How Differential Privacy will impact Connecticut Data

To protect the confidentiality of 2020 Census respondents, the US Census Bureau plans to use a "differential privacy" framework; it is a statistical technique to protect individual data from reidentification or reconstruction. To learn more, read our blog post.

The data users community, including CTData Collaborative, is concerned about the potential impact of this approach on the scope and quality of census data products.

To help data users assess the impacts of differential privacy, the Census Bureau has shared files that show the impact of differential privacy on 2010 decennial data. CTData Collaborative developed the interface to compare town-level data based on the IPUMS joint data release of the original 2010 decennial data and the differentially private 2010 datasets.

From the US Census Bureau: "Feedback on the 2010 Demonstration Products can be submitted to dcmd.2010.demonstration.data.products@census.gov. We are accepting feedback through Summer 2020, though earlier is better as work on improving and optimizing the Disclosure Avoidance System for greater accuracy is ongoing."

Tables

Variables

Select a census table from the dropdown below to see how differential privacy would change the values of variables in that table for a particular Connecticut town.

The same data is displayed in a bar chart, as well as the sortable table underneath. Generally, smaller communities are affected the most. Before DP refers to 2010 Census data; After DP refers to what the value would be if DP were applied.

1. Select a table

2. Select a town

Select a census variable from the dropdown below to see how differential privacy would affect that variable in all 169 Connecticut towns. Generally, smaller communities are affected the most.

Before refers to 2010 Census data; After refers to what the value would be if DP were applied.

1. Select a table

2. Select a variable

Town population