1

Session No.2

Paper No.4

Country: New Zealand

Use of Administrative Data

Ian Ewing

Statistics New Zealand

New Zealand’s contribution to the Technology session is presented in two papers.

The first paper, Technological Change and the use of Administrative Data in Economic Statistics-What does it mean for the Business Register?, focuses on:

  • How administrative data can enhance Business Registers’ capability to contribute to the quality and cost of economic statistics.

The second paper, Integrating Administrative Data into an Official Economic Statistics Framework- A Statistics New Zealand perspective, focuses on:

  • A more detailed discussion of the New Zealand experience, over the last ten years in the use of these data sources and their integration with our economic survey programme. The paper also discusses future directions for this work.

Session No.2

Paper No. 4a

Country: New Zealand

Technological Change and the use of

Administrative Data in Economic Statistics

- What does it mean for the Business Register?

Ian Ewing

Statistics New Zealand

Developments in information technology are enabling increased use of data derived from taxation and other government administrative records in the production of official economic and business statistics. In the last ten years in New Zealand, taxation data has been used to replace or complement data derived from direct surveying of firms, or to improve the coverage, accuracy and timeliness of survey-based data. In our programme of economic surveys for example, which covers over 300,000 enterprises over a five yearly cycle, only 15% are now surveyed directly. The data for the remaining 85% is derived from annual tax returns.

Use of administrative record data is likely to continue to grow in the future. This is partly to further exploit its statistical potential, but also because it is less expensive than direct surveying and imposes a lower compliance cost on businesses. Use of transactions records is likely to extend beyond government data to include private sector data bases, such as point of sale and bar-coded commodity information.

The increased use of administrative record data has implications for the Business Register. These are set out in Appendix 1 below. In summary:

Administrative data can enhance the register’s capacity to contribute to the quality of economic statistics and reducethe cost of their collection by:

  • improving the register’s coverage of economic activity
  • reducing the effort required to identify and classify firms
  • reducing the cost of resolving duplication of units
  • improving the capacity to track change through time in the unit structure, classification and economic significance of firms
  • provide information to enhance the capacity of the register to support the design of efficient cross-sectional and longitudinal panel surveys.

Use of administrative data alters the focus of register management:

Taxation data has been identified in New Zealand as the most valuable single administrative data source. This reflects its broad coverage, content relevance and frequency of contact with businesses. Tax numbers are captured in a range of public and private transaction records and are becoming the de-facto business identifier. Our commitment to taking full advantage of tax data has required a substantial initial investment in matching existing statistical units on our business register with the tax authority client register, and the ongoing operation of a maintenance process. The initial matching cost some NZ$100,000 over a 2 years period, and was only made worthwhile by the tax authority’s own investment in information technology. This investment enabled the tax authorities to capture and store the wealth of statistically valuable financial data supplied by firms as part of the taxation process.

Extending the use of administrative data has also shifted our register maintenance strategies away from a survey intensive common approach to all firms to one which is more integrated with the maintenance processes of the tax authority and designed to compensate for its deficiencies-from a statistical perspective. The role of the statistical agency needs to be more concerned with:

1/ Ensuring coverage of statistically significant transaction activity that is not well captured or identifiable within the tax system, including:

  • undisclosed or under-reported economic activity (the “illegal“ economy);
  • businesses whose scale of operation is too small for data to be captured more than once a year by tax authorities, but which are nonetheless economically significant in particular industries or sectors (e.g. farming businesses, accommodation providers and require more frequent or timely statistical monitoring; and
  • businesses engaged in particular classes of transaction that are not well captured by the administrative system - e.g. businesses engaged in Balance of Payments transactions, “Not for Profit“ transactions and small financial services firms.

2/ Offsetting the incomplete or biased application of statistical classifications and unit definitions:

  • taxation administration is not concerned with the industrial, regional or sector homogeneity of data captured as part of the tax compliance process, and is prepared to accept single consolidated returns for groups of enterprises. The statistical profiling of these firms remains a key role for the statistical agency; and
  • the application of industry classifications by the administrative agency may be subject to biases. In New Zealand, for example, industries are subject to differential employee accident compensation premiums - creating an incentive for firms to define their activity (on which the application of industrial classification is based) to fit the lowest possible premium category. The independent surveying of samples of firms is undertaken to identify any bias.

3/ Detecting and removing data associated with duplicate unit records, which may be quite legitimate in terms of the administrative system’s business rules, or which arise due to error in the management of the system.

4/ The integration of data from “secondary“ administrative data sources with data derived from tax records. For example, Agriculture, Overseas Trade and Point of Sales databases. Even where there is a common identifier (such as a tax number), different business rules can leave quite substantial residual problems in matching and maintaining transactor links at a unit level, and the option of less detailed integration of data (e.g. by region, industry or type of activity may be more cost-effective).

Physically, if not conceptually, the statistical business register, and the economic statistics based on it, become integrated with and dependent on the administrative systems which provide its source data. The quality and integrity of these statistics are exposed to the wider range political, administrative, commercial or technologically driven decisions affecting tax, customs, land record or point of sale systems. There is the added risk that public co-operation in statistical surveys and confidence in data quality could be eroded by negative perceptions of the association between statistical agencies on the one hand and authorities such as tax offices, or with private sector database owners with potentially conflicting commercial interests on the other. Managing these issues could be assisted by:

  • national privacy protocols that set a framework for the use and exchange of information about individuals. These protocols should recognise statistical analysis as a legitimate, worthwhile and non-intrusive use of administrative and transaction databases. It would be helpful if these protocols, which are usually applied to individuals, were formally adapted to business entities as well;
  • statistical legislation that embraces the use of administrative records for statistical purposes, and protects the confidentiality of unit data once it is “acquired“ for statistical purposes; and
  • protocols for the management of government held information that oblige database owners to take account of their statistical use both in the set up and management of the database and in providing policy advice on its use.

Finally, we should perhaps ask whether maintaining a business register needs to be a core function of a statistical agency? The close links between the Statistics New Zealand business register with the tax database, means that the process of register construction, and increasingly its "routine" maintenance work is undertaken as part of tax administration. The core function of the statistical agency is to provide the additional statistical layer, or “view“ on top of the tax file (or other underlying administrative file) that incorporates our own data and that derived from other administrative sources and compensates for the gaps and biases inherent in tax and other administrative processes. We also have a role in managing the risk to statistical quality of changes affecting the underlying databases.

A business register has value that extends well beyond any single government application, tax, statistical or otherwise, and more often than not the information that provides this value-comprehensiveness, currency, freedom from gaps and overlaps, industry, sector and regional classification and a general measure of size are those added by the statistical agency and are of limited commercial sensitivity. Managing access to this “public“ view of the register, could be independent of both statistical, tax or other agencies providing the source data.

In conclusion contemporary information technology has enabled “business registers“ to be built across the databases of government and even private sector agencies at a reducing cost and without expensive bureaucratic centralisation. This is changing the role of the statistical agencies with respect to frame maintenance. It will involve closer co-operation between government and private sector agencies to manage risks to data supply and establish the rules for data exchange, custodian-ship and access to register information.

Appendix 1: Use of Administrative Record Data in Statistics-Implications for the Business Register

Desired Characteristics of Economic Data / Role of the Business Register / Use of Administrative Record Data/Technologies
Cover all economic activity within the production boundary /
  • maintain up-to-date lists of significant transactors
  • identify the transactors from which data is collected. Enable their aggregate economic contribution to be measured, and allow coverage “gaps“ to be identified and estimated.
/ Tax or company registration files have broad coverage of businesses, and compliance rules are generally specified in institutional or financial terms (e.g. turnover; net income) which assists in measuring under-coverage. Currency of contact details can be problematic.
Be homogenous with respect to industry, sector, region or class of transaction / define transactor units and uniquely classify each with respect to standard definitions and classifications / Administrative files may wholly or partially meet this requirement. Tax or business registration systems focus on legal entities. For the majority of firms which undertake a single industrial activity at a single site this is not a problem, but for the more complex firms statistical unit profiling is needed.
Even for “single site/single activity“ firms administrative records may not hold sufficient information to allow industry classification.
Include no double counting / Resolve possible duplicate units /
  • There is a financial disincentive to be recorded twice in the tax system, however the business rules of administrative record systems can legitimately allow the same statistical unit to be represented more than once.
  • even when rules proscribe duplicates, they can continue to exist through error/oversight.
  • no single administrative source is usually sufficiently comprehensive for register, maintenance (e.g. Balance of Payments, Agriculture), necessitating reconciliation of multiple sources and duplicate checking.
  • availability of cheap and efficient commercial name/address matching technology is reducing the cost of reconciling multiple sources at a unit level, Geographic Information Systems can assist in matching regionally related aggregates.
  • economy-wide system identifiers such as tax numbers are becoming either legal or de-facto common reference points for units across data sources-assisting “hard“ matching.

Reflect real change in economic structure and in the level of economic activity through time-in a timely manner. /
  • add new transactors and delete deaths
  • track changes to units, through time ensuring changes to classifications reflect responses to economic change (changes in supply of inputs, changes in technology affecting processes or changes in demand for products) and are not driven by changes in business accounting, administrative re-organisation or reflect biases inherent in source data.
/
  • Registration with an administrative system is often a prerequisite for engaging in business-hence their long established use as a source of “births“. The uptake of electronic commerce is a further incentive for firms to capture key data such as tax numbers within transactions records, which could extend the timeliness and completeness of this data.
Note: Just because this data source exists does not necessarily mean that the frame will be more accurate.
  • Systems such as VAT or payroll tax which require periodic (monthly or quarterly) reporting can be used to monitor change in the level of activity or changes in address or ownership, or to remove “deaths“-.
  • The capture of economic information such as sales, expenses, assets and liabilities can provide size criteria that relate more closely to the economic significance of a firm than employment head counts and allow design of more efficient sample surveys based on the register.
  • co-operation with administrative agencies in the design of systems can enhance their statistical value (e.g. agreeing on monthly or quarterly reporting periods and the breakdown of data items),

Efficiently collected (i.e. at minimum cost to both the statistical agency and to respondents) /
  • maintain current contact details for units
  • map “real world“ reporting units to statistical units
  • provide access to relevant measures of economic significance for sample design and maintenance
  • track the involvement of firms in survey activity to control provider costs
/
  • see above.
  • investment in unit level matching of the tax file with the business register and the ongoing maintenance of that relationship can allow financial data from annual tax returns to be integrated with survey data within a single integrated collection-reducing direct collection by up to 85%.

Session No.2

Paper No. 4b

Country: New Zealand

Integrating Administrative Data into an Official

Economic Statistics Framework

-A Statistics New Zealand perspective

Ian Ewing

Statistics New Zealand

1. The Background

The ongoing process of economic and technological change requires statisticians to measure economic activity in an environment where administrative data is extensive yet imperfect. As such, it has become essential to realistically evaluate the statistical potential of administrative data and its increased use.

Administrative datahas the potential to provide a cost effective alternative to the direct surveying process. It can often provide us with information that can replace, complement or enhance this process, and reduce processing and compliance costs.

Potential savings in post out and respondent liaison costs, from the use of administrative data are offset to some degree by data management and editing costs.

Although the benefits of administrative data are real, it is important to also consider the risks of our increasing dependence on these data sources.

Keyword

Administrative Data

1.1 Introduction

The objectives of this paper are to discuss the following issues:

1. The current and future utilisation of administrative data by Statistics New Zealand.

2. The role of the Business Register (Business Frame) in the current and future utilisation of administrative data by Statistics New Zealand?

1.2 Definition

Administrative data

Administrative data is defined as information resulting from the interactions of individuals and groups with government agencies. Information is collected for, or as part of, a service delivery, an application, membership, regulation or registration.

"Cross sectoral" administrative data is information about an individual or group that has been assembled by integrating data collected through administrative processes occurring in different sectors of government (tax, customs, health, etc.). Cross sectoral data collection enables (when there is sufficient information) the matching of a variety of characteristics about the same individual or group. Administrative data is generally characterised by its comprehensiveness. However, there are often administrative rules (such as cut off thresholds like the NZ$30,000 turnover used for NZ's Goods and Services Tax (GST): Refer Glossary) and low levels of compliance in some areas (e.g. landlord compliance with NZ’s requirement to register tenancy bonds) that can reduce coverage.

Note: In this paper the term "administrative data" includes information resulting from transactions between a private sector business and its customers or suppliers (inventory data). This data is often characterised by large unsummarised data sets.

1.3 The Changing NZ Environment

Over the last decade, Statistics New Zealand (SNZ) has undergone considerable changes centred on efficiency and data quality improvements from investment in information technology. This has coincided with similar investments in major data gathering agencies of government most significantly, by the tax office (The New Zealand Inland Revenue Department (IRD)). Tax reforms have also worked to broaden the base of tax collection and increase the range and frequency with which tax compliance data is collected. SNZ is now able to utilise data sources that were previously unavailable, or were too vast to capture manually.

Over the last 100 years, SNZ (like most statistical agencies) has relied, to some extent, on administrative data, such as customs and tax records. However, these data sources are subject to policy and legislative changes. For example prior to 1984, SNZ used the Reserve Bank's Overseas Exchange Transactions data to produce our Balance of Payments statistics. This data source has been lost with the easing of foreign exchange controls and SNZ had to replace these sources with direct surveys.

Since 1984, successive New Zealand governments have implemented programmes of economic deregulation. Major public policy changes included the freeing up of foreign exchange controls and the introduction of a comprehensive value added tax (GST). This reformhad a major impact on the way SNZ collected data and provided an invaluable source of data for populating and partially maintaining its business register.