Difference between revisions of "Social Stratification Mappers Proposal"

From Visual Analytics and Applications
Jump to navigation Jump to search
Line 43: Line 43:
 
|}  
 
|}  
 
<!-- END OF INTRODUCTION--->
 
<!-- END OF INTRODUCTION--->
 +
 
<div style="padding: 17px">
 
<div style="padding: 17px">
 
</div>
 
</div>
Line 50: Line 51:
 
|  
 
|  
 
<div style="border-left: #56A5EC solid 10px;font-family: Helvetica; padding: 0px 30px 0px 18px; ">
 
<div style="border-left: #56A5EC solid 10px;font-family: Helvetica; padding: 0px 30px 0px 18px; ">
−
Our research and development efforts were motivated by the ongoing debates on social inequality but a general lack of “hard evidence” on the geospatial aspects of social inequality.
 
−
 
−
It aims to equip urban planners with a geospatial tool for visual discovering of social inequality across neighbourhoods based on the three dimensions of drivers of inequality. More importantly, it goes beyond highlighting “pain points” but dives into “solutioning”. It does so by equipping planners with the ability to visualise how the spaces and amenities they have built can serve as important touch points to promote social mixing and reduce inequality.
 
−
 
−
Specifically, it attempts to support the following analysis requirements:
 
−
<br>
 
−
1) To display social inequality’s geographic indices cartographically on an internet map such as Openstreetmap; <br>
 
−
2) To create a map and graphical visualisation that display social segregation and spatial point clustering as measures of social inequality; <br>
 
−
3) To create a map and graphical visualisation that display accessibility measures to amenities and spaces that can promote social mixing; <br>
 
−
|-
 
−
|}
 
−
<!-- END OF MOTIVATION & OBJECTIVES--->
 
  
−
<div style="padding: 17px">
+
<b>1. Discover Drivers of Inequality – Race, Age, Housing Type </b>
−
</div>
 
−
<div style="background: #56A5EC; padding: 20px; line-height: 0.3em; text-indent: 16px;letter-spacing:0.1em;font-size:26px"><font color=#fbfcfd face="Bebas Neue"> KEY OBJECTIVES</font></div>
 
−
{| class="wikitable" style="background-color:#FFFFFF;" width="100%"
 
−
|-
 
−
|
 
−
<div style="border-left: #56A5EC solid 10px;font-family: Helvetica; padding: 0px 30px 0px 18px; ">
 
−
<b>1. Identifying Demographic Clusters via Spatial Points Pattern Analysis</b><br>
 
−
First, we will analyse whether there exist clusters that could point towards social inequality and whether this is more pronounced in certain neighbourhoods, based on ethnic mix, age composition, and housing type. This will be done first by choropleth mapping at the subzone level, then applying the spatial segregation index at the dwelling unit level.
 
  
−
Next, we will perform spatial points pattern analysis using kernel density estimation as a density-based point pattern measure for visualisation of our clusters, followed by the L Function as a distance-based measure. To test the significance of our clusters, we will then apply Complete Spatial Randomness Test using the Monti-Carlo Simulation.
+
Our project was motivated by the ongoing debates on social inequality but a general lack of “hard evidence” especially on the geospatial aspects, of social inequality. Our dashboard aims to equip urban planners with a geospatial tool for visual discovering of social inequality across neighborhoods based on the three dimensions of drivers of inequality – race, age, housing type.  
  
−
|-
+
<b>2. Find Touch Points for Social Mixing – Amenities & Common Spaces </b>
−
|
 
−
<div style="border-left: #56A5EC solid 10px;font-family: Helvetica; padding: 0px 30px 0px 18px; ">
 
−
<b>2. Modelling Geographical Accessibility to Amenities & Spaces </b><br>
 
−
Next, we will map out the available touch points within neighbourhoods that could facilitate social mixing. This would include the ease of access to common spaces, amenities and opportunities for choice of education. This is because an important aspect of social inequality is having reasonably fair access to different resources.
 
  
−
We will use the Hansen Accessibility Index's REAT measure to assess the accessibility of residents within each neighbourhood to these public spaces. We will then plot and visualise these indices across neighbourhoods for comparison.  
+
We also map accessibility of each HDB postal code to the nearest school to see whether some households are disadvantaged in access to elite schools than mainstream schools. More importantly, we go beyond highlighting “pain points” but dive into “solutioning”. We hope to equip planners with the ability to visualize how the spaces (e.g. parks) they have built can serve as important touch points to promote social mixing within and across neighbourhoods.  
  
−
For amenities that exist segregation of access, such as schools, pre-schools and childcare centres, geographical accessibility will be assessed separately based on its class, that is GEP/SAP vs Mainstream Schools and Subsidised vs Privatised Pre-Schools/Childcare Centres.
 
−
|-
 
 
|}
 
|}
−
<!-- END OF KEY OBJECTIVES--->
+
<!-- END OF MOTIVATION & OBJECTIVES--->
−
 
 
  
 
<div style="background: #56A5EC; padding: 20px; line-height: 0.3em; text-indent: 16px;letter-spacing:0.1em;font-size:26px"><font color=#fbfcfd face="Bebas Neue">DATA SOURCES</font></div>
 
<div style="background: #56A5EC; padding: 20px; line-height: 0.3em; text-indent: 16px;letter-spacing:0.1em;font-size:26px"><font color=#fbfcfd face="Bebas Neue">DATA SOURCES</font></div>
Line 205: Line 179:
 
|}
 
|}
 
<!-- END OF DATA SOURCES--->
 
<!-- END OF DATA SOURCES--->
 +
 +
<div style="padding: 17px">
 +
</div>
 +
<div style="background: #56A5EC; padding: 20px; line-height: 0.3em; text-indent: 16px;letter-spacing:0.1em;font-size:26px"><font color=#fbfcfd face="Bebas Neue"> ANALYTICAL MAPPING</font></div>
 +
{| class="wikitable" style="background-color:#FFFFFF;" width="100%"
 +
|-
 +
|
 +
<div style="border-left: #56A5EC solid 10px;font-family: Helvetica; padding: 0px 30px 0px 18px; ">
 +
<b>1. Social Segregation Index at Subzone Level </b>
 +
 +
While there had been studies done on racial, age and housing mix, these were typically done at the aggregate level that measures a subzone’s composition of race, as compared to the national average. While this was informative, it only informed us where different population groups stay – and not whether they mix with one another. This provided little actionable insights for urban planners to turn into operations. What was more useful was whether there were sufficient opportunities for citizens of different race, age and housing type, to mix with one another. In spatial demography, measuring the extent of segregation between population groups would help urban planners to understand social cohesion and integration in our society [4].
 +
 +
We thus selected the entropy-based diversity index as the social segregation measure for race, age and housing type. This diversity index goes beyond measuring dissimilarity, isolation and interaction between two population groups (as is common for literature measuring Black-White segregation), and allows multi-group measure of segregation [5]. Living away from one another could imply that population groups are segregated [6]. We used this to measure whether citizens of the four national races experienced complete segregation, or no segregation; and do the same for housing type (an indicator of income) as well as age. This was then visualized in a choropleth map. Each subzone would have its entropy score, or diversity, defined as the follows, where πri refers to a particular racial/age/housing type group’s proportion of the population in subzone i.
 +
 +
<insert formula>
 +
 +
We also plotted the histogram of segregation to see the distribution of the diversity indices, which can be filtered by subzone and housing type. For detail on-demand, we also complement the above map with a parallel coordinate plot to view the composition of each demographic type across subzones. 
 +
|-
 +
|
 +
<div style="border-left: #56A5EC solid 10px;font-family: Helvetica; padding: 0px 30px 0px 18px; ">
 +
<b> 2. Spatial Point Pattern Analysis at Postal Code Level </b>
 +
 +
There has been work done on the visualization of demographic data by subzone. However, by and large, these data are typically analysed by the subzone level than by the postal code level. We analyse our data at the postal code level because social inequality is typically defined at the household level, especially since Singapore is a small country where analyzing by subzone is equivalent to assuming different regions develop unequally. But point symbol map would not give a good visualization due to the abundant amount of postal code data. Instead, we use the Kernel Density Estimation, a frequently used technique to determine hotspots of point locations, to easily identify spatial clusters – by race, age and housing type.
 +
 +
This density-based measure is visualized in a kernel density map. To test statistically and allow users to have a distance-based measure, we support the visualization with the Ripley’s K Function and the Nearest Neighbour histogram. The K-Function is a method to estimate the second-order properties of a point process by constructing a circle of radius h around each point event, I and counting the number of other events j that fall inside this circle. Under the assumption of CSR, the expected number of events within distance h of an event is: 𝐾 ℎ = 𝜋ℎ2, where ℎ = the radius of the circle. K(h) > 𝜋ℎ2 if point pattern is clustered. This is visualized as such; if the line falls above the confidence band, there is clustering. If it falls within, the clustering is statistically insignificant.
 +
|-
 +
|
 +
<div style="border-left: #56A5EC solid 10px;font-family: Helvetica; padding: 0px 30px 0px 18px; ">
 +
<b> 3. Accessibility to Schools and Parks  </b>
 +
 +
Using the Hansen Accessibility Model, we mapped out each household’s ease of access to schools – mainstream and elite schools. Accessibility refers to the ease of reaching a destination.  For our application, the origin zones would be the HDB blocks and the destination zones will be the amenities.  A higher Hansen Accessibility Index would mean that the HDB block is very accessible to that particular school type – elite or mainstream.
 +
 +
We also assessed accessibility to parks (which includes playgrounds); common spaces where we expect them to be easy-to-reach, open platforms that can promote social mixing within neighbourhoods. We wanted to identify areas that are underserved for urban planners to focus their attention on for future space planning.
 +
 +
The approach of this measure to the calculation of a zone's accessibility is to add together the opportunities available in each other zone, weighted by a function of the difficulty of reaching that zone. It is measured by:
 +
 +
<insert formula>
 +
 +
|-
 +
|}
 +
<!-- END OF MOTIVATION & OBJECTIVES--->
 +
 +
  
  

Revision as of 15:42, 7 August 2018

Exploring Inequality’s Geographic Dimension Across Neighbourhoods in Singapore: It's Driving Forces & Touch Points

OVERVIEW

PROPOSAL

POSTER

APPLICATION

RESEARCH PAPER

BACK TO HOMEPAGE

INTRODUCTION

Amidst the recent debate over growing social inequality in Singapore such as the distinct clustering of elite schools and varying access to resources, the dangers of hardening social mobility pose as a concern for a culturally diverse nation that has upheld its values of social cohesion and racial harmony. In bridging social divide, Singapore has put in place various community programmes to create shared experiences and promote inter-communities mixing.

Using geospatial techniques in R, the dashboard serves to explore the geographic dimension of social inequality, by mapping the extent of social segregation and accessibility to important spaces across neighbourhoods. This is done in three approaches. First, we analyse whether there exists social segregation across subzones using the Entropy-Based Diversity Index, based on three dimensions of inequality - race, age and housing type. Second, using spatial point pattern analysis at the HDB postal code level, we visualise whether there exist housing type clusters that could point towards social inequality and whether certain towns are overpopulated with a specific housing type. Third, using the Hansen Accessibility Model, we map out available touch points that could facilitate social mixing, particularly the ease of access to primary schools. We also see whether there is any variation in accessibility between the elite and mainstream primary schools for different housing types. This is because an important aspect of social inequality is having reasonably fair access to resources. Lastly, we move into solutioning and explore whether there exist sufficient common spaces that allow for social mixing, such as parks, and identify areas that are underserved for urban planners to focus their attention on for future space planning.

MOTIVATION & OBJECTIVES

1. Discover Drivers of Inequality – Race, Age, Housing Type

Our project was motivated by the ongoing debates on social inequality but a general lack of “hard evidence” especially on the geospatial aspects, of social inequality. Our dashboard aims to equip urban planners with a geospatial tool for visual discovering of social inequality across neighborhoods based on the three dimensions of drivers of inequality – race, age, housing type.

2. Find Touch Points for Social Mixing – Amenities & Common Spaces

We also map accessibility of each HDB postal code to the nearest school to see whether some households are disadvantaged in access to elite schools than mainstream schools. More importantly, we go beyond highlighting “pain points” but dive into “solutioning”. We hope to equip planners with the ability to visualize how the spaces (e.g. parks) they have built can serve as important touch points to promote social mixing within and across neighbourhoods.

DATA SOURCES

Our team's datasets are retrieved from https://data.gov.sg

Type

Format

Data

Source URL

Boundary (Polygon)

SHP

OSM Layer (Singapore)

OpenStreet Map

Boundary (Polygon)

SHP

Master Plan 2014 Subzone Boundary (No Sea)

https://data.gov.sg/dataset/master-plan-2014-subzone-boundary-no-sea

Demographics

CSV

Estimated Singapore Resident Population in HDB Flats

https://data.gov.sg/dataset/estimated-resident-population-living-in-hdb-flats

Demographics

CSV

Dwelling Units under HDB's Management, by Town and Flat Type

https://data.gov.sg/dataset/number-of-residential-units-under-hdb-s-management

Demographics

CSV

Residents by Age Group & Type of Dwelling, Annual

https://data.gov.sg/dataset/residents-by-age-group-type-of-dwelling-annual

Demographics

CSV

Land Area and Dwelling Units by Town

https://data.gov.sg/dataset/land-area-and-dwelling-units-by-town

Demographics

SHP

Singapore Residents by Subzone and Type of Dwelling, June 2016

https://data.gov.sg/dataset/singapore-residents-by-subzone-and-type-of-dwelling-june-2016

Demographics

SHP

Singapore Residents by Subzone, Age Group and Sex, June 2016 (Gender)

https://data.gov.sg/dataset/singapore-residents-by-subzone-age-group-and-sex-june-2016-gender

Demographics

SHP

Resident Population of Other Ethnic Groups by Age Group, Ethnic Group and Sex, 2015

https://data.gov.sg/dataset/resident-population-of-other-ethnic-groups-by-age-group-ethnic-group-and-sex-2015

Amenities

CSV

Primary Schools

https://data.gov.sg/dataset/school-directory-and-information

Spaces

SHP

Parks (including playgrounds)

https://data.gov.sg/dataset/parks

ANALYTICAL MAPPING

1. Social Segregation Index at Subzone Level

While there had been studies done on racial, age and housing mix, these were typically done at the aggregate level that measures a subzone’s composition of race, as compared to the national average. While this was informative, it only informed us where different population groups stay – and not whether they mix with one another. This provided little actionable insights for urban planners to turn into operations. What was more useful was whether there were sufficient opportunities for citizens of different race, age and housing type, to mix with one another. In spatial demography, measuring the extent of segregation between population groups would help urban planners to understand social cohesion and integration in our society [4].

We thus selected the entropy-based diversity index as the social segregation measure for race, age and housing type. This diversity index goes beyond measuring dissimilarity, isolation and interaction between two population groups (as is common for literature measuring Black-White segregation), and allows multi-group measure of segregation [5]. Living away from one another could imply that population groups are segregated [6]. We used this to measure whether citizens of the four national races experienced complete segregation, or no segregation; and do the same for housing type (an indicator of income) as well as age. This was then visualized in a choropleth map. Each subzone would have its entropy score, or diversity, defined as the follows, where πri refers to a particular racial/age/housing type group’s proportion of the population in subzone i.

<insert formula>

We also plotted the histogram of segregation to see the distribution of the diversity indices, which can be filtered by subzone and housing type. For detail on-demand, we also complement the above map with a parallel coordinate plot to view the composition of each demographic type across subzones.

2. Spatial Point Pattern Analysis at Postal Code Level

There has been work done on the visualization of demographic data by subzone. However, by and large, these data are typically analysed by the subzone level than by the postal code level. We analyse our data at the postal code level because social inequality is typically defined at the household level, especially since Singapore is a small country where analyzing by subzone is equivalent to assuming different regions develop unequally. But point symbol map would not give a good visualization due to the abundant amount of postal code data. Instead, we use the Kernel Density Estimation, a frequently used technique to determine hotspots of point locations, to easily identify spatial clusters – by race, age and housing type.

This density-based measure is visualized in a kernel density map. To test statistically and allow users to have a distance-based measure, we support the visualization with the Ripley’s K Function and the Nearest Neighbour histogram. The K-Function is a method to estimate the second-order properties of a point process by constructing a circle of radius h around each point event, I and counting the number of other events j that fall inside this circle. Under the assumption of CSR, the expected number of events within distance h of an event is: 𝐾 ℎ = 𝜋ℎ2, where ℎ = the radius of the circle. K(h) > 𝜋ℎ2 if point pattern is clustered. This is visualized as such; if the line falls above the confidence band, there is clustering. If it falls within, the clustering is statistically insignificant.

3. Accessibility to Schools and Parks

Using the Hansen Accessibility Model, we mapped out each household’s ease of access to schools – mainstream and elite schools. Accessibility refers to the ease of reaching a destination. For our application, the origin zones would be the HDB blocks and the destination zones will be the amenities. A higher Hansen Accessibility Index would mean that the HDB block is very accessible to that particular school type – elite or mainstream.

We also assessed accessibility to parks (which includes playgrounds); common spaces where we expect them to be easy-to-reach, open platforms that can promote social mixing within neighbourhoods. We wanted to identify areas that are underserved for urban planners to focus their attention on for future space planning.

The approach of this measure to the calculation of a zone's accessibility is to add together the opportunities available in each other zone, weighted by a function of the difficulty of reaching that zone. It is measured by:

<insert formula>



TIMELINE & MILESTONES
The following table shows the timeline and milestones of our project. Details will be updated progressively.

S/N

Task

Done by

Week

Status

1

Topic Brainstorming

ALL

2 & 3

Completed ✔

2

Drafting and refinement of Project Proposal

ALL

2 & 3

Completed ✔

3

Consultation with Prof Kam for Feedback on Proposal

ALL

4

Completed ✔

4

Finalization of Project Topic and Focus

ALL

5

Completed ✔

5

Compilation and Cleaning of Datasets
1. Geocoding and Categorization of Schools
2. Combine all Demographics datasets (Housing, Age and Ethnic) for analysis by Subzone
3. Convert all spatial files to WGS84 for OSM compatibility
4. Data Clean, Manipulate and Overlay the Amenities/Spaces with Demographics datasets

Grace
Yuing

6

Completed ✔

6

1st Wiki Content Update
(Proposal, Methodology & Storyboard)

Yuqing
Yuanjing

6

Completed ✔

7

Independent learning of R and R Shiny on DataCamp

ALL

6 & 7

In Progress ✔

8

Consultation with Prof Kam for Feedback on tools for Geospatial Analysis

ALL

7

Completed ✔

9

2nd Wiki Content Update

Yuqing

7

Completed ✔

10

Map Development
PART 1: Kernel Density Estimate: Grace
PART 2: Hansen Potential Analysis: Grace
PART 3: Common Spaces for Solutioning: Yuanjing

Grace
Yuanjing

7 - 9

11

Interface Development
User Interface (Web Layout): Yuqing
User Interface (Description) : Yuanjing

Yuqing
Yuanjing

8 - 9

12

Consultation with Prof Kam for Feedback on progress/techniques

ALL

9

13

Debugging and Analysis of Results

ALL

9

14

Consultation with Prof Kam for Feedback on final product

ALL

10

15

Creating and Submission of Poster

Yuqing

11

16

Uploading of App on Shinyapps.io (Artefact)

Yuanjing

12

17

Finalizing Wiki Page & Research Paper (Deliverables)

ALL

13

18

Poster Presentation / Conference

ALL

14


TOOLS & PACKAGES

The following is a list of tools to be adopted for the project’s scope.

  • QGIS
  • Open Street Map
  • OneMap Geocode
  • R Studio
  • R libraries
    • shiny
    • leaflet
    • rgdal
    • sf
    • sp
    • spatstat
    • dplyr
    • plyr
    • maptools
    • shinydashboard
    • spatialsegregation
    • REAT
    • SpatialAcc
    • ggmap
    • SpatialPosition
    • maptools
    • shinyBS
    • shinyJS


REFERENCES TO RELATED WORKS

1. We would like to credit the following referenced visualization works adopted in the design of our dashboard.

Centroid-Amenities: An Interactive Visual Analytical Tool for Exploring and Analysing Amenities in Singapore
Authors: Siew Xue Qian Jazreel, Tay Wei Xuan, Sean Koh Jia Ming
https://jazreelsiew.shinyapps.io/AppV2/

Map3.png


2. The following also lists the referenced literature considered in the design of our problem statement.

No

Title

Link

Author

1

Commentary: Inequality has a geographic dimension - between and within neighbourhoods in Singapore

https://www.channelnewsasia.com/news/commentary/inequality-in-singapore-exists-across-within-neighbourhoods-10276898

Leong Chan-Hoong

2

Lack of social mixing is a symptom of inequality, not a cause

https://www.straitstimes.com/opinion/lack-of-social-mixing-is-a-symptom-of-inequality-not-a-cause#main-content

Teo You Yenn

3

Class divide: Singapore in danger of becoming academic aristocracy

https://www.straitstimes.com/opinion/class-divide-singapore-in-danger-of-becoming-academic-aristocracy

Chua Mui Hoong

4

The Big Read: Social stratification — a poison seeping into S’pore’s housing estates and schools

https://www.todayonline.com/singapore/big-read-social-stratification-poison-seeping-spores-housing-estates-and-schools

Kelly Ng and Toh Ee Ming

5

COMMENT: Can Singapore's elite circle turn around growing social divide?

https://sg.news.yahoo.com/comment-can-singapores-elite-circle-turn-around-growing-divide-124724650.html

Nicholas Yong

6

This Is What Inequality Looks Like

Ethos Books (Publisher)

Teo You Yenn

Retrieved from ‘https://wiki.smu.edu.sg/1718t3isss608/index.php?title=Social_Stratification_Mappers_Proposal&oldid=11106’