The projections use 2010 census data as well as 2018 ACS data. 

Here is a more detailed explanation copied from one of the metadata files (keep in mind that their are block and block group level projections, so this metadata may not match up exactly with the file you are looking at): 

**Overview from Haystaq's metadata on their projections: **

These projections are at the 2010 Census Block Group level and the 2010 Census Block level, which is the most granular level of geography currently available. The projections update the current population data to match Census projections for overall statewide totals from 2020 and 2030. They project population nationally, from 2020 to 2030, with total population, as well as Census race groups and Hispanic origin categories. 

Although the US Census has generated population projections for 2020 and 2030, these projections are at the state level. While it would be simple to apply the statewide growth rate to each Census Block in a given state, we know that population growth is not uniform statewide. In order to generate projections at the Block and Block Group level, we need to estimate where in each state the population growth is occurring. A large share of population growth comes from new housing developments built in Census Blocks that were previously undeveloped and had zero population in 2010. 

We addressed the question of population growth in new housing developments using a combination of Census American Community Survey (ACS) Block Group level population estimates, and geocoded commercial data files from which we can find individuals currently living in Census Blocks that had zero population in 2010.