localities with as many as 50,000 residents. Our study tested four hypotheses: (a) Socioeconomic characteristics of the locality will be negatively correlated with the availability of social services (i.e., less advantaged localities will have more social services available to the population); (b) Reported child maltreatment rates will be negatively correlated with the socioeconomic characteristics of the locality (i.e., more advantaged localities will have lower rates of reported child maltreatment); (c) The availability of social services will be positively correlated with reported child maltreatment rates (i.e., localities with more social workers will have higher rates of reported child maltreatment); and (d) Overall reported child maltreatment rates will be negatively correlated with the overall status of the localities (i.e., if the status of a locality worsens, the rates of reported child maltreatment will increase). Using locality as our unit of analysis enabled us to consider a range of variables and data in the social, economic, religious, and national realms. These localities were also relatively homogenous, unlike larger, diverse cities in Israel, and thus served as a better alternative definition for communities. Dependent Variable: Rate of Reported Child Maltreatment Our study used official investigation reports on the maltreatment of children ages 0–14. In Israel, all of these investigations are carried out by special child investigators (see Sternberg, Lamb, & Hershkowitz, 1996, for details). We used official data, published annually (as an official report) by the child investigation unit of the Ministry of Welfare on the number of investigations (either substantiated or not). Because annual outliers in the number of reports can occur, we calculated the average number of investigations per locality from 2004 to 2006 and from 2009 to 2011. In addition, we calculated child maltreatment rates based on the average child population size (ages 0–14) for the same periods. Official reports of child maltreatment are often considered an imperfect source of data because of problems in recordkeeping, variability in definitions, and misidentification of cases. In that regard, investigation reports are considered more accurate than reports to child protection officers. Moreover, child maltreatment investigation reports are indicators of the distribution of, recognition of, and response to child maltreatment and have been usefully applied in past research (Coulton, Korbin, Su, & Chow, 1995) including earlier studies in Israel (Ben-Arieh, 2010c; Ben-Arieh & Haj-Yahia, 2006). Independent Variable: Availability of Social Services Based on official administrative data from the Israeli Ministry of Social Affairs, we calculated the number of full-time posts of social workers and child protection officers per locality. In Israel, the ministry allocates social workers and child protection officers (see Kop, 2007). We calculated the rates of social workers and child protection officers as follows: (1) Rate of social workers per 1,000 residents: The average number of social workers in each locality was divided by the average number of residents in each locality (for 2004–2006 and 2009–2011). (2) Rate of child protection officers per 1,000 children: The average number of child protection officers in each locality was divided by the average number of children ages 0–17 in each locality (for 2004–2006 and 2009–2011). Independent Variable: Socioeconomic Characteristics of Localities Literature on the etiology of child maltreatment has suggested that economic status, residential mobility, family structure, race or ethnicity, age distribution, locality size, and nationality are related macro structural factors. In our study, as in earlier studies (Attar-Schwartz et al., 2011; Ben-Arieh, 2010c; Swanson, 2001), we chose a set of 10 indicators that we believed were consistent with the aforementioned concepts and most of them enabled a dynamic measure during the two periods we studied. For each measure, we calculated an average for 2004–2006 and 2009–2011.
A. Ben-Arieh / Child Abuse & Neglect 41 (2015) 136–145
139
Economic Status. (1) We considered the number of adults receiving unemployment benefits relative to the number of working-aged adults (ages 20–64). We divided the average number of people receiving unemployment benefits in each month in 2004–2006 and 2009–2011 by the average adult population in the same periods for the overall unemployment rate. Residential Mobility. (2) We considered population loss or gain by calculating the number of new residents in a locality minus the number of residents who left the locality in a given calendar year. (3) We considered the rate of new immigrants in each locality by calculating the rate of new immigrants who arrived to Israel after 1990. Family Structure. (4) We considered the rate of children living in single-parent households by dividing the number of children in single-parent households by the overall number of children in the locality. (5) We considered the rate of families with five or more children by dividing the number of families with five or more children by the number of families with children in the locality. Age Distribution. (6) We considered the percentage of children ages 0–17 among the population in the locality. (7) We considered the dependency ratio by dividing the number of children ages 0–19 and adults older than age 65 by the number of working-aged adults (20–64 years old). (8) We considered the percentage of adults aged 75 and older among the population in the locality. Locality Size. (9) We considered the locale’s population. Nationality. (10) We considered the nationality of each locality using the following categories: Arab, Jewish, or mixed Jewish and Arab nationalities.
Data Collection Data were collected from the Israel Central Bureau of Statistics and from the service of child investigators at the Ministry of Social Welfare and the annual State of the Child publication in Israel (Tzionit & Berman, 2013). All data were merged into one database, which we used for analysis.
Data Analysis We conducted our data analysis in three stages. First, we performed a frequency analysis to characterize the localities included in our study. Second, we looked at the availability of social services in our localities at two points. We did so by calculating correlations (using the Pearson correlation test for continuous factors and analysis of variance [ANOVA] for categorical factors) between the independent variables. In the third stage we used a two-points-of time fixed effects regression analysis to explore changes in reported child maltreatment rates and the various independent variables.
Findings Reported child maltreatment rates increased from an average of 0.57% (SD = 0.43) from 2004–2006 to 0.69% (SD = 0.95) from 2009–2011. These figures represent approximately 2,735 investigations from 2004 to 2006 and 2,782 investigations from 2009 to 2011.
Localities Table 1 provides a summary of the localities, including community characteristics during the two periods.
Table 1 Overview of localities. Characteristics
Rate of unemployed adults Population loss or gain Rate of new immigrants Rate of children in single-parent households Rate of families with 5 or more children Nationality of town % of children Dependency ratio % 75 or older Size (in thousands)
M (SD) 2004–2006
2009–2011
1.48 (0.57) 40.93 (282.16) 14.54 (12.15) 6.94 (5.07) 10.57 (10.44) 1.82 (0.72) 38.26 (9.14) 95.49 (27.70) 2.45 (2.02) 14.12 (11.52)
1.50 (0.51) 38.24 (254.66) 13.29 (10.70) 7.31 (5.22) 8.79 (9.67) 1.49 (0.53) 36.80 (8.74) 92.54 (27.46) 2.78 (2.16) 15.6 (12.62)
140
A. Ben-Arieh / Child Abuse & Neglect 41 (2015) 136–145
Availability of Social Services The average rate of social workers was stable between periods at 0.51 (SD = 0.27) from 2004 to 2006 and 0.52 (SD = 0.22) from 2009 to 2011. But the average rate of child protection officers per 1,000 children grew substantially from 0.39 (SD = 0.29) in 2004–2006 to 0.54 (SD = 0.44) in 2009–2011. Finally, the rate of social workers was significantly correlated with the rate of child protection officers (r = .147, p < .000 from 2004 to 2006; r = .237, p < .000 from 2009 to 2011). Bivariate Analysis Results of a comparison of the independent variables are presented in Table 2. Only 3 of the 10 independent variables were correlated with both social service variables, and only one was correlated in both periods. The rate of children in single-parent households was positively correlated with both social service variables during both periods. In other words, a higher rate of children in single-parent households was linked with more social workers and child protection officers, both in 2004–2006 and 2009–2011. The percentage of children ages 0 to 17 in 2004–2006 was negatively related to both social service variables, and in 2009–2011, it was negatively related only to the child protection officers. The nationality variable was related to the social workers only in 2004–2006, and a post hoc analysis showed that in these years, the towns with mixed nationality had more social workers than Jewish and Arab towns. The nationality variable was related to the child protection officers only in 2009–2011, and a post hoc analysis showed that in this period, the Jewish towns had more child protection officers than the Arab towns. Four independent variables were correlated only with the rate of social workers in both periods. The variables of unemployment, new immigrants, and rate of residents aged 75 and older were positively correlated with the rate of social workers, so increases in these variables were linked to more social workers in the locality. The population loss or gain variable was negatively correlated with social workers (i.e., more social workers was correlated with a decrease in the population size). Two independent variables were correlated only with the rate of child protection officers during both periods. The dependency ratio and size variables were both negatively correlated with child protection officers (i.e., smaller localities and those with smaller dependency ratios had more child protection officers). Examining Both Periods Table 3 presents the changes in the independent and dependent variables between periods. Table 3 clearly illustrates that the rate of reported cases of child maltreatment increased between the periods. Yet the direction of change in the independent variables was mixed. Three variables remained stable (rate of social workers, rate of unemployed adults, and dependency ratio). We witnessed a decrease in three variables (rate of large families, rate of children, and population gain or loss). Finally, four variables increased (rate of child protection officers, rate of new immigrants, rate of children in single-parents households, and rate of persons aged 75 and older). Analyzing Two Periods of Time with Fixed Effects Regressions. We conducted this analysis between the independent and dependent variables to analyze differences across time resulting in differences in reported cases of child maltreatment. The results are presented in Table 4. Changes in six of the independent variables resulted in changes in the reported rates of child maltreatment across the two points of time, while the remaining four independent variables were not related. The economic status variable—the rate of unemployment—was not correlated over time with rates of child maltreatment. This is interesting given that when using a Pearson correlation in each period independently, we found a significant correlation (r = .60, p < .001 in 2004–2006; r = .46, p < .001 in 2009–2011). Both variables of residential mobility—population loss or gain and rate of new immigrants—were related across time but in opposite directions. Although population loss across time correlated with a rise in child maltreatment rates, a rise in the rate of new immigrants also correlated with a rise in reported rates of child maltreatment. Our study used two variables on family structure. First, a rise in rates of children in single-parent families was strongly correlated with a rise in rates of child maltreatment (for each unit of change in child maltreatment rates, there was a 0.06 change in rates of children in single-parent families). The second variable, children in families with more than 5 children, was not correlated across time with child maltreatment rates, even though when using a Pearson correlation in each period independently, we found a significant correlation (r = −.40, p < .001 in 2004–2006; and r = −.27, p < .001 in 2009–2011). We used three variables to analyze the age distribution relation to child maltreatment. None of these variables correlated with child maltreatment across time. Yet when looking at each period independently, all three variables were related and in the same direction. The percentage of children in the population was negatively related to child maltreatment rates (r = −.57, p < .001 in 2004–2006; and r = −.58, p < .001 in 2009–2011). The same was true for the dependency ratio (r = −.43, p < .001 in 2004–2006; and r = −.38, p < .001 in 2009–2011). In the same vein, the rate of elders (75+) in the locality was positively related with child maltreatment rates (r = .54, p < .001 in 2004–2006; and r = .70, p < .001 in 2009–2011). One possible explanation may be the slow changes in demographic trends and the short period between the two points of time. Locality size was positively correlated with changes in child maltreatment rates. Localities that grew between the two points witnessed an increase in child maltreatment rate, if only a moderate one. Each unit increase in the dependent variable
Table 2 Correlations between independent variables. Test
Rate of social workers 2004–2006
Rate of child protection officers 2009–2011
2004–2006
2009–2011
Rate of unemployed adults
Pearson
r = .47** ; p < .001; n = 167
r = .36** ; p < .001; n = 168
r = .04; p = .66; n = 141
r = .06; p = .45; n = 169
Population loss or gain
Pearson
r = −.32** ; p < .001; n = 167
r = −.39** ; p < .001; n = 168
r = −.13; p = .43; n = 141
r = −.09; p = .23; n = 169
Rate of new immigrants
Pearson
r = .37 ; p < .001; n = 94
r = .51 ; p < .001; n = 94
r = −.13; p = .27; n = 79
r = .00; p = .97; n = 95
Rate of children in single-parent households
Pearson
r = .46** ; p < .001; n = 134
r = .36** ; p < .001; n = 139
r = .33** ; p < .001; n = 118
r = .33** ; p < .001; n = 139
Rate of families with 5 or more children
Pearson
r = .−07; p = .41; n = 164
r = .05; p = .51; n = 165
r = −.15; p = .07; n = 141
r = −.13; p = .09; n = 166
**
Nationality of town
ANOVA
f = 16.17 ; df = 2.00; p < .001; n = 167
f = 2.76; df = 2.00; p > .05; n = 168
f = .19; df = 2.00; p < .001; n = 141
f = 4.16* ; df = 2.00; p = .02; n = 169
% of children
Pearson
r = −.17* ; p = .03; n = 167
r = −.13; p = .09; n = 168
r = −.21* ; p = .01; n = 141
r = −.26** ; p = .001; n = 169
Pearson
r = −.13; p < .11; n = 167
r = −.09; p = .26; n = 168
Dependency ratio
**
**
*
*
r = .−21 ; p = .02; n = 141
r = −.25** ; p = .001; n = 169
*
% 75 or older
Pearson
r = .20 ; p = .01; n = 167
r = .16 ; p = .03; n = 168
r = .07; p = .40; n = 141
r = .08; p = .32; n = 169
Size (in thousands)
Pearson
r = .02; p = .81; n = 167
r = −.01; p = .92; n = 168
r = −.49** ; p < .001; n = 141
r = −.30** ; p < .001; n = 169
* **
A. Ben-Arieh / Child Abuse & Neglect 41 (2015) 136–145
Independent variable
