2021-22 Kindergarten Placement Data
When Ronald Taylor and the Board of Education have presented how the III operates, they've emphasized how it uses proximity when it makes assignments, and that's true. In 2021-22, in fact, 60% of kindergarteners were assigned to their closest school.
HOWEVER, having 60% of kids at their closest school means that 40% of kids are not at their closest school, and sometimes those assignments can be up to 3.8 miles away, and stressful, difficult for work schedules, and dangerous if a child must walk or bike.
This is an analysis of how the Taylor-Maini algorithm works in real life, using home-to-school distance data for 2021-22 kindergarteners. My analysis overlaps with Michael Alves' own "Year 1 Recap Report," but focuses on students with the greatest logistical hardship and the most irrational placements. In sharing these data I want to underscore the importance of a student's Random Number, since it's the random number that determines placement. The BOE & Admin always call the system an "algorithm" but at its heart it is a lottery.
Of the 130 kids within 0.5 miles, 70 were placed by the algorithm. The others were placed via sibling preference, special education, or ELL.
Of those 70 algorithmic placements, 43, or 61%, were placed at the nearest elementary school, which almost exactly the 40% of algorithmic-placement kindergarteners overall who were assign to the closest school.
Ronald Taylor and BOE members have emphasized how the algorithm uses proximity to assign students, but I think what they have said is incomplete and therefore inaccurate.
Annemarie Maini says that the pre-III system for over-capacity placements was "arbitrary" but the district also tried to place those kids at the nearest school that had a seat available, and transfers were possible then after a single year.
Further, the Taylor-Maini algorithm relies on random numbers to place kids, and random numbers, by definition, are arbitrary.
1. Being extremely close to an elementary school does NOT make a child more likely to be placed there.
The SOMSD had 130 kindergarteners who lived 0.5 miles or less from an elementary school, of whom 35 were 0.25 miles or less.
Of the 130 kids within 0.5 miles, 70 were placed by the algorithm. The others were placed via sibling preference, special education, or ELL.
Of those 70 algorithmic placements, 43, or 61%, were placed at the nearest elementary school, which almost exactly the 40% of algorithmic-placement kindergarteners overall who were assign to the closest school.
Ronald Taylor and BOE members have emphasized how the algorithm uses proximity to assign students, but I think what they have said is incomplete and therefore inaccurate.
EG, as Annemarie Maini told the Village Green in August 2022:
The integration initiative actually takes proximity into consideration in the algorithm, so it is more fair to everyone equally than the previous system, which gave some people their zoned school and others an arbitrary school only because of space availability, and went against our community’s values by leading to segregation. Now we have a system that expresses our values by achieving integration while taking proximity into consideration.”
But what the Taylor-Maini algorithm assigns students to the (linearly) nearest school with an available seat at their diversity tier when their turn to be processed comes up, and their order in processing is determined by their random number. I take Maini's statement to imply that the algorithm uses proximity so that the closer a student is to a school, the more likely he/she is to be placed there, but that is not accurate and it is irrelevant for bad-number children.
Thus, the community has a misunderstanding of how proximity is used, where people think "the more proximate the more likely to be assigned." This common error was repeated by former BOE candidate Ritu Pancholy, who said at the Hilton debate. (See minute 41)
I've met parents who live literally across from the Clinton Elementary School and are being assigned across town. I don't understand the algorithm, proximity was supposed to be a factor. It does appear that the algorithm is kinda of getting some things not right... We need to hold the superintendent and the consultant [Michael Alves] accountable for ensuring that the algorithm uses proximty as a factor. I don't understand how you can live across from the school and be the parent and faced with this choice."
But the problem is the algorithm itself. Not that it is being incorrectly applied.
A student who is a tenth of a mile from the nearest school is just as likely to be sent to that school as a student who is a full mile from the nearest school. If a student is a tenth of a mile from their nearest school and their bad random number prevents them from being placed there, they will go to a school 1.5-3.5 miles away, which is worse than the pre-III system.
If you look at this table, five kids who were under 0.5 miles from a school were sent to schools over three miles away. Four of those kids lived by Seth Boyden and were sent to South Mountain. A fifth, who only lived 0.3 miles from South Mountain, was sent to Seth Boyden. A pair of students who were 0.2 miles from Clinton were sent to Marshall.
Annemarie Maini says that the pre-III system for over-capacity placements was "arbitrary" but the district also tried to place those kids at the nearest school that had a seat available, and transfers were possible then after a single year.
Further, the Taylor-Maini algorithm relies on random numbers to place kids, and random numbers, by definition, are arbitrary.
The role of a student's random number has been concealed by Taylor and the BOE. There is only a single SOMSD document that mentions it, the February 2021 Michael Alves memo, which only came out after the zero-choice assignment system was approved.
For instance, Taylor said in April 2021 in a letter to parents:
the District will be utilizing a modified Berkeley Approach for our Integration methodology, which includes the development of an algorithm that creates micro-neighborhoods and utilizes key variables:
● Parental education level;
● Parental income;
● Race;
● Sibling Preference; and,
● Proximity
Notice that the importance of a student's random number and the lottery-like reality of the Taylor-Maini algorithm is never disclosed.
2. There are actually very few students in the high-stress 1.5-1.99 mile distance (27) and not many in the 1.0-1.49 mile distance (88), so the failure to bus at least the kids over 1.5 miles was due to the BOE not wanting to bus them and not budgetary necessity.
2. There are actually very few students in the high-stress 1.5-1.99 mile distance (27) and not many in the 1.0-1.49 mile distance (88), so the failure to bus at least the kids over 1.5 miles was due to the BOE not wanting to bus them and not budgetary necessity.
Susan Bergin, semi-secretly, shepherded through the elimination of bussing for elementary schoolers below 2.0 miles distance saying that it would "break the budget" if we did not.
The Board of Education's elimination of "courtesy" bussing was unconditional: it made no attempt to evaluate the sub-2.0 elementary school families who could not drive their children, or whose distances were the closest to 2.0 miles. And, worst of all, the BOE did not follow-through on hazardous route designation or inform staff or parents of the impending change. As we've seen from the start of the school year, the busses that are running are mostly empty.
Bergin was wrong about the "break the budget" argument because there were only 27 kindergarteners placed at schools 1.5-1.99 miles away. Of those 27, only seven were between 1.9-1.99 miles away, the "ultra-high stress zone." Since there are so few kids at this distance, bussing them was/is affordable.
Moreover, most of the elementary schoolers placed 1.5-1.99 miles away were at Marshall and Clinton, for which we already have bussing networks. If the BOE and Taylor had cared, they could have arranged bus stops for those kids on busses that were carrying >2.0 mile and ESL students. The BOE could also have granted transfers to nearer schools, to further reduce costs.
The relatively small number of kids in the high-stress zone is a credit to the Taylor-Maini algorithm, but the failure to think of their transportation is yet another marker of how the BOE is either ignorant of implementation or it did know and was completely indifferent.
Thair Joshua claimed in July 2022 that the BOE talked about providing busses for 1.75-2.0 mile kids. There were only 13 2021-22 kindergarteners and presumably a small number of 2022-23 kindergarteners. Thair's statement that the district is unable to transport at least 1.75-2.0 mile kids cannot be true. It is a lie.
3. The characteristics of kids traveling over three miles.
Thair Joshua claimed in July 2022 that the BOE talked about providing busses for 1.75-2.0 mile kids. There were only 13 2021-22 kindergarteners and presumably a small number of 2022-23 kindergarteners. Thair's statement that the district is unable to transport at least 1.75-2.0 mile kids cannot be true. It is a lie.
3. The characteristics of kids traveling over three miles.
The Year 1 Recap Report revealed that 5% of kindergarteners who were algorithmically assigned had home-school distances over three miles. It also revealed that 11% of Black students had this distance, 6% of Asians, 4% of whites, 3% of Latinos, and 0% of multiracial students.
What I can add to the Year 1 Recap Report is likely a confirmation based on what we assumed from SOMA geography, which is that these long-distance journeys are mostly between South Mountain-area kids and Seth Boyden-area kids, with one exception of a child who is closest to Tuscan but is sent to South Mountain.
I don't include race in the table, five of the kids who live near Seth Boyden and are sent to South Mountain are white, which is contrary to the integrative goal of the III, due to how the algorithm actually uses SES, not race.
I do not include sibling placements in this analysis, but there were sibling-placement kids who live three+ miles from Seth Boyden who were placed at Boyden, so the family must have originally voluntarily opted in to Boyden. This underscores that some families do not object to long distances. Since families differ in preferences/needs, a better integration algorithm would use parental ranking instead of the Taylor-Maini proximity assumption.
Comments
Post a Comment