Special Characteristics
Definitions and Principles of Use for Not Elsewhere Classified (n.e.c) and Not Specified (n.s) Codes
Not elsewhere classified (n.e.c.) and not specified (n.s.) codes appear frequently in the CCRI code sets and play similar, but distinct roles. Both are general codes that act as catch-all categories for data that does not fit in any other column.
N.e.c is used in cases where a response to a particular census question is relevant (i.e. it is clearly an answer to the question, as opposed to something written in the incorrect column) and often quite detailed, but where no specific code exists to match the response. When a researcher sees n.e.c. in a code description they can assume that code captures values that are specific but do not have their own distinct codes.
For example, if 4 people responded “Gananoque” in the Other Languages Spoken column, we can establish that this person spoke an Aboriginal language. We can also assume Gananoque refers to the town in Ontario, which is located in the traditional territory of the Six Nations. Given the small number of people that gave this response, we do not want to create a separate code for “Gananoque.” Instead, these values would be mapped to the code Iroquoian, n.e.c.
N.s. is used in cases where a response to a particular census question is relevant (i.e. it is clearly an answer to the question, as opposed to something written in the incorrect column) but general, so that it is impossible to map it to a detailed code. When a researcher sees n.s. in a code description they can assume that code captures values that are too general to be captured by any of the more detailed codes in the set.
For example, hundreds of responses of “Native” or “Native Indian” were given to the question of Racial or Tribal Origin. It is impossible to tell exactly which aboriginal group or band these responses belong to, but they can be mapped to a distinct “Native, n.s.” code.
Note: N.s. codes are not the same as parent codes. Some parent codes are very general (i.e. Aboriginal Peoples), and the temptation is to code general terms to these parents instead of to the specific n.s. code. Parent codes group similar codes together for easy aggregation. In the CCRI Primary Codes, parent codes act as headings. As a result, no values should be mapped to parent codes. For example, codes for each province of Canada all have Canada as a parent code. General responses, like “Canada” should not be mapped to this Canada parent code – they must be mapped to a specific “Canada, n.s.” code. If general values are mapped to the parent code, researchers will not be able to exclude these values when aggregating data. For example, a researcher would not be able to look for all Aboriginal Peoples that gave specific responses to the Racial and Tribal Origins question excluding those that have general responses like “Native.”
9-series Codes
In each code set there is a series of codes that is eight digits long and begins with 99999. In the CCRI database, all verbatim responses must receive a code. This means that even in cases where responses were illogical, illegible, or blank, they needed to be coded. The 9-series codes deal with these cases. They identify illegible, illogical, or damaged data to improve the overall quality of the data set. Many of these codes are the result of decisions made by the Data Entry Operators (DEOs) or during the Cleaning process. Below is a brief desciption of the purpose of each 9-series code.
High-frequency 9-series codes:
99999001 – Blank
In cases where the column of the census was not filled in, the blank answer was noted and mapped to this code. This code can be used to determine how many blank responses exist for a given census question.
99999003 – Illegible
In cases where the response was not legible, Data Entry Operators either flagged the response as illegible, or attempted to decipher part of the response and used question marks to indicate letters they could not read. In either case, these responses were mapped to the Illegible code.
99999006 – Missing – Mandatory Field
In cases where the column of the census was not filled in and the question was one the Dominion Bureau of Statistics deemed “Mandatory,” the missing data was noted and mapped to this code.
99999011 – Unknown - Suggestion
When the verbatim response to a given question was “Unknown”, DEOs had the option of flagging the response as “Unknown” in the suggestion window. These values were then mapped to this code.
99999901 – None
When the verbatim response to a given question was “None” (and “n”, in some root variables), it was mapped to this code.
99999902 – Not Given
When the verbatim response to a given question was “Not Given” or a variation of that response, it was mapped to this code. Please note that this code is different from Blank and Missing codes, where no verbatim response was given.
99999903 – Unknown
When the verbatim response to a given question was “Unknown”, it was mapped to this code.
99999905 – Uncodable
When a response was not a logical or comprehensible answer to the question being asked, it was mapped to this code.
Low frequency 9-series codes:
99999002 – Damaged
In cases where the response was not legible because the census enumeration form was damaged, Data Entry Operators flagged the response as damaged. These responses were then mapped to the Damaged code.
99999004 – In Error
In some cases the software prohibited the DEOs from entering the enumerator”s response. For example, in a column where the software anticipated a numeric response like a weekly salary, but the enumerator wrote “Income,” the software did not allow the DEO to enter “Income.” Instead, the DEO flagged the response as “In Error” and recorded “Income” as an Alternate Value. The Alternate Value was then promoted as the cleaned value. If no Alternate Value was suggested, the response was coded to “In Error.”
99999005 – Suspicious
In cases where the DEOs believed an enumerator”s response was logically incorrect, they flagged the response as “Suspicious.” For example, if an individual answered “Male” in the Sex column and “Wife” in the Relationship to Household Head column, the DEO flagged the response “Wife” as suspicious. They had the option of entering an Alternate Value, which would be promoted as the cleaned value. If no Alternate Value was suggested, these responses were then mapped to the Suspicious code.
99999007 – Not Applicable
Responses mapped to this code indicate that the column in question did not exist on that particular census form for that particular individual. For instance the question “Infirmities” only appears on the 1921 Form C (for the Northwest Territories). Individuals captured from Form types A or B would have this question mapped to the Not Applicable code.
99999008 – Not Mapped
Responses mapped to this code are responses for which the extract software could not find a code that corresponds to the response. This indicates that this response was overlooked in the coding process; it has not been coded.
99999009 – Correction
In cases where a verbatim response was “corrected” or overwritten by census officials, the DEOs flagged the response as a “Correction.” They had the option of entering an Alternate Value (except for the Nationality, Racial or Tribal Origin, and Occupation questions, where they were asked to enter the “correction” in a separate field). However, the verbatim response was still promoted to the cleaned value, and coded. If no Alternate Value was suggested, these responses were then mapped to the Correction code.
99999010 – Suggestion
In cases where the DEO was unsure of the verbatim response, they could suggest an Alternate Value. This Alternate Value was promoted as the cleaned value. If no Alternate Value was suggested, these responses were then mapped to the Suggestion code.
99999904 – Invalid Value
Responses mapped to this code are those that have been “stuck” in the data promotion process.
