Agriculture Taxation Data Program (ATDP)

Detailed information for 2012

The Agriculture Taxation Data Program (ATDP) is designed to produce detailed estimates for the following variables: revenues and expenses of farms (preliminary; final); and farm and off-farm income of farm operators and farm families.

Data release - December 2, 2013 (First in a series of releases for this reference period.)

• Description
• Data sources and methodology
• Data accuracy

Description

The Agriculture Taxation Data Program (ATDP) samples unincorporated and incorporated taxfiler records annually to estimate a range of financial agricultural variables. The financial variables collected by the ATDP include detailed revenues and expenses of farms and off-farm income of farm operators and farm families.

Data are used by Agriculture and Agri-Food Canada, other federal and provincial departments, and various agencies to monitor the financial health of the Canadian agricultural sector and serve as a tool for farm-level policy analysis. The Farm Income and Prices Unit, Agriculture Division, Statistics Canada, relies primarily upon ATDP financial data to establish expense estimates for the Agricultural economic statistics (AES) and Statistics Canada's National Accounts. The annual off-farm operator and family income estimates are used to: measure the relative importance of farm and off-farm operator (family) income at different aggregation levels; assess the economic welfare of Canadian farm operators and their families; and facilitate farm policy development.

Reference period: Taxation year

Collection period: January to December (24 months)

Subjects

• Agriculture and food (formerly Agriculture)
• Farm financial statistics
• Farms and farm operators

Data sources and methodology

Target population

The target population consists of all unincorporated and incorporated farms in Canada. Since the 1993 taxation year, it has also encompassed all communal farming organizations in Canada.

For statistical purposes, the estimates presented cover both unincorporated farms and communal farming organizations with total farm operating revenues equal to or greater than $10,000 as well as incorporated farms with total farm operating revenues of $25,000 and over.

Instrument design

This methodology type does not apply to this statistical program.

Sampling

This is a sample survey.

From 1987 to 1989, the ATDP sampled both sectors (incorporated and unincorporated) for all of the non-Prairie provinces. For the 1990 taxation year and all subsequent years, the sample includes both sectors in all of the provinces.

For taxation year 2012, the number of useable records included almost 142,500 records (136,600 unincorporated farms and 5,900 incorporated farms).

The sampling rates for the unincorporated and incorporated sectors vary from a complete census in the territories and the Atlantic provinces (T2) or Newfoundland and Labrador (T1) to random samples selected in all the other provinces. There is also a pre-specified sample of farms selected based on particular characteristics.

The sampling frame for the unincorporated farms is stratified by province/territory and gross farm income. The predetermined initial sample size is allocated, using the square-root allocation algorithm for the sampled provinces, to ensure adequate representation of all provinces. Following the initial provincial allocation, additional records are added to the sample in some provinces to improve the quality of the estimates.

For further information about the sample size refer to the Whole Farm Data Base - Reference Manual - Provincial data availability and sample size by sector (Table C.1).

Data sources

Data are extracted from administrative files.

Data are extracted from administrative files supplied by Canada Revenue Agency (CRA). The Self-Employment File for Agriculture (SEFA) for T1s and the Corporation Tax Processing System (CORTAX) file for T2s, supplied by the Canada Revenue Agency (CRA) contain the ATDP universe for the unincorporated and incorporated sectors respectively. The Statistical Universe File--T3, also from CRA, provides the universe for communal farming organizations.

ATDP data are compiled from two separate sources: the SEFA (Self-Employment File for Agriculture), from which a sample of unincorporated farm tax returns is randomly selected; and the CORTAX (Corporation Tax Processing System) file, from which is selected a random sample of tax returns filed by incorporated operations classified as farms under the North American Industry Classification System (NAICS) with sales exceeding $25,000.

The source of data on the unincorporated sector is comprised of three different types of taxfiler returns: printed forms, electronic forms (since 1992) and joint AgriStability/AgriInvest-CRA tax returns (since 2007). There are three types of printed forms: traditional printed forms, printed forms completed using tax preparation software and forms completed using tax preparation software with a two-dimensional bar code (at the bottom of the first page). Unincorporated farm data on traditional printed forms or printed forms with no bar code are captured by CRA staff at several regional taxation centres and forwarded to StatCan in electronic format. Since 2007, data on printed forms with a bar code printed on the first page of the return are captured in electronic format by scanning the bar code and then forwarded to StatCan. CRA also supplies StatCan with the electronically filed returns and data from the joint AgriStability/AgriInvest-CRA farming returns throughout the year.

For the incorporated sector, Statistics Canada captured all financial data (detailed revenues, expenses) from corporate farm taxation returns up to and including the 1999 data year. Since the 2000 taxation year, corporate farming data have been supplied electronically by CRA from a file termed General Index of Financial Information (GIFI).

The T3 Trust Income Tax and Information Return is the source for communal farming organizations.

The T1 General--Income Tax and Benefit Return form serves as a source of off-farm income statistics: wages and salaries, net off-farm self-employment income, investment income, pension income, government social transfers and other off-farm income. Data from the Canada Child Tax Benefit File supplement data on off-farm income.

Error detection

We collaborate with Canada Revenue Agency (CRA) to adapt their capture screens and develop new variables.

Detailed edit programs identify among other things, errors, inconsistencies and extreme values in the captured data. Data that fail to meet the predetermined criteria are referred to subject-matter specialists for appropriate action. Then, the records of the 25 taxfilers that contribute the most for each income and expense item at the provincial level are analyzed further.

Imputation

Once all records have passed through the editing steps, those requiring imputation are identified and isolated. A process of donor imputation is used in cases where taxfilers failed to itemize (all or part of) their revenues and expenses. This involves the use of what is known as the "nearest-neighbour approach" to impute a value to a field. For example, if a farm taxfiler reports only a lump-sum figure for fertilizers, pesticides, and seed items, then an imputation will break down this aggregate figure into its component parts. The particular record is isolated and identified as a "recipient". A computer search is then made among the remaining records to identify the taxfiler that most closely matches the characteristics of the "recipient". This record would have reported values in the fields requiring imputation and have a "similar" farm type, geographic region and value of total farm expenses as the "recipient". For this example, the values reported by the donor for the three items specified above are summed and the proportion each represents of the summed value is calculated. This same proportion is then used to split the aggregate value reported by the "recipient" into the component parts.

Units with partial non-response in the unincorporated sector are imputed using the Banff generalized edit and imputation system. In the incorporated sector, they are imputed by a combination of donor imputation using the Banff generalized system and manual imputation using notes (financial statements and balance sheets) from the tax forms.

The majority of total non-respondents are dealt with through weight adjustments, i.e., the records are excluded from the sample counts and the weights of the other sampled records are adjusted to compensate for these non-responses.

Once the records have been imputed and the weights have been applied, the weighted top 25 contributors for each income and expense item at the provincial level are analyzed further. As a final check, the top 10 contributors by province and type of farm are reviewed. At this stage, the weights may be adjusted if records are added or removed.

Estimation

Farm revenues and expenses

Total farm revenue and expense items are estimated by inflating the in sample revenue and expense items using an estimation weight. To represent the entire population, each entity is assigned a weight, which reflects the proportion of the population actually observed in the sample, multiplied by the partnership share of the entity in the case of unincorporated farms. The pre specified units are self representing (estimation weight equals one) as they are included in the sample with certainty. The calculated weighted revenue and expense items are summed by domain to produce the total revenue and expense items. A domain is defined as a region, a type of farm, a revenue class or a combination of these variables.

Only in-scope sampled records are included in the estimates. Data for non farmers are excluded. Data for the three territories are also excluded.

For statistical purposes, the estimates cover both unincorporated farms and communal farming organizations (with total farm operating revenues equal to or greater than $10,000) as well as incorporated farms (with total farm operating revenues of $25,000 and over).

See the following publications (available through the "Publications" sidebar above) for more detailed information on the sample, methodology and data quality of the ATDP:

Statistics on Income of Farm Operators (21-206-X)

Statistics on Income of Farm Families (21-207-X)

Statistics on Revenues and Expenses of Farms (21-208-X)

Quality evaluation

We compare ATDP estimates with certain control totals and trends provided by other Statcan statistical data sources such as the Census of Agriculture and the Agriculture Economic Statistics series produced by Agriculture Division.

Data from the Personal Master File and estimates on income of all families produced by the Small Area and Administrative Data Division (SAADD) from the Family File (T1FF) are also used to assess the quality of the series on off-farm income.

If there are discrepancies, the micro-level data of individual records are analysed to determine if any differences are attributable to weights. Various tools are used to analyse data at the sub-provincial level, by farm type and revenue class including year over year percentage change, outlier detection by means of scatter plots and impact of top contributors on estimates.

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. The confidentiality provisions of the Statistics Act override the provisions of any other Act, including the Access to Information Act, to guarantee the confidentiality of reported data of individual respondents. 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.

For each of the tabulations produced, the estimated number of farms is rounded to the base 5 and the estimates of the other variables within that table are adjusted by a variable factor. The estimated number of farm families is rounded to the base 10. With regard to the estimated number of farm operators, it is rounded to the base 5 in the series of farm operators operating single unincorporated agricultural holdings and to the base 10 in the series of farm operators operating incorporated or unincorporated agricultural holdings. If the degree of detail required to answer user requests creates confidentiality concerns, the affected data or the entire table will be suppressed.

This method preserves the confidentiality of the data, without jeopardizing the quality of the actual estimates.

Revisions and seasonal adjustment

This methodological step does not apply to this survey.

Data accuracy

While considerable effort is made to ensure high standards throughout all stages of processing, the resulting estimates are inevitably subject to a certain degree of error. These errors can be broken down into two major types: non-sampling and sampling.

Non-sampling errors
Non-sampling errors can occur whether a sample is used or a complete census of the population is taken. Such errors can be introduced at various stages of data processing (such as coding, data entry, editing, weighting or tabulation) and include response errors introduced by the taxfilers as a result of misclassifications. Such errors are reduced through extensive edits and data analysis; however, some of these errors are outside the control of Statistics Canada. Specifically tax forms are designed for the collection of income data for tax purposes and not for survey purposes.

Sampling errors
Among the estimates produced by the ATDP, those that are derived from samples are subject to sampling errors. Such errors occur when observations are based only on a sample and not on the population as a whole. The size and design of the sample, the variability of the characteristic of interest in the population, and the estimation method all affect data quality. In sample surveys, inference is made about the entire population based on data obtained from a part of the population; therefore, the results are likely to be different than if a complete census was taken under the same survey conditions. The most important feature of probability sampling is that the sampling error can be measured from the sample itself.

Each estimate derived from the ATDP is assigned a coefficient of variation (c.v.) to measure its quality. As an objective statistical measure obtained through random sampling of the variation between each estimate and its "true" value, the c.v. indicates the degree of confidence that should be placed on a particular estimate. The users must determine if an estimate with a significant c.v. is appropriate for use.

The qualities of CVs are rated as follows:
- Excellent 0.01% to 4.99%
- Very good 5.00% to 9.99%
- Good 10.00% to 14.99%
- Acceptable 15.00% to 24.99%
- Use with caution 25.00% to 34.99%
- Too unreliable to be published 35.00% or higher

CVs were calculated for each estimate. The CVs are available upon request.