When Canadian police officers close an investigation, they give it a code to signify the outcome. One of those closure codes is "unfounded," a term that means no violation of the law occurred or was attempted. Once a complaint is classified as "unfounded," it is no longer considered a valid allegation. So for example, if a police service received 100 sexual-assault complaints in a year, and 20 were determined to be unfounded, the service would report that it had received 80 allegations of sexual assault. Unfounded statistics for sexual assault have historically been dramatically higher than for all other types of crime.
Fourteen years ago, Statistics Canada stopped publishing unfounded rates, over concerns about the quality of the data. Officials identified two main issues. One, police services weren't always documenting unfounded cases, and so they weren't being reported to Statistics Canada. And two, police services were using the code improperly; as a result, allegations that simply couldn't be proved had sometimes wrongly been classified as unfounded. At the time, researchers criticized the Statscan move, warning that by not releasing unfounded data, it would be impossible to determine how many victims were reaching out to police and how many cases officers were dismissing as baseless.
In "Unfounded," The Globe and Mail has tried to fill the gaps in the data. We sent 250 freedom-of-information requests to every Canadian police service, requesting unfounded data for both sexual assault and physical assault (some police services, such as the RCMP and Ontario Provincial Police, are responsible for more than 100 jurisdictions each). We received responses from 873 police jurisdictions, which represent 92 per cent of the population.
As did Statistics Canada, we found inconsistencies and identified problems in the data. Some cases are still being miscoded as unfounded – and, in fact, we found several examples of this. It is also likely that some cases of sexual assault were never documented – The Globe found at least two examples. Finally, to collect its crime data from police services, Statistics Canada uses the Uniform Crime Reporting Survey. In the case of the "Unfounded" project, we have relied on freedom-of-information co-ordinators at individual police services, all of whom may use varying methodologies to collect and record the data.
We asked each service for the number of sexual-assault complaints received in a year and the number of sexual-assault complaints determined to be unfounded. Some forces have calculated a total number of complaints by adding the number of valid allegations (founded cases) with unfounded ones. For example, an analyst might record 80 sexual-assault allegations and 20 unfounded cases, to determine there were 100 total complaints. Other forces have gone into their call systems and recorded the total number of complaints that came through dispatch and were logged as a sexual-assault complaint, then tracked the outcome from there. In some cases, this could result in different numbers than the previously mentioned method. There was also the possibility that an FOI analyst misread the request and sent us a total of complaints that does not include unfounded cases.
Finally, some analysts have warned us that our data is not comparable to what is reported to Statistics Canada through the Uniform Crime Reporting Survey. In some cases, data may include historical complaints. In others, an allegation of sexual assault or assault may have been deemed unsubstantiated and cleared by the investigator using a different offence code. Also, one large police service cautioned that there may be instances when multiple victims are connected to one General Occurrence Report, further skewing the data.
All of this is to say: Unfounded data was flawed in 2002 and it continues to be flawed today. Nevertheless, the unfounded situation is a matter of huge public interest that we believe should be subjected to scrutiny. Another point to keep in mind: If there are problems with the unfounded data, it should theoretically affect both sexual-assault allegations and assault allegations equally. The Globe has found that sexual-assault allegations continue to be closed as unfounded at a rate that is nearly double that of physical assault – 19.39 per cent compared to 10.84 per cent.
This project unfolded in several stages. In June, 2015, The Globe requested sexual-assault unfounded data from about 50 police services, initially targeting major cities, provincial and territorial capitals and some selected rural areas. It took about eight months to receive responses from the jurisdictions that agreed to supply the numbers. The preliminary results showed wide variations between police services, with no discernible pattern. It became apparent that in order to truly capture what was going on, we would need to collect data from every police jurisdiction.
The Globe obtained a list of what is known as "police respondent populations" (PRPs) for 2014 from the Canadian Centre for Justice Statistics, a division of Statistics Canada. The list included 1,119 jurisdictions. The PRP titles included the name of the jurisdiction, the province, and the type of police service – municipal, Royal Canadian Mounted Police, Ontario Provincial Police, Sûreté du Québec or Royal Newfoundland Constabulary. This was the basis for The Globe's freedom-of-information/access-to-information requests.
We decided that for the second batch of requests, which were mailed in April, 2016, we would add a question about total assault complaints and unfounded cases to serve as a benchmark for the sexual-assault statistics.
Each police service was asked to provide the following information for a five-year period, 2010 to 2014, broken down by year:
1) The number of sexual-assault complaints brought forward each year.
2) The number of sexual-assault complaints determined to be unfounded.
3) The number of sexual-assault complaints resulting in a charge.
4) The number of sexual-assault complaints cleared with a departmental discretion code.
5) The number of allegations of assault – not sexual in nature – brought forward.
6) The number of allegations of assault determined to be unfounded.
Police services we had never surveyed before were sent all six questions. To those services from which we had already obtained sexual-assault unfounded and charge data, we sent only the last three questions. In some cases, police services that received both requests sent us updated sexual-assault information, even though we did not request it. Whenever we received two sets of numbers, we used the newest set.
We chose not to ask for the data for 2015 because, given the delays we encountered in the first go-around, it was more likely data would be readily available for 2014 and that cases from that year would be complete. We chose to ask for data over five years, because a longer period of time would help avoid uncharacteristically high or low unfounded figures. To calculate the five-year national rate, we totalled the number of cases cleared as unfounded over that period of time and divided it by the total number of allegations made over the five years. This enabled us to include almost all the police respondent populations that responded to our request for sexual-assault unfounded numbers in our five-year national rate, even those that were missing data for one or more years.
In addition to calculating a national rate, The Globe has ranked PRPs based on their five-year, sexual-assault unfounded rate in the project. PRPs that did not supply all five years' worth of data were not included in these rankings. Extremely small jurisdictions – the smallest 5 per cent in terms of population – were also not included. This helped to eliminate outliers and volatility due to the low number of cases.
Many police services did not respond with departmental discretion codes, because they don't use that code, which is why we have not included that information. A handful did not include data on charges. Some provided information only for select years. In the cases where a zero was supplied, it was entered as numeric data. But in the cases where chart fields were left blank, it was interpreted as no data supplied. In instances where a service wrote "nil," we interpreted this as a numerical zero. In general, our policy was to trust that the data we received was correct. However, in some instances when there was an obvious mistake or the response was confusing, we contacted the analyst for clarification.
All the information is available on a jurisdiction-by-jurisdiction basis in the searchable data feature.
While the assault numbers are available on a PRP basis, we did not include them in our overall analysis beyond a national average comparison because many police forces did not provide assault numbers.
One of the challenges of sending freedom-of-information requests to so many police services was dealing with the starkly different practices in each jurisdiction. For example, many provinces charge a fee to file, ranging from $5 to $25, and then reserve the right to charge additional fees for the records themselves, based on the time required to assemble them. With a few exceptions, only police services within Ontario demanded additional payment for the statistics. Those fee estimates ranged from free to more than $1,360. In the end, The Globe agreed to pay between $30 and $120 for the information, depending on the size of the file and the resources available to the service. Most police services eventually agreed to lower their estimates. At the time that we filed these requests, Saskatchewan police were not subject to freedom-of-information legislation, so any municipal statistics from that province were volunteered by the individual police forces.
A reporter used a spreadsheet to track the progress for each individual police respondent population, adjusting from "sent" to "received" to "processing" to "complete." Once a file came in complete, it was passed off to the data team, where one individual entered it into an intake form. A second person proofed it. Another team member took a third look before publishing. Finally, a data specialist not connected to the project double-checked all the calculations.
In order to create the mapping aspect of this project, The Globe obtained another data set from Statistics Canada that included what are called RESP codes – a unique identifier assigned by Statistics Canada – for each PRP (police respondent population). The Globe then matched our original PRP list with the new one. There were a few examples where PRPs on our first list did not match up with our second list. In these few cases, The Globe got help from Statistics Canada to match them up. On Statscan's advice, The Globe used 2014 population statistics included in the second spreadsheet.
The Globe examined the correlation between the number of female officers and a police service's unfounded rate. We did this by comparing Statistics Canada's Police Administration Survey from 2014 (CANSIM table 254-0004) to our own data. The administration survey has data on female officer staffing for 458 municipalities. However, we could only use data from 234 of them for two reasons. One, some of the PRPs did not provide The Globe with data for 2014. Two, some of the jurisdictions included in the administration survey were not compatible with our data because of our need to group certain PRPs together. (We had to group some PRPs together because of the way police services responded to our request.)
The search function in the data feature uses a data set from Statistics Canada that matches the RESP codes to census subdivision areas (CSD). This allows for a broader list of places to be searched. For example, Halton Region (west of Toronto) includes Oakville, Burlington, Milton and Halton Hills. Any of those places, as well as Halton Region, can be searched in order to find Halton Region statistics. In many cases, CSDs change RESP codes over the years. Our search function is based on the RESP code that the CSD belonged to in 2014.
One challenge we faced in processing the replies was that police forces responded in several ways. Some answers came in chart form with a single number to represent each answer. Some came as a list of individual charges, which we needed to add up to reach a total. Further complicating matters was the fact that, while our request specified we only wanted "sexual assault" and "assault" complaints, the response of some police services noted that their data included other types of sexual offences, such as voyeurism, child luring and sexual interference. Others specified that their response included only the three levels of sexual assault. Most did not specify one way or the other.
In the end, we set a policy that we would include whatever information police services sent us in reply to our request, rather than try to make judgment calls about whether it would be appropriate to sever certain records. It also would not have been possible to sever some of the responses, where a single number was provided and the service indicated it included other sexual offences.
Experts who have reviewed our data warn that unfounded rates from police services that provided statistics for all sexual violations could have artificially low unfounded numbers, as the other, less-complicated charges in that category would dilute the figure. To test this theory, for the police services that gave us responses that included other types of sexual offences – and which also gave us a line-by-line breakdown, enabling us to identify only the sexual-assault allegations – we calculated the unfounded average if only sex-assault complaints were considered. Generally, there was little difference between the two unfounded averages – typically 1 per cent to 3 per cent. Some were in fact slightly higher once other sexual violations were excluded, but others were slightly lower. The biggest difference between the two averages occurred in Guelph, Ont., with a difference of 7 percentage points. Each police jurisdiction where this applies includes a note.
In rare instances, police forces sent a breakdown of offences that included files in which they were assisting other police forces. In keeping with our policy, we included these entries.
The other key hurdle we encountered was when police services – particularly the RCMP, Ontario Provincial Police and Sûreté du Québec – sent us data for jurisdictions that did not match the PRP list. This would be as a result of the service grouping together multiple PRPs into a larger jurisdiction. In these instances, The Globe combined the affected PRPs into one entry. All entries affected by these types of groupings have notes on them. For the most part, only small areas needed to be combined. The exception was Quebec. The Sûreté du Québec did not respond to our request for each of its 87 individual PRPs' statistics. Instead, the provincial police sent us total figures for five of its 10 districts. To deal with this, we matched up the PRPs to their respective districts. If you search for one of the PRPs in those districts, you will be directed to the district's numbers.
In general, any PRPs affected by something out of the ordinary include notes.