Farm Environmental Management Survey (FEMS)

Detailed information for 2006




Every 5 years

Record number:


This survey, focusing on both livestock and crop operations, will allow the establishment of base lines and development of updates for an expanded set of agri-environmental indicators, and generate the information to design effective and well targeted policy and program responses.

Data release - March 14, 2008


The national survey on farm environmental management practices is a key aspect of Agriculture and Agri-Food Canada's new National Agri-Environmental Health Analysis and Reporting Program (NAHARP). The information generated from this survey is needed to support the industry's environmental initiatives, to address federal and provincial policy needs and to guide sustainable development actions in Canada's agriculture sector. This voluntary survey, focusing on both livestock and crop operations will allow the establishment of base lines and development of updates for an expanded set of agri-environmental indicators. These indicators are needed to: determine the present status of farm environmental management across Canada; identify areas that are most in need of environmental management movements; and generate the information to design effective and well targeted policy and program responses. This survey is conducted in conjunction with Agriculture and Agri-Food Canada, to ensure that agriculture programs reflect the changing way resources are being managed on today's farms.


  • Agriculture and food (formerly Agriculture)
  • Land use and environmental practices

Data sources and methodology

Target population

The conceptual universe is made up of all active farms in the Agriculture Division's Farm Register. The following types of farms were excluded:
. Active farms with agricultural sales in 2005 of less than $10,000;
. Institutional farms (prisons, research stations, colleges);
. Farms located on Indian reserves;
. Farms for whom more than 50% of the gross income in 2005 came from sales of greenhouse, sod and nursery products;
. Farms without livestock inventory or crop area at the time of the 2006 Census of Agriculture;
. Farms located in the Yukon, the Northwest Territories and Nunavut.

Instrument design

The Farm Environmental Management Survey questionnaire was completely redesigned to include new data needs. To reduce response burden the redesigned questionnaire was divided into two modules: one for crop and one for livestock.

The Farm Environmental Management Survey questionnaires were designed by a project team made up of Statistics Canada and Agriculture and Agri-Food Canada employees and provincial experts assigned to the project. Questionnaire design specialists were consulted in Statistics Canada. In February 2006, one-on-one in-depth interviews were used to test draft versions of the questionnaires with 59 farm producers interviewed in five regions: Abbotsford, British Columbia, Saskatoon, Saskatchewan, London, Ontario, Trois-Rivières, Quebec and Charlottetown, Prince Edward Island. Participants represented various types of agricultural operations. The questionnaires were then revised based on questionnaire design specialist recommendations and a second round of consultation with Agri-Food Canada employees and provincial focal points or experts.


This is a sample survey with a cross-sectional design.

There was a requirement to report the survey results for each of the 27 sub-provincial regions. The combination of the 12 ecological regions (described below) and the 10 provinces produced 27 different sub-provincial regions defined by similar agronomic, climatic and soil attributes in the Canadian eco-stratification system.

The 12 regions by which the survey frame was stratified are as follows:

1) Atlantic Maritime ecozone
2) St. Lawrence Lowlands ecoregion
3) Manitoulin- Lake Simcoe - Frontenac ecoregions
4) Lake Erie Lowland ecoregion
5) Boreal Shield ecozone
6) Brown Soil Zone (ecoregions Mixed Grassland and Cypress Upland)
7) Dark Brown Soil Zone (ecoregions Moist Mixed Grassland and Fescue Grassland)
8) Black Soil Zone (ecoregions Aspen Parkland and Southwest Manitoba Uplands)
9) Lake Manitoba Plain ecoregion
10) Boreal Plains ecozone
11) Montane Cordillera ecozone
12) Pacific Maritime ecozone

The first step was to determine whether each farm in the sampling frame was a candidate for the Crop module (or questionnaire), for the Livestock module, or for both (a Mixed farm). The farms were assigned to one of the three groups (Crop, Livestock or Mixed) according to their contribution to cropland and to livestock units in their province. The population was then stratified according to the province and ecological region. The gross farm income and the crop type (wheat, grain, oilseed, potatoes, fruits, vegetables, hay, other field crop and mixed crop) and/or livestock type (cattle, hogs, poultry, mixed livestock) were also used, to ensure that each type of farm was represented.

Sample allocation
Based on the budget, the expected response rate and a targeted level of precision, the total sample size was set at 20,000 farms, roughly 10,000 for each module. Within a given stratum of the Crop module, the sample was split between farms with only crops and Mixed farms, in such a way that their distribution was the same as in the target population. The same strategy was used for the Livestock module.

Sample Selection
Farms were randomly selected within each stratum. The sampling strategy ensured that a farm could not be selected for both the Crop and Livestock modules. A further restriction was used in the selection process to reduce response burden: an attempt was made to eliminate overlap with a similar survey, conducted by the provincial department of agriculture in Quebec during the same period as the Farm Environment Management Survey.

Data sources

Data collection for this reference period: 2007-02-14 to 2007-03-31

Responding to this survey is voluntary.

Data are collected directly from survey respondents.

As suggested by participants during questionnaire testing, the collection period was chosen for data collection because it had the least effect on farming operations (before spring planting). The survey also tied in well with the 2006 Census of Agriculture completed in the summer of 2006. A Computer Assisted Telephone Interview data collection technique was developed for this survey. Due to the farmers' heightened awareness of environmental issues, support by several farm producer associations and historical participation of farm producers in regular agricultural surveys, the survey overall response rate was expected to be higher than 70%.

View the Questionnaire(s) and reporting guide(s).

Error detection

A systematic approach was used to identify missing or incomplete data as well as to identify outliers. There were consistency and deterministic edits applied. The Ag2000 data processing system was used for editing the data and for identifying outliers.


Missing values were imputed only if available from other sources for the same farm.


Prior to estimation, the data was verified for completeness and consistency. The data was further adjusted for outliers. Based on the clean response records the initial weights were adjusted to reflect the sample response and estimates were produced at the province and sub-provincial levels where there was sufficient response to produce reliable statistical aggregated data.

Quality evaluation

The data collected from the survey were compared to the Census of Agriculture data. Estimates were also compared to other comparable published estimates and analyzed by subject matter experts. The estimates at the provincial level were also validated by the provincial focal points who were involved in the design of the survey questionnaire and who are knowledgeable about the subject matter on a regional basis. Finally, the results of the survey were also compared, whenever possible, to the previous results of the Farm Environmental Management Survey.

Disclosure control

Statistics Canada is prohibited by law from releasing any data which would divulge information obtained under the Statistics Act that relates to any identifiable person, business or organization without the prior knowledge or the consent in writing of that person, business or organization. Various confidentiality rules are applied to all data that are released or published to prevent the publication or disclosure of any information deemed confidential. If necessary, data are suppressed to prevent direct or residual disclosure of identifiable data.

Tabular results were produced using the estimation software "ESTIMEX" developed for Agriculture Division. The tabulation system automatically applies the Statistics Canada standard rules for confidentiality and data not satisfying the rules are suppressed automatically. Manual residual disclosure analysis was also done to ensure that no confidential information is released.

To reduce response burden and to ensure more uniform statistics, Statistics Canada has entered into an agreement under section 12 of the Statistics Act with Agriculture and Agri-Food Canada, the Ministère de l'Agriculture, des Pêcheries et de l'Alimentation du Québec, the Institut de la statistique du Québec, the Ontario Ministry of Agriculture, Food and Rural Affairs and the Alberta Agriculture and Food Ministry. Under Section 12, Statistics Canada will not share any name, address or other identifying information. The information is required to be kept confidential and used only for statistical and research purposes. These various data sharing partners will only have access to respondents within their respective province who agree to share survey information.

Revisions and seasonal adjustment

Revisions can only be approved by subject matter experts from Statistics Canada and Agriculture and Agri-Food Canada.

Data accuracy

While non-sampling errors are difficult to quantify, sampling errors can be estimated from the sample itself using the standard error (SE) of estimated values also referred to as an absolute sampling error. For level estimates (e.g. totals and averages), a statistical measure called the coefficient of variation (CV) is normally used. The CV, defined as the standard error divided by the survey estimate, is a measure of precision in relative terms and is expressed as a percentage.

For level estimates, the CV is the appropriate measure of the sampling error. For proportions, however, an absolute sampling error -- such as the SE itself -- is preferred. In the case of the Farm Environmental Management Survey, most of the estimates deal with proportions.

The SE (which is a function of the population size, the sample size, and the estimate), along with the confidence level, can be used to calculate the margin of error. This measure is straightforward to interpret, since it is on the same scale as the estimate itself. For example, an estimated proportion of 80% might have a margin of error of 3%, meaning that we would conclude (with the appropriate confidence level, usually 19 times out of 20) that the true proportion is between 77% and 83%.

Suppose we want to estimate the proportion of Canadian livestock farms that store liquid manure. The estimated proportion is 14% with a standard error of 2.54. It can be deduced that the proportion of farms that do not store liquid manure is 86% and that the quality of the estimate is the same (i.e., the standard error is still 2.54). The standard error is an absolute error that applies to both the 14% and 86% estimates. The CV, being a relative error, would be different for the two estimates. It can even appear good for one proportion (86% for CV1) and bad for the complementary proportion (14% for CV2) as shown below:

CV1 = 100 * 2.54/86 = 3 (for farms which do not store liquid manure)
CV2 = 100 * 2.54/14 = 18 (for farms which store liquid manure)

Though the quality of the estimates is the same, the CV2 implies that the quality of the estimated proportion of farms which store liquid manure is much lower. In this case as with all proportion estimates, the CV can be misleading.

The following is a suggested CV rating system for level estimates, and a standard error (SE) rating system for proportion estimates:

CV Rating

0.01% - 4.99% A -- excellent
5.0% - 9.99% B -- very good
10.0% - 14.99% C -- good
15.0% - 24.99% D -- acceptable
25.0% - 34.99% E -- use with caution
35.0% and more F -- too unreliable to be published

SE Rating

0.01% - 2.49% A -- excellent
2.5% - 4.99% B -- very good
5.0% - 7.49% C -- good
7.5% - 12.49% D -- acceptable
12.5% - 17.49% E -- use with caution
17.5% and more F -- too unreliable to be published


Date modified: