Different number of blocks in data table and GIS file?

I am using the 1990 Block-level total population data of Alabama.

In the block-level GIS file (1990, basis: 2000 TIGER/Line +), there are 173,147 features.

However, the source table has 111,868 block-level observations.

Is anyone else seeing the same problem? Could anyone help explain the discrepancy?


I further checked the number of features at the block-level for Alabama in subsequent decennial years.

In 2000:
The source table has 175,220 observations.
The GIS shapefile (basis: [2000 TIGER/Line +]) has 175,118 features.

In 2010:
The source table has 252,266 observations.
The GIS shapefile (basis: [2010 TIGER/Line +]) has 251,720 features.

In 2020:
The source table has 185,976 observations.
The GIS shapefile (basis: [2020 TIGER/Line +]) has 185,741 features.

There is still a gap in feature count for 2020, 2010 and 2000, although the gap is smaller. The biggest gap is for 2000.

There are two causes for these differences…

  1. The 1990 summary files include no records for blocks that had no housing or population. (This was likely a space-saving measure, given the very large number of blocks, and the relatively limited digital storage capabilities of that era!) The NHGIS shapefiles, however, include polygons for blocks that have no housing or population. You can safely assume that any polygon with no match in the population data has zero population and housing units.

  2. NHGIS shapefiles exclude areas that lie entirely in coastal or Great Lakes waters. We do this so that maps of the data show the coasts and lakeshores accurately. Some blocks lie entirely in coastal water or Great Lakes. These blocks will appear in the 2000-2020 population tables but not in the NHGIS shapefiles.