Boston housing data provided by GeoDa center is a well-known dataset. GeoDa Center provides this data in shapefile format. A metadata file included in the download indicates that the data use projection UTM 19N. However the shapefile does not have a projection defined (no accompanying .prj file), and I am unable to load it correctly.

I would like to overlap the Boston data with the census tract data available at Census Bureau using ArcGIS for Desktop.

Do you know how I can do this?

  • @chris-w Can you remove the hold on this question, or do I have to edit the question first? The problem with the data is that the reported coordinate system does not match any of the ESRI predefined coordinate systems. I have developed an answer which I would like to share with the user. – Lee Hachadoorian May 21 '15 at 6:42
  • 1
    To give a preview of the answer, which I will post after the question is removed from on-hold status, the shapefile distributed by GeoDa Center does use UTM 19N, but the coordinates are in kilometers, whereas all of the ArcGIS predefined UTM coordinate systems assume the coordinates are in meters. The solution is to recreate the spatial data from the lat-long coordinates also included in the source download. – Lee Hachadoorian May 21 '15 at 6:54
  • @chris-w Got full answer all cued up for when/if reopen happens. Thanks for info on on-hold/reopen process. – Lee Hachadoorian May 21 '15 at 7:23

The shapefile distributed by GeoDa Center does not have a PRJ file, so normally you would have to define the projection in order to line it up with other spatial data. But it appears that while the shapefile distributed by GeoDa Center does in fact use the UTM 19N zone, for some reason the units are in kilometers, where UTM coordinates are usually expressed in meters. If you already used ArcGIS to define the shapefile's projection as, for example, NAD 1983 UTM Zone 19N (not immediately clear that NAD 1983 is the datum, but often a good guess for United States data), all of your points will be way off because ArcGIS will interpret the coordinates as meters where the underlying data are actually in kilometers.

Your best bet is to make use of the boston.txt file that is also included in the ZIP archive you downloaded from GeoDa Center to create your own shapefile from the coordinates. This file has the coordinates in both lat-long and UTM 19N, but since the UTM coordinates are in kilometers, you would need to divide all the values by 1000 to import it correctly into ArcGIS using any of the predefined UTM coordinate systems. It will be easier to just use the lat-long coordinates.

Steps to do so are as follows:

  1. Strip the first line from boston.txt (containing the text 506,23), so that header which was on line 2 is promoted to line 1.
  2. In ArcGIS go to File→Add Data→Add XY Data.
  3. Select the boston.txt file and set X Field = LON and Y Field = LAT. Note that since the file has fields named x and y, the dialog will sensibly default to assuming that the X Field is x and the Y Field is y, but these are the coordinates in UTM kilometers, which we don't want to use.
  4. Under "Coordinate System of Input Coordinates" click the Edit button, and set the coordinate system to Geographic Coordinate Systems→North America→NAD 1983.
  5. Right-click the boston.txt Events layer that appeared in the Table of Contents. Choose Data→Export Data. Name it and export it to the desired format (shapefile, file geodatabase, …).

If you follow the above procedures, the data align tolerably with 2010 vintage Census Bureau shapefiles, which also use the NAD 1983 lat-long coordinate system. Note that using NAD 1983 for the housing data is just a guess. WGS 1984 will produce virtually indistinguishable results (see http://mappingcenter.esri.com/index.cfm?fa=ask.answers&q=740). But if you need to do larger scale mapping, and find that the points don't seem to align with streets or other features of another layer you are using, you may have to do some research to determine the correct datum. (This last part taken from comments by @chris-w below.)

| improve this answer | |
  • I would have guessed WGS84 for the UTM datum initially, but WGS84 and NAD83 in the US will be largely indistinguishable. NAD27 would show a shift. I was thinking UTM had a specific datum as 'root'. This is mentioned in @whuber's answer at gis.stackexchange.com/questions/11345 but everything I can find suggests that 'real' UTM can use any of several datums, so having it specified along with the other parameters like zone could be important. Comparison/reference to known data is about the only way to go if you don't know. – Chris W May 21 '15 at 20:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.