data visualization for policy advocacyadvocacy group policy process . data & tools external...
TRANSCRIPT
![Page 1: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/1.jpg)
David Epstein Research Associate
Baltimore Neighborhood Indicators Alliance Jacob France Institute
University of Baltimore
Data Visualization for Policy Advocacy
Presentation for the National Low Income Housing Coalition Housing Policy Conference April 27-30 2014
![Page 2: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/2.jpg)
The visual representation of
data to improve cognition
Based on a definition by Dr. Eytan Adar, University of Michigan, 2010
Data Visualization:
![Page 3: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/3.jpg)
Data intermediary
Advocacy group
Policy process
Data & tools
External effort • Communicate persuasively
Internal effort • Understand issues • Develop stances & strategies
Data-backed narrative
Data visualization: Both internal & external efforts should be informed by an explicit theory of change that identifies the target audience of data visualizations (e.g. policy makers, domain experts, the media, or the general public).
Locating data visualization within advocacy
![Page 4: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/4.jpg)
Function of Visualization in Advocacy
See also: Schwabish J., “An Economists Guide to Visualizing Data”, Journal of Economic Perspectives, Winter 2014, page 227
Explore Explain
Purpose Reveal relationships between variables
Communicate specific points in a narrative
Form
Usually: many rough, static graphics Recently: interactive linked (“brush”) plots
Usually: a few polished, static graphics with text. Recently: a static or interactive infographic.
![Page 5: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/5.jpg)
Quick! What’s the pattern?
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Hypothetical housing data
![Page 6: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/6.jpg)
Based on Jacques Bertin’s “retinal variables” in Semiology of Graphics 1983
Human vision generally excels at seeing
Size Shape
Value (B/W)
Color (Hue)
Position
Texture
![Page 7: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/7.jpg)
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Color (hue)
Value (B/W)
While pretty, this conditional formatting does not improve cognition.
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Does this help?
![Page 8: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/8.jpg)
Type Count Vacant 3 Occupied 7
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Frequency table
Visualization of a single categorical (text) variable
There are roughly twice as many occupied houses as vacant houses.
![Page 9: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/9.jpg)
Type Count Vacant 3 Occupied 7
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Frequency table
Visualization of a single categorical (text) variable
Size
There are roughly twice as many occupied houses as vacant houses.
![Page 10: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/10.jpg)
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Visualization of single numeric variable
![Page 11: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/11.jpg)
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Visualization of single numeric variable
Split between small and large houses at around 1,700 sq/ft.
Many houses around 1,400 sq/ft (darker gray).
![Page 12: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/12.jpg)
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Visualization of single numeric variable
Binning
Careful when setting bin sizes!
![Page 13: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/13.jpg)
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Visualization of multiple variables
![Page 14: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/14.jpg)
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Vacant houses are all larger than occupied houses.
Visualization of multiple variables
![Page 15: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/15.jpg)
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Shape
Location
Size
Visualization of multiple variables
Scatter plots can produce simple maps without the need for GIS software
![Page 16: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/16.jpg)
Type X Y Size Vacant 0.5 0.8 2,100 Occupied 0.55 0.9 1,200 Occupied 0.53 0.85 1,400 Occupied 0.45 1.1 1,300 Vacant 0.7 0.75 1,970 Occupied 0.7 1.1 1,450 Occupied 0.72 1.25 1,610 Occupied 0.68 0.9 1,290 Occupied 0.7 1.2 1,370 Vacant 0.7 0.6 1,880
Shape
Location
Size
Vacant houses are not only larger than occupied houses, they are also clustered in the South.
Visualization of multiple variables
Scatter plots can produce simple maps without the need for GIS software
![Page 17: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/17.jpg)
http://mbostock.github.io/protovis/ex/cars.html
Multivariate techniques Parallel plots or “spaghetti” plots show all the data in a single view. Each line represents a single record through all variables (columns)
Filtering helps search for trends and outliers.
![Page 18: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/18.jpg)
Multivariate techniques Small multiples (“facets”) make basic plotting techniques more powerful.
Employment of individuals (with 95% confidence), entering, leaving, or staying in Baltimore City 2008-2012 Draft version (excerpt). BNIA 2014. Based on U.S. Census ACS Public Use Microsample (PUMS).
![Page 19: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/19.jpg)
The Baltimore Neighborhood Indicators Alliance (BNIA) is an
organization committed to promoting, supporting, and helping
people make better decisions using accurate, reliable, and
accessible data and indicators to improve the quality of life in
Baltimore City neighborhoods.
• 50+ data sources
• 150+ indicators
• Neighborhood level
• 12 years of data
• Baltimore city focus
• Specialized in mapping
http://www.bniajfi.org/
![Page 20: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/20.jpg)
CSA = community statistical area (neighborhood)
Implications for policy advocacy: Improving public transit may support attraction & retention efforts. Creating jobs where people live may reduce transit demand.
Travel Time 2011
Avg Growing CSAs 15.8% Declining CSAs 23.1%
![Page 21: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/21.jpg)
“Tracking the foreclosure
and home sales data
provided by BNIA-JFI has
enabled BHPC and our
members to better target
resources and outreach
efforts around programs
and other assistance,
and to measure our
progress.” --Baltimore Housing Preservation
Coalition, Vital Signs 11, page 37.
Foreclosures by neighborhood (Q1 2009)
Clustered points of the same color represent foreclosures in the same neighborhood.
http://foreclosures.bniajfi.org/
![Page 22: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/22.jpg)
Thank you!
![Page 23: Data Visualization for Policy AdvocacyAdvocacy group Policy process . Data & tools External effort • Communicate persuasively Internal effort • Understand issues • Develop stances](https://reader033.vdocuments.us/reader033/viewer/2022053119/60a02998e25a1d4c6115fffb/html5/thumbnails/23.jpg)
ggplot(data,aes(Type)) + geom_bar() + theme_bw()
Using the gglot2 package in R with the table named “data”
ggplot(data, aes(Size,0)) + geom_point(alpha=0.25, size=5) + coord_cartesian(ylim=c(0,0),xlim=c(0,2500)) + theme_bw() + ylab("") + xlab("Size in Sq/ft")
ggplot(data, aes(Size)) + geom_histogram(binwidth=500,fill='gray', colour='black') + theme_bw() + xlab("Size in Sq/ft")
ggplot(data, aes(Size,0)) + geom_point(alpha=0.25, size=5, aes(color=Type)) + coord_cartesian(ylim=c(0,0),xlim=c(0,2500)) + theme_bw() + ylab("") + xlab("Size in Sq/ft")
ggplot(data, aes(x=X, y=Y, shape=Type, size=Size)) + geom_point() + scale_shape_manual(values=c(15,0)) + coord_cartesian(ylim=c(0,1.5),xlim=c(0,.8)) + theme_classic() + scale_size(range = c(1, 6))