Monday, May 13, 2019

GSS Victimization - Territories and Geography

Question:
I have a question about microdata files available for the GSS (victimization) cycle 28, or previous. A researcher is interested in joining data from this survey with data from the Census and therefore requires more detailed geographic information than what is available with the DLI PUMF. In the PUMF, the geographic variables only 'go down' to the provincial level. The restricted dataset has many more geographic variables available, which would be better in this case.

However, the microdata files aren't consistent with their inclusion of data from the territories vs provinces. None of the DLI PUMFs contain data from the territories. The cycle 28 RDC dataset only contains variables for responses from the provinces, while the cycle 23 RDC dataset only contains responses from the territories.

Are there any microdata datasets (RDC/DLI) that contain responses from both the provinces and territories, that will have more specific geographic variables?

Answer:
I’ve received the following response from subject matter:

“It is never a good idea to ‘join data’ from two different surveys, given differences in in their populations, methodologies and time frames.

In general, the GSS is only carried out in the provinces, the exception being the Victimization cycle, last completed in 2014.

As of 2019, the GSS targets a sample size of approximately 20,000 respondents. Sometimes a cycle has a higher target sample size if funding has been received for an oversample, either in the form of a geographical sample top-up (i.e., adding more units in certain geographic areas),  a targeted oversample (e.g. focussing on immigrants, youth, or another population group), or a general oversample (i.e., increasing the raw sample size). With a final sample of 20,000 respondents, basic survey estimates are usually available for the national and regional levels, and for some provinces and census metropolitan areas. Depending on the survey topic, the sample size may be sufficient to produce estimates for certain population groups such as lone parent families, certain visible minority groups or seniors.”