Map Blue was written by Charlie and Joe Gunyon. It uses data from the 2010 US Census combined with survey data from the General Social Survey to predict likely political orientation based on some very basic demographic information.
You can configure the map to display the data in two ways. The first (and default) way is to show the net Democratic votes in each block compared to the block with the most net Democratic votes. For example, if the map is displaying 2 blocks, block 1 has 10 Democratic votes and block 2 has 5 Democratic votes, block 2 will be half as opaque as block 1. Of course, if there are more Republican votes than Democratic votes, the block will be red.
The second way is to show percentages, i.e. this block is 90% Democratic. This can be more useful as a rough overview, but in general, the first method is much more useful.
We use the 2010 Decennial Census because it contains information down to the census block level. This level is very high-resolution; any higher and we're talking individual addresses.
There are some downsides to this approach. Because of the scale of the Decennial Census, key demographic information, like religion, income and education level, is left out. Additionally, the data can be outdated.
We would like to use additional data sources (ACS data, SAVI, etc.) to try and augment the data from the Decennial Census, however, we lack the time and money to do so.
Map Blue uses some moderately advanced web techniques that require a modern browser. Your best bet is to use Google Chrome. Firefox will work, but tends to be slow.
You can speed things up by zooming in. The more blocks Map Blue has to draw, the harder it has to work. For example, zooming far out in urban areas might draw 50,000 census blocks. That's a tall order.
We've considered drawing census block groups or tracts when zoomed out past certain levels, and then using lower-resolution data (like ACS data). But, again, we lack the time and money to improve things past this point.