Leaving School - Results From a National Survey Comparing School Leavers and High School Graduates 18 to 20 Years of Age - January 1995
| Previous | Contents | Next |
Appendix A. SLS Methodology
1. Introduction
This appendix describes the methodology used by Statistics Canada in conducting the 1991 School Leavers Survey. It presents some of the problems associated with identifying and contacting a sufficient sample of school leavers to produce reliable estimates. These problems include choice of sampling frame and appropriate stratification. difficulties in tracing, and considerations for weighting.
The two primary objectives of the survey were:
- to develop comparative profiles of three groups of secondary school attendees, those who successfully completed secondary school (graduates), those still attending (continuers) and those who left school before receiving a diploma or certificate (leavers).
- to establish rates of leaving school before graduation in Canada and the Provinces.
The first part of this appendix provides a brief description of the reasons for the choice of sampling frame, the target population. stratification and sample size determination, as well as a description of how the 1991 School Leavers Survey was implemented. In the second part of the appendix, the contact and response rates are provided. The last part of the appendix indicates potential bias due to the tracing difficulties and the attempts to adjust for it.
2. Design
2.1 Sampling Frame
In the very early stages in the design of the School Leavers Survey. approval was obtained to use the Family Allowance Files as the survey frame. The Family Allowance program is federal and provides a monthly payment to a supporting parent or guardian of young persons in Canada. It covers Canadian citizens, landed immigrants as well as anyone legally residing in Canada until the month that the young person turns eighteen years of age. Payments will cease earlier than age eighteen when the parent or guardian ceases to provide financial support for the young person.
At the time the decision was made to use these files, it was believed that they were the most complete listing of young persons under age 15 in Canada. After the age of fourteen, there is an increasing decline in the coverage of each age group. Given that the basis for these files is as a payment to the supporting individual. it was believed that the address information would be kept up-to-date. There is also a variable on the files which gives the reason why payments would have stopped prior to age eighteen. The reasons "parent no longer supports child", "child is married. and "child has income" were thought to pre-identify those young persons who were more likely to be leavers than other individuals.
There are Family Allowance files for every year as well as individual ones for each province. One of the first steps in creating a frame was to link records from year to year based on an identification number that does not change as long as the individual remains within the same province. Unfortunately, when an individual migrates from one province to another, the same identification number is not used. There are codes that tell the user of the files that the individual has been transferred to a specific province. On the other hand, there are also codes on the provincial file where the individual next appears that indicate the province from which the individual moved. Using these codes and the name and age of the individual, matching between provincial files was performed.
Matching of records was also required if the support for the young person was transferred from one parent to the other, or from a parent to an agency, etc. These agencies were mainly orphanages. transition homes and children's aid societies.
There is however one potential problem in using the Family Allowance files. The survey was to be conducted with the young person for which the payment is provided. However, the personal information maintained on the files specifically for that individual is first name and birth date. Since the files are maintained mainly to send payments to the supporting person for the child, most of the information, such as surname and address, relates to the supporting person. The decision was to assume at first that the young person had the same surname as the individual receiving the payment. In most cases this would be the mother. Situations when mother and child do not have the same surname are a practical problem which the interviewers would later have to confront.
Five years of Family Allowance Files were used to generate a sampling frame of 18-20-year-olds.
2.2 Target Population
The ages of eighteen, nineteen and twenty were chosen as the target age group for two reasons. First to obtain enough leavers and graduates in our sample and to be able to evaluate the differences in their labour market experiences, the project team decided that a selected individual needed to be twenty years of age. This would allow these individuals to have the normal length of time to complete secondary school and be in the labour market for at least a short period of time. The team also decided to cover the age of eighteen for the opposite reason, that is to ensure obtaining enough continuers. The suitability of selecting these three ages to answer the required objectives was decided upon using the results of a test conducted in 1990.
Choosing these three ages created a problem in using the Family Allowance files. As mentioned earlier, the Family Allowance program supports individuals only up to the age of eighteen. Thus all of our selected persons would no longer be covered by the program and the addresses would be out-of-date. For instance, the twenty-year-olds would be last on a file three years prior to the survey year. To compound this problem, for those persons whose payments had stopped prior to age eighteen, their addresses would be even further out-of-date.
One property of the Family Allowance files helped in finding the most current address for the sampled persons. Families are maintained together on the files, and as long as payments are being made to a younger person within that family, the record of the sibling whose payments have stopped is kept on the file. Thus, it was possible to obtain a more current address for those individuals who had younger siblings.
All ten provinces were covered in the survey but the Yukon and Northwest Territories were excluded mainly due to cost considerations.
Two further adjustments were made to the target population. The first was to exclude individuals where the recipient of the payment was an agency. In addition, during the creation of our families. to find the most current address, it was discovered that there were some cases where there was a large number of individuals within the same family who were within eighteen to twenty years of age. It seemed quite unlikely that a mother could have as many as eight children within three years of age. Further investigations pointed out that this was happening within Saskatchewan mainly and it was believed that these cases were Hutterite communities where one woman receives the payments for all families in the community. To control response burden, the decision was made to select at most two individuals within the same family.
2.3 Stratification
Once the target population was decided on and the frame was created from matching of the 1986 to 1990 Family Allowance files, stratification was then performed. Variables that were available on the records to use for stratification purposes were age, province of residence and payment status (the variable that could potentially identify a greater proportion of leavers). Two tests were performed in 1990 to evaluate the ability of this chosen variable to pre-identify potential leavers. Sex was not available on the files to use as a stratification variable.
Table A-2 shows the population count for the frame created for the School Leavers Survey. Each age and province cell was subdivided using the payment status code variable to preidentify potential leavers to create two further breakdowns known as the "Potential. stratum (where they had one of the chosen codes) and the "Non-Potential. Stratum. The effectiveness of these codes to preidentify potential leavers will be shown later in the results of the survey.
2.4 Sample Size and Selection
To estimate school leaver rates, only the twenty-year-olds would be used. This decision was taken since there would be less chance of twenty-year-olds being continuers in secondary school. "Continuer. can be thought of as a "temporary state" which results in either "dropping out" or "graduating". All three ages. together would be used to analyze the differences in the characteristics of the continuers, leavers and graduates since there would not be enough graduates at age eighteen nor enough continuers at age twenty to use only one of the age groups.
Based on the results of the testing done in 1990, approximate numbers of leavers, graduates and continuers were obtained for each stratum. Approximate response rates and trace rates from the pilot test were also available. To provide national and provincial leaver rates for twenty-year-olds within a coefficient of variation of 16.5%, and to have continuers, leavers and graduates, each considered separately, possess some characteristic from the questionnaire estimated within a coefficient of variation of 16.5%, a final sample size of 18,000 was decided upon. Table A-3 provides the sample size for each stratum. A simple random sample was taken in each stratum, except of course for those strata which were take-all.
2.5 Obtaining Telephone Numbers
Telephone interviews were chosen as the method of collecting the data. It was felt that the length and complexity of the questionnaire would create problems for a mail survey both in terms of lower response rates and difficulty in being self-enumerated. As well. it was felt more efficient to conduct the tracing over the telephone and once contact was made using this mode, it would not be a substantial increase in costs to complete the questionnaire using the same mode. Personal interviews were ruled out because of the cost. Unfortunately, the Family Allowance files contain addresses but no telephone numbers are available. Telephone billing files are available at Statistics Canada and were matched to the addresses on the frame to assist in the tracing process. From testing done in 1990, it proved beneficial to provide the interviewers with a number of telephone numbers from which they could attempt to trace the selected individual In fact, five methods of matching between the sample and the telephone files were performed. Each of the five methods is given below along with a description of the reasons for its choice.
Match 1: by surname, exact address, postal code. This match should be the best method to contact the guardian or the selected young person.
Match 2: by surname, partial address, postal code. The removal of the civic portion of the address was done to accommodate slight coding differences between the Family Allowance and telephone billing files.
Match 3:by surname, postal code. This match was performed to allow the possibility of another family within the same small geographic area with the same surname knowing the whereabouts of the selected person or family. In this case we would be finding either relatives or persons who over time might have been getting each other's mail or phone calls. These individuals could know of the movement of the selected individual.
Match 4: by exact address, postal code. In this case, a new family could move into the same dwelling in which the selected person had lived and possibly would know of an address to which to forward mail to. As well, we could have used the wrong surname for matching. The surname used from the Family Allowance file in most cases would be the mother's surname while the telephone files could contain the father's surname. This was an attempt to allow for different surnames within the same household.
Match 5:by exact previous address, postal code. In the event that the matching of records for a family between provinces was incorrect, this last match was performed. This match also accommodated those situations where the support for a selected individual was transferred from one parent to the other and the previous parent knew the whereabouts of the selected individual, as well as the situation where the people who lived at the previous address of the family would know of the whereabouts of the selected individual.
The School Leavers Survey was conducted by telephone on a Computer Assisted Telephone Interviewing (CATI) system. The matched telephone numbers were loaded into the system and the interviewers would try each of the numbers provided in an attempt to trace the selected individual. Up to thirty-six telephone numbers were provided. If all telephone numbers proved unsuccessful, the interviewers would use other typical tracing procedures to locate our selected family or young person, such as phoning neighbours found through city directories or contacting directory assistance.
The first step that faced the interviewers was to confirm whether the selected individual was at the telephone number reached or was a member of the family living at another location with a different telephone number. If the response indicated that such a person was not known, then the interviewer tried to establish one of two scenarios. The first was to try the other surname provided for the spouse on the Family Allowance files as the surname for the selected young person and determine whether that person lived there or was known to the family. Failing this, the interviewer then tried to identify whether the correct family had been located but there was a problem with the selected person. To do this. interviewers were provided with names for up to three siblings as well as the information on the spouse available from the Family Allowance files. If it was confirmed that the correct family had been located but the selected person no longer resided at that household, the interviewers tried to obtain a telephone number so that the tracing could continue for the selected person. If this proved unsuccessful, then the case became one of "traced family but untraced respondent". The frequency of this occurrence is provided in section 3.2.
2.6 Implementation by Wave
The size of the sample for the School Leavers Survey was 18,00. selected individuals. To ensure that all cases had the same chance of being tried, a decision was made to implement the survey in a series of five waves, each wave to be in the CATI system for relatively equal amounts of time. The sample was divided into four parts, each part being representative of the total sample. The first part of the sample was put into the CATI system as wave one. When the time given for wave one had expired, the unresolved cases were removed except for those which had appointments for call-back. The second wave was then put into the CATI system along with the appointments for its allotted time. All four parts of the sample were handled in this fashion as waves one to four waves back into the system and further attempts were made for these four. The fifth and final wave then put all the unresolved cases from the first four waves back into the system and further attempts were made for these.
3. Contact Results
3.1 Hit Rates
The first result to be presented is that of a "hit rate". By this term, we mean the proportion of respondents that were classified as being leavers. One of the primary reasons for using the Family Allowance files was to be able to preidentify leavers as much as possible so that a sufficient number could be found for analysis. Table A-4 shows the results of this stratification by province for the three age groups collapsed. The term "ever left" refers to the fact that at one point in time, the individual had left but now could be classified as a continuer, leaver or graduate as some had in fact returned to secondary school. This table shows that a much higher rate of finding leavers was found in the potential stratum over that in the non-potential stratum. Thus the use of this file did in fact enable more leavers to be found than would otherwise have been possible using a different frame that did not provide such information.
3.2 Tracing and Response Rates
Interviewers were instructed to make all reasonable attempts to obtain interviews with respondents. For individuals who at first refused to participate in the School Leavers Survey, interviewers stressed the importance of the survey and the respondent's cooperation. For cases in which the timing of the interviewer's call was inconvenient. an appointment was arranged to call back at a more convenient time. For cases in which there was no one home, numerous call backs were made. Under no circumstances were sampled individuals replaced by other persons for reasons of non-response.
Table A-5 presents the tracing and response rates achieved at the Canada level for all three ages together. "Recipient. is used to indicate whether or not success was achieved in finding the family of the selected young person while "respondent. refers to the selected individual. This table shows that for 34. of the sample there was no success in finding either the selected individual nor the parent or guardian. However, the response rate once the selected individual is traced is high: of the 10782 youths traced, 9460 youths were completed responses and in-scope to the survey (88%).
One other point to make from this table is that for 6% of the sample, the parent or guardian was traced with success but the selected person was not traced. These cases were those where the family had lost contact with the young person.
Table A-6 shows the response rate for twenty-year-olds by province and potential versus non-potential stratum. The twenty-year-olds were the targeted age group to estimate the national and provincial leaver rates. The response rate for the non-potential stratum is higher than that for the potential stratum. Table A-5 shows that once an individual was traced, a response was obtained 88% of the time. From this one could say that difficulty in tracing was the major cause of the non-response. Thus the lower response for the potential stratum is due to the inability to trace the family.
One other factor to be taken into account is the fact that for the potential stratum, by definition, the Family Allowance payments had ceased. This meant that the addresses on the files were more out of date than those for the non-potential stratum. This factor on its own could be a major reason for the response differences.
4. Corrections for Potential Bias
During the data collection phase, it was noticed that the proportion of completed cases within a province was decreasing as one moved from the East coast of Canada to the West. It was determined that. because of the time zone differences, a larger proportion of the interviewer's shift was being applied to the eastern provinces than to the western provinces. The problem with this happening was that the provincial estimates from the West would not be as reliable as those from the East. Another danger was the fact that the non-response was linked to the inability to trace the selected individual, as discussed in section 3.2. It was also discussed in that section that the harder to trace individuals could, in fact, more likely be leavers resulting in the estimates of leaver rates for the West being too low. To rectify this problem, priority was given to the four western provinces during the final phase of data collection to increase their respective response rates. The final response rates for twenty-year-olds for the survey are given in Table A-6. Similar results were found for the other two age groups.
After completion of the survey, an interesting result was found in the leaver rates for twenty-year-olds by province. Table A-7 presents rates for "Ever left secondary school" for the twenty-year-olds by province. Again one notices a decreasing rate from the East coast of Canada to the West. A concern was raised that this decreasing leaver rate was somehow linked to the method of handling the problem described in the previous paragraph. It was hypothesized that because the priority given to the western provinces was not constant over the collection period, then the proportion of the resolved cases that consisted of the harder to trace individuals was not the same in the West as it was in the East. It seemed to be a reasonable assumption that the hard to trace were in fact more likely to be leavers thus producing the lower leaver rate. On the other hand, it was also a reasonable assumption that the estimated leafier rates were showing the actual situation in Canada. The possibility that a bias had been introduced to the estimates was a concern. Investigations were done to analyze and adjust for this potential bias. The next points briefly describe these investigations.
Wave Implementation: Section 2.6 discussed the fact that the School Leavers Survey had been implemented by wave. Two factors were considered to be reasons to look at the marginal differences in finding leavers by wave. First, the interviewers for this survey had very little experience with the CATI environment or in conducting any type of survey. Hence they might have been more successful in tracing leavers in the latter part of collection. The second reason was the priority given to the West during the last wave because of finding the lower response rates. The investigation attempted to find a marginal increase in the leaver rates for respondents by wave. This pattern did not emerge and so no adjustment for the potential bias was made using wave.
Number of Attempts/Calls: The CATI system provides counts of the number of attempted calls and the number of completed calls for each case. If one assumes that leavers are harder to contact and therefore require more attempted calls before a completed call is achieved, one could make similar assumptions about the school leaving status of the unresolved cases. That is, unresolved cases would be expected to include a proportionately higher number of leavers. This "estimate" of the number of additional leavers contained in the unresolved portion of the sample could then be used to adjust the bias in the estimates. However, analysis of the data from the sample which required numerous calling attempts before a contact was made showed results which did not support the original assumption, and therefore could not be used as a means of bias adjustment.
Sex: In section 2.3, mention was made of the fact that there was no sex variable on the Family Allowance file. However, the results from the survey showed that there was a difference between the leaver rates by sex. Post-stratification by sex was performed using the following method. From the respondent file, the probability of being male given a first name was calculated. For those names with small frequencies, manual judgement was used to assign either 100%, 50% or 0% values. Then for the non-respondents, since first name was on the Family Allowance file, the probability of being male for that name, calculated from the respondent file, was used along with a random number to assign a sex to the records.
Demographic Projections: Since the target population for the survey was to cover all individuals aged eighteen to twenty in the ten provinces, comparisons were made between the estimates of total population achieved from the survey with those from demographic projections. The numbers were in fact quite similar although not exact. It was decided not to adjust to reflect the demographic counts for two reasons. The first reason was that the province of interest for the survey was the province last in school and not that of current residence. Another reason, although not as significant as the first, was the fact that the demographic projections were based on the 1986 census and could be somewhat dated. The 1991 census figures were not available at the time.
To conclude this section on the corrections for potential bias. the only adjustment performed was to post-stratify by sex. The other investigations did not lead to methods that could adjust for a bias nor did they rule out that the bias did not exist. On the other hand, the estimates of school leaver rates may in fact reflect the reality of Canadian society in that there are differences between the Eastern and Western provinces. What one can say is that if there is any bias in the estimates, then these estimates are lower than the true rate. The magnitude of bias. if any, is unknown.
5. Weighting Procedures
5.1 Basic Weight
In a probability sample, the sample design itself determines weights which must be used to produce unbiased estimates of the population. Each record must be weighted by the inverse of the probability of selecting the person to whom the record refers. In the example of a 2% simple random sample, this probability would be .02 for each person and the records must be weighted by 1/.02=50.
5.2 Non-Response
Notwithstanding controls, some non-response is inevitable, despite all the attempts made by the interviewers
Non-response was compensated for by proportionally increasing the weights of responding youths. The weight of each responding record is increased by the ratio of the number of youths that should have been interviewed, divided by the number that were actually interviewed. This adjustment was done in each stratum including those created by the post-stratification by sex. It is based on the assumption that the youths that have been interviewed represent the characteristics of those that should have been interviewed. To the extent that this assumption is not true, the estimates will be somewhat biased.
5.3 Weighting
The principals behind the calculation of the weights for the SLS are the following:
- An original sampling weight
- An adjustment due to the post-stratification done by sex
- An adjustment to account for non-response to the SLS.
6. Sampling Error
The SLS produces estimates based on information collected from and about a sample of individuals. In sample surveys, since inference is made about the entire population covered by the survey on the basis of data obtained from only a part (sample) of the population. the results are likely to be different from the "true. population values. The true population values in this context refer to the values that would have been obtained when the entire population was enumerated under the same general survey conditions. The error arising due to drawing inferences about the population on the basis of information from the sample is termed sampling error.
Since it is an unavoidable fact that estimates from a sample survey are subject to sampling error, sound statistical practice calls for researchers to provide users with some indication of the magnitude of this sampling error. This section of the appendix outlines the measures of sampling error which Statistics Canada commonly uses and which it also urges users producing estimates to use.
The sampling error, in addition to the size of the sample, depends on factors such as variability in the population, sampling design and method of estimation. For example, the sampling error depends on the stratification procedure employed, allocation of the sample. choice of sampling units and method of selection employed.
The accuracy of estimates from sample surveys is affected by both variance and bias. Under the assumption of simple random sampling within each stratum and with the further assumption of absence of bias, the variance of an estimated characteristic value is a good indicator of its reliability. Since the true variance of the estimate depends, like the estimate itself, on the whole population. it must be estimated from the available sample.
A notable feature of probability sampling is that the quality of the estimates may be estimated from the sample itself. The estimated coefficient of variation (cv) is defined as the ratio of the square root of the estimated variance to the estimate itself. Guides to the potential size of sampling errors are provided by the estimated coefficients of variation. The quality of the estimate increases as the corresponding coefficient of variation decreases.
| Type of Estimate | CV (in %) | Guidelines |
| 1. Unqualified | 0.0 - 16.5 | Estimates can be considered for general unrestricted release. Requires no special notation. |
| 2. Qualified | 16.6 - 25.0 | Estimates can be considered for general unrestricted release but should be accompanied by a warning cautioning subsequent users of the high sampling variability associated with the estimates. Such estimates are identified by the symbol *. |
| 3. Confidential | 25.1 - 33.3 | Estimates can be considered for general unrestricted release only when sampling variabilities are obtained using an exact variance calculation procedure. Unless such variances are obtained, such estimates are deleted and replaced by dashes (-). |
| 4. Not for Release | 33.4 or greater | Estimates cannot be released in any form under any release OR circumstances. Such estimates are deleted and replaced by dashes (-) |
Note: These sampling variability guidelines should be applied to rounded estimates.
In this publication, where the cv is 0.0 - 16.5, the estimate is unqualified and there is no special notation in the text. For cv's 16.6 - 25.0, the estimate is qualified by an asterisk (*. which indicates that high sampling variability is associated with the estimate and it should be viewed with caution. For cv's above 25.1, the sampling variability is too high to release an estimate (Table A-1).
7. Summary
Much was learned from conducting the SLS. There may be a potential bias problem in the methodology used for this survey. More effective tracing methods or adjustments to correct for the bias should be examined before a similar survey is conducted. As well, the choice of the sampling frame should also be readdressed. One problem with the use of the Family Allowance files at this particular time in Canada is the federal government's announcement in the last budget of considerations to end the universality of the Family Allowance program. This would certainly reduce one's belief that the files were a complete census of young persons under the age of fifteen in Canada. Statistics Canada does have plans to look at alternative vehicles to conduct any future School Leavers Survey.
| Province | 18 Years Old | 18 Years Old | 19 Years Old | 19 Years Old | 20 Years Old | 20 Years Old |
| Potential | Non-Potential | Potential | Non-Potential | Potential | Non-Potential | |
| Newfoundland | 111 | 11,424 | 172 | 11,910 | 181 | 11,703 |
| Prince Edward Island | 25 | 2,003 | 32 | 2,195 | 54 | 2,122 |
| Nova Scotia | 132 | 13,576 | 163 | 14,167 | 175 | 14,357 |
| New Brunswick | 185 | 11,998 | 270 | 12,261 | 302 | 12,281 |
| Québec | 557 | 86,514 | 1,096 | 88,579 | 1,046 | 91,928 |
| Ontario | 1,618 | 135,231 | 1,922 | 137,478 | 2,006 | 142,403 |
| Manitoba | 144 | 16,174 | 200 | 16,413 | 215 | 16,939 |
| Saskatchewan | 277 | 15,283 | 329 | 15,392 | 356 | 15,563 |
| Alberta | 360 | 34,801 | 499 | 35,567 | 644 | 36,919 |
| British Columbia | 443 | 40,349 | 460 | 41,159 | 559 | 44,056 |
| Canada | 3,852 | 367,353 | 5,143 | 375,121 | 5,538 | 388,271 |
| Province | 18 Years Old | 18 Years Old | 19 Years Old | 19 Years Old | 20 Years Old | 20 Years Old |
| Potential | Non-Potential | Potential | Non-Potential | Potential | Non-Potential | |
| Newfoundland | 111 | 209 | 172 | 173 | 181 | 469 |
| Prince Edward Island | 25 | 105 | 32 | 108 | 54 | 566 |
| Nova Scotia | 132 | 228 | 163 | 217 | 175 | 475 |
| New Brunswick | 185 | 165 | 270 | 105 | 302 | 368 |
| Québec | 557 | 353 | 575 | 415 | 330 | 370 |
| Ontario | 865 | 480 | 742 | 523 | 330 | 370 |
| Manitoba | 144 | 241 | 200 | 205 | 215 | 435 |
| Saskatchewan | 277 | 163 | 253 | 177 | 356 | 344 |
| Alberta | 360 | 290 | 380 | 270 | 330 | 370 |
| British Columbia | 443 | 247 | 411 | 289 | 330 | 370 |
| Canada | 3,099 | 2,481 | 3,198 | 2,482 | 2,603 | 4,137 |
| Province | Potential | Non-Potential |
| Newfoundland | 93% | 24% |
| Prince Edward Island | 92% | 24% |
| Nova Scotia | 87% | 23% |
| New Brunswick | 89% | 20% |
| Québec | 86% | 23% |
| Ontario | 83% | 18% |
| Manitoba | 86% | 22% |
| Saskatchewan | 78% | 17% |
| Alberta | 87% | 17% |
| British Columbia | 82% | 17% |
| Non-Response | Out of Scope | Response | Total | |
| Untraced Recipient | 6065 | - | - | 6065 |
| Untraced Respondent | (34%) | - | - | (34%) |
| Traced Recipient | 1134 | 19 | - | 1153 |
| Untraced Respondent | (6%) | (0%) | - | (6%) |
| Traced Recipient | 1151 | 171 | 9460 | 10782 |
| Traced Respondent | (6%) | (1 %) | (53%) | (60%) |
| Total | 8350 | 190 | 9460 | 18000 |
| (46%) | (1%) | (53%) | (100%) |
| Province | Potential | Non-Potential | Overall |
| Newfoundland | 47% | 69% | 63% |
| Prince Edward Island | 43% | 59% | 58% |
| Nova Scotia | 38% | 60% | 54% |
| New Brunswick | 49% | 67% | 59% |
| Québec | 48% | 59% | 54% |
| Ontario | 43% | 65% | 54% |
| Manitoba | 41% | 59% | 53% |
| Saskatchewan | 29% | 65% | 46% |
| Alberta | 42% | 56% | 49% |
| British Columbia | 38% | 56% | 48% |
| Canada | 41% | 62% | 54% |
| Province | Estimated Rate | Coefficient of Variation |
| Newfoundland | 25.9% | 7.1% |
| Prince Edward Island | 26.8% | 6.2% |
| Nova Scotia | 25.0% | 7.7% |
| New Brunswick | 24.7% | 8.3% |
| Québec | 27.3% | 8.4% |
| Ontario | 24.7% | 8.6% |
| Manitoba | 23.5% | 8.8% |
| Saskatchewan | 17.2% | 11.7% |
| Alberta | 16.2% | 12.8% |
| British Columbia | 20.0% | 11.1% |
| Canada | 23.7% | 4.3% |
| Previous | Contents | Next |