Business Finance Survey: 2004

Survey background

The Business Finance Survey was sponsored by the Ministry of Research Economic Development (MED). It is the first survey of its type to be conducted in New Zealand, and was developed by Statistics New Zealand in collaboration with MED.

The core objective of the Business Finance Survey is to provide information on the capital structure of businesses in New Zealand, the sources of finance they use and their recent financing experiences. This will be used to help inform public policy decisions and help stakeholders (both public and private) to better comprehend the challenges and opportunities surrounding the provision of capital to New Zealand businesses.

The information will help government and other organisations develop a better understanding of the financing needs and practices of businesses in New Zealand. In particular, the results will:

• improve understanding of business financing needs, thus enabling better informed public debate
• give financing providers information that they need to gain a more comprehensive understanding of their clientele, thus better enabling them to design products and services that meet market needs; and
• facilitate the assessment of whether business financing needs are being addressed by the market, and help gauge the effectiveness of government policies and programmes.

Previous work on New Zealand’s financial system and its constituent parts (eg banking, capital markets, venture capital) has almost exclusively focussed on the supply side of the market. While understanding supply-side issues is critical, it cannot provide a complete picture of the state of business financing because these studies do not highlight the relative importance of non-intermediated or “private” capital flows which, according to the international literature, are an important source for all but the largest of businesses.

Without comprehensive information on business financing in New Zealand, it is very difficult for policy makers to assess the need for, and likely impact of, existing and potential government initiatives. Public sector initiatives to promote access to finance are best justified if market imperfections exist and result in the private sector not providing capital to businesses on competitive terms. In the absence of market failure, government intervention may cause distortions such as non-viable businesses being subsidised at public expense.

It is therefore essential to collect high-quality data on business financing to determine the extent to which particular categories of businesses, if any, are systematically disadvantaged with respect to financing. Better demand-side information will facilitate more effective public policy by fostering the development of a widely accepted and empirically supported policy framework around the notion of capital market imperfections. Otherwise, anecdotally based perceptions of specific types of financial market “gaps” may inappropriately drive public policy.

Data collection

The Business Finance Survey was a postal survey, sent out in late August 2004. The survey form was directed to the 'Managing Director'.

Information collected included: whether the business had sought or received finance over the previous 12 months; and the instruments, sources and uses of that finance. Financial information was also collected for the business's last financial year, to enable the relationship between a business’s current financial situation and its need for, or success in, attracting extra finance. A range of other individual and business characteristics were also examined to determine if non-financial factors also impact on a business’s ability to attract finance.

Target population

The target population for the Business Finance Survey was live enterprise units on Statistics New Zealand’s Business Frame at the population selection date which:

• were economically significant enterprises (those that have an annual GST turnover figure of greater than $30,000)
• had between 1 and 500 employees inclusive
• had been operating for six months or more
• were not subsidiaries, more than 50 percent owned by another business
• were classified to Australian and New Zealand Standard Industrial Classification – NZ Version 1996 (ANZSIC96) codes listed as in scope in Table 1 below
• were private enterprises as defined by New Zealand Institutional Sector 1996 Classification (NZISC96) listed in Table 2 below
  were classified to New Zealand Standard Classification of Business Types (BT96) codes listed in Table 3 below.

The final estimated population size for the Business Finance Survey was 84,000 enterprises.

---PDF BREAK---

Table 1
ANZSIC96 Codes in Scope

Table 2

ZISC96 Codes in Scope

---PDF BREAK---

Table 3

BT96 Codes in Scope

Sample design

Table 4

Groups Used in Sample Design

Measurement errors

The Business Finance Survey results are subject to measurement errors, including both non-sample and sample errors. These errors should be considered when analysing survey results.

Non-sample errors: Non-sampling errors include mistakes by respondents when completing questionnaires, variation in the respondents’ interpretation of the questions asked, and errors made during the processing of the data. In addition, the survey applied imputation methodologies to cope with non-respondents. Statistics New Zealand adopts procedures to minimise these types of error, but they may still occur and are not quantifiable.

Given the nature of the data collected, there are additional non-statistical limitations on the level of accuracy that can be expected from the survey. New businesses or those which have recently begun operations are of particular interest for this survey, but most of these businesses do not yet have a full set of annual accounts. Also, many smaller businesses are not in possession of full and/or up to date accounts, and may or may not use external accountants. In addition, for businesses of any size or age, records may not be kept in the exact form required for the survey, and some estimation may be required in these cases.

Sample errors: The accuracy requirements were specified at the following levels:

• published industry by size breakdowns for the categorical questions
  overall for the numeric questions.
• for: ANZSIC divisions; Rolling Mean Employment (RME) groupings; Age groupings

Final requirements were:

• for primary outputs by age, industry or size: 5 percent absolute sampling error for categoric questions
• for primary outputs by size and industry or age and industry: 10 percent absolute sampling error for categoric questions
• for key numeric questions (total debt and total operating revenue): the lowest relative sampling error possible with maximum post-out size of 6,000. Final maximum relative sample errors for numeric variables were:
  - 10 percent for total debt
  - 59 percent for total equity

Response rate

The target overall response rate from the survey was 80 percent for questionnaires returned. The survey achieved this overall rate of return, which represented 4,775 businesses.

Non-response and imputation

Unit and item non-responses

The unit non-response criteria were:
(a) the business did not return a form
(b) it returned a form but did not answer the key questions on debt and equity finance requests or current debt and equity
(c) it returned a form, answered the key questions, but did not answer 60 percent or more of the rest of the questions it was meant to answer (with regards to the routing).

These units were not imputed for, but were considered a complete non-response and had their weight adjusted within strata.

Units that did not reach the criteria for unit non-response, and did not respond to one or more questions were treated as partial non-responses and had missing variables imputed.

Imputation of item non-response – numeric variables

The numeric questions range from breakdowns of total equity and debt, to selling accounts receivable and value of assets.

---PDF BREAK---

Imputation cells and merging

Units were assigned to imputation cells for the calculation and assignment of imputation factors. Imputation cells were based on the design groups detailed in Table 4 above (which were determined by industry, RME, and age).

For each variable there was a minimum number and percentage of linked units within the imputation cell for the imputation method to run in that cell for that variable. This is to ensure the robustness of the imputation factor calculations. The minimum number of linked units was 10 and the minimum percentage of linked units was 60 percent.

If an imputation cell did not reach these criteria, it was merged before imputation, using the following merging rules:

• If there is insufficient response, merge the cell with the first merge preference.
• After the first merge, if there is still insufficient response (ie less than the required number or percentage of units linked) in the merged cell, merging continues with the next cell in the specified order.
• Merging continues until there is sufficient response (ie enough linked units in the merged cell) to calculate an imputation factor.
• Once sufficient response (ie enough linked units in the merged cell) is achieved during the merging process, the imputation factor is calculated from all of the linked units in the merged cell, but applied to non-respondents in the original (un-merged) imputation cell only.
• If there is still insufficient response once all the specified cells have been merged in order, the imputation factor is calculated across all linked units in the final merged cell (ie use the 'best available' imputation factor), and applied to non-respondents in the original (un-merged) imputation cell only.

---PDF BREAK---

Imputation of numeric variables

Weighted mean imputation was used to impute numeric variables. With this method, non-responding units were imputed with the weighted mean calculated from linked responding units within their imputation cell.

The weighted mean for the current period t is calculated as follows: 

Imputation of component questions

The Business Finance Survey questionnaire included routing and questions with many components. For example, total debt has 10 different components of debt to give a value for. For such questions, due to linecode-specific unlinking, it was possible for the mean imputation factor for the total to not equal the sum of the mean imputation factors for the components. To preserve additivity, imputed components were scaled to add to the total.

Unlinking

Influential responses were unlinked from the imputation factor calculations for numeric variables. There were three kinds of unlinking:

• automatic exclusion due to non-response or special treatment – before imputation
• automatic unlinking due to influence, ie units with undesirable influence on imputation factor calculations were automatically detected and unlinked (with the ability for manual declining of this) – after special treatment.
• manual unlinking due to influence, ie additional units with undesirable influence on imputation factor calculations that were not automatically detected were unlinked – after special treatment.

---PDF BREAK---

Automatic exclusions

A unit was automatically excluded from the imputation factor calculations for the imputation if:

• it was a non-respondent, or
• it was specially treated.

Automatic unlinking by outlier detection

Automatic unlinking (ie by the outliers process) and manual unlinking could be done separately for each variable (linecode).

Checks were done for automatic unlinking candidates using standard interquartile range tests. All checks were done at the imputation cell level or merged imputation cell level. In all cases, the ratio zi is the absolute percent change in the cell factor in the imputation cell or merged imputation cell if the unit is unlinked,

where the cell factor is the mean factor in the imputation cell or merged imputation cell.

The parameter k is used in the checks to determine the number of outliers identified.

---PDF BREAK---

Imputation of item non-response – categoric variables

Responses were imputed for categorical questions using nearest neighbour imputation. The nearest neighbour method involves finding a donor unit with the most similar set of responses to the unit that requires values to be imputed. The donor unit supplies responses for all variables requiring a response. If the donor unit does not respond to any of the variables requiring a response, then the next best donor unit would be used to supply the information. This is continued until all the variables have a response.

Example:

Unit A does not respond to four variables. Units B and D have the greatest number of matches with unit A. Unit D is chosen as the first donor variable, as it can supply the greatest number of responses required (it supplies responses for variables 4, 5, and 6). Unit C is the donor for variable 7 as units D and B cannot supply a response.

Therefore, unit A would have the following responses:

Definitions

The Business Finance Survey was designed to collect data in accordance with the following definitions and terminology:

Copyright

Liability

While care has been used in processing, analysing and extracting information, Statistics New Zealand gives no warranty that the information supplied is free from error. Statistics New Zealand shall not be liable for any loss suffered through the use, directly or indirectly, of any information, product or service.

Timing

Timed statistical releases are delivered using postal and electronic services provided by third parties. Delivery of these releases may be delayed by circumstances outside the control of Statistics New Zealand. Statistics New Zealand accepts no responsibility for any such delays.

Next release ...

A full Business Finance Survey report will be released on 30 June 2005.