Economic Survey of Manufacturing

Changes to the methodology for the Economic Survey of Manufacturing

See Economic Survey of Manufacturing: September 2015 quarter for changes we made to the methodology used in the Economic Survey of Manufacturing. The changes:

• make greater use of administrative data sources

• reduce respondent burden

• introduce a consistent methodology and processing system across the Economic Survey of Manufacturing, Wholesale Trade Survey, and quarterly Selected Services Survey

• enable the delivery of information at lower levels of detail for research and customised requests

• improve the quality of the published series.

Under the old design, we surveyed all the large businesses in each industry, plus a sample of medium-sized businesses. We supplemented this with modelled tax data for the smaller businesses.

Under the new design, we use administrative data (goods and services tax (GST) data, sourced from Inland Revenue) wherever possible, and supplement this by surveying only the largest and most complex businesses. With this new design, we have eliminated most small and medium-sized businesses from the survey entirely.

The methodology changes improve the quality of the series we publish. This is largely because we effectively have a full coverage of all businesses within an industry, rather than relying on a smaller sample to represent the entire population.

We have also reduced the number of variables being collected for the Economic Survey of Manufacturing. We no longer collect or publish information on manufacturing salaries and wages.

See Methodology changes to manufacturing, wholesale trade, and selected services statistics for more detailed information about the methods used to calculate each of the variables.

Population

The target population is all kind-of-activity units (KAUs) on Statistics NZ’s Business Register (BR) that are operating in New Zealand and are classified to:

• Australian and New Zealand Standard Industrial Classification 2006 (ANZSIC06) Division C – Manufacturing.

Statistical design

The series we publish for this survey are produced using GST data wherever possible. After extensive work on GST data, we established that it is a reliable measure of activity in these industries apart from the largest and most complex businesses.

We supplement the GST data for each series with survey data for large and complex businesses that meet the following criteria:

• a $100 million significance rule – if an enterprise, or group of enterprises linked by ownership, have an annual GST turnover of more than $100 million

• a 3 percent industry dominance rule – if an enterprise makes more than a 3 percent contribution to annual total income for an industry

• all enterprises that have a significant level of activity across multiple industries.

Sales and purchases

We have developed robust methods of transforming the GST data, which is submitted at different frequencies, to a quarterly frequency. In addition, we have developed methods of detecting and removing sales and purchases of large capital items, which can at times occur in the GST data. These are not part of the conceptual measure of sales and purchases required for national accounts purposes.

Where a business reports GST on behalf of other businesses (referred to as GST groups), we apportion GST data between these different businesses using data from Inland Revenue’s employer monthly schedule.

Stocks

Under the new design, we collect stocks data for large and complex businesses. However, no quarterly stocks data is available for other businesses from administrative sources.

We have a range of methods to estimate stocks for businesses that are not surveyed. Different methods are better suited to different conditions depending on the size of the industry and the contribution of the surveyed units. The aim is to use the best method available for each industry.

See Methodology changes to manufacturing, wholesale trade, and selected services statistics for more detail on the GST data assessment and methodology changes.

Non-response imputation

Postal data imputation

Although we attempt to achieve a 100 percent response rate, in practice this does not occur. We estimate values for these non-responding businesses using methods that include:

• historic imputation

• ratio imputation

• mean imputation.

Historic imputation involves multiplying the unit's response in the previous period by a non-response factor. The non-response factor is the average movement over the quarter for similar businesses.

Ratio imputation involves estimating the variable of interest from the unit's administrative data (GST sales), based on the relationship shown by similar businesses.

Mean imputation involves estimating a value for a unit by using the average value for a set of similar businesses.

Tax data imputation

In the administrative data (GST) we have late filers, which are not received in time for publication. We impute the GST data using the historic and mean methods described above.

We also use median imputation for a small number of units, where we take a median response from a unit's previous GST history.

Measurement errors

Model errors

Statistics NZ uses models to standardise the GST reference period to quarterly, which may include model errors. These errors measure the variability that occurs due to a statistical model being applied to produce estimates. It quantifies the cumulative effect of model 'imperfections'.

Other measurement errors

Errors can arise from biases in the patterns of response and non-response, inaccuracies in reporting by respondents, and errors in recording and coding data. The size of these errors is difficult to quantify. We revise data if significant errors are detected in subsequent quarters.

Seasonally adjusted and trend series

For any series, the survey estimates can be broken down into three components: trend, seasonal, and irregular. While seasonally adjusted series have the seasonal component removed, trend series have both the seasonal and irregular components removed. This reveals turning points and the underlying direction of quarterly movement.

We re-estimate seasonally adjusted and trend values quarterly when each new quarter’s data becomes available. Figures are therefore revised, with the largest changes normally occurring in the latest quarters. The seasonally adjusted and trend series are produced using the X-13ARIMA-SEATS package developed by the U.S. Census Bureau.

See seasonal adjustment within Statistics NZ

Seasonally adjusted series

Seasonal adjustment removes the estimated impact of regular seasonal events, such as annual cycles in agricultural production, pre-Christmas shopping, and summer holidays, from statistical series. This makes figures for adjacent periods more comparable.

For the ESM, removing the purchasing monopoly in the dairy industry in mid-2002 caused an abrupt change to seasonal variation in the meat and dairy industry. In response, we changed the calculation method for total sales from direct to indirect (whereby component industries are individually adjusted before being summed). We use both direct and indirect adjustment methods, according to appropriateness.

We use the following methods to seasonally adjust components:

Trend series

Trend estimation removes the estimated impact of regular seasonal events and irregular short-term variation from statistical series. Trend estimates reveal the underlying direction of movement in a series, and are likely to indicate turning points more accurately than are seasonally adjusted estimates.

Standardising dairy industry quarters

Before December 2008, we calculated data for most dairy values on a non-standard quarter. This meant that the June quarter, for example, included dairy values for the months of March, April, and May, while the standard June quarter includes April, May, and June. From the June 2011 quarter onwards, we publish standard quarter data, revising previously published data back to December 2008.

Use in national accounts

A key use of the ESM is in the quarterly gross domestic product (GDP) for calculating manufacturing ‘value added’ (value of output after the cost of input materials and services has been deducted). GDP base-year manufacturing value added is moved forward using volume indexes that we calculate from ESM sales and finished-goods stock changes (deflated by sub-indexes of the producers price index – published as part of Business Price Indexes as of March 2015 quarter).

We supplement ESM volumes with quantity production data for the following industries: meat and dairy product manufacturing

• petroleum and industrial chemical

• manufacturing

• basic metal manufacturing.

The ESM is also used in the expenditure measure of GDP for compiling stock-change values at current and constant prices.