Author: Lynda Sanderson
The Longitudinal Business Database (LBD) links together firm-level data held by Statistics New Zealand from a combination of administrative and survey data sources. This linking has opened up a wide range of research opportunities. In many cases, there are now multiple sources of information on a single area of firms' activities, often capturing different aspects or a different population, requiring a choice to be made about the most appropriate source of data for any given research question. This note outlines the available firm-level data sources for research on foreign direct investment, describes the data available, and sets out some of the opportunities and challenges for using these data for research purposes. It also provides detail on maintenance procedures for one of the core elements of the LBD, the Longitudinal Business Frame, which is applicable to a wide variety of firm-level research topics.
I would like to thank Geoff Mead, Ranjith Nanayakkara, Brian Torrey, Upul Paranawithana, Mark Gordon and Mike Faire for their assistance in understanding the intricacies of the data, and Richard Fabling, Hamish Hill, Talo Talosaga and James Beard for feedback on the paper. This paper was completed as part of the author's secondment to Statistics New Zealand.
The views, opinions, findings, and conclusions or recommendations expressed in this Working Paper are strictly those of the author. They do not necessarily reflect the views of the New Zealand Treasury or the New Zealand Government. The New Zealand Treasury and the New Zealand Government take no responsibility for any errors or omissions in, or for the correctness of, the information contained in these working papers. The paper is presented not as policy, but with a view to inform and stimulate wider debate.
This paper was undertaken while the author was on secondment to Statistics New Zealand. The results in this paper are not official statistics, they have been created for research purposes from the Integrated Data Infrastructure prototype (IDI) managed by Statistics NZ. The opinions, findings, recommendations and conclusions expressed in this paper are those of the author. Statistics NZ takes no responsibility for any omissions or errors in the information contained here.
Access to the data used in this study was provided by Statistics NZ in accordance with security and confidentiality provisions of the Statistics Act 1975. Only people authorised by the Statistics Act 1975 are allowed to see data about a particular person, business or organisation. The results in this paper have been confidentialised to protect individual people and businesses from identification.
Careful consideration has been given to the privacy, security and confidentiality issues associated with using administrative data in the IDI. Further detail can be found in the Privacy Impact Assessment for the IDI available from www.stats.govt.nz.
The results are based in part on tax data supplied by Inland Revenue to Statistics NZ under the Tax Administration Act 1994. This tax data must be used only for statistical purposes, and no individual information may be published or disclosed in any other form, or provided to Inland Revenue for administrative or regulatory purposes. Any person who has had access to the unit-record data has certified that they have been shown, have read, and have understood section 81 of the Tax Administration Act 1994, which relates to privacy and confidentiality. Any discussion of data limitations or weaknesses is in the context of using the IDI for statistical purposes, and is not related to the data's ability to support Inland Revenue's core operational requirements.
The Longitudinal Business Database (LBD) links together firm-level data held by Statistics New Zealand from a combination of administrative (primarily Inland Revenue) and survey data sources. This linking has opened up new research opportunities, allowing firm characteristics to be identified and tracked over time. In many cases there are now multiple sources of information on a single area of firms' activities, requiring a choice to be made about the most appropriate source (or sources) of data for any given research question.
This paper focuses on one specific area of research interest - foreign direct investment - and describes the alternative sources of information available at the firm level. In the LBD, information on firms' international investment activities can be identified from four main sources:
- The Longitudinal Business Frame (LBF)
- Company tax returns (IR4)
- The Business Operations Survey (BOS)
- The Quarterly and Annual International Investment Surveys (QIIS/AIIS)
As each of these datasets has been collected to address a different need, they differ in terms of both sampling methodology and the definition of international investment. This paper outlines each source, including the purpose of the collection, the population of interest and sampling strategy, and the aspects of international investment which are captured. This information is intended to give new and potential users a broad understanding of the pros and cons of the four sources of firm-level investment data, enabling them to identify the best source(s) for their purpose and to make informed decisions about data treatment.
For example, while the IR4 data is likely to be most appropriate for studies which require both a large population and annual data, use of the IR4 alone limits attention to Limited Liability Companies. Combining IR4 data with that from the LBF expands the potential research population to other firm types, but continues to leave gaps in information about small and medium-sized firms for which LBF data is not regularly updated. In contrast, BOS provides annual data on both inward and outward investment for a representative sample of firms, but may not have sufficient coverage to examine relatively rare events such as foreign acquisitions and divestitures. Finally, QIIS and AIIS provide by far the most detailed information about firms' international investment activities, but create substantial challenges for micro-empirical research due to the sampling method, which focuses on firms which are known to have substantial international assets or liabilities. As such, the challenge for researchers interested in understanding international investment behaviour at the firm level is to find ways to make appropraite use of data which has been collected for other purposes.
While the primary focus is on international investment indicators, the paper also provides background on the maintenance procedures for the Longitudinal Business Frame (LBF). As the LBF lies at the core of the LBD, forming the spine to which all other firm-level data sources are linked, an understanding of the sources of LBF data and changes in survey sampling over time is valuable for researchers across a wide range of firm-level research topics.
The Longitudinal Business Database (LBD) links together firm-level data held by Statistics New Zealand from a combination of administrative (primarily Inland Revenue) and survey data sources.  It constitutes the firm-centered component of the recently developed Integrated Data Infrastructure (IDI), which links firm-level data from the LBD with individual-level data from the Linked Employer-Employee Database and a range of other administrative sources. The development of the LBD has opened up new research opportunities, allowing firm characteristics to be identified and tracked over time. In many cases there are now multiple sources of information on a single area of firms' activities, often capturing different aspects or a different population, requiring a choice to be made about the most appropriate source (or sources) of data for any given research question.
This paper focuses on one specific area of research interest - foreign direct investment. Two forms are considered: inward foreign direct investment (FDI) in which a non-resident individual or firm holds complete or partial ownership of a firm located in New Zealand; and outward direct investment (ODI) in which a New Zealand firm has an ownership interest in a business located offshore.
In the LBD, information on firms' international investment activities can be identified from four main sources:
- The Longitudinal Business Frame (LBF)
- Company tax returns (IR4)
- The Business Operations Survey (BOS)
- The Quarterly and Annual International Investment Surveys (QIIS/AIIS)
As each of these datasets has been collected to address a different need, they differ in terms of both sampling methodology and definition of international investment. This note outlines each source, including the purpose of the collection, the population of interest and sampling strategy, and the aspects of international investment which are captured. It is designed to give new and potential users a broad understanding of the pros and cons of the various sources of investment data, in order to better identify the most appropriate source(s) for their purpose.
Section 2 describes each of the four data sources. A summary reference table is also provided in Appendix B. Section 3 compares the three enterprise-level data sources (BOS, LBF and IR4) and examines the aggregate implications of using alternative sources to identify foreign investment. Section 4 briefly describes potential changes to the availability of FDI data due to the upcoming redevelopment of the Business Frame. Section 5 recaps the key themes.
Throughout the paper we provide examples and descriptive statistics from the LBD. For the majority of these examples we define the population of interest as all private-for-profit firms with current labour input (either employees or working proprietors) and a current industry classification. Table 1 reports the total number of in-scope firms per year and total employment in those firms.
-  Fabling (2009) outlines the content and development of the Longitudinal Business Database.
-  These firm characteristics are drawn from the Longitudinal Business Frame and the Linked Employer-Employee Database (LEED). Firms are excluded from the population if they have ever been in the central or local government sector (with the exception of SOEs), or if they are ever classed as private non-profit organisations. These restrictions lead to the exclusion of around four percent of employing firms, or 19 percent of total employment. This is mainly due to the exclusion of Central Government, which accounts for 64 percent of the excluded employment across the period. Incorporated and unincorporated societies and associations and Charitable trusts jointly account for a further 23 percent.
-  All firm counts reported in this paper have been random rounded base three and employment counts have been graduated random rounded in accordance with Statistics New Zealand confidentiality protocols. As the sampling of the QIIS and AIIS surveys is based on groups of parent and subsidiary firms, rather than individual enterprises (see discussion in section 2.4), the standard population criteria are not applied to those data.
2 Sources of international investment data
2.1 Longitudinal Business Frame
The Longitudinal Business Frame (LBF) forms the backbone of the LBD, to which all other firm-level data sets are linked. It captures basic information on all economically significant firms in the New Zealand economy since 1999, including information on employment, location, industrial activity and parent-subsidiary relationships. The LBF is a longitudinal representation of Statistics New Zealand's primary sampling frame, the Business Frame (BF). That is, it unwinds the historical information available in the BF to create a longitudinal (panel) dataset. The primary purpose of the Business Frame is to maintain a representation of all firms in the economy from which relevant survey populations can be identified (eg, for the Agricultural Production Survey or the International Trade in Services and Royalties Survey). It is also used directly for the construction of business demography statistics.
The LBF provides a single indicator of foreign ownership (bop_ownership_rate)  based on firms' responses to the question “Does any individual or business located overseas hold any ownership interest or shareholding in this business?” and the follow-up “Please give the total percentage.”
The LBF also includes a binary indicator of firms' outward investments (bop_ownership_ind), based on the question “During the last 12 months did this business hold any ownership interest or shareholding in any overseas located business (including its own subsidiaries, branches or sales offices)?” Until 2008, firms were also asked to provide an indication of whether they owned any tangible assets located overseas, such as land, buildings or equipment (excluding stock held overseas) (bop_asset_ind).
The BF indicators of foreign investment are primarily collected to enable the production of national accounts, particularly Balance of Payment statistics. As such, maintenance procedures for collecting and updating firm-level information emphasize firms which will have significant influence on the aggregate value of investment stocks and flows. Information on ownership is originally sourced from the Birth Survey or Monthly Frame Update Survey (MFUS), and is updated over time through the Annual Frame Update Survey (AFUS).  These two surveys, together with administrative data from Inland Revenue and the Companies Office, provide the core information needed to maintain the BF.  The frequency at which firms are surveyed is determined by the maintenance tier (size) of the enterprise. A minimum response rate is then applied to ensure that respondents represent at least 85 percent of industry GST sales.
Since 2004, large and complex (Tier 1) firms have been surveyed at birth and receive an update survey annually. Medium sized firms (Tier 2) are surveyed at birth and updated at least once every three years. Smaller firms (Tier 3) are surveyed at birth only if they are registered for both GST and PAYE (ie, they are above the mandatory GST filing threshold or voluntarily register for GST, and have at least one employee), and are not re-surveyed unless they reach the threshold for Tier 2. As such, foreign ownership information is never updated for most small firms (where “small” is currently defined as those with less than $1 million in annual turnover), and may never be collected if the firm is not employing when they are birthed onto the Business Frame.
Table 2 describes three snapshots of the conditions for membership of each maintenance tier, indicating how these have changed over the period covered by the LBD. In addition to changes in maintenance tier boundaries in 2007 and 2009, there have also been a range of adjustments to the sample, largely reflecting increased use of administrative data. These changes have resulted in substantial reductions in the AFUS sample over time (table 3).
Tier 1: Large and complex businesses
|Number of businesses||45,000||22,000||8,000|
|Share of businesses||0.10||0.05||0.02|
|Share of total GST turnover||0.82||0.75||0.58|
Tier 2: Medium-sized simple businesses
|Number of businesses||120,000||90,000||78,000|
|Share of businesses||0.30||0.20||0.17|
|Share of total GST turnover||0.13||0.20||0.32|
Tier 3: Small simple businesses
|Number of businesses||227,000||329,000||377,000|
|Share of businesses||0.60||0.75||0.82|
|Share of total GST turnover||0.05||0.05||0.10|
Note: Proportions of firms and turnover are approximate. RME = rolling mean employment.
(A) All enterprises with multiple geographic units (multi-geo enterprises), and members of an enterprise group.
(B) Businesses with RME > 7 that do not file their own GST return (due to group-filing arrangements).
(C) Multi-geo enterprises, except those below the tier 1 GST and RME thresholds with 2 or 3 geos that are all in the same Territorial Local Authority. Businesses with RME>20 that do not file their own GST return.
(D) Businesses with 2
(E) Enterprises that don't qualify for tier 1, but are members of an enterprise group, have more than 1 geographic unit, or are identified as having an unusual reporting system for employment (eg, group-level payroll reporting). Businesses with 5
|AFUS postal sample||LBF population||Proportion surveyed|
LBF population defined as the count of live, economically significant enterprises as at February.
Outline of changes to AFUS sample over time
2003: Extended use of Inland Revenue data to maintain the Business Frame reduces AFUS sample, and increases total number of businesses on BF through better coverage of small businesses (realised in 2004).
2006: Postal sample reduced due to increased use of Companies Office data to establish company ownership information, and reduced surveying of large, single location businesses with stable attributes which are included in other Statistics New Zealand Surveys.
2007: Tier boundary changes, and exclusion of superannuation funds, commercial and residential body corporates, and non-complex (single location, non-employing, not part of a group with common ownership) commercial property operators and developers.
2008: One-off exclusion of around 7,000 Agriculture firms because firms surveyed for the Agricultural Production Survey (APS) are excluded from AFUS, and the 2007 APS was a full-coverage census.
2009: Tier boundary adjustments come into effect.
2010: Dropping of “tax triggers” for tier 2 enterprises. Previously, a substantive change in GST activity or employment triggered a firm to be included in AFUS when they otherwise wouldn't. Changes to industry classification in the Inland Revenue register still trigger inclusion in AFUS.
Source: Business Frame metadata: Annual Frame Update Survey Postal Sample Size.
Recording conventions for AFUS respondents have also changed over time. Prior to 2004, AFUS and MFUS responses were coded manually by Business Frame operators. In 2004, Statistics New Zealand shifted to automatic scanning and coding of returns. At this time, the update rules were changed so that zero values were consistently entered for AFUS respondents that indicated that they had no foreign ownership, many of which had previously been left blank. This resulted in a substantial increase in the apparent coverage of the FDI indicator (table 4, columns 1 & 2). In contrast, most MFUS respondents that report no foreign ownership continue to have their ownership status left blank.
|Non-missing FDI indicator||Non-missing ODI indicator|
Meanwhile, the apparent coverage of the outward investment indicator increases slowly over time, and is broadly representative of the population in terms of firm size (table 4, columns 3 & 4). Unfortunately, rather than reflecting a more consistent capture of AFUS and MFUS responses, this is due to assumptions imposed within the Business Frame processing system. Specifically, firms which were born after the ODI indicator was introduced in February 1998 and did not complete an MFUS survey have a “no” response imposed on them at birth. This assumed value remains with the firm until they complete an AFUS survey. Thus, while “yes” values can be taken as an actual firm response, “no”s may indicate either an actual response or simply that the firm has never completed a survey.
These changes imply that while the proportion of firms with a recorded ownership status on the BF has increased, this recorded status has become less up-to-date over time for all but the largest firms. Moreover, it is not clear from the data whether a firm received and returned an AFUS form, so the vintage of the FDI information recorded for any given year and the veracity of recorded “no” responses to the ODI question are unknown.
Further reductions in coverage of the AFUS and MFUS surveys are intended in future as Statistics New Zealand moves towards more extensive use of administrative data. In particular, a new Business Register is under development, which will replace the current Business Frame from 2014 onwards. This change is discussed in section 4 and is likely to increase the available information on foreign ownership among smaller firms as additional administrative data sources come into use. In addition, a clean-up of the historical data is proposed as part of the refresh process which would, among other things, repair the anomalies in the ODI data.
-  Although the definition of economic significance covers all firms that have non-zero employment and/or annual turnover of at least $30,000, there may be some undercoverage of firms with turnover below the mandatory GST filing threshold, which increased from $30,000 to $60,000 between 1999 and 2011.
-  The LBF also includes one variable which, at face value, appears to provide foreign ownership information. However, this is not generally the case. The country_name_text field reports the final line of the respondent's contact address as provided to Statistics New Zealand, rather than the source country of foreign investors. Among firms which are reported as being majority foreign owned, only 20 percent provide an offshore address. When firms do report a foreign contact address, it is not possible to discern whether this relates to the global head-office or to a regional branch (eg, an Australasian head-office).
-  Prior to 2004 these questions were worded as “Does any overseas located enterprise or individual directly hold any ownership interest or shareholding in your enterprise?” and “What is the total percentage (of your enterprise's shareholding or ownership) directly held by overseas located enterprises and individuals?”
-  This question has been asked consistently since 2004. A follow-up question requests that firms provide the name and country of any subsidiaries in which the business has greater than 50 percent shareholding, but this information is not available in the LBF. From 1998 to 2004, this indicator of overseas shareholdings appears to have been derived from the question “If this business or enterprise has a greater than 50% shareholding in any other businesses or enterprises in New Zealand or overseas, please list the name(s) below (or attach own list) i.e. subsidiaries.”
-  Where firms transition into or out of foreign ownership between surveys, these changes are recorded as occuring at the date of the survey, not the date of the ownership change.
-  There is also some manual updating based on media reports and other publicly available information.
-  Firms are legally required to complete official statistics surveys under the Statistics Act 1976.
-  Update surveys may also be triggered if the business shows a significant change in tax flows.
-  Administrative data (eg, IR4s) are not currently used to update LBF investment indicators.
-  Firm counts on the LBF exceed those used in this paper due to the exclusion of non-employing firms and certain industries and firm types as discussed in section 1.
-  Firms born prior to 1998 have a null value for this variable unless they have subsequently completed an AFUS survey.
-  Work is currently underway to provide more information about AFUS updating to LBD users. Initial exploration of this data confirms that LBF direct investment data is indeed outdated for a substantial share of firms. For example, of the 155,000 firms which employed in either 2000 or 2001 as well as in 2010, 48 percent had not responded to an AFUS survey since before 2002 (including those that had never responded). Restricting to those with a recorded FDI indicator in 2010, 20 percent had not responded to AFUS since 2005 or earlier, implying that their FDI indicator may be more than 5 years out of date.
2.2 Company tax returns (IR4)
A second source of data on inward FDI is the annual company tax return, or IR4. IR4s are completed annually by “all active New Zealand resident companies” (with the exception of look-through companies) including body corporates and unit trusts. As the IR4 is designed purely for determining the appropriate tax treatment of responding firms, the indicator of foreign investment reflects the tax legislation. Respondents are asked a series of disclosure questions, relating primarily to the firm's international transactions and interests, including “Is the company controlled or owned by non-residents?”.
This question is used to identify situations where companies may be able to transfer profits internationally through related-party transactions such as transfer pricing. The definition of non-resident control is quite broad, including both New Zealand subsidiaries owned by foreign companies and also New Zealand companies owned or controlled by non-resident natural persons.  Non-resident ownership/control is distinct from identification of the tax residence of the company itself. As firms are expected to indicate any form of non-resident ownership or control, the IR4 indicator is less clearly defined than other sources of FDI data, where firms are asked to provide percentage shareholdings (LBF and BOS) or a dollar value (QIIS/AIIS).
As completing the IR4 is a compulsory annual requirement for all active companies, it is more regularly updated than other sources in the LBD. Response rates for the population of interest are consistently high - around 94 percent among employing registered limited liability companies (LLCs), 79 percent among co-operative companies and 74 percent among branches of companies incorporated overseas.
The data presented here focus on limited liability companies, as LLCs account for 40 percent of all in-scope observations (72 percent of in-scope employment), compared to less than 0.5 percent for the other two groups combined. The high annual response rate and wide coverage makes the IR4 data more suitable for looking at transitions into and out of foreign ownership/control in the population than other sources. In contrast, other sources have either a relatively small sample (BOS) or a selective sample with only partial coverage (LBF, AIIS/QIIS).
From a population perspective, coverage of the IR4 data is driven primarily by the extent of corporatisation. This differs substantially across industries, from over 70 percent in Mining, Wholesale Trade, and Finance & Insurance to less than 20 percent in Agriculture, Forestry and Fishing, and Communication Services (table 5). In addition, the population of LLCs has changed markedly over time due to increasing corporatisation. Between 2000 and 2010, the share of LLCs in the population increased from 28.8 percent to 52.0 percent (63.2 to 77.9 percent of total employment, table 6). This is due both to transitions into corporate structures by existing firms (primarily sole-proprietorships and partnerships) and higher entry rates of LLCs over time (table 7). Of the 176,661 employing LLCs observed in 2010, four percent had previously transitioned from other legal structures, 69 percent entered the population as LLCs over the past 10 years, and 27 percent were continuing LLCs over the full period. As such, while the IR4 data provides a more comprehensive means of identifying transitions into and out of foreign ownership among LLCs, comparisons both within the LLC population and aggregate statistics calculated from this data need to take account of changes in the population itself.
|Agriculture, Forestry and Fishing||0.156||0.297|
|Electricity, Gas and Water Supply||0.620||0.454|
|Accommodation, Cafes and Restaurants||0.579||0.797|
|Transport and Storage||0.437||0.732|
|Finance and Insurance||0.713||0.813|
|Property and Business Services||0.434||0.673|
|Health and Community Services||0.370||0.727|
|Cultural and Recreational Services||0.357||0.639|
|Personal and Other Services||0.331||0.564|
|Share of Firms:||Share of Employment:|
Sole: sole-proprietorship. Part: partnership. LLC: Limited Liability Company.
Population includes all in-scope firms that employed in 2000 and/or 2010. Non-emp: Firm not employing (ie, captures entries/exits). Sole: sole-proprietorship. Part: partnership. LLC: Limited Liability Company. Other: other firm types.
These changes in coverage may, however, be less meaningful than they appear at first glance. In particular, the definition of sole-proprietorships (which account for around half of the non-LLC population of firms - table 6) requires that a sole trader must control, manage and own the business, and is personally entitled to any profits and liable for any debts or taxes incurred by the business. As the business itself is not separable from the owner/manager (and the business is tax resident in New Zealand), there is very little scope for a sole-proprietorship to be foreign owned or controlled. The scope for foreign involvement is greater among partnerships, particularly since the passing of the Limited Partnerships Act 2008. This act enables businesses to be registered as Limited Partnerships or Overseas Limited Partnerships, with two classes of partners - General Partners (the owner/managers, responsible for the debts and liabilities of the business) and Limited Partners (who are liable only to the extent of their capital contribution), and was specifically introduced to facilitate venture capital and private equity investment in New Zealand firms (Companies Office, 2010). However, reported foreign ownership rates from LBF and BOS are low among both sole-proprietor and partnerships (effectively zero for sole-proprietorships and less than 0.5% for partnerships among firms with 6 to 10 employees, compared to five to six percent among similarly-sized LLCs). This, combined with their relatively small size and the constraints of their legal structure, suggests that treating sole-proprietors and partnerships as domestically owned is likely to be a close approximation of the truth.
-  This covers a range of situations including non-resident companies, majority shareholdings by non-residents (including their own and associated persons' direct and indirect interests) and situations where a non-resident has "control of the company by any other means".
-  A company is defined as tax resident in New Zealand if: it is are incorporated in New Zealand; the head office or centre of management is in New Zealand; or the directors control the company in New Zealand. If none of these conditions holds, the company is non-resident for tax purposes (Inland Revenue, 2012).
-  Changes in legal status are one event which may trigger a firm to be given a new enterprise number by Statistics New Zealand. Enterprise number breaks have been repaired where possible, using the method developed by Fabling (2011). However, some breaks likely remain, particularly for firms employing less than 3 people, leading to an upward bias in the number of 'entering' LLCs and 'exiting' sole-proprietor and partnerships.
2.3 Business Operations Survey (BOS)
The Business Operations Survey (BOS) is a large-scale sample survey administered annually by Statistics New Zealand since 2005. The survey generally consists of three modules. Module A covers general questions on Business Operations and is conducted annually. Module B alternates between Business Use of ICT (even years) and Innovation (odd years). Module C is a contestable module, sponsored annually by one or more government departments.
The primary source of FDI/ODI data in BOS comes from a set of questions on firms' “international presence” in Module A. These questions have been asked consistently through nine waves of BOS (2005-2013), and provide information on percentage foreign ownership in the New Zealand enterprise, and the existence and method of ODI (eg, joint ventures, acquisitions of existing firms, greenfields investment).
In addition, the contestable modules for 2007 and 2011 focused on firms' international engagement, capturing information on the types of international activities firms were engaged in as well as their motivations for, and barriers to, those activities. The modules are organised around the types of activity the firm undertakes, distinguishing between generation of overseas income, overseas production, and overseas purchasing. For the former two categories, firms were also asked about the means through which the activity is undertaken (eg, whether production is undertaken through ODI, a joint venture, or via agents/contractual arrangements), providing an indication of the broad purpose of ODI.
The advantage of BOS over other firm-level foreign investment data for the production of longitudinal statistics is the consistent coverage of small to medium-sized firms. The BOS survey population includes all private-for-profit firms with six or more employees (roughly 35,000 firms). Of these, a sample of around 7,000 firms is surveyed each year, with a required minimum response rate of 80 percent. The survey is approximately full coverage for large firms, with a stratified random sample of small and medium firms. This sampling strategy allows for comparison of the aggregate proportion of firms with FDI and/or ODI over time across a clearly and consistently defined population.
In addition, Statistics New Zealand's sampling strategy leads to substantial repeat sampling of firms over two or more years. This is enhanced by an explicit panel element in the BOS, through a top-up sample which re-surveys all respondents to the 2005 survey, whether or not they would otherwise have been surveyed in later years. Table 8 reports the longitudinal coverage of the BOS data, showing the number of respondents in each year conditional on responding to the 2005, 2007 and 2011 surveys respectively (columns 1 to 3), and the total number of responses per firm, conditional on responding in those same years (columns 4 to 6). For example, column 1 shows that of the 7,440 BOS respondents in 2005, 4,218 also responded in 2008. Column 4 divides those same 7,440 2005 respondents into groups according to the total number of BOS surveys they have completed – 1,734 only have information for the 2005 year, while 2,271 responded in each of the 7 years to 2011. This panel component provides an opportunity to examine characteristics associated with transitions into and out of foreign ownership, albeit for a relatively small population.
|Number of firms:||Number of BOS observations:|
Includes all responses which meet a threshold quality standard. Population restrictions outlined in section 1 are not applied to this table. The imposition of these restrictions (in particular, the exclusion of firms which are ever in the government or non-profit sectors) reduces the population used elsewhere in the paper by approximately 50 observations per year. Imputation rates for simple FDI indicators are relatively low (
-  See Appendix A for BOS investment questions and response categories.
-  Survey results, broken down by industry and firm size, are available at http://www.stats.govt.nz/browse_for_stats/businesses/business_growth_and_innovation/business-operations-survey-info-releases.aspx.
-  While these data are potentially of interest, there are some concerns about their quality due to internal inconsistencies in responses. In particular, in the 2007 survey almost half the firms which identified as having an ownership interest in an overseas production facility in Module C did not identify as having overseas shareholdings in Module A. Similarly, firms which report ODI in Module A may validly not respond to Module C questions if their overseas investments are not currently generating overseas income or producing goods or services overseas (eg, research facilities).
-  Further information on the sampling methodology can be found in the annual Hot Off the Press releases at http://www.stats.govt.nz/browse_for_stats/businesses/business_growth_and_innovation/business-operations-survey-info-releases.aspx.
-  The panel component was re-set in 2012. This will reduce the overlap with earlier years, but improves the representativeness of the forward-looking panel.
-  Producing longitudinal statistics at the aggregate level (eg, transitions into and out of FDI/ODI) would also require creation of longitudinal survey weights which are not currently provided by Statistics New Zealand.
2.4 Quarterly and Annual International Investment Surveys (QIIS/AIIS)
The Quarterly and Annual International Investment Surveys (QIIS & AIIS) provide significant detail about the international investment stocks and flows of New Zealand firms which: are wholly or partly foreign owned, including New Zealand branches of overseas businesses; have ownership interests in businesses located overseas, including branches; and/or have financial asset and liability positions with overseas residents. These surveys form the basis of the Balance of Payments (BoP) and International Investment Position (IIP).
Information on the value of FDI and ODI stocks are captured in both surveys, while QIIS also provides details of the composition of changes in investment values (financial accounts, market price revaluations, exchange rate adjustments, and other changes). Investments are distinguished according to source or destination country, and are divided into equity capital, re-invested earnings, and “other capital”, which includes permanent or long-term loans to and from overseas parent or subsidiary companies.
The survey methodology of QIIS and AIIS is designed to capture the value of New Zealand's international balance sheet in a cost effective manner. The sample strategy therefore focuses on those firms known to account for the majority of the overseas assets and liabilities. These firms are primarily identified through the Business Frame maintenance process, with AFUS including a series of questions on firms' international investments and transactions, as discussed in section 2.1. Additional firms may also be identified using administrative data or through other information sources.
QIIS is conducted quarterly and is designed to capture 95 percent of New Zealand's total international financial assets and liabilities. It collects data on the breakdown of foreign investment in New Zealand and New Zealand's investment abroad by country, industry, type of instrument, maturity, currency and by type of asset and liability. The QIIS population includes banks, corporate enterprises, the Reserve Bank and the Treasury. A response rate of 100 percent of “key” and 80 percent of “non-key” enterprises is required.
AIIS is conducted annually as a stratified sample survey to estimate the remaining five percent of international assets and liabilities, with a census of the known QIIS/AIIS population conducted every three years. While QIIS collects income earned from international investments (profits, dividends and interest), as well as stock measures of international assets and liabilities, AIIS collects only stock values. Table 9 reports the number of AIIS and QIIS respondents in each year, distinguishing firms which respond to at least one QIIS survey and those which respond to all four quarterly surveys within the year.
Both surveys are sampled at the group level, rather than the individual enterprise. The survey goes to the “New Zealand Group Top Enterprise” (NZGTE) - the highest enterprise within the group which has a geographic unit located in New Zealand - and that firm is expected to respond on behalf of all its subsidiaries, including those located offshore. For example, in figure 1, firm B would respond on behalf of itself and firms D through J.
- Figure 1: QIIS structure
This survey method provides the extensive detail required to produce the IIP and BOP aggregates efficiently, reducing both processing costs and respondent burden, as well as the potential for double-counting of assets and liabilities by multiple members of an enterprise group. However, it creates substantial challenges for micro-economic research. First, any analysis using linked data from other sources must take account of the grouping structures in the data, including the potential for restructuring or mergers and acquisitions to affect the structure of respondent firms. Second, as the sampling method is focused on firms which are known to have substantial international assets or liabilities, it cannot be accurately used to identify a comparison group of firms without international holdings or to examine transitions into and out of international investment activity. Finally, due to the small number of firms with international investments, the level of detail which can be reported is limited to protect the confidentiality of respondents.
However, there are still a range of firm-level estimates which could be produced using QIIS/AIIS responses. These include comparisons of characteristics of firms with investment from different source countries, examination of within-firm investment dynamics, and associations between investment and other international assets and liabilities.
-  QIIS and AIIS also contain data on non-direct forms of international assets, including portfolio investment and reserve assets.
-  Unlike BOS, where sample weights are provided for each respondent based on the inverse probability of being included in the sample, weighting in QIIS and AIIS is applied through a macro adjustment to the aggregate figures.
-  For example, in recent years, almost 30 percent of AIIS responses and 45 percent of QIIS responses come from enterprises with no recorded employment in New Zealand. Even at the group level, as many as 15 percent of AIIS respondents appear to be from non-employing groups. However, many of these groups appear to employ at some point over the sample period, and the share of non-employing groups is lower in early years (5-8 percent), suggesting that respondent enterprise identifiers may not be kept fully up to date with the changing structure of the group.
3 Comparisons across data sources
This section compares coverage and implied FDI and ODI rates across the three enterprise-level data sources: the LBF, IR4 and BOS. QIIS/AIIS is excluded due to the nature of the sampling and substantive differences in the range of applications for which the data are suitable.
As all three enterprise-level sources have higher coverage rates among large firms (by design in the case of the BOS and LBF samples, and due to the characteristics of LLCs in the case of the IR4 data), there is substantial overlap in the availability of FDI data. In particular, almost all firms which have BOS data also have at least one of the other sources, and most have all three (table 10). Taking the 2010 year as an example, 27.7 percent of firms have a recorded ownership status in the LBF, with approximately two-thirds of those (18.4 percent) also having ownership information sourced from IR4s in that year (panel A, columns 3 & 5). A little over a third of all employment (33.7 percent) is in firms which respond to BOS, of which nearly all (32.6 percent) also have LBF indicators available (panel B, columns 4 & 6). Notwithstanding issues of timeliness discussed above (section 2.1), since 2005 at least one indicator of foreign ownership has been available for between 48 and 58 percent of firms (over 80 percent of total employment). The share of firms which are not included in any of the three sources fell from 72.0 percent in 2000 to 42.2 percent in 2010, with the employment share of those firms falling from 40.4 percent to 13.7 percent (panels A & B, column 1).
Coverage rates are strongly correlated with firm size. Figure 2 plots the share of firms covered by each data source for each percentile of the population employment distribution. Coverage rises rapidly with firm size in both the IR4 and post-2004 LBF, reaching 80 percent for firms with more than 10 employees. The impact of improvements in data capture post-2004 are clearly visible in the LBF data, with substantial increases in coverage across the population but particularly in larger firms. Similarly, increasing corporatisation over the decade shows up in the increasing level of IR4 coverage in the smaller size groups.
The substantial overlap allows for an assessment of the consistency of responses across sources. Restricting to the population of firms for which both sources are available, table 11 reports the comparison of banded ownership percentages from the LBF and BOS data, while table 12 compares the IR4 indicator variable with each of the BOS and LBF bands.
Comparing LBF and BOS responses (table 11), 90.2 percent of all observations fall within the same ownership percentage band across the two sources. This consistency is driven almost entirely by two groups of firms - those that record no FDI in either source (79.6 percent of observations) and those that record 100 percent foreign ownership in both sources (8.3 percent of observations). Of the 10 percent of observations where the two sources are inconsistent, rather than being concentrated in cells which are immediately off diagonal (ie, where the reported FDI percentages are close but not identical), the most common situation is for firms to report positive levels of foreign investment in the LBF, but not to report having FDI in BOS (5.3 percent of observations).
Table 10: Availability of FDI data sources
- Figure 2: Apparent coverage of FDI indicators, by employment percentile
Plot of rolling mean employment percentiles (x-axis) against the percentage of firms within that percentile for which the relevant data source is available (y-axis).
Table 12 reports the share of firms in each ownership band from the LBF and BOS that also report foreign ownership or control in their IR4 returns. Here the responses are most consistent for ownership levels above 50 percent. For example, among firm-year observations with 100 percent recorded foreign ownership in the LBF, 67.8 percent also indicate foreign ownership or control in IR4. In contrast, recorded ownership rates below 25 percent tend to elicit negative responses to the IR4 question, while recorded foreign shareholdings of 26 to 50 percent are associated with moderate probabilities of a positive response in corporate tax returns. The level of consistency is higher between the two annual data collections (BOS and IR4). However, there are still a large number of firms which report conflicting information regarding their foreign ownership status, even in data collected annually.
Outward investment indicators from the LBF and BOS show a similar degree of inconsistency (table 13). While in total over 96 percent of observations are consistent across the two sources, this is primarily due to the large share of firms which report no ODI in either source. In contrast, if we compare those firms which report ODI in at least one source, only a little over half are consistent across the two sources (3.8 percent of all observations, compared to 3.6 percent with conflicts).
Although it is not possible to determine which source is more correct in any particular case, due to conceptual differences in the questions as well as the potential for differences in interpretation and for respondent error, less frequent updating through AFUS surveys means the LBF is less likely than other data sources to pick up transitions between foreign and domestic ownership. Table 14 suggests that this is likely to be one source of difference in the FDI data, with the combined foreign-to-domestic and domestic-to-foreign transition rate observed in the LBF data (0.9-1.8 percent, depending on definitions) substantially lower than that for the BOS or IR4 (1.6-3.2 percent and 2.9-3.0 percent respectively).
Aggregate estimates of ODI propensity from LBF and BOS suggest that the imposition of “no” indicators for new firms and the relatively infrequent updating of ODI indicators for small and medium firms may also be affecting the aggregate recorded ODI propensity in these firms. Table 15 reports the aggregate share of firms with six or more RME with a recorded ‘yes’ response to the LBF and BOS indicators. While the LBF statistics report a fairly constant proportion of firms with outward investments and share of employment accounted for by those firms since 2004 (at around 2 percent and 10 percent respectively), BOS results suggest a slight increase in both the share of firms with ODI (from 3 to 4 percent) and the employment share of those firms (from 12 to 15 percent of employment in 6+ RME firms). The common sample comparison is also consistent with delayed updating of the LBF data, as inconsistencies between the two sources are more commonly due to a positive response in BOS for firms with no recorded ODI in the LBF.
|LBF||0||1 - 25||26 - 50||51 - 75||76 - 99||100|
|1 - 25||0.019||0.010||0.000||0.000||0.000||0.000|
|26 - 50||0.009||0.000||0.006||0.001||0.000||0.000|
|51 - 75||0.018||0.001||0.001||0.005||0.001||0.027|
|76 - 99||0.001||0.000||0.000||0.000||0.002||0.001|
Proportions of all firms by foreign ownership percentage bands (0, 1-25, 26-50, 76-99, 100), as shown by BOS responses and LBF indicators. Sample restricted to those firms which have a reported percentage in both sources in a given year.
|1 - 25||0.019|
|26 - 50||0.107|
|51 - 75||0.533|
|76 - 99||0.415|
|1 - 25||0.055|
|26 - 50||0.264|
|51 - 75||0.516|
|76 - 99||0.516|
Proportion of firm-year observations in each foreign ownership percentage band (in BOS and LBF) for which firms also report foreign ownership or control in IR4. Sample restricted to those firms which have a reported status in both sources in a given year.
Proportion of firm-year observations by recorded ODI status in LBF and BOS. Sample restricted to those firms which have a reported status in both sources in a given year.
A: 50%+ foreign ownership
B: Any foreign ownership
Transitions reported for observations of firms which have all three sources of data in two consecutive years. Panel A: Foreign ownership defined as > 50% ownership in BOS, LBF, or a positive indicator in IR4. N=17,016. Panel B: Foreign ownership defined as > 0% ownership in BOS, LBF, or a positive indicator in IR4. N=17,262. IR4 shares differ slightly across panels due to firms which report having some FDI in BOS but do not specify the level of foreign shareholding.
|Share of firms:||Share of employment:|
Missing and “don't know” responses treated as “no”.
-  As discussed above, “coverage” for the LBF indicates that information on foreign ownership has been collected at some time in the past, but this information may not be recent.
-  Correlation coefficient for the raw percentages is 0.882.
-  While differences in timing between the BOS survey and filing of IR4 returns may account for a portion of this gap, this cannot be the major explanation, as transitions between domestic and foreign ownership are rare.
-  Measurement error may also have an amplified effect on transition rates in annual data. For example, if a firm gives an inconsistent response in one year of the BOS or IR4 this would show up as a transition into, and then out of, foreign ownership. While there is no reason to expect measurement error to be more of a problem in BOS/IR4 than in the AFUS survey, the same level of misreporting would have less impact on observed transitions in the LBF data as the average inter-reporting period is longer. For this reason, Fabling and Sanderson (2013b) perform sensitivity tests to ensure their results are not strongly affected by “transitory” foreign acquisitions, while Maré et al. (forthcoming) adjust indicators of foreign ownership to allow for some misreporting in IR4 responses.
-  Firms with missing or “don't know” responses treated as “no”.
3.1 Implications of sampling method for aggregate FDI statistics
The combination of within-firm inconsistencies and differences in sampling strategies leads to significant differences in the observed share of foreign-owned firms in each source. Table 16 reports the proportion of foreign ownership among those firms which have a non-missing value for the FDI indicator. As discussed above, manual capture of AFUS and MFUS responses prior to 2004 often failed to record zero values, significantly increasing the apparent share of foreign-owned firms. Similarly, ongoing corporatisation lead to a fall in the apparent share of foreign ownership or control reported in the IR4 over time, as more small (owner-operated) domestic firms entered the LLC population. In contrast, BOS data, which are collected on a consistent basis over time, present a more consistent statistic, displaying moderate growth in foreign ownership since 2005.
As the main differences across sources are driven by sampling, restricting to something akin to a consistent population and variable definition provides a more consistent picture of foreign ownership both across sources and over time. Table 17 reports the implied aggregate level of foreign ownership where missing values in the LBF (ie, firms which have never responded to an AFUS or MFUS survey, or which submitted a negative response prior to 2004) and null values from the IR4 (primarily sole-proprietors and partnerships) are assumed to be under domestic ownership. Columns 1 and 2 report implied proportions from the LBF and IR4 across all firm sizes, columns 3 to 5 restrict to firms with at least six RME for consistency with the BOS population, while columns 6 and 7 impose a majority ownership definition on the LBF and BOS data, to provide an alternate comparison with IR4s.
Finally, table 18 reports a “best guess” estimate of the share of firms and employment under foreign ownership or control, combining information from LBF and IR4. Firms are classed as foreign-owned based on either a positive response to the IR4 question on foreign ownership or control, or a majority foreign shareholding recorded in the LBF (following (Fabling and Sanderson, 2013a)). These numbers are closely comparable to BOS results for those firms with 6 or more employees, though the combined population estimate is slightly higher than that from the BOS survey.
|Share of firms:||Share of employment:|
Population for each column is the firms which have non-missing FDI information for that source (null and “don't know” values are dropped). BOS column reports weighted share of in-scope respondents and may differ slightly from official estimates due to population restrictions.
|All firms:||6+ RME:||6+ RME, 51%+:|
|Share of firms:||Share of employment:|
|All firms||6+RME||All firms||6+RME|
- There is no clear way to distinguish between these two cases. Inferring a 'no' response from the presence of other AFUS-only variables is of limited value as (a) these variables are also likely to not be recorded if the response was negative or may have been assumed negative in the case of new births and (b) in some cases these variables may have been updated via manual changes based on media reports or responses to other surveys
- Future work could integrate information on the vintage of AFUS responses to guide the choice of which source to prefer in each year. Maré et al. (forthcoming) take this approach in defining foreign ownership status over time (distinguishing between firms which are 'always', 'never', or 'sometimes' foreign owned).
4 Future developments
As noted above, Statistics New Zealand are currently engaged in a redevelopment of the Business Frame (BF). This process is designed to reduce the cost of maintaining a full-coverage sampling frame, while maintaining an appropriate level of coverage and data quality for the production of National Accounts and other official statistics. One of the key aspects of this redevelopment is formalising the use of administrative data as a core means of maintaining the BF in future.
Statistics New Zealand has already made substantial moves in other areas (such as the auto-coding of industry classifications), shifting the balance away from direct data collection towards the use of administrative data sources collected by other agencies (especially Inland Revenue). This is clear in the reduced use of AFUS and MFUS surveys. As administrative data is able to provide much of the same information as surveys, this does not imply a reduction in data quality for many variables of interest. However, in the case of foreign investment, currently used administrative data sources do not hold information about FDI or ODI.
As part of the redevelopment, Statistics New Zealand is looking to make greater use of shareholder data collected by the Companies Office. This could provide more detail on the characteristics of foreign investors currently captured only by the IR4 indicator, including information on investor source countries and percentage of foreign ownership, for the population of registered companies. Integration of the available historical data would be particularly valuable to support empirical research.
The development of the LBD has dramatically expanded the options available to empirical researchers with an interest in international investment. Linking existing data sources opens up a wider range of potential questions and provides new challenges to make appropriate use of the data available.
Recent firm-level research has utilised a range of different data sources, variable definitions and methods to account for sample selection, depending on the research question at hand. These include restricting attention to limited liability companies (IR4 filers) where large samples and frequent updating are required, such as when examining the performance impacts of foreign acquisition (Fabling and Sanderson, 2013b); combining LBF and IR4 annual data to provide a simple indicator of foreign ownership with wide coverage, as used in examinations of differences across foreign and domestically-owned firms in effective marginal tax rates (Fabling et al., 2013) and exchange rate pass-through behaviour (Fabling and Sanderson, 2013a), and research on productivity spillovers from foreign-owned to domestic firms (Doan et al., forthcoming); and categorizing firms by ownership status over all time to allow for differing vintages of LBF information, as used by Maré et al. (forthcoming) in their examination of the earnings impacts of employment in foreign-controlled firms.
This paper has set out some of the factors that researchers need to consider before embarking on firm-level research on FDI and ODI. How each of these factors affect methodology and feasibility depends on the specifics of the research question at hand. However, some broad themes emerge:
- From 2006 onwards, over half of all employing firms, employing over 80 percent of workers, have at least one indicator of their foreign ownership status (though the recorded information is not necessarily timely).
- Coverage of small firms is limited, with only IR4 data being collected across all firm sizes. Where data is available, sole-proprietors and partnerships are seen to have very low reported rates of FDI, suggesting that an assumption of domestic ownership in the non-corporate small-firm population is likely to be close to the truth.
- Transitions into and out of FDI, required to examine the dynamics of acquisition and divestment, are best considered through either the IR4 data (if a broad definition of foreign control is sufficient) or BOS (if a more specific understanding of the level of ownership is required).
- Information on ODI is more limited than FDI data, particularly where smaller firms are concerned. Limited updating and the imposition of of a negative indicator among firms birthed since 1998 suggests that current LBF indicators may underestimate the extent of ODI by New Zealand firms. While this problem is expected to be partially mitigated as part of the redevelopment of the Business Frame, the BOS is likely to remain the most useable source of ODI data in the foreseeable future.
- The recent addition of AIIS and QIIS data to the LBD opens up new opportunities for research on international investment dynamics. However, given the sampling structure (small, non-representative sample), these are largely limited to exploration of within-respondent changes in investment activity, and will involve a range of challenges associated with small samples (including both generalisability of results and the ability to design research questions which do not conflict with confidentiality requirements).
Companies Office (2010). Introducing limited partnerships. available from http://www.business.govt.nz/companies/learn-about/other-entities/limited-partnerships/introduction. Accessed 30 May 2013. Last updated 17 June 2010.
Doan, T., K. Iyer, and D. Maré (forthcoming). Linkage spillovers of foreign direct investment on domestic firm productivity in New Zealand industries. Occasional paper, Ministry of Business, Innovation and Employment.
Fabling, R. (2009). A rough guide to New Zealand's Longitudinal Business Database. Global COE Hi-Stat Discussion Papers No. 103, Institute of Economic Research, Hitotsubashi University.
Fabling, R. (2011). Keeping it together: Tracking firms in New Zealand's Longitudinal Business Database. Working Paper 11-01, Motu Economic and Public Policy Research.
Fabling, R., N. Gemmell, R. Kneller, and L. Sanderson (2013). Estimating firm-specific effective marginal tax rates and the user cost of capital in New Zealand. Working Paper 13/XX, New Zealand Treasury.
Fabling, R. and L. Sanderson (2013a). Export performance, invoice currency and heterogeneous exchange rate pass-through. Working Paper 13/03, New Zealand Treasury.
Fabling, R. and L. Sanderson (2013b). Foreign acquisition and the performance of New Zealand firms. New Zealand Economic Papers DOI: 10.1080/00779954.2013.773600.
Inland Revenue (2012). Company tax return guide 2012. Technical Report IR4 GU, www.ird.govt.nz, Inland Revenue.
Maré, D., L. Sanderson, and R. Fabling (forthcoming). Earnings impacts of foreign direct investment. Working paper, Motu Economic and Public Policy Research.
Appendix A: Survey questions
- Figure A.1: Business Operations Survey Module A questions
- Figure A.2: Business Operations Survey Module C questions, 2007 and 2011
Small changes to these questions were made between the 2007 and 2011 surveys. The questions above are from the 2011 survey.
Appendix B: Summary of data sources
|Percentage foreign ownership||✔||✔|
|Stock value of investment||✔||✔|
|Decomposition of investment stock values||✔||✔|
|Decomposition of investment flow values||✔|
|Form of investment||✔|
|Stock value of investment||✔||✔|
|Decomposition of investment stock values||✔||✔|
|Decomposition of investment flow values||✔|
LBF - All firms covered in principle, but maintenance aimed at larger firms. Indicators may be out-of-date for small/medium sized firms, and for larger firms which have not responded in one or more years.
BOS - Annual stratified sample survey of approximately 7,000 firms with 6+ RME.
IR4 - Annual administrative data for almost all limited liability companies and a substantial share of other corporate entities.
QIIS - Quarterly survey of approx 500 firms, responsible for 95% of known international assets and liabilities.
AIIS - Annual sample and triennial census of firms responsible for the remaining 5% of known international assets and liabilities (based on AFUS indicators).