The evolution of administrative data use for the - The evolution

By Ronald Jackson,2014-07-10 12:06
14 views 0
The evolution of administrative data use for the - The evolution ...

    The evolution of administrative data use for the Canadian Business Register (BR)

    Gaétan St-Louis

    June 2008


    More than twenty years ago, the Canadian Business Register (BR) only included businesses with

    employees and thus, only covered a subset of the Canadian economy. At that time, the employer‘s

    account data was in effect the only reliable and administrative data source available to identify and

    maintain businesses on the Register. In the late 1980‘s, it was supplemented with Tax records but

    with no integration between the two sources. In the late 1990‘s, the Canadian Revenue Agency

    (CRA) introduced the Business Number (BN) in order to administer their various programs and this

    greatly increased our potential use of administrative data. The BN permitted Statistics Canada to

    link multiple administrative data sources together, significantly enriching information on the BR

    frame for the economic survey program. During the past ten years, the administrative data has

    become a key component of the Canadian BR. The Register now covers the majority of Canadian

    businesses. These data are still not fully exploited and Statistics Canada is working with other

    Canadian departments to further develop the usage of administrative data in order to maintain the

    BR and its relevance.

    1. Introduction 1.

    Canada‘s statistical system is heavily dependent upon administrative data for the regular production

    of statistics. The computerization of administrative programs in the 1960s and 1970s made it

    possible to increase usage of the resulting data files for statistical purposes. The motivation for this

    interest in administrative data stemmed primarily from three factors. Firstly, tightening budgets

    caused a search for alternatives to the relatively expensive collection costs of statistical surveys and

    censuses. Secondly, an increasing concern about the burden of statistical enquiries on respondents

    led to a search for alternatives. Thirdly, increasing demands for small area data that could not be

    obtained from sample surveys motivated the examination of administrative sources as potential

    supplies of such data. In recent years, further technological advances have facilitated and reduced

    the cost of manipulating large administrative files and have encouraged the increased use of

    administrative data.

Use of administrative data can be divided into four broad categories: direct tabulation, substitution

    for survey data, indirect estimation and the creation and maintenance of a survey frame (Business

    Register). Most statistical applications of administrative data fall into one of these four categories.

In this paper we will focus upon the evolving use of administrative data in the development and

    maintenance of the Canadian Business Register and its benefits to the statistical program. In effect,

    administrative data are an excellent source for creation and maintenance of a frame database. It is

    commonly used to automatically create and update the tombstone data (legal name, operating

    name, address, telephone number and contact name), geography and business size variables such

    as revenue and number of employees. Although it is an excellent source for identifying newly

    birthed units on the frame, it does not meet Statistic Canada‘s needs for the identification of inactive


2. Role of the Business Register (BR) 2.

    The role of the Canadian Business Register (BR) is to provide statisticians, engaged in the

    production and analysis of economic statistics, with the highest quality frame in terms of coverage

    and data elements. It encompasses some of the fundamental concepts of the System of National

    Accounts and provides facilities to store, browse, maintain and retrieve frame information. It is also

    responsible for storage of the contact name, address and telephone number and the survey

    questionnaire identifiers. The Register, upon request, can generate an accurate list of contacts

    which is required for the survey data collection process. It monitors the level of response burden


imposed upon individual Canadian businesses by Statistics Canada and provides relevant

    information to effectively manage response issues. Finally, it provides statistical information

    regarding the composition of the population of businesses in Canada in terms of organisational

    structure, industrial activity, size and geography.

Today, the Canadian Business Register contains approximately 2.3 million active businesses of

    which 99% is composed of single unit businesses. There are about 22 000 businesses that have

    more than one operating entity (complex businesses) and they account for more than 60% of the

    Canadian economy. Different types of organizations are defined as a business such as: a

    corporation, a self-employed individual, a government entity, a non-profit organization, a partnership,

    or a financial fund.

    3. A look in the past 3.

    At the beginning of the 1970‘s, there was an attempt to create a central Business Register based

    upon administrative data with the intention of fulfilling frame requirements for all economic surveys

    at Statistics Canada. For a variety of reasons, this role was not realized. The coverage was not

    fully controlled by the Business Register and the systems were fragmented over several divisions

    and databases. The frame for economic surveys was divided into three major components:

    a) The Business Register based primarily on Payroll deduction (PD) information from

    Revenue Canada for small businesses and survey feedback for the large ones;

    b) Income tax data files based on corporate (T2) and individual tax (T1) returns from

    Revenue Canada;

    c) Frame files for specific surveys or groups of surveys in various divisions.

During these years, the treatment of administrative data at Statistics Canada was decentralised and

    its usage was limited. The BR relied primarily upon a single administrative source, the PD account

    files. The BR program was not meeting the requirements of economic survey programs in terms of

    coverage, timeliness and tools. It became obvious that the BR must be redesigned in order to

    integrate the data, improve the coverage, upgrade the technology and reduce the costs.

The creation of a central Business Register at Statistics Canada

In 1984, the business survey redesign project formally came into existence and was scheduled for

    completion by March 1988. It was a major undertaking and touched upon virtually every major

    economic survey. At the heart of the project, the redesign of the central infrastructure involved

    concepts, systems and the procedures for acquisition and use of frame and income tax data. One

    of the major objectives was to increase the use of administrative data in order to reduce response

    burden, the amount of frame maintenance and data collection costs.

There was an initial attempt to integrate and link different administrative data files into a Central

    Frame Data Base (CFDB) during the redesign of the business survey program in 1984. Due to

    Canadian tax practices and the fact that these files did not carry common identifiers, it was neither

    technically nor financially feasible to completely integrate these files into a single frame. The PD file was driven by a PD number, the T2 file was indexed by corporation number and the T1 file

    contained a unique social insurance number which identified unincorporated businesses. This

    forced Statistics Canada to maintain two alternate views of the business universe, the PD account

    view and the income tax view. In addition, it was decided that the BR would be organized into three

    categories: an ‗Integrated Portion‘ (IP), a ‗Non-Integrated Portion‘ (NIP) and an Out-of-Scope


The Integrated Portion (IP) consisted of unique and unduplicated lists of large and complex

    businesses with the associated structural information (legal, operating and reporting units). It was

    derived through a complete linkage and unduplication of all administrative and non-administrative

    input sources. During a process known as ―business profiling‖, BR employees delineated the legal

    and operating structures and assigned the industrial classification, the revenue and the number of


employees to all operating entities. Extensive manual intervention was required in order to maintain

    IP frame.

The Non-Integrated Portion (NIP) contained two overlapping sets of statistical units. The first set

    was the ‗income tax-based‘ units that were derived from the T1 and T2 tax files. The second set

    was the ‗PD-based‘ units that were obtained from the employer payroll deduction source. No

    systematic attempt was made to integrate or unduplicate these two sets. They provided alternative

    frames for different types of surveys: the tax-based frame for annual surveys and the PD-based

    frame for the sub-annual surveys. Although profiling activities did not extend to the NIP frame, the

    BR was able to assign the Standard Industrial Classification code to each PD account using the

    activity description supplied at the time of registration.

The Out-of-Scope portion contained businesses that fell below size-based thresholds. These units

    are defined as zero employees for Payroll Deduction and less than $25,000 for income tax. These

    units were out-of-scope for a survey. In summary, the survey frame was constructed with units from

    one of the two NIP frames together with units from the IP frame.

     Central Frame Data Base (CFDB)

     Integrated Portion (IP) Includes large businesses IP Units fully linked and classified

     T1 Unincorporated Data NIP Units not linked nor are Payroll Deduction T2 all the units classified (PD) Data Incorporated Data

     Threshold Out of scope for surveys

     Income tax alternate frame Payroll deduction The intent of dividing units into IP, NIP and out of scope categories was to focus manual frame alternate frame maintenance resources on the IP units, which accounts for the largest portion of the Canadian

    economy. The other categories were for the most part maintained by simple, automated processes

    using administrative data files.

Limitation of administrative data integration in the first redesign of the BR

Along with the many advantages associated with the usage of administrative data, there were some

    drawbacks. Administrative data were, in many situations, an excellent source for updating existing

    units and birthing new units but a poor source to identify businesses that had ceased operations.

The PD frame was a good source of births because in Canada, employers were required to

    withhold income tax at source on a monthly basis for their employees. However, employers could

    open multiple PD accounts for a single legal entity and this could potentially result in duplication on

    the frame. To partially resolve this situation, resources were allocated to the IP portion to manually


    link the PD account with a legal unit. No such efforts were made for the NIP portion of the frame because its large number of units would have made this investment impractical. In addition, Frame units were simply added without having a formal process in place that could remove them from the business universe. Again, this caused over-coverage of businesses on the PD frame.

    Because the NIP PD frame was not profiled, the units did not contain any indication of the size of the business in terms of revenue or number of employees. For production surveys, a measure of size is desirable in order to properly stratify the population. By way of compensating for this lack of information, Statistics Canada developed a ratio model to derive the Gross Business Income (GBI) and the number of employees for all PD accounts on the NIP population. These size variables

    were derived on a monthly basis. The frame was updated for all new births as well as for existing

    PD accounts for which there was a significant change in terms of calculated GBI or number of employees. Although this model was very useful, it had its limitations. While providing excellent

    estimates for some sectors of activity, estimates for other sectors were of questionable quality.

The income tax frame had some advantages over the PD frame. It had a one-to-one

    correspondence with the legal unit which eliminated the problem of duplication. It contained employer and non-employer businesses which reduced the under-coverage for self-employed businesses. Contrary to the PD frame, the income tax frame contained an exact size measure for business income.

    However, the income tax frame also had some drawbacks when compared to the PD frame. Income tax returns are filed long after many business events and therefore a poor source for birthing units on the base compared to PD accounts. Unlike the PD frame, only a sample of the income tax frame was assigned a Standard Industrial Classification code. In addition, the tax files contained only a small subset of the reported items on machine readable forms and so there was no explicit means of quality control of captured data.

    Maintaining two views of the NIP frame portion was complicated and resource intensive. It resulted in a duplication of effort for the manual coding of the industrial activity since the PD and Tax frames are not mutually exclusive. The two frames did not include the same population of businesses so survey results between monthly and annual surveys within the same sector of activity were harder to compare. As a result, the survey programs had to take measures to adapt their sampling plan, data collection and estimations.

    As an example, the sub-annual surveys used the PD frame to draw their samples. The duplication of PD accounts on the NIP PD frame was problematic because it could potentially have an impact on the total estimates. To manage this situation and to eliminate this potential bias, respondents were queried about any other PD accounts they might have. This information was used to deflate the survey data provided on the questionnaire.

    For the annual surveys, to overcome the missing industrial classification on the tax income frame, a two phase sample design was adopted. The first phase was used to select and assign an industrial classification to the tax returns. These businesses were maintained on the BR and became the basis for the second sample phase.

    It is relevant at this point to mention that only four sub-annual and five annual surveys used the BR at that time. They were predominantly mission critical surveys for Statistics Canada‘s economic

    program. Other economic survey programs were still using their own frames or a combination of BR frame with other lists.

Goods and Services Tax (GST)

    In 1991, the government introduced the goods and services tax (GST). The GST is a tax that applies to the supply of most goods and services in Canada. Three provinces (Nova Scotia, New Brunswick, and Newfoundland and Labrador, referred to as the participating provinces) harmonized


their provincial sales tax with the GST to create the harmonized sales tax (HST). The HST applies

    to the same base of taxable goods and services as the GST.

    GST/HST is paid upon the purchase of taxable supplies of goods and services. Although the

    consumer pays the tax, businesses are responsible for its collection and remittance to the

    government. Businesses collect the GST/HST from their sales and pay the GST/HST on purchases

    they make to operate their business. Businesses with sales of more than $30,000 must remit

    collected taxes to the government on a monthly, quarterly or annual basis depending upon the size

    of sales.

    This new source of administrative data in electronic format had the potential to supply the Register

    with an earlier size indicator in terms of sales for employer and non employer businesses. It was

    also an excellent source for early birthing units on the frame. Unfortunately, the GST number did

    not link to the PD account nor to the tax data and this lack of a common key meant that we could

    not use this information to maintain the BR. We needed the ability to uniquely link all the

    administrative information for a given legal unit. It would take the introduction of the Business

    Number (BN) registration process before we could take advantage of this administrative information.

    Introduction of the Business Number (BN) at CRA

    Commencing in 1995, there was a federal government initiative to simplify the way businesses dealt

    with government. Given that businesses and organizations already dealt with CRA on a regular

    basis, it was logical for that department to introduce this initiative. Therefore, the Canada Revenue

    Agency (CRA) began a phased-in implementation of the Business Numbering (BN) system. The BN

    was introduced by CRA in order to simplify and streamline the way businesses deal with the federal

    government in Canada. It is based upon the concept of one business -- one number. The Canada

    Revenue Agency (CRA) required all businesses to register and to provide the social insurance

    number (SIN) of at least one owner/director and to describe the business activity at the time of

    registration. As of January 1997, the BN became mandatory for all transactions covered by the

    government‘s programs.

    The 9-digit Business Number (BN) identifies the business and is assigned at the time of registration.

    This registration number is used by businesses when communicating with CRA about pre-existing

    accounts or in order to register new program accounts. To support these programs, a radical of 6

    digits is simply added to the BN. Effectively, the 15-digit BN consists of three partsthe 9-digit BN

    registration number, the account identifier and the account number. For example, a 15-digit BN

    could look like this: 123456789RP0002

    ? The first nine digits are the registration number and they identify a specific business. The

    registration number is unique and it will remain constant no matter how many or what types

    of CRA accounts the business may have.

    ? The two letters identify the type of account, a payroll deduction (RP) account in this example. ? The last four digits identify the account number, a second (0002) payroll deduction account in

    this example.

    If a business expands, it may need to open more accounts. For example, if they open offices in

    different cities across the country and with separate payroll offices in each city, they will need more

    payroll deduction accounts. Each payroll office will have the same business registration number and

    the same account type but a different account number. In this way the 9-digit BN ties together all

    the accounts for a given business.

    The introduction of the BN increased our capability to link multiple taxation administrative data files.

    The linkage process was documented and submitted to Statistics Canada's internal Confidentiality

    and Legislation Committee for their review and approval. It should be noted that Statistics Canada

    has developed a policy regarding record linkage. The policy recognizes the benefits of record

    linkage to statistical programs as well as the concerns that unconstrained record linkage creates.

    This process also ensures that record linkage is not undertaken lightly and only implemented where

    it is well justified and in the public interest. The Business Register Division received authorization to use and link multiple taxation data files in order to create and maintain the central BR.


    During that period, Statistics Canada embarked upon a new project, the objective of which was to substantially upgrade the reliability and detail of provincial economic statistics. These improvements arose from the federal-provincial agreement to harmonize the Goods and Services Tax with provincial sales taxes in Newfoundland, New Brunswick and Nova Scotia. The single harmonized sales tax took effect on April 1, 1997. Statistics Canada was to be the primary source of data for the formulas which allocated the pooled revenues amongst the participating governments. For this reason, detailed and reliable annual provincial data was required.

    The Program to Improve Provincial Economic Statistics (PIPES) affected a wide range of statistical programs throughout the Bureau. The Business survey programs, including the BR, were substantially affected by that initiative. One of the objectives of the program was to have all business survey samples, without exception, drawn from the central business register. This meant that the BR had to improve its coverage and make better use of administrative data. Since the BR was a list of all employer businesses in Canada, businesses with no employees (some unincorporated farms, corner stores and small owner-operated service businesses) were not included.

     In 1998, following the introduction of the BN and the PIPES initiatives, Statistics Canada proceeded with a massive conversion of the BR frame. The BN had become the driving force to update and maintain the BR. With this unique identifier, it was now possible to link the PD account frame data with the BN data. Income tax data and the GST data would complement the frame in order to increase the coverage for the non-employers. Other frame sources used by individual surveys were reconciled with the Business Register. Extending the BR to cover these units was important for accurate measurement in smaller provincial economies.

    Rather than filing their income taxes under their BN, the unincorporated business file their income tax return (T1) under their personal Social Insurance Number (SIN). For this reason the T1 was not included in the conversion and was not integrated with the frame.

     Central Frame Data Base (CFDB)

     Integrated Portion (IP) IP Units fully linked and classified Includes large businesses

     NIP Units fully linked and BN Data

     classified Payroll Deduction (PD) Data

     GST Data

     T2 Data

     Threshold BN data below threshold or with no indication of economic Out of scope for surveys activities or unclassified

    Until now, the BR had used the payroll deduction account as a principal source of information on new entrants and on changes to businesses in Canada. This was therefore limited primarily to businesses that were employers. Using the BN as the identifier for businesses ensured a far more complete coverage as virtually all businesses were registered with CRA.


    Effectively, the BN registrant files became the source by which to identify new businesses and to create and update tombstone information. The appearance of a new BN is a signal that a new business may have been created. If the new BN is accompanied by a new payroll deduction account, then this is the birth of a potential employer. If it is only accompanied by a new GST account, then it is the birth of a non-employer. In either case, the new business can be added to the BR soon after its registration in the real world.

    Even with the massive conversion to the BN and the significant improvement of coverage there remained limitations on how much administrative data could be integrated. The CFDB was still divided into Integrated and Non-Integrated portions. Automation of the processes using administrative data was limited to the NIP frame. The size variables on the frame were still predominantly driven by the modeled GBI and number of employees from the PD accounts. GST sales were assigned to BNs with no PD accounts. If no GST sales were detected then the T2 business income was assigned. In order to fully exploit the administrative data on the BR and meet

    the requirements of survey programs, it was necessary to rethink the conceptual framework of the BR. We would have to wait until 2005 before receiving the mandate to review the BR‘s concepts,

    systems and operations.

4. Todays BR 4.

In 2005, it was recognized that the limitations of the BR environment seriously hindered the BR‘s

    ability to respond to new challenges over the medium and long terms. The Central Frame Data Base (CFDB) was designed and implemented in the mid-1980s and despite modifications over the years the Register was unable to implement the necessary improvements to the database and to our products in order to meet the clients‘ changing needs. Furthermore, the computer environment

    was outdated and work methods did not reflect the changes that Statistics Canada had undergone in recent years.

    Thus an ambitious project to redesign the BR began in 2005/06 and was completed in January 2008. Its main objectives were to:

    ? integrate the legal unit, the operating and statistical structures and the BN into a single structure

    in order to support survey activities;

    ? employ more modern and user-friendly technology;

    ? simplify the task distribution system (Workbench);

    ? make increased use of tax data to improve the coverage and quality of the information on the

    Register; in particular, use tax data in place of derived modelled variables (e.g., GBI) to obtain

    the size variables used for stratification and allocation of samples; and ? improve response burden management

    The current BR model uses the BN as the basis by which to identify businesses on the frame. The other administrative sources such as T1, T2, GST and PD are used to complement the BN data. We use the BN master file from CRA to create and update tombstone data and to identify potential new businesses. In addition, for new and updated BNs, Statistics Canada assigns the NAICS (North American Industrial Classification System) code classification. Fifty percent of these descriptions are classified automatically by the system and the other fifty percent are classified by Statistics Canada employees. This process is run on a monthly basis.

    In contrast to the old CFDB model, we verify if there is an administrative source of data for each BN. This could be a corresponding income tax record, GST sales and/or a PD account. In all cases, we assign the reported value from the corresponding administrative data source to the BN. As an example, a corporation with employees could have a BN with T2 business income, annual sales from GST and a value for the number of employees from the PD file.


    It is important to remember that CRA requests unincorporated businesses to file their income tax return (T1) under their SIN and not their BN. Statistics Canada creates a concordance file between the T1 and the BN by matching on name and address. This BN-SIN link is used to supplement the BN with T1 business revenue information. This is particularly relevant when looking at improving the frame‘s coverage. Effectively, the GST file only covers the employer and non-employer businesses

    with sales of more than $30,000. In addition, roughly ten business sectors that are exempt from the GST program (e.g. real-estate and lessors) were under covered. The T1 added the unincorporated with less than $30,000 of revenue, with no employees and these excluded sectors.

    Because all the information concerning a business is now integrated, administrative data can be used to automatically maintain and update the tombstone and size information of legal units (BNs) for all businesses on the frame. In the case of a complex business, the profiler will use tax schedules to determine the ownership link between legal units and the provincial distribution for revenue before contacting the business. A profile is still required to confirm the ownership link between legal units and to delineate the operational structure and its tombstone and size data. Using administrative data together with profiling activities increases the coherence between legal and operating data.

    As previously mentioned, the use of administrative data is a cornerstone of the business statistics program at Statistics Canada. The intent of PIPES was to use administrative data, particularly GIFI (General Index of Financial Indicators) data, to replace direct collection of financial information from simple enterprises in Canada. GIFI data consists of T2 tax schedules and financial reporting on electronic format. It was introduced by CRA in 1998 and this initiative greatly improved the quality of income tax data.

    In order for the survey program to use administrative data in replacement of survey data, there must be a clear link between the survey units, which are based on the Business Register, and the administrative data units. Such a link exists for corporations via the Business Number, but is still in development for T1 data. The integration of data for complex businesses is extending potential use administrative data in place of survey data for those businesses.

     Business Register (BR)

     Complex Businesses 22,000 Alive and active Businesses The BR integrates the: BN Data Payroll Deduction (PD) Data T1 Unincorporated Data Simple Businesses T2 Incorporated Data GST Data 2,300,000 Alive and Active Businesses

     Statistical Delineation

    Inactive/Dead/unclassified BNs BNs without an economic activity 6,000,000 BNs and/or classification code are

    Out of scope for surveys


    Although the BN registration system works well to identify new businesses, it is not able to systematically identify businesses that have ceased operations. In order to meet Statistics Canada requirements, BR has developed a methodology to identify businesses that ceased operations. The BN account files, together with the income tax files (T1 & T2), the GST file and the PD file now make it possible to identify inactive businesses. Using the account status (open/close) in conjunction with the last remittance dates, we are able to identify inactive businesses with 99% confidence. The drawback with this methodology is that it takes between 13 to 36 months after the fact to inactivate or to death a BN.

    We also use another administrative source to identify amalgamated BNs. CRA produce a monthly file on amalgamated BNs with the successor BN. This process is fully automated and it permits us to inactivate amalgamated BNs.

    The integration of administrative data has improved frame coverage, coherence of the data and the quality of the business profile. It has reduced manual intervention, increased the use of administrative data on the BR and increased the likelihood that survey programs will substitute survey data with administrative data.

5. Looking in the future 5.

    The statistical program has become quite dependent upon these administrative data sources. Further work with our partners remains to be completed in order use administrative data more efficiently to maintain a central Business Register.

    Until we have a solid T1 link with the BN, Statistics Canada will continue to maintain the concordance file between the Social Insurance Number (T1) and the Business Number (BN). This type of link depends upon matching names and addresses and it is not 100% effective. Ideally, if an unincorporated business is registered with CRA, it should submit its T1 tax return indicating their BN. The second portion that is not currently represented on the BR is the T1 businesses with no BN. Effectively, it is still possible to run an unincorporated business with no employees, without obtaining a BN from CRA. Extending the BR to cover these smaller units is important for accurate measurement within certain provinces and specific sectors of the economy.

    Today, a person may request a BN with the intention of starting a new business in Canada. There are no requirements for that individual to report if he/she decides not to operate that business. The same situation applies for an established business. There are no rules that require a business to report that operations have ceased. It would be desirable if such a rule could be implemented. It would greatly facilitate the BR‘s ability to remove inactive businesses in a timely fashion and

    improve the accuracy of the base.

    Statistics Canada is responsible for assigning the North American Industrial Classification code (NAICS) to all registered BNs. There are two issues with the current method used at BN registration time. The activity description is often not well reported, particularly when the business has registered via the web. Only fifty percent of the activity descriptions supplied at registration time can be classified automatically by a system at Statistics Canada. The other fifty percent must be performed manually by Statistics Canada staff. This is a resource intensive activity that requires follow-up with respondents. Secondly, there are no rules that force a business to report their new description should they change business activity. The NAICS is very important because the existence of the new business is only made known to the BR user community once the business is classified that is, when a valid industrial activity code is assigned and have there is evidence of

    economic activity.

    In order to compensate for the lack of information, Statistics Canada runs an ongoing Quality Assurance Survey in order to evaluate the quality of the NAICS classification and the volatility of the information. Ideally, the BN registrant would use a tool to autocode their NAICS at the time of registration and legislation would required the business to report any changes in their activity description.


In most provinces, a business can register through CRA or through their province of residence. As

    a result, the registration process is inconsistent. In both situations they will receive a BN supplied

    by CRA. However, some provinces may use their Provincial Business Numbers to administer the

    provincial programs which undermines the BN link. This makes access to provincial data difficult

    and reduces the efficiencies inherent in the use of administrative data. It would be preferable if the

    registration process and the use of the BN was standardized among the provincial and the federal


6. Conclusion 6.

    Up to the 1970‘s, business surveys were conducted almost entirely as individual surveys. Thus all processes, from determining the frame to disseminating estimates, were controlled and performed

    by one subject matter unit. Control of some survey processes have and continue to be transferred

    to infrastructure divisions. The Business Register has increasingly become the source of frame

    information for business survey programs.

The movement to standardise practices among various departments and programs greatly

    improved Statistics Canada’s efficiency in using administrative data. In the beginning, the BR was composed of employer businesses only which resulted in an under coverage of the thbusiness population. At the end of the 20 century, the Business Number was introduced.

    This unique identifier harmonized related programs such as the GST, PD and the income

    tax programs which resulted in significantly increased coverage of non-employer

    businesses. The BN was a huge step forward in the use of administrative data allowing the

    Register to be more efficiently maintained while at the same time greatly improving its

    coherence. The BN has helped to reduce respondent burden through the replacement of

    survey data, improve the quality of estimates and harmonize the business survey program.

    Administrative data are now playing a pivotal role not only in the survey cycle but also in

    the maintenance of a central BR in Canada.


I would like to say a special thank to Heather Campbell for her excellent work in reviewing and

    editing this paper. I would also like to thank Philippe Gagné, Mario Ménard and Donald Poulin for

    their helpful comments in preparing the final version of this paper.


Brackstone G. Statistics Canada (1987) ― Statistical uses of administrative data: Issues and


Colledge, M. Statistics Canada (1987) ―The business survey redesign project implementation of a

    new strategy at Statistics Canada‖

Colledge M. and Armstrong G. Statistics Canada (1988) ―Statistical units, births and deaths at

    Statistics Canada after the business survey redesign‖ Prepared for the Third roundtable on

    Business Survey Frames, Auckland, New Zealand

Dodds D. and Colledge M, Statistics Canada (1988) ―Using the new Business Register at Statistics

    Canada: Problem and solutions‖


Report this document

For any questions or suggestions please email