Introduction

The crime variable was derived from data supplied by the Winnipeg Police Department, Crime Statistics Unit.

Contents of the data include both 1999 and 2000 calendar data for the crimes against persons, crimes against property, drug crimes, and other crimes. Calls for service have also been included for both years. Only the 1999 crime data was analyzed here.

Before using these data make sure you are aware of the limitations of the agreement made with the Winnipeg Police Service. Key points:
• The use of the data only for projects approved by the REB
• Data will not be released to any other person or group without consent
• Prior to publication the Winnipeg Police Service will be informed of any publication and will be provided a copy.

Crime Data

These data were provided as point data in Microsoft Access 2000 and MapInfo MIF format (for the year 2000). The MS Access data is aggregated at the level of enumeration areas.

The data was provided as 4 tables for 1999 and was aggregated at the following level:
• Violation
• Date
• EA

Calls for Service

The calls for service data has been aggregated for 1999 by call type. Call types include the types found in the Police Automated Records System (PARCS) these include: ALMH - holdup alarm, ASLWA -attempt murder, etc… Calls for service data can also be subdivided into several types (persons, property, disorder, other), very similar to those used for crimes.

Geographic Aggregation

Because police information for 1999 came according to EA, an analysis at the EA level could be done, if such fine-grained geographic divisions are required. There were few discrepancies at the Community Centre Area level because almost all EA's are divided at the same major boundaries as are the Community Centre Areas. The same is true for the Neighbourhood Cluster level and Community Area level. Division into the 228 neighbourhoods of Winnipeg may not be ideal, since the EA boundaries and 228 neighbourhood boundaries do not coincide as well.

See SAS code below (internal access only) for geographic aggregation at the 72 CCA level.

Principal Components Analyses

Principal component analyses look for the commonalities amongst the variables that are included. It may be used to determine whether a single underlying construct is determining the scores on all the variables, or if the variables can be considered truly independent of one another.

As an alternative to using the individual rates for 3 different crimes types (persons, property, other) and 3 different call for service types (persons, property, disorder), a principal components analysis indicated that a single score may be more appropriate. Principal components analyses of these data, using several different geographic aggregations: Community Areas (12) , Neighbourhood Clusters (23) , and Community Centre Areas (72) , indicate that a single principal components score (R ² = .91 - .98) may be used as the crime variable rather than using the individual measures of Crimes Against Persons, Crimes Against Property and Other Crimes plus the calls for service. In other words, all the different crime types will indicate the same relative differences in levels of crime between these geographic areas. If there are more property crimes in St James than Fort Garry, for example, then there will also be more drug crimes, other crimes, and crimes against persons.

See SAS code below (internal access only) for principal component analysis.

Table 1 presents the range of principal component scores for CA's and NC's. Figure 1 displays the distribution of principal component scores for the 72 Community Centre Areas.

As an indication of how the crime scores correspond with other indicators, correlation matrices for the Crime scores, premature mortality rates (PMR), ADG25 rates (a clinical grouper encompassing substance abuse diagnoses and many psychiatric diagnoses) and SEFI scores (socio-economic factor index) are also presented (see Tables 2 , 3 and 4 ).

See also Crime / Police Data Dictionary below (internal access only).