This presentation explored key recommendations in the Annie E. Casey Foundation's publication, "A Child Welfare Leader’s Desk Guide to Building a High-Performing Agency," including strategies for collecting and analyzing data about disparities.
2. Tracey Feild
Managing Director
Child Welfare
Strategy Group
Presenters
1
Robert
Matthews
Senior
Associate
From the Annie E. Casey Foundation
Morgan
Cole
Program
Associate
Paula
Gentry
Senior
Associate
Katrina
Brewsaugh
Senior
Associate
3. 2
• Technical difficulties? Visit
www.aecf.org/webex. Or use the chat or
Q&A window. Or contact WebEx
technical support at 1-866-229-3239
• No Q&A window? If the Q&A icon at the
top of that column is not blue (like image
at top left), click the icon, and the window
should appear
• Questions? Type questions for
presenters in Q&A window at any time
• We’re recording! The webinar is being
recorded and will be available after the
presentation
Questions?
4. Poll: Audience
What is your role in the child welfare field?
A. Administrator
B. Supervisor
C. Caseworker
D. Advocate
E. Foster parent
F. Service provider
G. Other
3
5. 4
Casey’s desk guide: A key tool for agency improvement
• Two-volume Child
Welfare Leader’s
Desk Guide describes
10 practices for
building a high-
performing agency
• Practice #4: Measure
and address racial
and other disparities
6. 5
This webinar is for administrators,
community members, data analysts and
others. We will discuss:
• How agencies can address unequal
outcomes for children of different
races and ethnicities in their care
• Scholarship on race, poverty and
maltreatment
• Disparity measures and formulas
Today’s discussion
7. 6
• Collect data on race and ethnicity from hotline to case closing
and for re-entry cases
• Share results of data collection in user-friendly reports and
dashboards
• Use data to describe the experiences of all kids served,
using the Disparity Index and Relative Rate Index (described
later)
• Translate lessons learned from data analysis into policy and
practice changes, regularly testing solutions, measuring
results, improving outcomes and addressing disparities
Casey recommends: An overview
8. Agreeing on common language
Disproportionality
exists when the
representation of one
group is larger or
smaller than the same
group’s representation
in the general
population
Disparity
is the difference in
outcomes that
children experience
based on race
7
9. Comparing two terms
Equality
The quality or state
of being equal.
Having the same rights,
social status, etc.
Equity
Fairness or justice
in the way people
are treated based on needs
8
10. Dimensions of racism
Dimensions of Racism* Definition
Internalized
Private, individual beliefs about race that are
subconscious and may result in bias, prejudice,
oppression and privilege
Interpersonal
Public expressions of racial prejudice, hate, bias
between individuals
Institutional
Institutions use discriminatory practices and adopt
policies and practices that result in inequitable
opportunities and outcomes
Structural
Racial bias across institutions and society that has
systemically privileged one group and disadvantaged
another group
9
11. 10
• Kids of color are disproportionately represented in many systems
and agencies must address unequal child outcomes. To do so:
− Use the most up-to-date quantitative methods
− Analyze data
− Develop local, specific solutions
with measureable results
− Use common language
− Focus on institutional levers
It is possible to measure and address
unequal child outcomes
12. 11
• Decision points
How did the initial purpose of the decision
process lead to the unintentional
consequence of disparity?
• Policies and practices
What practices or policies contribute to
this problem?
• Organizational structures
What organizational norms or myths justify
or maintain the disparity?
• Solution development
How can solutions address the disparity’s
cause and advance systemic change?
Analyzing contributing factors
13. 12
• Define terms such as equality, equity, disparity, disproportionality,
institutional racism
• Share data with clear explanations about what it does
and doesn’t mean
• Define the strategy to address disparity, including what
outcome you are aiming to improve
• Identify each staff member’s role in moving the agency’s
equity work forward
• Measure, then measure again
Moving from recommendations
to results
Sample tool: Casey’s Racial Equity Impact Analysis
http://www.aecf.org/resources/race-matters-racial-equity-impact-analysis/
14. 13
What is the area of greatest racial/ethnic disparity in your
jurisdiction’s child welfare system?
A. Removal from home
B. Group vs. family placement
C. Length of stay in care
D. Reunification
E. Reentry
F. Don’t know
Poll: Who do you serve?
15. Does poverty drive disproportionality and disparity?
• Correlation is not causation
• As the poverty rate goes up, so do victimization rates
— but race matters (Wulczyn, 2011)
14
16. Wulczyn’s two findings: One that was expected
— and one that seems counterintuitive
15
Overall, maltreatment rates increased with poverty rates
SOURCE: Wulczyn, 2011
17. For white kids, as poverty rates increase,
so do victimization rates
16
SOURCE: Wulczyn, 2011
18. For black kids, higher black child poverty rates were tied
— even if weakly — to lower maltreatment rates
17
SOURCE: Wulczyn, 2011
19. Key lesson: Keep asking, “What’s the story?”
• Whether you are an administrator,
a caseworker or a data specialist,
be open to factors other than
poverty
• Avoid overgeneralizing.
Don’t assume poverty is the
sole explanation for differences
• Race and specific location
matter. The effect of poverty and
race often varies considerably
within jurisdictions
18
20. 19
Critical considerations
• Adoption & Foster Care Analysis & Reporting System (AFCARS)
reporting rules
• Race is a social construct
• Ask the question: Disparate compared to whom?
Measuring race and ethnicity is complex
21. 20
• Be clear in how you define groups
• Look at all geographic levels
• Examine ethnic groups that make sense for your location
It is possible to manage this complexity
Numbers alone are not enough!
22. We will review four measures:
• Disproportionality
• Disproportionality Metric (DM)
• Disparity Index (DI)
• Relative Rate Index (RRI)
Choose your measure carefully
21
23. • Proportional means percentages on left and right would match
• Measure can be sensitive to population changes
Disproportionality answers the question: Are groups
represented at the same rate in both populations?
White
80% White
65%
Black
20% Black
35%
0
20
40
60
80
100
General Child Population Foster Care Population
NOTE: Fictional data. Illustrative only.
22
24. • Disproportionality compares the ratio of the foster care
population to that of the general population for the same racial
group — it is not a comparison of two racial groups
• Underrepresented < 1 < Overrepresented
Disproportionality Metric examines
just one racial group in comparison to itself
% White
% White% Black
% Black
Foster Care Population General Child Population
Ratio
Ratio
23
NOTE: Fictional data. Illustrative only.
26. # 𝑔𝑟𝑜𝑢𝑝 𝐼𝑛 𝐶𝑎𝑟𝑒 ÷# 𝑡𝑜𝑡𝑎𝑙 𝐼𝑛 𝐶𝑎𝑟𝑒
# 𝑔𝑟𝑜𝑢𝑝 𝐺𝑒𝑛 𝑃𝑜𝑝 ÷# 𝑡𝑜𝑡𝑎𝑙 𝐺𝑒𝑛 𝑃𝑜𝑝
=
% 𝑜𝑓 𝑔𝑟𝑜𝑢𝑝 𝐼𝑛 𝐶𝑎𝑟𝑒
% 𝑜𝑓 𝑔𝑟𝑜𝑢𝑝 𝐺𝑒𝑛 𝑃𝑜𝑝
𝟕𝟓÷𝟓𝟓𝟎
𝟐𝟓,𝟎𝟎𝟎 ÷𝟓𝟎𝟎,𝟎𝟎𝟎
=
𝟏𝟑.𝟔%
𝟓%
= 2.728
Calculating the Disproportionality Metric
In Care (%) Gen Pop (%)
Black 75 (13.6%) 25,000 (5%)
Non-black 475 (86.4%) 475,000 (95%)
Total 550 (100%) 500,000
(100%)
25
Fictional data from Shaw et al. (2008). Illustrative only.
27. # 𝑔𝑟𝑜𝑢𝑝 𝐼𝑛 𝐶𝑎𝑟𝑒 ÷# 𝑡𝑜𝑡𝑎𝑙 𝐼𝑛 𝐶𝑎𝑟𝑒
# 𝑔𝑟𝑜𝑢𝑝 𝐺𝑒𝑛 𝑃𝑜𝑝 ÷# 𝑡𝑜𝑡𝑎𝑙 𝐺𝑒𝑛 𝑃𝑜𝑝
=
% 𝑜𝑓 𝑔𝑟𝑜𝑢𝑝 𝐼𝑛 𝐶𝑎𝑟𝑒
% 𝑜𝑓 𝑔𝑟𝑜𝑢𝑝 𝐺𝑒𝑛 𝑃𝑜𝑝
𝟕𝟓÷𝟓𝟓𝟎
𝟐𝟓,𝟎𝟎𝟎 ÷𝟓𝟎𝟎,𝟎𝟎𝟎
=
𝟏𝟑.𝟔%
𝟓%
= 2.728
Finding: The proportion of black kids in care is 2.7 times greater
than their proportion in the general population
Calculating the Disproportionality Metric
In Care (%) Gen Pop (%)
Black 75 (13.6%) 25,000 (5%)
Non-black 475 (86.4%) 475,000 (95%)
Total 550 (100%) 500,000
(100%)
26
Fictional data from Shaw et al. (2008). Illustrative only.
28. Theoretical Ceiling Effect means the maximum DM result is
determined by the make up of the general population. Changes
in measure may reflect changes in general population, not practice.
This is a point-in-time measure.
Limitations of the Disproportionality Metric
In Care
(%)
Gen Pop
(%)
Rate/
1000
Disp. Metric
Blacks
County A
75
(100%)
25,000
(5%)
3.0 100%
5%
= 20.0
Blacks
County B
750
(100%)
250,000
(50%)
3.0 100%
50%
= 2.0
Theoretical
Ceiling Effect
27
Fictional data from Shaw et al. (2008). Illustrative only.
When one racial group makes up a large percentage of the general
population, the DM is biased toward being a smaller number
29. A real world example of the DM limitations
Native
American
Representation
In Care
(%)
Gen Pop
(%)
Disproportionality
Metric
Rate/1000
Washington
713
(7.5%)
23,828
(1.5%)
7.5%
1.5%
= 5.0 29.9
Alaska
934
(51%)
32,674
(17.3%)
51%
17.3%
= 2.9 28.6
SOURCE: AFCARS 2011
28
The DM cannot reliably be used to compare jurisdictions because of
its extreme sensitivity to population size
30. Disparity Index compares the likelihood that
two groups will experience the same event
• DI compares ratio of rate in one group to rate of a different
group — in this example, the likelihood of being in foster care
• The DI is not affected by population changes
29
% White
% White% Black
% Black
Foster Care Population General Child Population
Ratio
Less likely < 1 < More likely
32. 𝑅𝑎𝑡𝑒 𝐺𝑟𝑜𝑢𝑝 1
𝑅𝑎𝑡𝑒 𝐺𝑟𝑜𝑢𝑝 2
Entry rate per 1,000 in general population:
White 3.8 Black 10.6
𝟏𝟎.𝟔
𝟑.𝟖
= 2.79
Calculating the Disparity Index
31
33. 𝑅𝑎𝑡𝑒 𝐺𝑟𝑜𝑢𝑝 1
𝑅𝑎𝑡𝑒 𝐺𝑟𝑜𝑢𝑝 2
Entry rate per 1,000 in general population:
White 3.8 Black 10.6
𝟏𝟎.𝟔
𝟑.𝟖
= 2.79
Finding: Black children are 2.79 times more likely than
white children to enter foster care in this state
Calculating the Disparity Index
32
34. A limitation of the Disparity Index
Indian Black Asian Hispanic White
Referrals 2.92 1.89 0.48 1.34 1.00
Accepted Referrals 3.05 2.02 0.51 1.44 1.00
1.00Initial High Risk 3.31 2.17 0.50 1.41
Removed From Home 4.56 2.29 0.49 1.48 1.00
1.00Placements Over 60 Days 4.96 2.24 0.41 1.45
Placements Over Two Years 6.29 2.79 0.41 1.37 1.00
SOURCE: Marna Miller. (2008). Racial disproportionality in Washington State’s child welfare system.
Olympia: Washington State Institute for Public Policy, Document No. 08-06-3901.
A State’s 2004 Entry Cohort
33
The Disparity Index can obscure which decision points in a
system are contributing the most to disparities
35. • The RRI (also known as
decision point analysis) uses
the same formula as Disparity
Index but focuses on one
specific decision point,
indicating disparities only for
that one decision point
• Make sure you carefully
specify which sub-population
you are studying
The Relative Rate Index answers the question:
Where in the system does disparity occur?
General Child Population
Abuse Hotline Report
Screened-In for
Investigation
Substantiated
Enter
Foster
Care
60 Days
in Care
Exit to
Family
Enter
Group
Home
34
36. * Compares percent of black children at each point with percent of black children in general population.
Does not rate for black children to white children.
An advantage of the Relative Rate Index
General Child Population
Abuse Hotline
Report
Screened-In for
Investigation
Enter Foster
Care
60+
Days in
Care
35
RRI can help identify disparities at specific decision points
37. 36
• As you change practice and policy to address
unequal outcomes for children, measure as you go
• Learn from what does
and doesn’t work
Measure as you go
38. 37
What actions will you take first after today’s webinar?
A. Use the RRI to determine which decision points
in my system may contribute to disparities
B. Learn more about the racial/ethnic composition
of my area and clients
C. Bring together stakeholders to discuss how
to address disparities in my system
Poll: What action will you take?
39. 38
At the end of this presentation,
find links to:
• Casey’s desk guide,
• Measuring Disparity,
• Primer on Entry Cohort
Longitudinal Data, and
• Materials on addressing
racial inequities
Casey resources can help
40. 39
Endnotes and additional resources
• Resources on measuring disparity
– A Child Welfare Leader’s Desk Guide to Becoming a
High-Performing Agency (2015)
– Measuring disparity: The need to adjust for Relative Risk
(2015)
– Primer on Entry Cohort Longitudinal Data (2015)
– Research Is Action: Disparity, Poverty, and the Need for
New Knowledge, Wulczyn, F. (2011)
• Resources on addressing disparity
– Race Matters Toolkit, the Annie E. Casey Foundation
41. • For print copies of the desk guide, please email:
dortiz@aecf.org
• Next desk guide webinar:
June 23: Getting to Permanence
Next steps
40
Please share your ideas and promising practices
Casey will update the desk guide in 2017.
What should be included? Do you have a promising practice to share with the field?
Please email your feedback and ideas to Morgan Cole at mcole@aecf.org.
[insert statement about desk guide]
Our discussion today is on practice #4 in the Desk Guide: measure and address racial and other disparities. This is, of course a huge topic.
This webinar is for administrators, community members, data analysts and others tasked with assessing and addressing where children of different races or ethnicities fare better or worse than other children in their child welfare system, with a goal of learning how to improve outcomes for all children and families on our localities.
We will discuss the importance of agencies addressing unequal outcomes. We will briefly discuss scholarship on the relationship between race, poverty and maltreatment. We will then examine various methods of measuring disparity, including each measure’s strengths and weaknesses, and the need to measure results as you take steps to address unequal child and family outcomes in your agency.
The desk guide provides several recommendations on steps your agency can take to address unequal outcomes. First of all, it’s hard to define and measure something if you don’t collect data about it. We recommend collecting race and ethnicity data at multiple points throughout the life of a case. Most systems already collect this data at the points of investigation and entry into out of home care due to federal requirements. When possible, we recommend collecting this data from the very first point of contact – the hotline – and updating the data at multiple points during a case.
Data are only useful if they are understandable. Present results of analyses in user-friendly formats. There are many resources online that can provide inspiration in designing dashboards and visuals that increase comprehension of data.
Casey strongly advocates using entry cohort longitudinal data when assessing outcomes as it describes the experiences of all children served. We specifically recommend using the Disparity Index and Relative Rate Index for analyzing disparities, which we will discuss later today.
Finally, completing the quantitative analysis is only the first step in addressing disparity. The process is iterative, requiring continuous review and changes to policies and practices and listening to key stakeholders.
Let’s start by understanding what the terms disproportionality and disparity mean for child welfare systems. Our assessment of differential outcomes experienced by children of different races—and our ability to choose and evaluate strategies to address them—is dependent on our ability to properly measure at each step of the way.
Many people use the terms disproportionality and disparity interchangeably. But they actually refer to two distinct concepts. Disproportionality compares a group’s representation in the general population to their representation in child welfare or some other systems. By comparison, disparity looks at the differences in outcomes between different populations. In child welfare, we want to measure disparity. That is, we want to examine the likelihood that a child will experience different outcomes than their counterparts of a different race. What do we mean by outcomes? Outcomes are the measures that tell you how well the children are: are they achieving permanency and stability in a family; are they safe; are they happy. While process measures such as home visits and assessments are important, they aren’t outcomes because they don’t really tell you how well children are.
Disproportionality is a “snapshot.” Similar to a photograph, it tells you: this is what the children in our area and our system look like. In contrast, disparity tells you who is more or less likely to have a specific experience, such as entering care or living with kin.
As we begin to examine issues surrounding racial equity, a common understanding of frequently used concepts and basic terms is critical.
Racism can manifest in different ways.
Internalized – if a person dislikes his/her physical features because of what dominant culture deems attractive or acceptable.
Interpersonal – Hate crimes are an examples of this type.
Institutional – Such as zero tolerance policies or different entry/acceptance policies that create disparities.
Structural - Such as, education, juvenile justice, and child welfare, interacting with each other in ways that create disparate outcomes.
Just like racism can manifest in different ways, you must address disparities in multiple ways.
We have already talked about the importance on common language.
And while you may have strategies for various types of racism observed in the system, we believe that beginning with an institutional-level analysis will lead to the greatest outcomes. Remember institutional levers focus on one agencies practices as opposed to an individuals personal beliefs.
In a minute, Katrina will walk us through the best ways to measure quantitative data such as using a disparity index and relative rate index.
I am going to focus on how we can make the step from measuring quantitative data to understanding the story behind it and ways to improve outcomes.
READ HEADER – this essentially means, it is not enough to look at quantitative data alone. You need to understand which factors contributed to creating particular outcome. You need a qualitative analysis that helps unpack the story behind the numbers.
There are three main factors that impact child welfare outcomes.
The first is an agency’s decision points. While our decision processes were not created to support inequities, a lack of attention to them can create disparities. For example, if you look at your quantitative data and see a disparity in first placement type you should review the decision processes related. If you us a team meeting to make that decision, ask questions such as: Are meetings happening consistently for all families? Are families and young people included in all meetings? Etc…
The second review area is policy and practice. For example, if your data shows a high initial removal rate for children of one race/ethnicity, you may want to review your licensing and kin placement policies. Do your policies and practices take in to account cultural differences in living situations, such as whether or not sharing rooms is acceptable?
Organizational structure reviews include intra, or inter- agency dynamics. One inter-agency example is disparities that may occur due to school related to discipline. Perhaps school policies result in more young people in truancy court, followed by referrals to child welfare. The child welfare agency is not responsible for the school system. But your agency still may be able to have a conversation to limit those situations.
An intra-agency example can be seen when a whole county or unit of staff may use out-of-date practices based on long-standing office culture, not necessarily intentionally but this can lead to decisions based on subconscious bias.
SOLUTION DEVELOPMENT
As you assess each of these factors, you will be able to consider appropriate and specific solutions. In the same way that we encourage a thoughtful analysis of each factor, it is similarly important to analyze your solutions to make sure they will directly impact the outcome you are trying to change and won’t have unintended consequences.
The best analysis of the areas and solutions will include conversations with those most effected – families and young people.
We suggest using a race equity impact assessment tool. It helps focus the analysis and navigate what can feel like uncomfortable conversation.
Once you have a plan, there are 5 ways to support it’s implementation.
Agency leaders should clearly
communicate their commitment,
set the common language, and
identify each staff’s role in addressing disparities
Use and share data, at all levels of the agency.
It gives the conversation a focus, establishes common knowledge of the problem and begins conversations.
It is important to provide explanations of the data so you don’t enhance stereotypes or lead to further disparities.
Define and broadly communicate the processes that will be used to measure and analyze the system.
Such as - measurement strategies, the tools you will use and timelines for the work.
This supports staff in practicing being an equity-focused system and understanding their specific role.
Measure often. Set a regular schedule to measure and analyze the data and solutions you are implementing so you can assess and adjust your approach as necessary.
Now that you know that many agencies have disparities outcomes, and there are ways to address these challenges –
We are posting a poll for you to answer. If you haven’t pulled data based on race yet, feel free to make your best call.
While we are waiting for answers, and as we move in to the understanding of the measures let me just say, as someone who had an aversion to math, Katrina is masterful as demonstrating the nuances so that you will leave thinking you are a math and stats professor!
Before we proceed to discuss specific measures of disproportionality and disparity, however, it is important to address a topic that frequently raises its head when discussing race in child welfare. The question is, to what degree is poverty the driver of disproportionality and disparity?
Clearly, there is a correlation between race and economic status in this country as a result of the historical legacy of institutionalized racism. But correlation is not causation. And to design practice, policy or programmatic solutions to inequity, you want to get as close as possible to understanding causation.
As you surely know, many researchers have tried to untangle the race-poverty knot. One example can be found in research completed by Dr. Fred Wulczyn at the Chapin Hall Institute in 2011. He used data from 40 states to answer the question: Does the poverty/maltreatment relationship for children differ by race? In other words, if a state has a higher rate of child poverty, do they also have a higher rate of racial disparity in maltreatment?
Each dot on this chart represents a state. A simple look at child poverty rate and victimization rate shows that, unsurprisingly, maltreatment increases as child poverty increases. But, to determine the role of poverty vs. race, we need to next examine differences in this same relationship separately for white and black children.
This graph shows the same relationship but only for white children. It shows the same pattern as before: as the rate of white children’s poverty increases, the rate of white children’s maltreatment also increases. The slope of the line indicates that this relationship is strong:
This graph shows the same data just for black children. Now the conclusions are different. Rates of black children’s maltreatment are actually lower in states with higher black child poverty rates. Indeed, the line is almost straight, indicating a very weak relationship between poverty and maltreatment for black children. Also notice how there is a lot of variation among the states, with markedly different rates of victimization between states with the same black child poverty rates.
These data highlight that linking disparity to poverty is too simplistic. While there is a relationship, it’s not as simple as higher poverty equals higher racial disparity. The wide variation in states reinforces the need to examine disparity within the local context.
There are some basic rules for how race should be reported in AFCARS, set forth by the Administration for Children and Families.
Race is to be determined by the child (if old enough) or the child’s parents.
Clients are allowed to select multiple racial categories.
Each race category is reported separately and marked as either a yes or no for each client. The 7 race/ethnic categories in the AFCARS are: American Indian/Alaska Native, Asian, Black/African American, Hawaiian/Pacific Islander, White, Hispanic Origin, and unable to determine (used when neither the child nor their parents are able to provide this information)
Despite these rules, it is important to acknowledge that race is a social construct – there are no hard and fast scientific rules that define who is white, black, Hispanic, and so on. If you’ve ever looked at historical census records, the fluid nature of racial definitions is readily apparent. . Recent studies have found that race is not a stable construct, particularly for multi-racial individuals. The race that a person selects on a form may change over time based on life experiences. A 2015 study of a sample of teens in child welfare found that 1 in 5 changed their racial identification over a one year period. The study also found high rates of discordance between how the youths’ self-identified and how the child welfare system categorized them. The findings highlight the need to continually check-in with children regarding their racial and ethnic identification as opposed to asking only at entry (or only asking their parents).
White children are most frequently used as the reference group when analyzing disparity (white vs black, white vs Native American). However, some theorists argue that this assumes that the level of services offered to white children is the correct level of services for all children. They advocate that instead one group should be compared to the whole of all other groups, such as black vs. non-black. Others counter that doing so can reduce the appearance of disparities and ignores the historical impact of institutional systems that favor whites. Whichever direction you take, be sure it is clearly stated.
Because AFCARS allows for multiple racial groups to be indicated for an individual, analysts will often attempt to collapse these fields into a single race variable with each category being mutually exclusive. While this may make quantitative analysis of disparity more efficient, it is important that the rules and assumptions made are clearly spelled out. For example, the CWSG codes as Hispanic any child for whom the field Hispanic Origin is indicated yes, regardless of any other racial category that may be indicated. Other analysts may instead create 2 categories for white-Hispanic and non-white-Hispanic. Regardless, it is important that these decisions are clearly stated before doing an analysis.
Disparity varies greatly by geography. Measures may show little disparity at the state level but when taken to the county level or lower, large differences emerge. This is because racial groups are not evenly distributed by geography, leading to an aggregation bias that hides variability.
It’s also important to know what ethnic groups are relevant for your area. For example, the category of Asian is comprised of many different ethnicities and research has shown that not all Asian ethnicities have the same outcomes. A location with a high proportion of Asians in the general population should look at disaggregating this group by Chinese, Vietnamese, Cambodian, and so on as the child welfare system may have very different impacts on each group.
Finally, analyzing disparity requires quantitative and qualitative exploration. Quantitative analysis will only tell you what things look like: who’s more likely to experience an event. What it won’t tell you is why. For that, you need to listen to stakeholders and critically examine policies and practices.
As with most data analysis, there are several different ways to assess disproportionality and disparity. This is why multiple articles on the issue can each report different sets of numbers when describing the level of the problem. Today, I’m going to discuss 4 commonly used methods for measuring disproportionality & disparity, how they’re calculated, and what they tell us.
Disproportionality is a very commonly reported measure, and is easy to calculate. It simply answers the question “are groups equally represented in both populations?” Essentially, do the percents on the left match the percents on the right. It can be sensitive to increases or decreases in the total population. If you have 100 people, it takes 20 additional people to go from 20% to 40%. However, if you have only 5 people, a change from 20% to 40% means going from 1 person to 2 people. This means that if you live in an area with very few people of a certain background, it doesn’t take too many more/less of them to make a large shift in proportions.
The next method, the disproportionality metric, is one that is less common but has been used in a few national reports. Some people also use the term disproportionality index to describe this method. The disproportionality metric compares the representation of a group in the foster care population to that SAME group in the general population. It does not compare the difference between two races. The metric would not tell you if blacks are more likely to be in care than whites. Instead, it tells you the rate at which black children are represented in foster care based on their representation in the general population. If the result of the ratio is less than 1, than it means that group is under-represented. A result of 1 means they are equally represented. A result greater than 1 indicates over-representation.
So, how do you actually calculate the disproportionality metric? I’m so glad you asked!
There are 3 steps in calculating the disproportionality metric. 1) Take the number of the group in care and divide it by the total number of all children in care. This gives you the percent of the group in care. 2) Take the number of the group in the general population and divide it by the total number of all children in the general population. This gives you the percent of the group in the general population. 3) Last, you divide the % of the group in care by the % of the group in the general population.
Alright, let’s try it out with some fictional numbers. This table gives us all the data needed for the disproportionality metric. We have the number of black children both in care and in the general population. We also have the total number of all children in care and in the general population.
To determine the disproportionality metric for black children, let’s follow our 3 steps. 1) Take the number of the black children in care – 75 – and divide it by the total number of all children in care – 550. This gives us the percent of black children in care – 13.6%. 2) Take the number of black children in the general population – 25,000 – and divide it by the total number of all children in the general population – 500,000. This gives us the percent of black children in the general population – 5%. 3) Last, you divide the % of blacks in care – 13.6% - by the % of the group in the general population – 5%. This results in a ratio of 2.728.
This number tells us that black children are represented in care at a rate 2.7 times greater than in the general population. Thus, we can state that black children are over-represented in foster care based on their proportion of the general population.
This is a Theoretical Ceiling Effect. Because black children make up more of the general population in County B, the theoretical ceiling effect artificially results in a much lower DM. The DM makes it appear that County A is significantly worse than County B, when in reality black children enter care at the same rate in both counties.
The theoretical ceiling effect means that the DM is biased to being a smaller number when the racial group comprises a large percentage of the general population. This makes it difficult to make comparisons between locations or changes over time because the differences may reflect changes in the general population, not changes in foster care practice.
Since it’s important to connect theory to practice, let’s use a real-world example looking at the difference in Native American disproportionality between Washington and Alaska. When using a disproportionality metric, it appears that Washington has a much bigger problem with Native American disproportionality than Alaska. However, their rates/1000 are only 1.3 different (not very much). Because Native Americans comprise a much larger percentage of the population in Alaska, the DM has a lower maximum value possible.
The disparity index is very common. It provides a ratio of the rate in one group to the rate of a different group. Recall that the DM compared the rate of black to black, or white to white. The Disparity index compares black to white. The result tells you whether one group’s chances of experiencing an event are different from another group. It answers the question: are black children more or less likely to be in foster care than white children. A result less than one means less likely, 1 means equally likely, and greater than one means more likely. The calculation works in a way that cancels-out the population numbers, so the disparity index is not biased due to fluctuations in population sizes. It does use the general child population as the basis for all calculations. Why don’t we try it out?
To calculate the disparity index, you divide the rate in one group by the rate in a second group.
In this state, the foster care entry rate per 1,000 is 3.8 for white children and 10.6 for black children. So, we divide the black rate – 10.6 – by the white rate – 3.8 – to get a disparity index of 2.79.
This means that black children are 2.79 times more likely than white children to enter foster care in this state
The DI has the advantage of putting the rates on the same scale so it is not as significantly impacted by population differences, like we saw with the DM. This makes the measure more stable and thus you can compare different jurisdictions or look for change over time.
When you calculate the Disparity Index, the group used as the comparison group will always equal 1 (because a number divided by itself equals 1). In this example, each group was compared to White children.
In most cases, the general population is used as the denominator when calculating the rates at all decision points in a system. One impact of this is that the amount of disparity can “accumulate” as you look deeper into the system. You’ll notice that for Indian and Black children, the rates of disparity increase the further in the system. This is because the disparity present at the referral point is essentially carried forward to the next point due to the use of the general population throughout. This can make interpretation of a measure like disparity in exiting confusing. In the Disparity Appendix to the Desk Guide, a DI calculation for black children exiting care is 2.02, which would be interpreted as black children exiting 2 times faster than whites! Unfortunately, we know from practice that this is not often true.
The last method we’re going to cover is called the relative rate index, helps to solve this issue of accumulating disparity. It is also known as decision point analysis
The math is the same as for DI. However: Key difference is that the population used as the denominator changes in order to reflect the unique disparity at each stage.
RRI takes into account the reality that not ALL children are at-risk of experiencing each event. Not ALL children are at risk of entering foster care. Only those who have had a hotline call, that was investigated, that was then substantiated are at risk for foster care entry. This chart stops at 60 days in care, but could continue infinitely to 90 days, 1 year, 5 years, and so on. It can also include different outcomes such as exiting to permanency or re-entry to care.
RRI responds to one of the measurement issues inherent in the previous measures: the compounding disparity at each stage. When the general population is used as the denominator for all points in the system, the disparity at any point includes the disparity at the earliest point in the system. For instance, in some states there is a disparity in hotline calls, with black children more likely to be the subject of an abuse report than white children.
The flow of children is not always in a neat, linear fashion. The entry population can be the risk pool for any number of events such as exiting to family, length of stay in care, or placement in congregate care.
In the same fashion, one outcome can have multiple risk pools. Which one you use depends on the question you are trying to answer.
For children in care at least 60 days, what is the disparity in exit to family? Entry into CC?
For children who enter CC, what is the disparity in exit to family? That they will be in care for 60+ days?
The key with the RRI is being clear in which group of children is immediately at risk of the outcome under study.
This graphic shows some of the differences between the three measures. Again, the DI numbers continually increase as you go deeper into the system. However, the RRI treats like children alike. Using the RRI and looking only at children who entered care, black children are almost equally likely to be in care over 60 days as white children – which is good! What stands out is the disparity at entry, indicating this may be a point to target assessment and interventions.