Lecture 10 - Geodemographics Flashcards
Geodemographics
Geo - geography / place
Demo - people
Graphics - Writing / analyising / research
What is geodemography?
- Science of profiling / analysing people based on where they live
- You are where you live (birds of a feather flock together)
- Where you are says something about who you are
2 principles of geodemography
- people living in areas close to each other will have more in common, people farther away will have more differences
- you can also group people based on their characteristics no matter where they live (example: students)
Charles Booth
- 1889
- looked at how people in London cluster based on income, # family members, employment, etc
- used this information to map where different classes clustered
Concentric Zone Theory
- 1925
- Explain urban social structures with concentric rings
- CBD in the middle, then factory zone, transition zone, working class zone, residential zone, commuter zone
-each zone is a geographical area distinguished by both physical individuality and by social, economic, and cultural characteristics of population
Commercial examples
CACA, Claritas, Experian
- classify people based on where they live, analyse and sell results to companies, governments, etc.
- help decide where to build a shop or business, etc
- geodemographic segmentation assumes that consumer behaviour can be predicted by ‘who you are & where you live’
Geodemographic segmentation example (PersonicX)
profile market segments
- profile your customer base (behavior, demographic, lifestyle)
- identify market segments
- track performance of products or customer segments
- recognise customer risks
- target market spending to get returns on advertisement investing
- analyze campaign performance
- expand business with existing customers
- explore customer insights through third party data
Geodemographic classification (PersonicX)
-classify groups and sub-clusters
-group / cluster example:
group = big ethnic families
cluster B05 = young renting families
cluster B07 = kids & comfort
-profile example:
includes house income, education, driving, spending, percent of the population, defining features, etc
Geodemographic cluster scale (PersonicX)
- NZ’s smallest scale for census is the meshblock
- PersonicX gathers more detailed information for households/properties within each meshblock
How to create geodemographics (6)
- gather data (from census or with surveys)
- area level variables
- evaluate input variables (go there to see if data representative of real world)
- cluster ‘socially similar’ neighbourhoods together
- optimisation process & manual intervention
- form a class hierarchy & label
Gather data
- often census is the sole data source
- in data-rich countries census supplemented by property registers, electoral registers, car registration, etc
- non-census data is useful, gets info from more privileged people, available at finer aggregation, can fill gaps between censuses
- link into a single level of geography
Create area level variables
- more variables from different sources = more meaningful clusters
- relate ‘count’ variables to ‘base’ count to get rates (total cars divided by adult population)
- group counts:
- # of residents in different age groups
- employment by industry
Evaluate input variables
- is the variable skewed (not a normal bell curve)
- some variables may not be deemed appropriate for use in clustering
Cluster ‘socially similar’ neighborhoods together
- depends on cluster method
- 2 types: hierarchical and non-hierarchical (k-means)
Hierarchical
each neighborhood forms a separate cluster, then each cluster merges sequentially on similarity, reducing # of clusters in each step until 1 cluster left
Non-hierarchical (k-means)
Neighborhoods partitioned into a predetermined number of non-hierarchical clusters based on similarity
Optimization process & manual intervention
- are any 2 clusters too homogeneous? (combine them)
- are any 2 clusters too heterogeneous? (split them)
- map clusters and verify with other experts
- take photos in the field and compare
Cluster hierarchy & label
- homogeneous countries (Ireland, Hong Kong, Peru) = 25-30 distinct cluster types
- heterogeneous (UK, USA) = up to 60-65 cluster types (smaller number of neighborhood groups)
- merge cluster pairs causing least variability loss in the original dataset
Examples
Redlining - companies that sell houses or car insurance and create maps of ‘red’ areas where they won’t accept new customers from or won’t sell homes to certain people
Credit cards - refuse new banks in certain areas because they don’t want certain customers (group people by area and assume they are all the same)
Ecological fallacy
Geodemographics can contribute to ecological fallacy which can have policy implications
- group people and assume that all individuals have the same characteristics as the group
- the mean of a group that is very similar could be the same as the mean of a group that is very diverse so wrong assumptions can be drawn
- geodemographic classifications simply a complex geographic reality to make the decision-making process easier, faster & more understandable to stakeholders
