0% found this document useful (0 votes)
43 views15 pages

Child Care Project Report

This project develops a linear integer programming model to optimize the allocation of resources for eliminating child care deserts in New York State, focusing on increasing the availability of child care slots, particularly for children under five. The model considers various constraints such as budget limitations, distance requirements, and fairness in access across regions to propose scalable solutions for expanding existing facilities and constructing new ones. By utilizing data on population, employment, and income, the project aims to provide a strategic framework for addressing child care shortages and enhancing social well-being statewide.

Uploaded by

abhihere019
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
43 views15 pages

Child Care Project Report

This project develops a linear integer programming model to optimize the allocation of resources for eliminating child care deserts in New York State, focusing on increasing the availability of child care slots, particularly for children under five. The model considers various constraints such as budget limitations, distance requirements, and fairness in access across regions to propose scalable solutions for expanding existing facilities and constructing new ones. By utilizing data on population, employment, and income, the project aims to provide a strategic framework for addressing child care shortages and enhancing social well-being statewide.

Uploaded by

abhihere019
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Eliminating Child Care Deserts in New York State through Optimization: A Case Study of

New York State's Expansion Plan of Available Child Care Slots


Shiyun Zhao, Yiran Ge, Fanxi Wang, Abhinaya Anil Menon
Professor Yaren Bilge Kaya
IEOR 4004 Sec 2
1 November 2024

Abstract: This project addresses the problem of child care deserts in New York State (NYS) by
utilizing linear integer programming (LIP) with the Gurobi optimization solver to develop a
dynamic model for determining the optimal allocation of resources to eliminate these deserts. The
model adjusts the required number of child care facilities and slots based on regional demand,
including population, employment rates, and income levels, while considering the cost structure
for both expanding existing facilities and building new ones. By factoring in complex constraints,
such as distance limitations between facilities and the increasing cost of expansion, the model
suggests scalable solutions that prioritize the creation of new facilities and the expansion of
existing ones in a cost-effective manner. This approach provides a strategic framework for
addressing the child care shortage in NYS while maintaining fairness and maximizing social
well-being across regions.
Keywords: child care deserts; resource allocation; facility expansion; social coverage index

1.1 Introduction
1.1.1 Introduction
Access to affordable, high-quality child care is crucial for child development and workforce
participation, yet many areas of New York State (NYS) face significant shortages in child care
availability. These "child care deserts," where demand far exceeds supply, are a growing concern,
particularly for children under five, whose early childhood experiences are critical for long-term
development (Adams & Henly, 2020). In these regions, working parents often struggle to find
adequate care options, which can hinder their ability to fully engage in the workforce. To address
this issue, the state of New York aims to eliminate child care deserts by increasing the capacity of
child care facilities across the state. The focus is especially on children under five, given their
heightened need for early care and education. NYS plans to achieve this goal through a
combination of expanding existing facilities and constructing new ones, but these efforts must be
carried out within a series of constraints, including a $100 million budget, distance limitations
between facilities, and fairness requirements to ensure equitable access across regions.
This project seeks to help NYS meet these goals by developing an optimal strategy for resource
allocation. The goal is to increase the availability of child care services in underserved areas while
ensuring fair access across regions, prioritizing the needs of younger children, and remaining
within the allocated budget.

1.1.2 Background
Child care is essential for both child development and family economic stability. It not only
supports children’s cognitive and social growth but also enables parents, particularly mothers, to
participate in the workforce, enhancing family financial well-being (Bradley & Vandell, 2007).
However, many families in New York State (NYS) and across the U.S. struggle to access
affordable, high-quality child care due to widespread "child care deserts." These are regions where
the demand for licensed child care far exceeds the available supply, a situation affecting over half
of all children under five in the U.S. (Adams & Henly, 2020; Center for American Progress, n.d.).
In NYS, these shortages disproportionately affect low-income and rural communities,
exacerbating social inequalities and limiting workforce participation.
To address this challenge, NYS has committed to increasing the capacity of child care facilities,
focusing especially on children under five. The state aims to expand existing facilities and build
new ones, prioritizing these younger age groups due to their critical developmental needs.
However, this expansion must be carried out within several constraints, including a $100 million
budget, distance requirements to prevent over-concentration, and fairness criteria to ensure
equitable access across regions. Specifically, the state mandates that the difference in child care
coverage between any two regions should not exceed 0.1, ensuring no area is left significantly
underserved.
This project aims to support NYS in meeting these goals by developing optimization models to
guide resource allocation. These models take into account the costs of expanding existing facilities
and constructing new ones, as well as constraints such as facility distance and fairness in
coverage. The models will help identify the most efficient and equitable ways to eliminate child
care deserts, with a focus on maximizing child care access for younger children, thereby providing
NYS policymakers with a data-driven strategy to close child care gaps and improve access
statewide.
2. Method
2.1 Model Assumptions
Facility Construction Locations: New child care facilities can be constructed at any available
location within New York State, subject to a minimum distance requirement of 0.06 miles between
any two facilities (new or existing).
Expansion Limits: Existing facilities can be expanded up to a maximum of 1.2 times their current
capacity, with a cap of 500 additional slots per facility. The model assumes that this expansion is
possible at any existing facility that has the physical space and operational capacity to
accommodate growth.
Cost Structure: The cost of expanding a facility is based on the size of the expansion. Costs are
linear within defined brackets (small, medium, and large expansions), with progressively higher
costs per slot as the expansion scale increases. Specifically, expansion costs rise sharply after the
10%, 15%, and 20% thresholds are reached.
Priority for Younger Children: Given the developmental needs of children under five, the model
prioritizes increasing child care slots for this age group. Additional costs for specialized
equipment are assumed for slots catering to children aged 0-5.
Fairness Constraint: The model assumes a fairness requirement that the disparity in the ratio of
child care availability to the child population between any two regions should not exceed 0.1.
Budget Constraint: The total budget for all construction and expansion activities is limited to
$100 million, which includes both new facility construction and the expansion of existing
facilities.
Social Coverage Index: The model seeks to maximize the social coverage index, which weighs
child care availability for children under five more heavily than for older age groups, reflecting the
greater importance of early childhood care.
No Over-Concentration: The model assumes that no two facilities (new or existing) can be
located within 0.06 miles of each other, to prevent over-concentration in specific regions.
These assumptions help to define the scope of the problem and establish a feasible framework for
resource allocation, guiding the development of optimal strategies for addressing child care
deserts in New York State.

2.2 Dataset
2.2.1 Child Care Data (new_child_care.csv)
This file contains information about existing child care facilities, including their capacities (infant,
toddler, preschool), types, locations (latitude and longitude), zip codes and specific IDs.
It’s necessary to extract and categorize capacity information for different age groups and aggregate
the data based on zip codes to assess existing resources in each area. This data will inform
decisions on where additional facilities or expansions may be required.

2.2.2 Employment Data (new_employment.csv)


This file includes employment rates for various areas, broken down by zip codes.
It’s necessary to integrate this data to analyze demand levels, as employment rates can be
correlated with child care demand. Areas with higher employment rates might require more child
care support.

2.2.3 Income Data (new_income.csv)


This file provides average income data by zip codes, which can help in determining demand levels
across regions.
This data can be used to identify regions with lower average incomes, as they have a higher need
for subsidized child care. This will help in the allocation of child care resources fairly based on
financial need.

2.2.4 Population Data (new_population.csv)


This file includes population statistics segmented by age groups belonging to different zip codes,
helping to estimate the number of children in specific age brackets who need care.
It’s necessary to calculate the estimated demand for child care based on the population data for age
brackets.
2.2.5 Potential Locations Data (new_potential_loc.csv)
This dataset contains information about possible new child care facility locations, including
latitudes, longitudes and zip codes. Potential capacities are informed as well.
This dataset can be used to identify viable new locations based on unmet demand. It will help
determine where to prioritize new facilities or expansions to meet demand and reduce "child care
deserts."
2.3 Data Preprocessing
The Data Preprocessing stage in this project prepares datasets to support the optimization model
for budgeting and capacity planning for child care facilities. Here’s a detailed explanation of what
each step accomplishes, based on the objectives of the problem and the model formulation:

2.3.1 Pre-Calculation
[Link] Child Population Calculation
The total number of children requiring care is estimated by summing up relevant age groups
in the population data. The child population is scaled based on age-specific proportions to
reflect child care needs, especially for ages 0-12.
To estimate the number of children from 2 weeks to 12 years old by proportion, we multiply
the total population under 14 by 13/15.

[Link] Current Capacity Calculation


a. Current Capacity for 0-5 ages Children
In order to have a clear picture of existing slots available for children under 5 to
compare to the population’s needs, a new capacity column (-5_capacity) is created by
summing the existing capacities for infants, toddlers, and preschool-aged children from
Child Care Data.
b. Current Capacity for 5-12 ages Children
From child care data, we already know the total capacity for each facility in each zip
code.
By grouping the data via zip code, we can aggregate the total existing capacity in each area,
simplifying the capacity constraint checks during optimization.

2.3.2 Demand Indicator


By merging employment dataset and income dataset, it is possible to identify areas with different
levels of demand in child-care facilities using thresholds Employment rate ≥ 0.6 or Average
income < 60,000, which allows identification of zip codes where the need for child care services
may be higher. The value 1 is assigned to that zipcode if it is high-demanded, 0 if not.

2.3.3 Desert Identification


Zip codes with insufficient childcare slots are flagged as "deserts" with a value 1 if:
Zip code belongs to high-demand area, child care capacity ≤ 1/2 of need in the zipcode;
Zip code belongs to low-demand area, child care capacity ≤ 1/3 of need in the zipcode;
If zip codes already meet the minimum required number of total slots, a value 0 is assigned. This
aligns with the constraint in the formulation that targets underserved areas, focusing resources on
zip codes that meet the "desert" conditions.

2.3.4 Insufficient Slots Indicator for Ages 0-5


A “0-5 slot not satisfied” column has a value 1 indicating areas where capacity for ages 0-5 falls
below two-thirds of the local population under 5, 0 if that area has an adequate number of slots for
children under 5.

2.3.5 Calculating Additional Slots Needed


a. Additional slots needed for the whole 0-12 ages children
New columns estimate the additional slots needed to meet minimum requirements: calculates
the total slot shortfall for high-demand desert zip codes and low-demand desert zip codes.
b. Additional slots needs for the 0-5 ages children
-5 slots needed focuses on deficits in care slots for children under 5, directly supporting the
optimization constraint related to 0-5 capacity.

2.3.6 Final Datasets:


After finishing the whole data preprocessing, we put all needed information into two dataset:
a. temp2
Including “Desert Identification”, “Insufficient Slots Indicator for Ages 0-5
”, “Additional Slots Needed” for each zip code.
b. cr_filter
Including the whole information about capacity for each facility in each zip code.
2.4 Model Construction
2.4.1 Problem 1: Budgeting
[Link] Decision Variables
𝑖∈𝑧𝑖𝑝𝑐𝑜𝑑𝑒: a zipcode which a particular facility belongs to.
𝑗∈𝑓𝑎𝑐𝑖𝑙𝑖𝑡𝑦: facilit y_id in the new_child_care sheet.
𝐸𝑥𝑝𝐵𝑖𝑗: # of expansion slots of 0-5 years-old children of facility 𝑗 in zipcode 𝑖.
𝐸𝑥𝑝𝐴𝑖𝑗: # of expansion slots of above 5-years-old children of facility 𝑗 in zipcode 𝑖.
α𝑖𝑗: A binary indicator variable for each facility 𝑗 in zipcode 𝑖.

α𝑖𝑗 = 1 if 𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗≥1, meaning expansion of facility 𝑗 in zipcode 𝑖 took place.

α𝑖𝑗 = 0 if 𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗 = 0, meaning expansion of facility 𝑗 in zipcode 𝑖 didn’t take place.

𝑆𝑖: # of small facilities newly built in zipcode i.

𝑀𝑖: # of medium facilities newly built in zipcode i.

𝐿𝑖: # of large facilities newly built in zipcode i.

[Link] Model Constraints


Available Slots Constraints:
When expanding existing facilities, the maximum increase in capacity is 1.2 times the current size,
while the total number of slots cannot be larger than 500 if that facility is to be expanded. In order
to make sure that the number of increasing slots and total slots meet the requirements, we limit the
maximum number of increasing slots and total slots, as shown below:
(1)​ 𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗≤0. 2𝐸𝑥𝑖𝑖𝑗, ∀𝑖, 𝑗

(2)​ (𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗 + 𝐸𝑥𝑖𝑖𝑗)·α𝑖≤500, ∀𝑖, 𝑗

In which new variable is introduced that can be obtained from dateset:


𝐸𝑥𝑖𝑖𝑗: # of existing slots of facility 𝑗 in zipcode 𝑖.

Binary indicator variable α𝑖 is also introduced, which implies that we ignore cases where the 𝐸𝑥𝑖𝑖𝑗

is greater than 500 if the expansion of facility 𝑗 in zipcode 𝑖 did not occur.
Besides, in high-demand areas, an area is considered a child care desert if the number of
available slots is less than or equal to half the population of children aged two weeks to 12
years. In normal-demand areas, an area is classified as a child care desert if the available slots
are less than or equal to one-third of the population of children within the same age range. In
order to make sure that the number of available slots in both high-demand areas and
normal-demand areas meet the requirements, we limit the minimum number of total available
slots, as shown below:
1
(3)​ ∀𝑖∈𝐻𝑖: ∑(𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗 + 𝐸𝑥𝑖𝑖𝑗) + 100𝑆𝑖 + 200𝑀𝑖 + 400𝐿𝑖 ≥ 2
𝑃𝑜𝑝𝑖
𝑗

1
(4)​ ∀𝑖∉𝐻𝑖: ∑(𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗 + 𝐸𝑥𝑖𝑖𝑗) + 100𝑆𝑖 + 200𝑀𝑖 + 400𝐿𝑖 ≥ 3
𝑃𝑜𝑝𝑖
𝑗

Beyond addressing overall demand, the NYS government ensures that children under the age of 5
have sufficient access to care. In order to make sure that the number of available slots for children
aged 0-5 must be at least two-thirds of the population of children aged 0-5, we set the minimum
2
number of total available slots for children aged 0-5 to 3 ·𝑃𝑜𝑝𝐵𝑖, as shown below:
2
(5)​ ∑(𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑖𝐵𝑖𝑗) + 50𝑆𝑖 + 100𝑀𝑖 + 200𝐿𝑖 ≥ 3
𝑃𝑜𝑝𝐵𝑖, ∀𝑖
𝑗
(6)​ 𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗≤𝑀×α𝑖𝑗, 𝑖∈𝑧𝑖𝑝𝑐𝑜𝑑𝑒, 𝑗∈𝑓𝑎𝑐𝑖𝑙𝑖𝑡𝑦, 𝑀 = 1, 000, 000
This constraint ensures that the expansion slots 𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗 are only non-zero if the binary
variable α𝑖𝑗 = 1. If α𝑖𝑗 = 0, this constraint forces 𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗 to be zero. The constant 𝑀 is
a large number (set to 1,000,000).
α𝑖𝑗
(7)​ 𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗 ≥ 𝑀
, 𝑖∈𝑧𝑖𝑝𝑐𝑜𝑑𝑒, 𝑗∈𝑓𝑎𝑐𝑖𝑙𝑖𝑡𝑦, 𝑀 = 1, 000, 000
This constraint ensures that if α𝑖𝑗 = 1, the expansion slots 𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗 must be at least a
minimal non-zero value. When α𝑖𝑗 = 0, this constraint does not enforce any additional restriction
on 𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗, allowing it to be zero.
(8)​ α𝑖𝑗∈{0, 1}

(9)​ 𝐸𝑥𝑝𝐴𝑖𝑗, 𝐸𝑥𝑝𝐵𝑖𝑗≥0


(10)​ 𝑆𝑖, 𝑀𝑖, 𝐿𝑖≥0

[Link] Optimization Model


To determine the minimum amount of funding (in total) needed to meet their target for each area,
the optimization model for minimum total funding can be ultimately formulated as an objective
function shown as below:
𝑀𝑖𝑛(
𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗
∑[ 𝐸𝑥𝑖𝑖𝑗
(20, 000 + 200𝐸𝑥𝑖𝑖𝑗)] + ∑(100𝐸𝑥𝑝𝐵𝑖𝑗) + ∑(65, 000𝑆𝑖 + 95, 000𝑀𝑖 + 115, 000𝐿𝑖)
𝑖,𝑗 𝑖,𝑗 𝑖

)
There are three main cost components: The first term represents the expansion cost per facility slot
for each facility 𝑗 in zipcode 𝑖; the second term represents the direct expansion cost specifically for
0-5-year-old slots in each facility; the third term represents the fixed construction cost for building
new facilities.

[Link] Output
a. Expansion
Total slots expanded for children under 5: 102,887.
Total slots expanded for children aged 5–12: 4,143.
b. New Construction
Small Facilities:
Number built: 281.
Slots created: 28,100.
Medium Facilities:
Number built: 261.
Slots created: 52,200.
Large Facilities:
Number built: 2,267.
Slots created: 906,800.
c. Total Costs:
The total budget spent: $353,774,438.68.
The breakdown indicates that new construction costs significantly outweigh expansion costs.
Figure 1. Comparison of Different Facility Costs in Problem 1

2.4.2 Problem 2: Realistic Capacity Expansion and Distance


[Link] Decision Variables
Newly Defined Decision Variables:
𝑘∈𝑧𝑖𝑝𝑐𝑜𝑑𝑒: The potential location where new facility can be built in 𝑧𝑖𝑝𝑐𝑜𝑑𝑒.
β1, β2, β3: Binary indicator variables represent distinct levels of facility expansion relative to the

existing capacity in a facility.


𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗
β1𝑖𝑗={1, 𝑖𝑓 𝐸𝑥𝑖𝑖𝑗
∈(0, 10%] 0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗
β2𝑖𝑗={1, 𝑖𝑓 𝐸𝑥𝑖𝑖𝑗
∈(10%, 15%] 0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗
β3𝑖𝑗={1, 𝑖𝑓 𝐸𝑥𝑖𝑖𝑗
∈ (15%, 20%] 0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

[Link] Model Constraints


a. Newly introduced constraints:
𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗
𝐸𝑥𝑖𝑖𝑗
≤0. 1×β1𝑖𝑗 + 0. 15×β2𝑖𝑗 + 0. 2×β3𝑖𝑗
𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗
𝐸𝑥𝑖𝑖𝑗
≥0. 1×β2𝑖𝑗 + 0. 15×β3𝑖𝑗
β1, β2, β3∈{0, 1}
β1𝑖𝑗 + β2𝑖𝑗 + β3𝑖𝑗 = 1
The first constraint ensures that if a particular β variable is set to 1, the left-hand ratio does not
exceed the corresponding threshold (10%, 15%, or 20%).
The second constraint sets a lower bound for the ratio. If β2𝑖𝑗 = 1, the ratio must be at least 10%;
If β3𝑖𝑗 = 1, the ratio must be at least 15%.
The final constraint ensures that only one of the β variables can be 1 at any given time, thus
𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗
assigning the ratio 𝐸𝑥𝑖𝑖𝑗
to only one of the defined intervals: (0%, 10%], (10%, 15%], or
(15%, 20%].
b. Distance Requirement
Since the minimum distances between facilities is 0.06 miles, there are two separate groups we
need compare:

1.​ The distance between each potential location and each existing child care center in each
zip-code area:
( )( )
𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 = ℎ𝑎𝑣𝑒𝑟𝑠𝑖𝑛𝑒[ 𝑙𝑎𝑡𝑖,𝑗, 𝑙𝑜𝑛𝑖,𝑗 , 𝑙𝑎𝑡𝑖,𝑘, 𝑙𝑜𝑛𝑖,𝑘 , 𝑢𝑛𝑖𝑡 = 𝑈𝑛𝑖𝑡. 𝑀𝐼𝐿𝐸𝑆]≤0. 06
∀𝑘∈𝑘𝑖, ∀𝑖, 𝑗

2.​ The distance within the group of potential locations in each zip-code areas (i.e. for
each pair of potential locations):
( )( )
𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 = ℎ𝑎𝑣𝑒𝑟𝑠𝑖𝑛𝑒[ 𝑙𝑎𝑡𝑖,𝑘, 𝑙𝑜𝑛𝑖,𝑘 , 𝑙𝑎𝑡𝑖,𝑘', 𝑙𝑜𝑛𝑖,𝑘' , 𝑢𝑛𝑖𝑡 = 𝑈𝑛𝑖𝑡. 𝑀𝐼𝐿𝐸𝑆]≤0. 06
'
∀𝑘, 𝑘'∈𝑘𝑖, 𝑘 < 𝑘 , ∀𝑖

The first set of comparisons is relatively simple and three for loops are applied: for each
potential location i, calculating its distance to each child care center j, and comparing the
distance to the threshold 0.06 miles in each area k. If the distance between each location i
and any of child care centers is less than 0.06 miles, location i is dropped from the list of
potential locations.
On the other hand, the distance within the group of potential locations is a bit more complex.
All potential locations that are not too close to any child care centers in each zip-code area
are extracted first, then the within-group comparison is initiated. A distance matrix is thus
created for all potential locations in each zip-code area, with each row i and column j
denoting the distance between potential location i and j in that area. Meanwhile, it is worth
mentioning that only the upper half of each matrix needs to be analyzed as the lower half is
repeated. Thus, the algorithm will simply be: in each zip-code area, for each potential
location i in that area, calculate the distance between i and all the potential locations j
starting from i + 1 in that list, and compare these distances with the threshold 0.06 miles. If
the distance is smaller than 0.06, one of the potential locations is dropped. With the above
algorithm, a binary decision variable “x_i” is attached to each location in each area (1 if the
location is finally chosen, 0 if not), and another optimization problem comes up, which
includes the objective function to maximize “x_i” in each area. If the distance between
location i and j are too close, x_i + x_j <= 1. Such a model would help to find out the
maximum number of locations chosen while each pair of locations in each area has a
distance no less than 0.06.
The eventually selected number of potential locations are displayed below:
c. Location Requirement
𝑆𝑖 + 𝑀𝑖 + 𝐿𝑖≤𝑡𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑜𝑡𝑒𝑛𝑡𝑖𝑎𝑙 𝑙𝑜𝑐𝑎𝑡𝑖𝑜𝑛 𝑤𝑖𝑡ℎ𝑖𝑛 𝑡ℎ𝑒 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑟𝑒𝑞𝑢𝑖𝑟𝑒𝑚𝑒𝑛𝑡
[Link] Optimization Model
Similarly, to determine the minimum amount of funding (in total) needed to meet their target for
each area, the optimization model for minimum total funding can be ultimately formulated as a
Linear Programming model shown as below:
𝑀𝑖𝑛(
𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗
∑[ 𝐸𝑥𝑖𝑖𝑗
(20, 000 + β1𝑖𝑗·200·𝐸𝑥𝑖𝑖𝑗 + β2𝑖𝑗·400·𝐸𝑥𝑖𝑖𝑗 + β3𝑖𝑗·1, 000·𝐸𝑥𝑖𝑖𝑗)] + ∑(100·𝐸𝑥𝑝𝐵𝑖𝑗) + ∑(65, 000·𝑆𝑖 +
𝑖,𝑗 𝑖,𝑗 𝑖

)
The first term calculates the cost of expanding existing facilities, scaled by the proportion of
𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗
expansion slots relative to existing slots 𝐸𝑥𝑖𝑖𝑗
​. The cost per slot is determined by which
interval the expansion ratio falls into, as defined by the binary variables​.

[Link] Output
a. Expansion
Total slots expanded for children under 5: 94,793.
Total slots expanded for children aged 5–12: 1,401.
b. New Construction
Small Facilities:
Number built: 279.
Slots created: 27,900.
Medium Facilities:
Number built: 266.
Slots created: 53,200.
Large Facilities:
Number built: 2,270.
Slots created: 908,000.
c. Total Costs:
The total budget spent: $354,695,816.39.
The breakdown indicates that new construction costs still significantly outweigh expansion costs.
Figure 2. Comparison of Different Facility Costs in Problem 2

The result has a small difference compared to the question1. This is because problem 2 gives the
piecewise function on expansion cost. However, a lot of cost incurred is from new facility
construction, and there is not too much expansion on existing facilities. Therefore, the changes on
the expansion cost in problem2 would not influence the final result a lot.

2.4.3 Problem 3: Fairness


[Link] Decision Variables
Newly Defined Decision Variables:
(100·𝑆𝑖+200·𝑀𝑖+400·𝐿𝑖+𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗+𝐸𝑥𝑖𝑖𝑗)
𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖 = 𝑃𝑜𝑝𝑖
: The child care coverage ratio in zipcode 𝑖,

defined by number of available child care slots divided by population of children in zipcode 𝑖.
(50·𝑆𝑖+100·𝑀𝑖+200·𝐿𝑖+𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑖𝑖𝑗·)
𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝐵𝑖 = 𝑃𝑜𝑝𝐵𝑖
: The 0-5 years-old children coverage ratio in

zipcode 𝑖, defined by number of available child care slots for 0-5 years-old children divided by
population of 0-5 years-old children in zipcode 𝑖.
𝑆𝐶𝐼𝑖 = 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝐵 ·2 + 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖: Social Coverage Index, based on the weighted sum of
𝑖

child care coverage for two groups.


[Link] Model Constraints
Newly introduced constraints:
𝐸𝑥𝑝𝐵𝑖𝑗+𝐸𝑥𝑝𝐴𝑖𝑗
∑[ 𝐸𝑥𝑖𝑖𝑗
(20, 000 + β1·200·𝐸𝑥𝑖𝑖𝑗 + β2·400·𝐸𝑥𝑖𝑖𝑗 + β3·1, 000·𝐸𝑥𝑖𝑖𝑗)] + ∑(100𝐸𝑥𝑝𝐵𝑖𝑗) + ∑(65, 000𝑆𝑖 + 95, 0
𝑖,𝑗 𝑖,𝑗 𝑖
This constraint represents a budget limitation for the fairness problem. It ensures that the total cost
of expanding existing child care facilities across different regions does not exceed the allocated
budget of $1,000,000.
𝑦𝑖 = 100·𝑆𝑖 + 200·𝑀𝑖 + 400·𝐿𝑖 + 𝐸𝑥𝑝𝐵𝑖𝑗 + 𝐸𝑥𝑝𝐴𝑖𝑗 + 𝐸𝑥𝑖𝑖𝑗 − 𝑃𝑜𝑝𝑖: Represents the balance
between the number of available slots and the population in a specific region 𝑖.
𝑧𝑖: A binary variable indicating whether the coverage ratio constraint is met in region 𝑖.
{𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖 ≤ 𝑧𝑖 + 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖 × (1 − 𝑧𝑖) 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖 ≥ 𝑧𝑖 + 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖 × (1 − 𝑧𝑖)
If 𝑧𝑖 = 1, this ensures 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖 = 1, which means full coverage is met;
If 𝑧𝑖 = 0, this ensures 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖 = 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖.
1
{𝑦𝑖 ≤ 𝑀 × 𝑧𝑖 𝑦𝑖 ≥ 𝑀
− 𝑀 × (1 − 𝑧𝑖)
This enforces that 𝑦𝑖 can only have a meaningful value when 𝑧𝑖 = 1; otherwise, 𝑦𝑖 = 0.
𝑀 = 1, 000, 000: The constant 𝑀 is a large number (set to 1,000,000).
{𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖 − 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖' ≤ 𝑡𝑖 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖' − 𝐶𝑜𝑣𝑒𝑟𝑅𝑎𝑡𝑖𝑜𝑖 ≤ 𝑡𝑖
𝑡𝑖≤0. 1
These constraints ensure that the difference in child care coverage ratios between any two
locations (𝑖 and 𝑖′) is minimized.
[Link] Optimization Model

𝑀𝑎𝑥 𝑍 = ∑ 𝑆𝐶𝐼𝑖
𝑖
[Link] Output
The output of the model is infeasible or unbounded.
The potential reasons of infeasibility of the model are presented below:
a. Overly Strict Fairness Constraint for Uneven Distribution of Child Care Availability:
The maximum allowed difference in coverage ratios (≤0.1) between regions is too restrictive,
especially for areas with significant disparities in child care resources. This makes it difficult to
balance fairness without excessive expansion or construction.
b. Overly Tight Budget Limitation:
The budget constraint does not provide enough resources to build or expand sufficient facilities to
meet both demand and fairness requirements.

Figure 5. Output of Problem 3


3. Conclusion
This project analyzed three critical problems related to child care facility planning and
optimization: Budgeting, Realistic Capacity Expansion and Distance, and Fairness. The results
provide key insights into the trade-offs and challenges in addressing child care needs across
different regions.
In The Problem of Budgeting and The Problem of Realistic Capacity Expansion and
Distance, the optimization achieved a feasible solution within the budget constraint. However, the
majority of the costs were directed towards building new facilities. It suggests that it’s extremely
difficult to improve the current situation of child care availability within the current child care
facility system.
Expansion of existing facilities is cost-effective but insufficient to address the total demand,
necessitating significant investment in new builds. However, the cost of constructing new facilities
is much higher, which means more resources are needed to meet the total demand requirement
especially with tight fairness constraint.
In The Problem of Fairness, the strict constraints (coverage difference≤0.1) led to infeasibility of
the model. This highlights the inherent challenge of balancing fairness with budget and resource
limitations, especially in areas with highly uneven child care availability distribution.
This project demonstrated the complexity of optimizing child care facility planning. Policymakers
must carefully weigh the competing goals of cost-effectiveness, accessibility and fairness, while
taking into account the limitations of available resources.
Works Cited
Adams, G., & Henly, J. R. (2020). Child care subsidies: Supporting work and child development
for healthy families. Retrieved from [Link]
Bradley, R. H., & Vandell, D. L. (2007). Child care and the well-being of children. Archives of
Pediatrics & Adolescent Medicine, 161(7), 669–676. [Link]
Center for American Progress. (n.d.). Child care deserts series. Retrieved October 9, 2024, from
[Link]
New York State Office of Children and Family Services. (n.d.). Child care deserts overview.
Retrieved October 9, 2024, from [Link]
Procare Software. (n.d.). 7 daycare center expenses: How much does it cost to start your own
daycare business? Retrieved October 9, 2024, from
[Link]
-your-own-daycare-business/

You might also like