Two pieces of news that I haven’t had time to post about.
First, the Bar determined that the request for an ethical review of the GAL Program’s use of attorneys needs to come from one of those attorneys, not an outside party. So, no investigation. That’s good. Those attorneys weren’t doing anything wrong. If a GAL/CBI attorney feels they’re being asked to do something ethically questionable, they can request a written opinion here. If a GAL/CBI attorney on specific case is doing something ethically questionable, you can file a Bar complaint against them here. If you just have problems with the way the Program is run, you can write long blog posts about it or something.
Second, DCF has settled a challenge to the new EFC rules brought by Florida’s Children First. The settlement is downloadable below. It states that DCF will offer kids in jail access to the Aftercare program in lieu of EFC and will work with the Department of Corrections on serving those youth. It also agrees to change rule language on how disabilities are documented and whether a youth who moves in with a TPR’d parent can be categorically denied EFC. All of that is good, too.
And unrelated: Today is the first day of school here at Miami Law. Our clinic is welcoming 23 law students who have signed up to represent foster kids and young adults for two whole semesters. Here’s to a great year!
The Guardian ad Litem Program started as a scrappy community of advocates and gadflies who sought to bring attention and change to the dependency system. It is now a state agency that’s been appropriated over $600 million in the last 15 years. There has never been a comprehensive study to determine whether the Program accomplishes its goal of improving the lives of foster children. I looked at the Program’s performance numbers, first out of irritation, then curiosity, and ultimately the realization that we need more fact-based information about the “Second DCF” to answer the core questions surrounding its continued funding and structure. The questions explored here are whether the GAL Program is ethical, effective, and good for children? The answer to all three questions, it turns out, is the same.
An email spurred me to start working on this post. I thought we were past the days when GAL Program leadership and their friends accused children’s attorneys of wanting kids to die. And yet here we are. I got this letter in my inbox a few weeks ago from GAL:
That email made me very sad. I worked at the Guardian ad Litem Program and I love the people there. I got started blogging in child welfare through defending the Program against unfair criticism and I still defend them when the criticism is unfair. I will do it again in this post, when appropriate. I have always answered GAL’s calls, accepted referrals for their kids, spoken at their events, and participated in their trainings. Yet, it’s getting harder to ignore the messaging coming out of its Tallahassee office. The accusation that representing kids means blindly throwing babies into unsafe situations is nothing new. I am still sad, though, because I know the people who wrote this letter know better.
I understand that people may not understand how direct representation of kids works in Florida, since it is the minority model here. What is described above is not it. I will briefly explain.
Client-direct child representation, in a nutshell
Let’s take the situation described in the letter where the parent’s home is unsafe and the child wants to go back. This describes almost every case where a child is removed — and the law says we’re supposed to try reunification first. In the best interest model, the GAL investigates the situation and comes up with a position. If the GAL decides that reunification is not their preferred path, then the GAL attorney makes no further efforts and waits for the parent to fail. Alternatively, a GAL could actively thwart the family’s efforts at reunification by filing motions to limit visitation or to pile on extra services.
In the direct representation model, on the other hand, we start with the position of the child. If the child wants to pursue reunification, then the attorney must take steps to try to make reunification possible. The law requires the home to be safe, so the attorney aims to make it so. That could include seeking services for the parents, holding the Department accountable if it doesn’t make reasonable efforts, filing injunctions to get abusive people out of the home, and a lot of other efforts to try to meet the legal standard. The client-directed attorney must also communicate with her child client and counsel her about the likelihood of success and other options. Maybe the child changes her mind about going home at some point; maybe the parent is never able to create a safe home. The client-directed attorney may never achieve the client’s goal and reunification may never happen.
The difference in the two models is that the directed attorney’s efforts are proportional to the child’s desire to go home, while the best interest advocate’s efforts are proportional to their own value system and beliefs (which none of the other parties or even the judge has any way of really exploring). The attorney providing direct representation to a child cannot lie to a judge, withhold information on known harms in a way that is tantamount to fraud on the court, or pick up children and throw them into burning houses. If a child’s attorney does any of that, please report them to the Bar.
So what happened that we’re even talking about this?
The email above mentions a request for an ethics opinion from a Florida Bar special committee. The committee was looking at legal representation in the child welfare system. I wasn’t on the committee, but I was part of a community of advocates who thought the committee was a good idea. There is a growing national movement around empowering parents’ and kids’ attorneys to strongly advocate to overcome the legal and social factors that drive up child maltreatment rates. To me, this committee was a step toward bringing that discussion to Florida. Knowing Florida, though, I doubt it even got close to that topic without devolving into exactly the nonsense that this whole post is about.
I’ve heard that the question of children’s representation came up frequently during those meetings. I’ve also been told that the GAL Program somehow wound up presenting on its model of representation and how it advocates for kids. I’ve seen the GAL powerpoint from that presentation and it seemed relatively straightforward. Apparently what was said during the presentation was not so clear.
I don’t know the intended goal of the GAL Program’s presentation to the committee, but multiple sources have told me it did more harm than good. People with no strong feelings about the GAL Program walked away from the presentation scratching their heads over what exactly is going on there. The email above from its leadership doesn’t say “sorry we botched it,” but that appears to be, in part, what happened. When the Florida Bar president is quoted as saying, “How could [an investigation] hurt?” that suggests the public education campaign didn’t go well.
The email above is meant to mobilize GAL volunteers to reach out to powerful people they are connected to. It lays out the talking points, and I suppose that’s good self-advocacy. The email includes a link to the proponent’s letter, but does not attempt to explain the situation or even defend the Program’s acts that the letter raises questions of. I’m going to try to explain and hope I get close enough.
Is the GAL Program Ethically Using Its Attorneys?
The special committee raised two main issues with the way the GAL Program uses its attorneys. Despite the tone of indignation in the GAL Program letter, the issues are what anyone employing in-house counsel should consider when setting up an office irrespective of whether that office is in a private corporation or a government agency. The questions require rational and objective application of the legal ethical rule that covers lawyers who represent organizations: Florida Rule of Professional Conduct 4-1.13. Also, despite the doomsday rhetoric, I don’t find the resolution of either question particularly fatal to the working of the GAL Program. In fact, I think the problem is the Program’s marketing, not its structure.
The first question revolves around who the attorney takes direction from in the Program. Organizations are required to designate a person or group of people to direct the attorney’s work. This person or group is called the “constituent.” The GAL Program, as stated in their powerpoint, has designated a three-person team: the staff or volunteer GAL, the Child Advocacy Manager, and the CBI Attorney. That last part is the potential problem — the attorney can’t also be the constituent. It’s an easy fix – just clarify that the attorney counsels and takes direction from the GAL and CAM. Then add that all attorneys directly report to a supervising attorney. It’s more or less what the powerpoint already says, so done and done.
Unfortunately, the Program doesn’t seem to want to do that because of the second issue. Apparently, the GAL folks were dead set on making the case that the GAL Program, and thereby its attorneys, actually represent thechild. That’s not clear from the powerpoint, but it apparently came up in the oral presentation a lot. It also came up later in the GAL Program’s written responses to questions from the committee. The question below is asking if the GAL Program has rules in place about how its attorneys communicate with the kids. Attorneys owe special duties to unrepresented parties, so that should be an easy answer — “we follow the rule.” Instead, they wrote this:
So, let’s see if I can flesh that out. In the GAL Program’s view, the child is represented by the GAL (who is not a lawyer) and is therefore not unrepresented, so the GAL Program attorney does not owe the kids the same duties that all attorneys owe to unrepresented parties, even though the attorney actually represents the GAL who represents the child, but does not represent the child who is already represented by their own client. This is the sort of ethical contortionism that got them recommended for an investigation. What could it hurt, indeed.
Here’s why that one sentence “The children are represented by the GAL” doesn’t make sense under the Rules of Professional Conduct.
One, the GALs (as in the volunteers or staff) are not lawyers and therefore cannot “represent” the kids in the legal ethical sense. I know the Florida statute that appoints them says “represent,” but they advocate for the kids. There’s nothing wrong with the word advocate. If they think the statute really means represent, then they have a much bigger problem on their hands because of my next point.
Two, an organization isn’t a lawyer and therefore, under the Rules of Professional Conduct, can’t “represent” someone in the legal ethical sense, either. Law firms and lawyers represent people. Programs do not. And the GAL Program is not a law firm. The GAL Program advocates for the best interests of children, but they do not, in the legal ethical sense, “represent” children with all the duties and obligations that come with that word. If the statute requires them to “represent” the kids, then it requires them to reorganize into a law firm. Many people would love to see that, including many people who work at the GAL Program.
And last, but probably most importantly, the child is actually a separate party in dependency cases. The GAL Program would be asking their attorney to somehow represent the GAL and the child — two parties at the same time — without both parties giving a full conflict waiver on the record. The ethics rules don’t allow that, either.
My eye-roll at this whole debate is that the problem is not real. The attorneys at the GAL Program do not take direction from the kids and they don’t actually go rogue on their teams and sell them out without getting fired, which GAL attorneys certainly have been over the years when they stood their ground against their bosses. The attorneys know good and well who their client is because Tallahassee calls them on the phone and reminds them when they get out of line on a case. Nobody in a local program office likes getting those calls.
What appears to be happening is that the GAL Program leadership desperately wants to say that kids have attorneys (specifically GAL Program attorneys) for some fundraising or political reason. They know it’s not true because when you look at the GAL Program’s pay plan, which is the official document that designates the role of each employee, there is no ambiguity: the attorneys represent the GAL Program and provide counsel to the Program. That means all the normal legal ethical rules apply, including Rule 4-1.13. End of story.
I really wish that were the end of this story
I was actually going to end this post with some recommendations about how I would love to see the GAL Program start to use its attorneys because there’s no law that says they have to be used any specific way. (I gave a presentation at the disability conference a few weeks ago suggesting that the GAL Program use some of its attorney resources to do antipoverty work and school advocacy, especially since many of the GALs are now educational surrogates.) I was going to acknowledge in this post that there are, in fact, people who say that the GAL Program is harming children and who would like to see it shut down or substantially reformed (that part of the letter is accurate). And I was going to list out my own thoughts on ways that the GAL Program’s statewide model was holding the local programs back from meeting the needs of their communities better.
But in the middle of drafting those ideas out, this happened. Another letter, another claim that attorneys for kids will stand by as children are murdered in the damn streets. Who is writing these things?
This second letter was in response to the First Star Institute report on child representation. The First Star report declares that Florida does not adequately provide representation to children in foster care. Since the whole point of First Star is to grade states on their provision of direct representation, the conclusion was more or less a given. We’re actually up to a grade of C from an F in the last report, so that’s progress, I suppose.
All the GAL Program had to say was “they’re grading us on something we don’t do” and that would have been enough. Instead, they went further. This is the paragraph that got me.
Kids don’t decide to go home — judges have to order it. And, despite what is written above, the legislature specially appoints attorneys to the disabled kids to protect their legal interests at a level of vigor that GALs cannot generally do. I’m not making that assessment up: One, it’s codified in section 39.01305; and two, the GAL Program has acknowledged it publicly. In its analysis of the Foster Kid Bill of Rights last session, the GAL wrote that if foster kids have clearly defined rights, then the Program will need a lot more attorneys. See below.
The problem is that the Bill of Rights didn’t add any new rights — it was just a collection of pre-existing laws and regulations. So, apparently, nobody is advocating for them “expansively” now. That’s a problem.
The email above was different than most of the others that come out of the Program, though. I usually ignore ad hominem fundraising attacks on me and everyone who does the work I do. I’ve sat out of this debate for years because I believe there is room for GALs and AALs in the world and some kids need one or the other or both.
But I can’t sit this one out, because this particular email went and dragged math into it. I quote:
The email suggests that, based on the outcomes, the GAL Program is provably, mathematically superior to kids having attorneys represent them directly. That felt like a challenge. If it’s provably true, then let’s try to prove it.
Is the GAL Program Effective?
In our modern child welfare system, the fundamental question for any intervention is whether it works. This is the legacy of the early child welfare reformers who sought to create a science of caring. We have become an evidence-based system, full of actuarial instruments and six sigma streamlining.
The tragic irony is that the child welfare system is itself, at best, a mixed bag of results. By forcibly removing kids from poverty-stricken homes, we often exchange one set of harms for another. Any potential long-term benefits of intervention get drowned out by the much stronger pulls of community neglect, trauma, and social isolation.
GAL Program leadership clearly knows that it has to maintain a sense of evidence-based legitimacy. That’s why the First Star report referenced in the letter is so threatening. The casual reader may mistake the report for a study on outcomes. It’s not. It’s a measure of how each state is living up to the particular values of the organization that put out the report. They’ve been publishing that report for years, and the GAL Program responds almost every time. Nobody likes to get a “C” and the Program wants to challenge its grade.
This year the response was different, though. The letter does not just espouse the GAL Program’s values. It suggests that GAL Program model is provably effective. As evidence, it identifies some high-graded states that use an attorney model and points out that they have lower rankings than Florida in many federal outcome measures. That isn’t how math or logic works.
What does the research on best interests advocacy show?
There are no studies that I know of directly comparing best interest advocacy versus any other kind of theoretical model. There are studies, however, on GAL, CASA, and other external citizen review models. These models often employ best interests advocacy and the studies on them show mixed, but usually neutral or positive, results.
The studies are severely limited, though. First, it should be noted that many of the most positive studies come from the 1980s and 1990s during a time when child welfare looked very different from today. Back then, even some state agencies did not have attorneys in the courtroom, so the addition of a dedicated advocate for the child would have been revolutionary no matter what model they used. As an example, old studies often include a measure for the percentage of kids that had a performance plan — federal law requires that number to be 100% today and it generally is. Old studies are of limited use to determine current efficacy.
Most of the studies of CASA/GALs have had serious limitations. Lawson and Berrick found that the rate of selection bias and poor research methods in the CASA research field was so prevalent that existing studies could not be relied on for or against the model. There has only been one true experimental study in this area, largely because research ethics and court rules wouldn’t allow it easily, and that study had other serious design flaws. Lawson and Berrick concluded that CASA/GAL representation is not an evidence-based practice, but may have other positive values.
The only studies I could find directly on the Florida GAL Program were qualitative and involved volunteer retention. A 2015 study looked at the reasons GAL volunteers stop working with the Program. The volunteer turnover rate in the circuit studied was approximately 30%. The short answer: 44% of the ex-volunteer group reviewed did not meet the work requirements (i.e., got let go or was fired), 24% relocated, 14% had life events interfere, and 10% just didn’t want to continue. A 2019 study on GAL Program recruitment showed that most GALs signed up after hearing about the Program through word of mouth and many were motivated to do so because they were going through their own life transition (e.g., retiring).
Studies on the impact of the GAL Program are possible, either at the child or system level. Kimberly Huggins-Hoyt out of Georgia State University has done interesting work on the statewide effects of privatization in child welfare. Someone could do similar research on the effects of representation models. Or someone could do an actual experiment and see what happens. There are statistical tools for measuring what we’re talking about here. Cherry picking some states that suck worse than Florida and happen to be in a report you don’t like is not a validated statistical method.
It was shortly after reading the second letter above that I remembered that the GAL Program actually does post stats on its performance. I found it odd that it didn’t cite them in their First Star reponse letter. Do those stats show any effect at all? Let’s find out.
The structure of the Statewide GAL Program
The Statewide GAL Program, as you probably know if you’ve read this far into the post, is a Florida state agency that provides support and oversight to the local guardian ad litem program offices. Its director is appointed by the governor and its staff are mostly state employees. I joined the Program in 2007. My paycheck came from one of the Program’s non-profit support organizations that raise money for the Program, and I moved over to the state payroll a year or two later. At the time, the legislature was not particularly interested in funding more attorneys. They currently employee around 170 attorneys, handling around 120 cases (not kids) each. Each case probably has 1 to 3 kids. The attorneys make around $45,000 to $50,000 per year.
The local programs had been around for decades, housed inside the court system. The independent Statewide GAL Program was born in 2003-2004 through some impressive advocacy by champions of the cause and some ethical questions around whether the GAL could be a party and a part of the court at the same time. It was, in a sense, born of an ethics complaint. Its volunteer-centric model was heralded as both cheap and effective, with an initial budget of just over $10 million.
The legislature set a goal for every child to have a guardian ad litem by no particular date, and tasked the Statewide Program with making that goal a reality (thus the studies on volunteer recruitment above). Over the last 15 to 16 years, its budget has grown by 500% and its representation numbers have gone up and down, but are currently down somewhere between 3% and 15% from 2007 (depending on which dataset you use to count kids). Since 2003, over $600 million has been appropriated for the GAL Program (this does not include funds raised by its direct support organizations). It has never, however, hit the full statewide representation rate of 100%.
The GAL Program is primarily a legal advocacy organization, which sets it apart from most child philanthropy endeavors. Volunteers do the field work, but the administrative work (and much of the report writing) is largely done by staff. The courtroom work has always been the hardest part for most volunteers. Early advocacy from the GAL Program involved seeking a statutory hearsay exception for their reports so that the GALs did not have to testify directly. It was not successful.
The use of staff attorneys was expanded to keep the Program on equal footing with DCF and parents, both of whom had attorneys. It also provided protection and support for GALs in court. The GAL program attorney position is not required by statute — it’s a policy and spending decision the Program’s leadership make to enhance its effectiveness. It has recently expanded its appellate program through recruitment of pro bono attorneys to handle appeals.
The GAL Program is therefore unusual and possibly unique. The Program has full party status under Florida law; it has trial attorneys in the trenches and an appellate team on standby at all times; and it has a full statewide reporting structure that goes all the way to Tallahassee, where it pushes for funding and legislative changes each year. It is probably the most deeply embedded, well-funded, and legally authoritative expression of best interests advocacy in the country. If best interests advocacy can change a child welfare system for the better, then the GAL Program should be doing it.
The Data and Design
Consistent with its courtroom focus, the GAL Program publishes its monthly performance data by court circuit. These Performance Advocacy Snapshot (PASS) reports track the percentage of kids appointed to the GAL Program, the percentage of active volunteers, the child to volunteer ratio, the volunteer retention rate, and GALs that are qualified to provide transportation and education advocacy. The Program also publishes monthly representation reports that break down the county and circuit numbers in more detail, and a few other reports with charts and graphs.
The PASS Reports lists outcome measures as “GAL Influence on Child Welfare Outcomes.” Those measures are:
Percent of Kids in Care 8 Days to 12 Months with No More than 2 Placements,
Children Achieving Permanency in 12 Months of Entering Care,
Children Not Re-Entering Out-of-Home Care within 12 Months of Achieving Permanency,
Foster Foster Youth Ages 19-22 with a Diploma or GED, and
Percent of Adoptions Within 24 Months.
This is the data we’ll use to test for an effect. Despite some obvious problems with experiment design here, this is a legitimate endeavor because the GAL Program holds these numbers out as evidence of its impact. As the main test variable, I’ll use the percentage of kids in each circuit who are appointed a GAL, because that is the measure that the legislature has set as the goal and the main measure the GAL Program uses in its reports.
The plan is to look for significant correlations between GAL appointment rates and the “influence” measures listed above. This would normally be the time to say that correlation does not prove causation, and in most circumstances we would need to be very careful with that. But here, the causation has already been declared and we’re working backwards to test whether it is justified given the existing data. The main questions is: Does the public data support a claim of causation, correlation, or “influence”?
To make the calculations more manageable, I had to make the following adjustments:
I’m using the combined Circuit 18 numbers, instead of separating them out by CBC, because the focus here is on the GAL model and not the CBC model.
I am separating out Orange and Osceola counties because Orange county’s model is significantly different in that it uses attorneys instead of lay GALs. That gives us 21 circuits.
I inverted the re-entry measure in the PASS Reports to the percentage of kids who did re-enter care in 12 months, because the negative in the original was confusing.
I’m using the simple average of the last six months of 2018 as the time frame so that we’re not victim to an abnormal month. I realize there was a hurricane during that time, but those are real-world conditions in Florida and we’ll check for outliers as we go. The GAL Program also doesn’t seem to have PASS reports on their website for the first half of 2018.
Some GAL and DCF reports are in months and others are in quarters, so the measure will be the average of six months or two quarters.
I did not run every standard test on every pair of variables. They appear to be mostly normal and I’ll do outlier tests on a case by case basis. I’m also not labeling partial correlations using the rpartial notation. It’ll be clear from context when it’s a partial.
Finally, we are looking at the effect of GAL representation rates on a circuit’s outcomes. We cannot say anything about individual kids from this data.
Here is a spreadsheet with all of the raw numbers for the calculations. I hope someone will check the math and go even further with it. I’ve tried to be as careful as possible, but this is a blog post and not a peer reviewed article. If something is wrong, I’m sure someone will tell me and we will celebrate that we had a fact-based conversation about the benefits of the various representation models. Maybe one day this post will turn up on Lawson and Berrick’s list of bad research. I would be honored if it resulted in anyone doing good research to rebut it.
Here we go.
GAL Appointment Rates Aren’t Uniform Statewide
Despite the statewide goal of 100% representation, the circuits showed a large variation in the percentage of kids appointed a GAL. Over the last six months of 2018, the circuits had an average GAL representation rate of 79.7% with a standard deviation of 12.4%. The high was 96.68% in Circuit 14 and the low was 54.75% in Circuit 20. The distribution is normal. This gives us a way to test for an effect using a natural experiment. We’ll try to control for the factors that could be affecting GAL appointment rates.
A visual check of the scatterplot below shows that the relationship between the GAL representation rate and circuit size (as measured by its out-of-home care population) is linear. There is a significant, strong negative relationship between the number of kids in a circuit and the percent of them that have a GAL, r(19) = -.659, p = .001. (That’s even true with Circuit 20 in the mix, which is clearly an outlier in terms of GAL representation.) This means that 81% of the variability in GAL representation rates can be explained by circuit size. The bigger the circuit, the lower the percentage of kids who are assigned to the GAL Program. That makes sense because, as the studies discussed above suggest, volunteer recruitment and maintenance is hard.
What about that other 19% of variability? I suspected two possible sources: demographics and economics.
As the simplest demographic explanation, I suspected that circuits with bigger populations would have higher representation rates because they had more people to pull volunteers from. However, there was no significant correlation between the percent of kids with a GAL and the circuit’s total population, r(19) = -.271, p = .235.
I suspected that circuits with fewer kids in out-of-home care per capita would have higher representation rates, on the theory that there is some volunteer penetration rate that the OHC per capita rate would play against. But again, there was no significant correlation, r(19) = -.219, p = .341.
Reaching to the studies that show that GALs tend to be older and female, I used the American Community Survey from 2017 (the latest year I could find) and calculated the weighted percentage of the population of each circuit who was female and age 50+. There was again no correlation to the percentage of kids with a GAL and the percentage of older women in the circuit, r(19) = -.247, p = .280.
There are other studies that show that GALs tend to be white(r than the kids they represent, at least). There was no correlation between GAL rates and the whiteness of the circuit, r(19) = -.281, p = .217.
I gave up on demographics. Size matters.
As for economics, I started with the theory that circuit income would correlate with volunteer rates. I created income indices based on the weighted average of each circuit’s county income per capita and the out of home care population. No correlation was found with GAL representation, r(19) = -.073, p = .753.
I next tested whether circuits with higher philanthropic giving rates would have higher GAL representation rates. I created a proportional Giving Index based on the size of the counties and the counties’ Giving Ratio in the How America Gives study. There was no significant correlation, r(19) = .361, p = .108, but it was suggestive (which I’m defining as p < .200). A look at the scatterplot (not shown here) shows that two circuits may be outliers, but it’s hard to say whether they can be validly removed.
Looking at Income Inequality, I calculated a weighted Income Inequality Ratio (IIR) for each circuit based on its county IIR and OHC population. The county IIR is defined at the 80th percentile income divided by the 20th percentile income, giving the ratio of the highest and lowest earners in the county. The result was just shy of significant and positive, r(19) = .431, p = .051. Controlling for OHC population size, the IIR is strongly correlated with GAL rates, r(18) = .535, p = .015. It explains 75% of the GAL rate variability after controlling for size.
Along the same lines as Income Inequality, when controlling for OHC size, a circuit’s OHC Racial Disparity, r(18) = .468, p = .037, and Violent Crime Rate, r(18) =.518, p = .019, are both positively correlated with GAL representation rates. That’s the holy trinity of injustice and overrepresentation in punitive systems.
So, for the rest of this post we’ll occasionally control for OHC population (size), Income Inequality (socioeconomics), OHC Racial Disparity (race), and Violent Crime Rate (community safety) to see if GAL representation has any correlation beyond those. Because we’re assuming strong causation, the GAL representation rate effect should survive controlling for correlated variables. Not all of these controls make sense for every effect. There will need to be some reasonable theoretical connection.
Oh, and since we’re assuming causation here, I do want to state that I do not think that GALs are pushing up community variables like Income Inequality or Violent Crime. The OHC Racial Disparity correlation, however, should give us pause. It’s defined as the ratio of over-representation of non-white kids to the under-representation of white kids in the foster care system — it’s not a community variable, and bigger numbers mean more racial disparity. Racial disparity is the sort of thing GALs have been accused of having very unfortunate effects on, and a fact that National CASA takes very seriously. Seeing a correlation here is not good.
Does the variation in GAL representation correspond to variations in circuit outcomes?
So the theory is that higher GAL representation rates should correlate with better circuit outcomes. The difference between 50% and 95% of kids having a GAL should result in some kind of measurable effect. To test that, I ran a Pearson product-moment correlation using SPSS on the GAL’s representation rate during the time period and the “influence” measures listed in the PASS reports. What did I get?
The correlation chart is below. Significant correlations are highlighted in yellow. There’s no significant correlation shown between GAL appointment rates and any of the measures listed in their PASS Reports.
The only significant correlation is the somewhat unobvious negative correlation between kids who get their diploma after aging out and placement stability. (Maybe circuits with more teens have more instability and conversely work harder on education goals?)
There is one suggestive correlation between the GAL representation rate and the percentage of kids who are adopted in under 24 months, r(19) = .300, p = .186. A check of the scatterplot for the adoption statistic shows that Circuit 4 may be an outlier. That’s Jacksonville, a circuit that’s notorious for fast adoptions. That’s reason to believe we should omit it from the analysis.
If we remove Circuit 4, we get a moderate positive, statistically significant correlation between GAL appointment rates and adoptions under 24 months, r(19) = .485, p = .030. That’s a good stat for the Program.
Notice, however, that the size of the out of home care population is also correlated with the adoption statistic. If we control for the size of the circuit’s out of home care population, then the correlation between a circuit’s GAL representation rate and its percentage of kids who are adopted in under 24 months disappears, r(18) = -.004, p = .986.
Summary: The data does not show a significant correlation between GAL appointment rate and adoption timing when controlling for circuit size.
Controlling for Size, Race, Inequality, and Community Safety
The next step is to run the correlations again while controlling for the demographics discussed above. First, OHC Size and IIR.
We again get a suggestive correlation, this time with the Percent of Kids Re-entering Care in 12 Months of Discharge, r(17) = .446, p = .056. According to DCF’s dashboard, which this PASS measure seems derived from, the 12-month period is counted from the time a case fully closes (“termination of supervision” or “TOS”). This means the GAL Program would be discharged and no longer working the case. The correlation is not in the direction I would have guessed: controlling for circuit size and IIR, GAL representation rates may be positively correlated with kids reentering care after TOS.
Again, that’s only a suggestive correlation, so we should look more closely at it. Below is the scatterplot for GAL Rate, controlling for OHC size and IIR. It’s hard to say what the outlier is here. I’m aware that Circuit 20 has very low GAL appointment rates for its own reasons, so removing it makes sense. Circuit 14 was a hurricane circuit, so maybe it can be omitted reasonably as well. Removing both outliers gives us a significant, strong correlation between GAL appointment and re-entry rates, r(15) = .593, p = .012. This is not good. We don’t want cases coming back in.
There’s no additional correlations that pop out from controlling for Violent Crime or Racial Disparity.
So what does that all mean? First, it means we cannot say from the circuit level data that the percentage of kids who are appointed a GAL has any direct correlation with — much less any influence on — the proposed circuit outcome measures.
Second, by controlling for circuit size and socioeconomic factors (and by omitting two circuits that seem to be reasonable outliers), it appears that the GAL representation rate has a positive correlation with the percentage of kids that re-enter care in 12 months. Size and economics being equal, circuits with more GALs on cases have more kids come back into care after TOS.
We don’t know if those re-entries are from failed reunifications, guardianships, or adoptions, but we know the placements didn’t last. That’s not a good statistic for the GAL Program which is supposed to be achieving the “best” outcomes. If the kids are coming back into care at higher rates in circuits with more GALs, that could be a problem.
What about DCF’s Outcome Measures?
I found it odd that the GAL Program PASS Reports suggest that GALs influence the number of kids who age out and get diplomas (unlikely), but don’t include a lot of other DCF measures that the GAL could reasonably have a much larger effect on, like medical and dental care, sibling separation, placement stability, or abuse in care rates. So let’s look at those.
I calculated 15 measures from the DCF dashboards and correlated them against the GAL representation rate. Here’s what I found.
The only direct significant correlation with the GAL appointment rate was found for the number of kids placed in their removal circuit, r(19) = -.511, p = .018. This seemed like another circuit size effect, and sure enough there is no correlation when controlling for out-of-home care population, r(18) = -.280, p = .231. Cross this one off.
There is a suggestive correlation between GAL appointment rates and the percent of kids in foster care in a circuit, r(19) = -.377, p = .092. This time a review of the scatterplot shows that the Southern Region may be the outlier. Regardless, the correlation does not survive controlling for circuit size, r(18) = -.208, p = .379. Cross this one off, too.
Abuse in Care
This next section looks at how the GAL rate correlates with various abuse investigation measures. Turns out it is strongly negatively correlated with both the number of investigations and number of foster care referrals (abuse in care).
Using counts instead of percentages usually runs into a size effect, though, and these both seem related to the size of the circuit. Sure enough, they melt away when controlling for OHC (below). Cross them off.
When controlling for circuit size, the closest suggestive correlation was between GAL appointment rates and the number of verified abuse allegations per 100,000 days in care, r(18) = .414, p = .069.
The scatterplot below suggests that Circuit 20 is again an outlier due to its low GAL rate. Removing it from the analysis, we obtain a strong positive correlation between GAL appointments and abuse rates in care, r(16) = .634, p = .005.
Why would GAL representation result in more kids being verified abused in care? There is an obvious hypothesis: maybe GALs call in the abuse reports. Unfortunately for that theory, when controlling for OHC size, there is no correlation between GAL representation and the total number of abuse investigation intakes or the total number of foster care abuse investigations.
That leaves the possibility that the rate of calls doesn’t go up, but the verification rate goes up when there is a GAL on the case. Either GALs are providing convincing evidence or argument to investigators or the situations are just worse. We cannot test that hypothesis using circuit level data. It’s worth looking into.
I expected GAL rates to have the most effect on the next group of measures. Instead, they had none. Circuit level permanency timing, dental care, medical care, and the reunification and adoption rates all showed no significant correlation to the GAL appointment rates.
Even taking into account OHC, Racial Disparity, Violent Crime Rates, and Income Inequality, there was still no correlation.
There are two suggestive correlations with the Reunification and Adoption rates. These are very important numbers, so we should look closer. The reunification plot appears linear, but the adoption plot may not be. When you control for OHC size, it’s even more scattered. GAL representation rates do not seem to have any significant correlation to the percent of kids adopted or reunified in a circuit. To the extent they’re suggestive, GALs may depress reunification and accelerate adoption.
Let the lack of correlation sink in
What struck me most is the long list of circuit measures that had no correlation with GAL appointment rates. Here they are below, with their r and p values (r, p). A p-value less than .05 is significant and most of these are nowhere close. I thought for sure that circuit-level GAL appointment rates would correlate with medical and dental care rates, or with siblings being placed together, or with minimizing the number of moves. They didn’t.
What’s worse, if you want to measure “influence” instead of correlation, a higher GAL Appointment rate in a circuit was more strongly correlated with worseresults in five measures when controlling for OHC size and IIR. That was shocking to me: GALs correspond to worse outcomes. Only the correlation for RTC rate got stronger and kept a “better” direction, but it is still very weak and not statistically significant.
Summary: There were no significant direct correlations between GAL representation rates and any of the DCF Dashboard measures that survived controlling for circuit size. Controlling for size and socioeconomics and removing one outlier circuit resulted in GAL representation being positively correlated with abuse in care.
Furthermore, if non-significant “influence” is how we’re measuring and we’re assuming causation (two things you SHOULD NOT DO in a normal situation), then the GAL Program is failing and possibly even bad for kids on four additional measures that appear to be core to its mission.
So what measures do correlate with child welfare outcomes?
It would be tempting to think that circuit level data is just too coarse to find a real correlation. Not so. I gathered socioeconomic data on the circuits in question from the County Health Rankings & Roadmaps data for 2018. Because it doesn’t include circuit data, I created a proportional circuit value by taking the weighted average of the county data and the number of kids in care from each county in that circuit. It’s not perfect, but it works.
Here’s what I found. If you control for circuit OHC size and IIR, the partial correlations below are significant. This list is going to be overwhelming to read so I’ve included plain English summaries after each section. The point is that lots of things correlate strongly to the outcomes we’re seeking to measure. For most, though, GAL rate just isn’t one of them.
Reminder: all of these are partial correlations controlling for OHC size and Income Inequality Ratio.
The percent of kids re-entering care in 12 months is…
Positively correlated with the circuit population, r(17) = .598, p = .007;. the unemployment rate, r(17) = .477, p = .039; and the percent of kids living in poverty, r(17) = .464, p = .045.
Negatively correlated with the number of mental health providers per 100,000 people, r(17) = -.489, p = .034; excessive drinking rate, r(17) = -.456, p = .049; per capita income, r(17) = -.457, p = .049.
Suggestively correlated with racial disparity (+), the chlamydia rate (-) , the reporting rate of poor or fair health (+), the food environment index (+), the number of PCPs/dentists/mental health providers per 100,000 people (-), the number of community associations per 10,000 people (-), and the county surtax rate (-).
Suggestively correlated with the GAL representation rate, r(17) = .446, p = .056; but strongly correlated when you remove two outlier circuits. [edited for clarity here.]
Summary: community health effects, poverty, and access to health care all appear to have significant correlations to the rate at which kids re-enter care. The GAL Representation rate correlation is even stronger, however. This is not good for GAL.
Percent of kids having two or fewer placements is…
Positively correlated with the percent of children in single parent households, r(17) = .514, p = .024;
Suggestively correlated with the weighted county surtax in the circuit (-); unemployment rate (+); excessive drinking (-).
Not significantly correlated with the GAL representation rate, r(17) = .044, p = .857.
Summary: circuits with lots of single-parent households also have fewer kids that have stable placements.
Percent of kids reaching permanency in 12 months is…
Suggestively correlated with excessive drinking (-).
Not significantly correlated with the GAL representation rate, r(17) = .075, p = .761.
Summary: there’s no community effect that correlates with the percentage of kids reaching permanency in 12 months. This seems totally driven by system factors.
Percent of kids who aged out and got a diploma or GED is…
Positively correlated with the circuit’s high school graduation rate, r(17) = .469, p = .043
Suggestively correlated with excessive drinking (+); mental health providers per 100,000 people (+); violent crime rate (+); kids living in single parent households (-); the county surtax (+); and the percentage of elderly women living in the circuit (-).
Not significantly correlated with the GAL representation rate, r(17) = -.010, p = .967.
Summary: Circuits with higher graduation rates also have higher graduation rates in their IL populations.
Percent of kids who were adopted in under 24 months is…
Suggestively correlated with the racial disparity index (-); low birth weight rates (+); discharges to ambulatory care (i.e., use of ER for routine care) (+); and the teen birth rate (+).
Not significantly correlated with the GAL representation rate, r(17) = -.282, p = .243.
Summary: there’s no community effect that correlates with the percentage of kids who were adopted in 24 months. This seems totally driven by system factors.
Percent of kids receiving Dental Care in Last 7 Months is…
Strongly positively correlated with the percentage of elderly females in the circuit, r(17) = .768, p < .001; the percentage of the circuit that is white, r(17) = .618, p = .005; the accidental death rate, r(17) = .670, p = .002; Rural Urban Continuum Code (less urban circuits had more timely dental care), r(17) = .524, p = .021.
Strongly negatively correlated with the chlamydia rate, r(17) = -.629, p = .004; the violent crime rate, r(17) = -.599, p = .007; and the PM2.5 Pollution Measure, r(17) = -.635, p = .003.
Suggestively correlated with the number of mental health providers (-),
Interestingly, it was not correlated with the number of dentists per 100,000 people in the circuit, r(17) = .167, p = .494.
Summary: circuits that are whiter, older, and less polluted have higher rates of timely dental care. Air pollution corresponds with industrialization and other environmental health factors, so this may not be crazy. Also, the increase in the elderly population may indicate an increase in Medicare/Medicaid providers.That’s worth looking into.
Percent of kids receiving Medical Care in 12 Months is…
Strongly positively correlated with the percent of the circuit that is white, r(17) = .568, p = .011; the percent of the population that is female and 50+, r(17) = .610, p = .006; the accidental death rate, r(17) = .635, p = .003; the Rural Urban Continuum Code (less urban had more time medical care), r(17) = .593, p = .007.
Negatively correlated with the PM2.5 Pollution Measure, r(17) = -.467, p = .044; and the violent crime rate, r(17) = -.546, p = .016.
Suggestively correlated to the percent of the circuit that is white (+), the chlamydia rate (-), the food environment index (-), mental health providers per 100,000 people (-), number of community associations per 10,000 people (+), and percent of households with a severe housing problem (-).
Not significantly correlated with the GAL representation rate, r(17) = -.076, p = .756.
Summary: circuits that are whiter, older, less urban, less polluted, and less violent give timely medical care more often. Again, it’s worth looking into the number of Medicaid providers to see if that’s holding up the rates.
Percent of kids in RTC Care is…
Strongly positively correlated with the Proportion of Circuit Reporting Poor or Fair Health, r(17) = .520, p = .023; the Proportion of a Circuit’s Average Poor Health Days, r(17) = .670, p = .002; the Proportion of a Circuit’s Kids in Poverty, r(17) = .538, p = .018; the teen birth rate, r(17) = .556, p = .014; average poor mental health days, r(17) = .560, p = .013; discharge from ER to ambulatory care, r(17) = .539, p = .017; percent of kids in poverty, r(17) = .541, p = .017; Rural Urban Continuum Code (less urban had more RTC usage), r(17) = .488, p = .034.
Negatively correlatedwith a circuit’s Per Capita Income, r(17) = -.727, p < .001; access to exercise, r(17) = -.690, p = .001; households with at least one severe housing problem, r(17) = -.460, p = .048.
Suggestively correlated with the percent of low birth rate (+), food environment index (-), percent uninsured (-), high school graduation rate (-), kid living in single parent households (+).
Not significantly correlated with the GAL representation rate, r(17) = -.163, p = .505.
Summary: poorer, unhealthier circuits have greater usage of RTC care.
There are interesting symmetric correlations with adoption and reunification.
The Percentage of Kids Exiting to Reunification is strongly negatively correlated with a Circuit’s Reported Number of Poor Mental Health Days, r(17) = -.558, p = .013.
Conversely, the Percentage of Kids Exiting to Adoption is positively correlated with a Circuit’s Reported Number of Poor Mental Health Days, r(17) = .627, p = .004; number of poor mental health days, r(17) = .630, p = .004; lack of access to exercise, r(17) = .473, p = .041; percent insured, r(17) = .485, p = .035; and pollution levels, r(17) = .462, p = .047. [Note: I’ve flipped the signs on some of these measures to align them.]
The adoption correlations around mental health are stronger than the one reunification correlation.
There was no correlation between the GAL representation rate and adoption or reunification rates.
Summary: circuits with poorer community mental health and environmental conditions reunify kids at lower rates and adopt kids out at even higher rates.
To sum that all up: poverty levels, racial makeup, mental health, physical health, healthcare access, and even the environmental quality of the circuits all have a stronger correlation to the DCF Dashboard outcome measures than the GAL representation ratesdo.
So the bottom line is that the GAL representation rate in a circuit doesn’t significantly correlate with anything that the GAL Program suggests it does. If the effect is too small to measure at the circuit level, then that’s evidence against the original premise that GAL representation is a significant factor in positive system outcomes. The aim for 100% GAL representation of all kids in Florida may not be the right goal.
To the contrary, the GAL representation rate, when partially correlated for OHC size and IIR, was associated with two negative outcomes: more kids coming back into care after TOS and more kids being found to have been abused while in care. If that’s because kids are winding up in worse placements, we need to know that. If that’s because GALs are raising more complaints after they’re in a bad placement, we need to know that, too. While the detection of problems may be a net positive, the supposed power of best interest advocacy is to avoid problems, not just identify them after the fact. We probably need better studies into the quality and timing of the recommendations that GALs make.
We also have to accept the possibility that best interests advocacy, even armed to the teeth, may be an insufficient tool to affect outcomes in the face of widespread community poverty. It appears that there are community health effects that far overpower the GAL representation rate as a correlative factor in outcomes. There are plenty of studies that suggest that CASA/GAL advocacy results in higher chances of adoption over reunification on individual cases (even if we didn’t see it at significant levels here), but I don’t know of any study that compares this to other community effects like poverty, health access, and environmental factors. Comparing GALs to AALs may not be the right test. If it turns out that the number of community doctors and therapists who take Medicaid has a stronger and broader impact on foster care outcomes than GALs do, then we could have spent the $600 million GAL Program budget in very different ways.
It seems counterintuitive that circuits with higher GAL rates would have (non-significant) lower rates of timely medical and dental treatment. This raises the possibility that GALs create a moral hazard in the system. The existence of an assumed-effective watchdog like a GAL may incentivize the system to act only when that watchdog alerts to a problem. If the quality of the GAL advocacy is poor, the system may mistake silence for a lack of problems, which then reproduces more problems. Recall that nearly 45% of GALs that separated from the Program were terminated for not meeting the expectations of the job. Forget GALs who may have acted badly — if those GALs just did nothing while on the job, that alone could explain how higher GAL rates could result in worse outcomes. We need to look more closely at GAL work processes and outputs.
The ultimate question, of course, is whether GALs actually offer meaningful and sustainable solutions to the system. It’s not about kindness or good intentions; it’s about quality. There have been a few sociological and legal studies looking at the gap in socioeconomic status between the families in the system and the people who volunteer as GALs. These studies highlight the disconnect that can happen when well-meaning people try to give advice to others with significantly fewer resources and different experiences than they have had. Maybe GALs just don’t know how to navigate our families’ worlds. That’s ok — they can help in other ways.
So, after all this, the question of whether the GAL Program is ethical, effective, or even good for children is all the same: we don’t know. And before we spend another $600 million, we should probably find out.
This post introduces a new public FSFN dashboard on permanency timing in Florida’s child welfare system. If you want to just play with the dashboard, you can find it here. All but one of the graphics in this post come from the dashboard.
Every year, in legislatures across the country, well-meaning people propose bills to speed up permanency for foster kids. Permanency is a psychological concept focused on attachment, belonging, and community. Those are hard to legislate, so people focus instead on procedural definitions. The legal meaning of permanency is to close the court case and get the state out of a family’s life for as long as possible. In the process, hopefully leaving the child better than the system found them.
That closure could happen by returning a child home to a parent, placing the child in a guardianship, or having the child adopted. It could also mean a child aging out. The end result is largely the same to the state: one less case on its docket, and varying ongoing financial obligations depending on the way the child exited care. The path a case takes can change a child’s life.
Current law prefers the two most labor-intensive forms of permanency (reunification and adoption) over the procedurally easier ones (guardianship and aging out). Current law also sets timeframes and conditions on when certain permanency goals can be achieved: reunification should be accomplished in 12 months, guardianships cannot be entered until a child has lived in a placement for 6 months, APPLA (the “plan to age out” goal) should be accepted after age 16, and adoptions require the termination of a parent’s rights and appeal periods, which — on paper at least — would require about 18 months.
The question for this post is not what the right policies should be. I want to know how things are operating under the current policies. How are we doing? Using the public Florida Safe Families Network placement database, we can trace the permanency path of individual kids down to the day.
Part #1: How kids exit the system
The chart below shows the approximately 184,000 children who exited foster care in Florida from January 2008 to April 2019. I’ve consolidated some of the categories for easier reading. You can see that around 49% of children were reunified, while 22% entered guardianships, 22% were adopted, and 7% aged out. The remaining <1% had their cases dismissed, administratively transfered or closed, or they passed away.
A handful of cases had no exit code marked and are omitted here.
Slightly over 21,000 kids (13%) statewide exited and reentered care, but only 2,900 kids (2%) did so more than once and only 405 (<1%) did so more than twice. Of the failed parental reunifications, approximately 35% successfully reunified later. Of the failed adoptions, the number was the same: 35% were later reunified with the adoptive family successfully.
In the next chart, you can see the final placement setting for each exit type (e.g., all children who were reunified out of a relative placement or who aged out from a group home). Reunifications and adoptions happen almost evenly between relatives and non-relative; guardianships come almost entirely from relative placements. Very few kids age out of a relative placement.
There’s another way to look at this data. You can stack the boxes by the type of custodian the child exited to and see the actual face of foster care: relatives. Non-relatives and foster parents, while important for every child who needs them, only account for a fraction of the total exits. The vast majority — nearly 80% — of kids went back home to their parents or to relatives. Slightly under 11% of kids were adopted by non-relatives. That’s only four points higher than the 7% who aged out. Foster care is temporary care. Forget all the saving children rhetoric — the real slogan of DCF should be “We’re here when you need us. But we really hope you don’t.”
Part #2: When kids exit the system
With the exception of aging out and death, a child’s exit from foster care happens via judicial action, usually at a court hearing. The timing of those hearings therefore matters. Using the placement database, we can look at the permanency question differently and calculate exactly wheneach child exited the system down to the exact number of days they had been in out-of-home care. For simplicity, we’ll look at months instead.
The chart below shows the number of kids exiting the system by the number of months they spent in care. I calculate the month by dividing the number of days in care by 30 and rounding down to the nearest whole number. That means Month 0 is 0 to 29 days, Month 1 is 30 to 59 days, etc.
If you know the dependency procedure timeline, this shape makes perfect sense:
Cases should go to trial in the first four months, but most parents take pleas at their arraignments, which is 28 days after the removal. That means a sizeable percentage of kids exit care before the time has passed for a trial in month 3. In fact, the highest single month for exits is the first month of a case, Month 0. You might wonder if those kids needed to be removed in the first place.
The first judicial review must happen within 3 months of the disposition or within 6 months of the shelter. The curve spikes sharply in Month 6 — presumably at that first judicial review. A quarter of kids exited by the first JR at month 6.
The permanency hearing must be heard within twelve months of the shelter and there’s another spike. Half of kids exited care by the end of the twelfth month.
The rest is an exponential decay curve as cases close out with no set timelines in the second year. Almost all, 95% of children, exited care by month 50.
To make more sense of the exit chart, we can break it down by exit type. The chart below shows when exits occurred for Reunifications, Guardianships, Adoptions, and Age Outs, Other, and Deaths. The patterns are very distinctive. This is the chart that made me do this whole post.
Let’s start with Reunifications (the blue-green-ish line). Reunifications were at their highest monthly rate at the beginning of cases. There could be lots of reasons for this. It could be that case management and parents were able to help remedy the circumstances that necessitated the removal where the CPIs did not (or could not in the time required for an investigation). It could be that lawyers or judges looked at the facts and decided that a removal was not legally justified. Whatever the cause, 25% of reunifications occurred in those 3 months.
The Reunification line flattens out around Month 3 and hovers through the case plan phase until Month 12. It then enters a period of exponential decay. About 25% of reunifications took longer than 12 months. There were over 26,000 children reunified after 12 months in care from 2008 to 2019.
The red Guardianship curve is similar but different in important ways. By statute, a case cannot close in guardianship until a child has been placed with that custodian for 6 months (meaning some of those placements may not be strictly legal). Not surprisingly, the chart above shows that guardianships spike to their highest point in Month 6. They bump up again slightly in Month 12, and then enter a period of exponential decay similar to Reunifications. Over 20,000 kids entered guardianships after being in care for a year.
Next, the blue Adoption line is different from either of the previous two. It is one of the few bell curves in child welfare: adoptions peaked in month 24 and were spread out evenly around that point. Adoptions overtook reunifications and guardianships as the primary exit type around Month 17.
Why so late? Because, unlike guardianships, adoptions require termination of parental rights. Unless a parent surrenders or defaults at trial, that means an entire failed dependency proceeding (approximately one year), an entire successful TPR proceeding (approximately six months), and then an entire adoption proceeding (which can also take months, especially if there are competing adoptions or complex subsidy issues). Insert appeals between all of those, and the curve makes perfect sense. Adoptions have a lot of moving parts.
I’m saving the Administrative Closures and Deaths for another day. That leaves the orange Age-out curve that is always present beneath the others. Kids whose parents cannot get them back, whose relatives will not take them, and who do not fit the profile that strangers want to adopt sit in care until they exit naturally. These kids experience the most placement instability of all of the groups. Some of them were orphaned by TPRs that never led to permanency. Others spent significant time in institutions, medical programs, or jail.
There is one more group worth talking about that is not separated on the graph by line-color but by time. That long tail of kids that reached permanency after four years is notable. There are reunifications, adoptions, and guardianships all through that period. They are proof that we should never give up completely on any of the goals.
Part #3: The system is changing over time
Most permanency tweaking is aimed at speeding up adoptions, the most procedurally complex of the exits. That’s a hard curve to move. One, adoptions have a lot of variability, so it’s hard to even know how to speed them up across the board. Two, they also have a lot of interdependence with the other goals: it’s difficult to know how putting more resources and emphasis toward adoption would affect the other, larger exit types. And lastly, even if we somehow halved the time to adoption for all kids from a median of 24 to 12 months, that would only reduce the system-wide time to permanency by about 3 months. There’s more variation than that between years, placement types, and agencies.
In fact, inside the decade average hides a different trend: permanency time has been going up across the board. The chart below shows the median number of months to each type of exit by the discharge year. The out-of-home care population size is in the background for context. Adoptions were at their fastest in 2012 at 22 months — just when the system was regearing for an expansion. All three exit types have been slowing since 2014 as the system grew larger. Overall, the median time to exit is 2 months slower than 2015.
Part #4: It matters who adopts
There is another way to speed up adoptions: find a relative. The charts below show the curves for adoptions by relatives and non-relatives. Adoptions by relatives are, on average, a staggering 6.7 months faster. That is, however, largely because relatives tend to adopt in the first 4 years or never. New non-relatives come into a child’s life all the time, and eventually one clicks, pushing the blue line out farther to the right.
Limiting the curve to the first 48 months (below), the Relatives still adopt about 3 months faster than the Non-Relatives.
It’s notable that the relative curve (above) is heavier at the front end. You can see it more clearly in the next graph, which shows the Relative to Non-Relative ratio over the months of a case. Relatives outpace Non-Relatives by up to 5x in the first months of a case. Why? Probably surrenders.
After 24 months, however, Non-Relatives begin to outpace Relatives. These are mostly failed case plan and contested TPR cases. Remember that only 11% of cases ended in a Non-Relative adoption and half of those happen in under 24 months. That means we’re looking at 5% of all cases. We could do a lot more to prepare parents, children, and custodians to work together in these situations.
There are slight differences even among the Relative adoptions. I’m not showing the graphs here, but grandparents adopt, on average, 1.5 to 2 months faster than aunts, uncles, or cousins. (The other family relations don’t show any significant differences.) Grandparents adopt a whopping 8.6 months faster than non-relatives — again in part because they adopt in the first few years or not at all.
Part #5: It also matters who takes custody for reunifications
Placement setting made a difference in reunifications as well. The median time to exit was slightly faster for non-relatives over relatives — 7 months, instead of 8 for relative placements. Institutions exited kids much faster, likely because institutions are time-limited placements, kids in them tend to be older, and there may be behavioral or other issues that make non-relatives less willing to accept the child. What was different was the shape of the curves indicating when those exits happened.
The chart below shows reunifications from each of the placement settings. (This only considers the child’s last placement setting — the child could have been in multiple other placements before exiting care.) You can see the so-called “Short Stays” in the blue line: over 6,000 kids were reunified from foster care in under 30 days. A smaller, but notable, effect is seen in group homes and institutions. Relatives, on the other hand, do not show the same sharp short-stay pattern.
What the relatives and non-relative curves share is the drop-off at Month 12. This is likely a result of the permanency hearing where judges are asked to make a major decision on the case: attempt reunification or go to TPR. The same drop isn’t seen in group homes and institutions, suggesting those (usually older) kids aren’t presenting the same choices. They may already be in APPLA.
Part #6: There are also differences in the CBCs (and circuits)
The database tells us which lead agency made each placement, so we can also tell differences in how CBCs exit kids from care. It’s important to note that courts are the ultimate deciders of when kids exit care. The database doesn’t tell us which judge is assigned to a case, so we are left with comparisons by CBC.
The first difference was surprising to me: different CBCs have drastically different rates of reunification and adoption. The chart below plots each CBC’s rate of reunification against its rate of adoption. The linear relationship is expected: a higher percentage of one means a lower percentage of the other. (Guardianships and aging out aren’t included here, so they account for some of the other differences).
What struck me was the significant difference in how kids exit each agency. Agencies in the top-left quadrant use adoptions at higher rates: over 28% of their exits were to adoption, and under 42% of their exits were to reunification. The lowest reunification rate was nearly 30%, and the highest adoption rate was over one third. These agencies tend to be a little smaller and in the northern regions of the state (though not exclusively). In contrast, agencies in the bottom-right quadrant use reunifications at higher rates.
The only (active) agency squarely in the top-right is Families First Network. It appears to be using adoption and reunification at higher than average rates (at the expense of guardianships and aging out). Essentially no agency is found in the bottom left quadrant.
We might expect CBCs that focus more on adoptions to have slower median times to permanency. We’ll see they don’t.
The next chart breaks down the CBCs by size (throughput, actually — the number of children who exited care during the time period) and median months to permanency. There’s no direct correlation. CBCs can be any combination of big, small, fast, or slow. Eckerd runs two big agencies in neighboring counties and one is much faster than the other. ChildNet in Broward is two months slower than its sibling ChildNet Palm Beach.
It is also notable that smaller does not always mean faster, but bigger almost always means slower.
The next chart looks at how adoption rates relate to the median time to exit. There is no significant linear correlation. Instead, it’s a parabola. Big agencies tend to be in the middle, and smaller agencies tend to take the extremes. Those extremes are Family Support Services, which handles a high percentage of adoptions very quickly, and CBC of Brevard, which handles fewer adoptions but still moves cases quickly.
Two agencies stand out as extremes on the chart above in terms of size to adoption ratio: Family Support Services in the bottom right and ChildNet (Broward) in the top left. How can FSS be so fast and do so many adoptions, while ChildNet (Broward) is so slow and does so few? We can look at their curves for a clue.
The chart below shows FSS’s adoption curves. The orange line specifically shows adoptions from relative placements. The far left of it looks more like the guardianship line at other agencies. FSS (or its courts system) is somehow getting relatives to adopt in the first six months of cases. That means parents are surrendering without even doing a full case plan. The first judicial review is at 6 months and full adoptions are happening at significant rates before that. The median number of months for a relative adoption by FSS is 14 months compared to 24 months statewide. That’s insanely fast. I’m not convinced it’s a good thing. (Note: the trend seems to have ended after 2015, which suggests a policy or personnel change.)
On the other hand, ChildNet (Broward), has a fairly normal time for relative adoptions at 24 months. It’s the non-relative adoptions that take longer: 29 months. ChildNet’s non-relative reunifications also take longer, so there appears to be something going on with foster placements — either at ChildNet or in the judiciary. The institution and group home numbers are also high, so it might be worth exploring whether the use of congregate placementis is contributing to delayed permanency.
The median adoption times across the state are very, very different. Below is a map of adoptions by county of last placement. You can see the streak of green counties starting in Duval and working down to Levy. Maybe someone who works in those areas can explain it, but those are the fastest adoption counties over the last decade.
Meanwhile, reunifications also show significant differences. The median time to reunify can take up to 50% longer in adjacent counties. For example the median time in Lake County is 5 months, while in Polk it’s 8 months. Two different CBCs and court circuits, very different experiences.
Part #7: What does it all mean?
I started by saying that this post isn’t about what we should do, and it isn’t. There are still some policy nuggets worth mining in the numbers. Here are the top five.
It’s not good that we don’t have a statewide advocacy group for parents and relatives. They make up 81% of the exits in foster care, and their concerns for the kids and the system are valid. Parents get attorneys in their cases, but that’s not the same as an advocacy group to shape policy. Why aren’t they being organized and heard?
A year is a long time in the life of a child? Not if the alternative is needlessly losing their parents and family forever. There are very few child welfare slogans that rile me up like that one, because it is almost always uttered before offering some policy change that would risk cutting off the 30,000 kids who got reunified or placed in a guardianship after 12 months. It’s also misleading: it lumps all kids together — some want to go home and are willing to wait, others aren’t; and it doesn’t recognize that different cases have different barriers and different communities have different resources. We should spend more time examining the differences around Florida that lead to vastly different permanency times before we set increasingly restrictive statewide mandates on parents who are not given even playing fields.
There are lots of kids we could probably avoid removing if we tried. The idea of short-stayers is nothing new. We see above that over 6,000 kids were removed to foster care and returned in under 30 days in Florida. That seems like needless trauma. This is worth talking more about.
Adoptions are actually pretty okay. I’m comfortable with the fact that adoptions are a bell curve and that they take about 2 years. Those kids are generally really young and in super-stable, loving homes. We can continue to speed up the paperwork and administrative delays on the post-TPR backend without robbing parents of the chance to reunify at the front. We could also speed up some cases by closing them in Permanent Guardianships with visitation rights instead.
We need a real plan in the second year. On the other hand, I’m not comfortable with the fact that reunifications and guardianships just fizzle out in the second year. This year a bill was proposed that had language requiring permanency hearings every 60 days after the first year. (The language got dropped later.) It didn’t require any programmatic changes, resource investments or even evaluations to determine what steps were needed to help a parent be successful. It would have just forced everyone into court every two months until the judge got tired and set the case for TPR. I get the motivation, but we need a meaningful Year Two Plan.
This post is already too long. There are a lot of conversations we can have around this type of information, especially about the differences that families experience by geography. All of the diagrams and charts in it are from the Plotting Permanency Tableau. You can filter down to the CBC and county level of many of the graphs. As always, if you find any errors or something that doesn’t look right, please let me know.
This post introduces a new public FSFN dashboard: the Placement Provider Info Dashboard. If you want to jump straight there, feel free. You should click the fullscreen button in the bottom right corner. Below is the why and how of it.
There is a well-meaning bill working through the legislature that would exempt the names of foster parents from Florida’s public record laws. (Current law exempts their addresses, financials, and the floorplans of their houses.) The bill cites four “public necessities” to bar access to foster parents’ names: (1) it will help keep foster children’s names confidential, (2) it will prevent “unwanted contact” by the press, (3) it will prevent “unwanted contact” by the child’s relatives [i.e., parents], (4) not doing so would compromise foster parents’ privacy. The reasons don’t really stand up to scrutiny. More importantly, public access to information on foster placements is actually a good thing.
The elephant in the room is named Candi Johnson
Let’s start by acknowledging that Candi Johnson, the mother of two children in foster care, orchestrated the shooting of an elderly foster parent in Miami. She went to the foster home with her teenage son and demanded the children. When the foster parent fought back, the son shot her and fled with Candi Johnson and the kids. The foster parent is a hero for defending the kids even when she had no idea who was after them. The media reports that Candi Johnson had a long history of violence and had absconded with her children before. She is currently pending trial for attempted murder, kidnapping, armed burglary, and interfering with child custody.
The public records exemption would not have prevented Candi Johnson and her son from shooting the foster parent. The list of foster parents in the FSFN database has nearly 68,000 people on it. Candi Johnson’s kids would have aged out before she figured out which provider was caring for her children that way. More importantly, Candi Johnson did not use public records to find the foster home. Everyone in the neighborhood knew the woman was a foster parent. Everyone on the case knew Candi Johnson was violent. The foster mother didn’t know who Candi Johnson was when she banged on the door — or else she wouldn’t have opened it and maybe wouldn’t have accepted the placement. Having foster caregivers meet with parents in a supervised setting when they first take in kids could have actually prevented this. Further increasing the separation between them would not.
The bill isn’t about the kids
Candi Johnson stirred up a lot of latent anxieties that some (but certainly not all) foster caregivers feel about the families of the children they take in. The sponsor of the bill says that DCF received calls from “several” foster parents that they would quit if their names were not protected. I received a comment from one foster parent saying the same. The problem is that the bill doesn’t protect foster parents from the people they (rightly or wrongly) are afraid of. It does, however, make it harder to identify wrongdoing by DCF or other foster parents towards the kids they care about.
Let’s start with the bill’s first goal: protecting children’s privacy. The bill doesn’t actually do that. Having the names of foster parents does not tell me the names of the kids in their homes. If we want to keep foster children’s information confidential then we would also include provisions making it illegal for foster care providers to post about the child online, including pictures and over-sharing facebook posts. That’s how most parents in the system find their kids — a mutual friend spots the pictures and forwards them.
If we were serious about not disclosing a child’s foster care status, the bill would have provisions aimed at school personnel who tell a child’s classmates and protective investigators who question neighbors and disclose more than they should. We would also shut down National Adoption Day and Heart Gallery events where kids are brought to one place with giant signs that say “foster care” and television cameras rolling. Nobody is particularly worried about any of that.
A public records request on foster parents would currently give you a name and maybe a zip code, but nothing on the individual kids. You can get that much from a google search (and more). There is actually a much bigger and more immediate leak of information about foster children: the court hearings are open to the public. I have been in countless hearings where essentially this exchange happened in front of a room full of strangers waiting on other cases:
CLERK: Calling the Case of [insert actual name of the child or parents]. All parties please announce.
[everyone, including the parents and children go around and say their actual legal names]
JUDGE: Are the foster parents here? Please just use their initials.
[nobody mentions that there is no law that says foster parents get to be anonymous in court hearings]
FOSTER PARENT: J.M. Good morning, Your Honor.
JUDGE: Ok, we’re here today for a status on medical treatment. Did the child go to the gynecologist for an STD check? She was sexually assaulted. I am very concerned that you did not take her sooner.
CASE MANAGER: Yes, the child is present and can report. She tested negative.
We accept that the hearings are open because we believe the system is better when it works in public. In that context, foster parents’ names are not more sensitive than a child’s history of abuse. Foster parents are good people who largely volunteer to help kids and families that need it. They do not, however, have stronger privacy interests than the actual children and families in the system. When you sign up for this work, you sign up for it in public.
Second, the bill seeks to limit foster parents’ names because disclosure could lead to “unwanted contact” from the child’s “relatives.” This also won’t work and is actually a bad policy goal. Foster parents, parents, relatives, case managers, guardians ad litem, and therapy dogs all sit outside court together, sometimes waiting hours for the case to be called. In most courthouses it’s impossible to hide. And they shouldn’t want to hide: if foster parents are following the Quality Parenting Initiative co-parenting guidelines, then they use that time to talk with the parents and relatives to get to know them and the children better. I hope this sounds as direct as I mean it: a foster parent who doesn’t want contact with a child’s relatives should look for other ways to help children. Fostering isn’t for you.
There may be times when it’s not safe to engage in co-parenting with the child’s family, such as in Candi Johnson’s case. In those situations, court orders and injunctions directed at the parties on the case are the right remedy. Limiting the public’s access to information about foster care providers doesn’t solve the problem: the child’s family already has the foster caregiver’s identity information, or can easily get it by reading the Case Plan or by waiting in the parking lot for 20 minutes. In cases where there are serious safety concerns, there should be serious security responses. A general public records exemption is not a cure.
Finally, I’m not a First Amendment scholar, but “unwanted contact by the press” is why we have public records laws — the press and other watchdogs are supposed to investigate and sometimes that investigation is unwanted. The foster care system is a billion-dollar-a-year government industry. The fact that it recruits and underpays volunteers to perform some of its essential functions does not insulate it from scrutiny.
Now for why it’s actually good to make this information public.
Public information helps make better decisions
After publishing the Visualizing Foster Care Instability project, I received a lot of comments asking for a dashboard that gives information about the foster care providers. I attended a Florida Youth Shine quarterly meeting where a young person still in foster care said in a session, “There should be some way for us to know about placements before we go there. We should know as much about them as they do about us.” That resonated with me. We wouldn’t stay at a hotel without reading the reviews, but we expect foster kids to just show up at a house in the middle of the night and take it on faith that it will be safe.
So I made something: The Florida Foster Care Provider Dashboard. (I’m not really good at naming things.) The goal is to put everything we know about foster care providers in one place so that advocates and the public can make better decisions on how the system operates.
It’s functional, not pretty. I recommend viewing it in full screen because it has a lot of parts. Here’s what it looks like:
Here’s what you can learn from it:
How long do kids stay with ____? This chart in the top-right shows the distribution of how long a provider’s placements lasted. A provider’s average placement length may be skewed due to a few kids they kept for years. This chart shows the real breakdown. You can set it to measure in days, weeks, months, or years. Above, you can see that this provider had 315 placements in total and 134 that lasted less than a week. The average placement length was 45 days, but the median was only 9 days. This is not a stable placement for most kids who go there.
What kind of placement is this? The two boxes in the bottom left break down DCF’s own designation of the placement type. Above you can see that this provider was almost exclusively a foster care provider (orange bar), but spent 16 days as a relative placement, and 4 as a group home. The Service Type shows that this home was mostly for kids aged 13-17.
Why do kids leave this placement? The third box on the bottom left shows the reasons that placements with this provider ended. You can see above that 101 placements ended “in accordance with the case plan,” which usually means pursuant to a court order, while 61 placements ended because the child ran away. Twenty kids aged out of this home.
How has it changed over time? The box in the bottom right corner show the complete placement history for this provider. You can see that they started fostering in 2003 and had their last placement in 2018. Over time, placements with this provider have gotten shorter and shorter. That’s fairly normal for the long-time placements. The first few placements are usually the longest.
What about some summary stats? Right in the middle are the summary stats that I’ve calculated for the provider: number of children, number of placements, average placement length, median placement length, average miles kids moved from the last placement, and average concurrent kids (meaning how many children were placed there simultaneously on average).
I’ve also created Provider Flags that alert you to certain questions you might want to ask about a placement before putting a child there. They are found in the pink box right in the middle of the dashboard. The flags are based on the objective criteria below.
Death of Child: The provider had at least one placement where the end reason was “Death of Child”.
High Reunification, Adoption, Age Out, or Guardianship: The provider was placement to at least 6 children and more than 50% of them went on to reach the stated permanency goal. This does not mean the children exited care from the provider directly.
High Runaway: The provider had at least 6 placements and more than 25% of placements ended in the child running away.
High Turnover: The provider had at least 25 placements and more than 50% of placements ended in under 30 days.
High Disruption: The provider had at least 25 placements and more than 50% of placements ended because the provider requested a change, the child requested a change, or the placement “disrupted.”
High Concurrency: Children with this provider had an average of 10 or more other children placed there concurrently. This could be either because the provider’s capacity is 10 or more children or because the high turnover rate caused 10 or more children to pass through the provider.
High Mileage: The provider had at least 25 placements and the average child moved more than 50 miles from their previous placement. Miles are calculated from the center of a provider’s zip code region.
High Hospitalization: The provider had at least 6 placements and more than 25% of placements ended due to hospitalization of the child.
First Run Warning: The provider had at least 6 placements and more than 10% of children placed there ran for their first time while with the provider.
Baker Act Warning: The provider had at least 6 placements and more than 5% of placements ended because the child was Baker Acted. Baker Acts were calculated by finding children whose next placements were for “Routine/Emergency Mental Health Services”. Note that for small placements, even one Baker Act will raise this warning.
Arrest Warning: The provider had at least 6 placements and more than 5% of placements ended because the child was arrested. Arrests were calculated by finding children whose next placements were for “Correctional Placement” or whose placement end reason was “Incarceration/Detention”. Note that for small placements, even one arrest will raise this warning.
Night to Night Warning: The provider had at least 25 placements and more than 25% of placements were for 2 or fewer days.
I joked that this would be like Yelp for Foster Care, so I went ahead and added three more tabs to make that a reality:
School Map: This tab allows you to click on a provider and see the greatschools.org map for the school in its zip code.
Walk Score: This tab allows you to click on a provider and see the walkscore.com ratings for the zip code.
Yelp: This tab allows you to click on a provider and see the yelp.com most popular places in the zip code.
What can we learn from this?
The Herald Tribune did a story on the public records bill a few days ago. In it, the sponsor is quoted as saying:
“The foster parents are not the people who have been suspected of doing anything wrong,” Roach said. “It’s the parents themselves. … Those are the people that need scrutiny, not the foster care parents.”
Hold on now, nobody said bio parents shouldn’t be scrutinized — and they are heavily scrutinized, in the form of evaluations, classes, supervised visits, and home inspections. The question is whether a higher-than-zero level of public scrutiny of foster parents is warranted.
Yes, it is.
If we didn’t know who they are, we wouldn’t know that foster parent “Lor. Hic” has taken in 525 kids for 626 placements and asked for their removal 360 times (“High Disruption”, “Night to Night Warning”). If she’s agreeing to be a night-by-night placement, then she’s running a a shelter with a foster care license. We need to talk about that practice.
We wouldn’t know that foster parent “Tif. Gip.” has an average placement length of 29 days, but a median placement length of only 2 days (“High Turnover”). The average child placed with Tif. Gip. would experience over 7.5 roommates during their time there. That’s essentially a group home with a foster home license. We need to talk about that, too.
We wouldn’t know that foster parent “Sha. Rob.” experienced the Death of a Child in 2002, or know whether the teams involved in placing kids there for the next three years were aware of that fact.
We also wouldn’t know that foster parent “Kat. Mel.” had over 10% of her 138 kids run for the first time while in her care (“First Run Warning”). Same for “Pat. Fau.”, “Ann. Gre.”, “Gen. Zie.” and many others. What’s going on in these homes?
If the confidentiality gets expanded to institutional providers, then we wouldn’t know that the Hibiscus Vero Group Home has six flags: Arrest Warning, High Turnover, High Concurrency, High Mileage, First Run Warning, and the Death of a Child in 2012. It may be a perfectly lovely place, but I would want to ask questions about all of those issues before I sent a child there.
I’ve requested the provider payment database from DCF as well, and it’s pending. I plan to add an overlay on how much providers get paid. An earlier version of the payment data that I have showed that there was a foster parent in Miami (Jef. Hor.) that received $15,000 per month to care for one child. That is not a typo. If foster parent info is exempted from public records, we wouldn’t know that. The public has the right to scrutinize how the government spends its money. Fifteen-thousand a month is either excessive or what everyone else should get.
I’m also working on a version that overlays the Florida Sex Offender database. The foster home for “Tra. Dav.” that I used in the first example above is in a zip code with nearly 250 registered offenders. Most zip codes have a fraction of that. Maybe we want to think about that when we place kids there, especially certain kids.
Beyond just risk factors, there’s also the problem of the foster parents who have literally hurt kids. If their names are exempt from public records, we don’t know who they are unless they kill a child. (And kids are placed with providers even after others die, as seen above.) I appreciate what the sponsor is saying, but the statewide Guardian ad Litem Program was created and funded with taxpayer money and then significantly strengthened in response to a foster parent murdering a child. The potential for misconduct with our most vulnerable children warrants constant vigilance regardless of who the caregiver is. Trust should not be blind.
Okay Robert Latham, you’re a hypocrite for not publishing the foster parents’ names
Here is where some astute commenter sends me a pointed note that I have chosen to use initials instead of names. I must, therefore, agree that publishing the names would be dangerous.
I believe that publishing the full names would kick the anthill and close down access to a valuable source of public information. I’m using the initials so that readers can focus on the importance of the information and not the hypothetical problems that can come from releasing it (even though it’s been public literally forever). For the public version, I think the three-letter initials are sufficient to find a specific foster parent you’re looking for if you already know the name. I am happy to share the unredacted version with anyone working directly in the system who could use the information.
Public information helps us make the system better, but we can’t do that if we’re not allowed to know things.
Christopher O’Donnell and Nathaniel Lash at the Tampa Bay Times recently published an outstanding investigative piece on the harmful number of placement changes some kids experience while in foster care. They write:
Foster care is intended to be a temporary safety net for children at risk of neglect and abuse at home. Those children, many already traumatized, need love and stability to recover and thrive, child psychologists say.
But thousands of Florida’s foster children were put at risk of further psychological damage by an overburdened system that repeatedly bounced them from home to home and family to family, a Tampa Bay Times investigation found.
Times reporters analyzed more than one million child welfare records recording the movements or placements of about 280,000 foster children under Florida’s care between 2000 and 2017. They show that thousands of foster children led transient lives, many staying only a few nights in one place before being moved on to the next foster family or group home.
For those of us working in the system, placement instability isn’t news (many professionals are numb to it). But it is news for the rest of the world, whose picture of foster care is based on the heartstrings marketing of charitable agencies or the five o’clock stories of deaths and abuses seen in the news. The daily pains and indignities of foster care are rarely discussed by a public who doesn’t have the information or language to talk about them. I was so happy for the Times article because it gave people a new idea: many foster kids move around a lot and that’s a bad thing.
This blog has a different audience, though. The readers here know about the system, often from deep in the weeds, handling cases or overseeing agencies and programs. We have seen placements disrupt both in 30-person staffings and via unexpected text messages that our client’s been kicked out of a home we thought would last — if not forever, at least for a week. We need no emotional priming on this topic. Short of telling a child he can’t go home, the hardest thing we sometimes have to say to them is they “can’t stay there anymore.”
It’s awful. Hold on to that feeling for these next parts. I want to show you placement instability from a thousand miles up, where the people look like ants. I want to multiply that gut-wrench feeling by 17,000 to break through the numbness and help you remember that this is not okay.
The database that the Times used in its reporting is a public record in Florida. I don’t know that any newspaper had ever written a story using it before, and I commend them for doing so. I also have it. I’ve been reluctant to share it largely because it is 77.8 million data points and completely overwhelming. The article and the discussion around it, though, made me believe that it’s time.
Instead of presenting the data with statistics and aggregates, I’m giving it to you how I first began to really understand it: as maps. Every dot on the map is at least one placement for a child. The colors show what type of placement: blue is foster home, purple is relative, orange is group homes, and so forth. The size of the dot shows the length of time the child spent there, and the lines show the moves from placement to placement. Sometimes there are breaks in the lines when run episodes, visitations, or administrative entries intervene. For the most part, though, it’s one continuous path from a child’s first removal placement to their last.
Here is an example. Below is the child with the second most number of placement entries in the database: 286 lines out of a million. He was removed twice: once in 2009 and once in 2011. He spent 1,211 days in institutions, 678 with relatives, 543 in group homes, and 201 on run. Only 22 days were spent in foster care. This child averaged a new placement every 9.3 days, and was moved over 3,700 miles from placement to placement, back and forth along the I-4 corridor. He had approximately 817 DCF roommates over the years and his last entry in the database was a juvenile facility in Orange County. He’s probably long gone now.
Every subsequent placement dot on the map means another “you can’t stay there anymore.” It means leaving your things behind or taking what you can carry. It means a new house with new people and rules, including other foster kids who may have already staked out their territory. You have to learn a new way to turn on a shower and hope there’s a toothbrush for you if you didn’t bring one. You get a new time to eat dinner, go to bed, and wake up — and if you don’t adjust fast enough then you might get kicked out just for that and start all over again. Imagine if you woke up with a different family every 9.3 days for years. That is not okay.
When I was working on this project last year, I showed a former foster youth his map. It was complicated, with lots of dots and lines crisscrossing Florida. He stared at it quietly for a while, looked up and said, “I remember every one of those places.” He asked me to print it out, and now he keeps it in a folder and takes it out when he wants people to understand what foster care was like for him.
I’m publishing the maps for 17,305 anonymized kids using Tableau. Instead of showing all 280,000 children’s maps, I’ve instead created groups of children by notable categories. For best results, I suggest opening it on a computer or tablet and hitting the full-screen button in the bottom right corner. If you’re interested in the details on each category or the database itself, there are tabs at the top of the Tableau with more information.
Below is a list of the categories you can view using the drop-down menu on the Tableau. I’ll do write-ups on them later, but I hope you will take the time to explore through the maps and imagine what life was like for the kids in these groups. It’s important to note that most kids in foster care have 3 or fewer placements and reach permanency in reasonable times. Those aren’t the kids we’re looking at here.
10+ Baker Acts
10+ Correctional Placements
Incarcerated Over a Year
Substance Abuse Programs
10+ Night-to-Night Placements
Longest time in care
Top Movers – No Admin
Top Movers – Post Privatization
Mom & Baby Placements
Most Non-relative Placements
Went to Camp
Failed Reunifications (<30 days)
Group Home Dwellers
Before I end, there is one more map below that captures what it can mean to be in foster care. Child 310000648701 came into care on February 11, 2005. We don’t know his exact age (or his gender actually), but his placements are marked as “Traditional 0-5” foster homes. (In 2010 he makes the transition to “6-12” homes, so he has to be on the young side of 0-5 in 2005.) By the end of February, five foster parents had kicked him out. He had two placements in March, three in April, two in May, and only one in June — that one was a non-relative placement and lasted 41 days before they kicked him out, too. Then another placement for two days; then one for one day. Then he was placed with a relative who kept him for 918 days — that’s two and a half years — before the placement ended “in accordance with case plan” (which I think means pursuant to a court order) and he went back to foster care.
He bounced around some more through regular and therapeutic foster homes, landed briefly in a group home in 2010 for eight days of “respite” care, and was finally placed (in entry 42) in a non-relative placement that adopted him after 175 days.
This child had 36 placement providers and only one was a group home. Families kicked him out, again and again, and for much of that time he was under the age of five. He was with relatives for two and a half years without permanency, and then removed presumably by a judge. After six or so years, it ended in adoption, which is good. We can celebrate the adoption while simultaneously asking hard questions about his experience with 34 other families who failed to make that connection or possibly even try.
These maps tell stories that placement stability statistics cannot. Over the next few weeks I’ll share examples and more thoughts on the categories above. I hope they will have the same impact on others as they did on me.
A video of a child being forcibly removed from his mother has been in the news lately. It’s brutal to watch. A group of police officers and security guards yank at the one-year-old while another swings a taser wildly around the room at anyone who gets too close. The woman is on the floor. Her sin is apparently trespassing, i.e,. sitting on the floor instead of standing when there were no seats available in the four-hour line. The charges are later dropped because she was trespassing at a government office (to extend daycare for her child) where people are actually allowed to be, and people sit on the floor in airports all the time and nobody rips their kids from them. The harm’s already done to the child, though. Nobody in the crowd intervenes — they don’t want to get arrested, shot, or lose their place in line — but they document it for the world to see.
(For added absurdity and likely thanks to the word “baby” in the title, the video I watched was preambled with an ad for Zales’s Enchanted Disney collaboration wherein a rich-looking white lady finds a diamond ring on a table, puts it on, and imagines she’s become a Disney Princess®. She has no idea whose ring it even is, but it’s hers now (hey she deserves it). Nobody arrests her for theft and wrestles her baby out of her arms. She does not spend a few nights in jail for having the audacity of self-worth. She, in fact, lives happily ever after.)
Children are removed from their parents every day, and it frequently looks just like that video. It is traumatic for everyone (especially the child) and is supposed to be only for very exigent reasons. In Florida, we know that anywhere from 1,000 to 1,500 kids enter the foster care system each month. The video got me thinking about when those removals happen.
I had a lot of assumptions. I’ve heard that removals go up around the holidays, and that they go down in the summers when kids are home. I felt sure that fewer removals would happen on the weekends, but also that there should be no reason for that because kids would be in more danger when not in school. Removals are supposed to happen when they are unavoidable, not when they are convenient to the investigator.
And I thought: how many kids are actually removed on Christmas? That would be horrible.
As part of a project we are working on, our office came into possession of the entire Florida DCF placement database (anonymized and unforgivably massive). This database includes the removal dates and details of 280,839 kids going back to some who entered care in the 1980s. Looking only at the removals from January 1, 2007 to December 31, 2017, we have data on 156,357 kids. Some of those kids came into care multiple times, so the time period covers 181,799 removals. (Caveat: as with all real-world records, the data is only as good as it is. Given the large numbers here, it is reasonable to assume that errors are evenly spread out and not biased in any given direction.)
How many of those removals were on a holiday?
Since we have the dates for all the removals, it isn’t hard to count them. On a normal, non-holiday day, the average number of removals is about 46. For Christmas, the average is six. For Thanksgiving, it’s eight. Only 67 kids were removed on Christmas Day in Florida from 2007-2017.
There are a few takeaways there. First, holidays seem to suppress removal numbers, at least on the day of the holiday itself. The only holiday that appears above average is Columbus Day; but, with year-to-year variations, it — along with President’s Day and New Year’s Eve — is not statistically different than a normal non-holiday. There is no holiday where more kids are removed than average (though there are periods of the year when removals are up, discussed below).
So what about the theory that more kids are removed around the holidays? If you thought (like I did) that there would be a giant spike in removals before or after, say, Christmas — well, there isn’t. In fact, removals bump up slightly and then start dropping off around the week before Christmas (a.k.a. now). The slight bump isn’t enough to call a correction. A glance at DCF’s dashboard on investigations shows that Decembers are usually high points for closing investigations during the year, so these numbers are even more pronounced. Removals don’t pick up again until January when school is back in session.
Here is the year-round chart. This confirms that removals go down in the Summer and rise again in the new school year. If you thought sentimentality was what kept DCF from removing kids on Christmas, think again. You see similar holiday drops on the 4th of July and Veterans Day. The dips would be more pronounced for MLK Day, Labor Day and Memorial Day, except those holidays don’t happen on the same calendar day each year. Kids don’t get removed on holidays because investigators are on vacation.
So that’s the question: if only 15 kids had to be removed on the 4th of July, why did 48 have to be removed on the 7th of July?
What about the weekends?
On average only 14 kids per day were removed on Saturdays and Sundays. The highest day of the week for removals was Thursday — probably because court is held one day after a removal and nobody wants to go to court on a Saturday, for sure. At 14 removals per day, the weekend was on par with the 4th of July.
And, while we’re at it, what about time of day? Kids get removed during business hours. The later in the day, the higher the number of removals. Below you can see three distinct peaks in the following hours: 9:00am, 1:00pm, and 5:00pm. Only 416 kids in this dataset were removed in the 6:00am hour. Maybe children are safer before dawn? (Those 5,580 kids removed at midnight include entries without a valid date — i.e., “00:00:00”. Don’t read into that spike.) I’m not including a graph because it’s messy, but if you look at the whole week the highest rates of removals happened during the 5:00pm hour on Thursdays. No surprise.
So removals happen all the time, except holidays, weekends, and usually not outside of business hours. And they especially happen when investigators first get to work, after lunch, and right before they go home for the day. Something feels very wrong about that. It’s as if “risk” is also a product of convenience, which is not how child protection is supposed to work.
If the averages hold, around six (and up to twelve) kids will be removed on Christmas Day this year. Let those removals be necessary and kind.
The First DCA published statistics on its caseloads and decisions. But notably (as appellate judges like to say), the length of time they take to resolve cases was not reported. It motivated me to update the How Long do Appeals Take tableau.
The answer? Probably 120 to 170 days for a Dependency case, 260 to 576 days for a Criminal case, and 345 to 603 days for a Civil case.
The Florida Supreme Court issued a child welfare opinion, so it seemed like a good time to do a case law review. Here we go…
What guardians ad litem don’t do
The Florida Supreme Court addressed the issue of whether the appointment of a guardian ad litem on a dependency case tolled the statute of limitation on a civil suit based on injury to the child. The answer: no, because guardians ad litem have no authority to file civil suits.
Twin children were (allegedly) injured at the hands of the biological mother while she was under the care of developmental disability support services. The children entered foster care and were eventually placed in a permanent guardianship with their grandparents. When the grandparents learned of the injuries, three years after being appointed permanent guardians, they filed suit against the agencies involved. The agencies moved to dismiss the suit as barred by the four-year statute of limitations.
The Florida Supreme Court held that, because a dependency guardian ad litem has no authority to file a separate lawsuit on behalf of a child, the four-year limitations period for filing a suit was tolled until the child’s grandparents were appointed as permanent guardians. D.H. v. Adept Cmty. Services, Inc., SC17-829, 2018 WL 5660595 (Fla. Nov. 1, 2018).
For all you Latin lovers, the term “ad litem” translates to “for the suit.” Litem is the accusative singular of the word lis, which is a feminine third declension noun that means lawsuit, quarrel, or strife. It is the root of the English words litigate, litigation, litigious, you get the point. It is the actual word in the legal phrase lis pendens, which means “suit pending.” If you’re appointed to multiple lawsuits for the same person/child and want to be grammatically snobbish, you could call yourself the guardian ad lites. I wouldn’t.
What guardians ad litem do do
Apparently they prevent DCF from falling on swords. The Fifth District affirmed the TPR of parents despite DCF’s concession of error. The GAL Program argued the TPR for material breach of the case plan was proper. The Fifth held that the trial court properly considered a father’s lack of efforts on a previous case plan when determining whether he was likely to comply with the latest one. W.D. v. Dep’t of Children & Families, 5D18-2241, 2018 WL 5603041 (Fla. 5th DCA Oct. 5, 2018).
The district courts are not very consistent on the question of what constitutes substantial compliance. The Fourth District affirmed the TPR of a mother who argued that she had remedied the circumstances that brought the case in by finding appropriate housing. The operative holding is that “[m]ere completion of services is not equivalent to substantial compliance with a case plan.” A bit of explanatory dicta expands that:
Moreover, it is important to acknowledge that what is initially recognized as a cause for sheltering children is more often than not a symptom of a larger underlying problem which—by definition—must be addressed. Here, the mother’s poor decisions relating to the children prior to DCF intervention were a side effect of her own trauma-based issues. This much was evident by the time the case plan was established. Further, the evidence at trial established as much in that the psychologist transparently linked the mother’s mental health issues with her ability to parent.
The Third District affirmed the denial of TPR and grant instead of a dependency ruling. The appeal arose out of an egregious conduct TPR petition against parents of a child with severe insulin-dependent diabetes. The Third affirmed, noting the trial court’s finding that it had to “consider the possibility that this is a case of a lack of clear communication with the medical professionals perhaps due to a language barrier or a case of some other complex mental health issue that was left unexplored by the Department of Children and Families.”
As an interesting side-note, the trial court permitted an attorney for the hospital where the child was treated to be present during the trial. The hospital’s doctors testified in the trial. The Third held that this was error, but harmless in this case. Termination proceedings are closed to the public.
And finally a bunch of cases that could have easily been avoided.
The Third District reversed an adoption by foster parents in which DCF failed to give grandparents notice. The grandparents had been actively involved in the case and — it appears — were left off of the filings that would give them notice that an adoption petition would be considered by the court. Berenyi v. Florida Dep’t of Children & Families, 3D18-922, 2018 WL 5624250 (Fla. 3d DCA Oct. 31, 2018).
The Third District reversed a shelter order that was entered at a hearing without the parent’s counsel present. M.K. v. Dep’t of Children & Families, 3D18-1802, 2018 WL 4761795 (Fla. 3d DCA Oct. 3, 2018).
How long do dependency appeals take?
I’ve been keeping track of how long appeals take for a few months and the numbers are very stable. Half of dependency appeals I’ve reviewed took 124 days or less from filing to opinion. Most cases resolved in month 4 or 5. If your appeal goes into month 6, it’s getting long. By comparison, half of non-dependency cases took 292 days or less, with a much wider range. Only 9% of dependency appeals won, compared to 13% of appeals in general.
In July of this year, the Florida foster care system did something unseen since February 2014: it shrunk. For the first time in over 50 months, the year-over-year (YOY) change in out-of-home care numbers went down by 45 children. By August it was down 118, and the reports out this month for October show a contraction of 165.
While a reduction of 165 kids does not seem like much in a system with over 24,000 children in it, the slowing actually started in January 2016, when the system was growing at a staggering YOY rate of 2,540 kids. Just the month before held a YOY increase of 2,683 — the fastest growth since data is available in 2003.
There’s no official definition of a contraction period, or any way to tell if one is real or a blip. I actually sat on this post for a few months to make sure the trend was stable — we’ve had hurricanes, elections, resignations, and other unusual events recently, so I wanted to let those pass.
In reality, though, changes from positive to negative OOHC growth (expansions to contractions) do not happen quickly and appear largely driven by intentional policies and not outside events. The expansion under Secretary Hadi in 2004-2006 lasted 19 months and ended abruptly with the Secretary’s resignation from office in December 2006. The subsequent contraction during the Butterworth and Sheldon administrations lasted 50 months and never wavered until three months into Secretary Wilkins’ term. That change of direction occurred in March 2011, right in the middle of the public hearings and media frenzy on the Barahona case, though the contraction had been slowing since 2009 and was well on the way to reversing course even without the public outrage to speed it along. (That is, media frenzy tends to reinforce — not set — existing child welfare policy positions.)
Oddly, Secretary Wilkins’ DCF changed its expansionary course by August 2012 and entered a contraction period that continued sharply until the month that he resigned in July 2013. (I’ve never heard a good explanation for that period.) The tide immediately turned back toward expansion, continuing through Interim Secretary Jacobo and halfway through Secretary Carroll’s tenure. Growth peaked in December 2015 and then precipitously fell, flattened, and then fell again. (Note that steady growth is still growth — the chart above shows change. The charts below show the actual counts.)
Even though the system as a whole tends to move in unison, not every geographic area shifts course at the same time. The current contraction has been driven largely by sharp decreases in OOHC in three circuits — 17 (Broward), 11 (Miami), and 18 (Brevard) — which shrunk a total of 670 children over the previous year in October. The top three growth circuits — 1 (Pensacola), 7 (DeLand), 9 (Orange/Osceola) — only grew by 127 kids in all.
Decreases were clustered largely, but not exclusively, in the southern regions. Here are the changes by county.
The contractions appear driven largely by reductions in removals. (I’ve chosen to use seasonal trends below to make the changes over time more clear. The actual numbers for removals and discharges have large but regular oscillations month to month due to seasonal effects like summer and national adoption day. The raw numbers are much harder to read.)
You can see the same decreases in removals statewide. Here’s the statewide seasonal trend graph.
Here are the seasonal trend graphs for all circuits. If you notice anything interesting or know why any of these charts look the way they do, let me know.
Sad to learn of the news of George Sheldon’s passing today. He was unassuming yet tough; determined and passionate. Our hearts are heavy. We lost a mentor, teacher, advocate and dear friend. I will miss the time we spent together. He was taken too soon. pic.twitter.com/TIyxOwWj7n
———- Forwarded message ———
From: Alejandro Alamo <email@example.com>
Date: Thu, Aug 23, 2018 at 10:19 PM
Subject: URGENT MESSAGE FROM KEITH WARD AND MICHAEL WILLIAMS
It is with great sadness that we inform you of the passing of George Sheldon earlier this evening at Mount Sinai Hospital.
This is a devastating loss for all of us; and a time for us to mourn together our good friend, George.
Below, please see a statement from George’s family and friends which will be released later this evening. Tomorrow morning, we will be prepared to meet with staff and share this sad news; which will require us all to be strong as they receive this news.
In times of grief like this, it is nearly impossible to include every individual who appropriately deserves to be on this message. For those who we forgot to include, please forgive our oversight. We encourage you to distribute this message.
George H. Sheldon
George H. Sheldon, an attorney, long-time Florida public servant and president and CEO of Our Kids of Miami-Dade/Monroe, a child placement agency, died Thursday, Aug. 23, at the age of 71, his family announced.
Mr. Sheldon’s public service included appointments as Florida Deputy Attorney General, Secretary of the Florida Department of Children and Families, and Assistant Secretary at the United States Department of Health and Human Services, Administration for Children and Families. He also served as a member of the Florida House of Representatives.
He died peacefully at Mt. Sinai Hospital in Miami Beach surrounded by family and friends following post operative complications due to a neck injury he sustained while exercising.
A statement by the family said: “George was beloved as a brother and uncle and we are greatly saddened by his loss and we will miss him dearly.”