Statistical processing

Source data

The main sources for compiling supply and use tables in current prices are the following:

• Accounts Statistics for Non-Agricultural Private Sector (SBS)
• Sales by manufacturing industries (PRODCOM)
• General enterprise statistics
• Manufacturers' Purchases of Goods and Services
• Accounts for public corporations
• Economic Accounts for Agriculture
• Government finance statistics
• Financial industries and FISIM compiled on a NA basis
• Accounts statistics for non-profit institutions serving households (NPISH)
• Energy accounts
• Balance of payments and external trade statistics
• Household budget survey

The main sources for the compilation of supply and use tables in previous year prices are:

• Producer price indices for goods and services
• Producer price indices for construction
• Construction cost indices
• Consumer price indices and net price indices (consumer price indices excluding taxes)
• Price indices for agricultural sales
• Wage indices
• Unit price indices from import and export statistics
• Volume indicators: Energy supply and use, transportation, final Individual consumption expenditure of government

The balancing process of the SUTs serves as a central control of the quality and liability of the data. The compilation of SUTs is not solely to provide a reliable base for IOT, but also to provide a confrontation between estimates from different sources at a very detailed level. See also 18.4 data validation.

The input-output table is almost entirely based on the SUT's. The only exception is the the elements of Gross Value Added; Other taxes on production, Compensation of employees and Gross surplus and mixed income that comes from other sources in the national accounts.

Frequency of data collection

The supply- and use tables play a central role in estimating the National Accounts Main Aggregates. Supply and use tables are therefore compiled every year in full detail like every year is a benchmark year. This implies that source data are collected and processed every year in “level” not using extrapolations. The Danish supply and use tables released with a delay of 2.5 years, i.e. year t-3.

Preliminary annual supply and use tables are calculated every year for year t-2 at a lower level of detail and to a large extent based on extrapolations and structures from the final year SUT.

The input-output tables are prepared once a year and disseminated at the same time as the final annual national accounts, and data for this process is collected as soon as the supply and use tables are available in a final version.

Data collection

There is no specific collection of primary data to build the supply and use tables. Data is collected in other parts of Statistics Denmark and delivered to the national accounts in datasets that have already been debugged and prepared.

Data for the input-output tables are available as prepared datasets as well.

Data validation

The initial data validation are formal checks including format of the files, internal consistency of the data and checking for any changes in the content of data. Then follows the main steps in the data validation process, which are the following:

• When targets (at the industry level) for total output, input, GFCF etc. are compiled a number of problems are revealed, investigated further and corrected if necessary.

• When the first version of the – unbalanced – SUT file is put together it is necessary to compare the initial estimates of supply for certain product groups with estimated use. A closer study often reveals errors in primary statistics or in the internal compilation process.

• Problems will usually be solved in cooperation with the relevant statistical sources domains. The inquiries are of considerable importance for the quality of the SUT and hence the national accounts.

• In recent years many of the problems have been related to globalization. A “large cases unit” (LCU), established in 2015, monitors the large multinational corporations and their complex business models (profiling) and confront data reported by these units before they enter the national accounts.

Generally, it can be said that balancing of supply and use tables as a mean of data validation, provides a systematic confrontation of data that captures a lot of inconsistencies. One of the strongest cross-checks is comparing information on supply and demand at the product level (or groups of products).

The complete integration of the input-output tables with the final national accounts and supply and use tables means that the data validation of the input-output tables primarily consists of ensuring that all data found in both statistics are the same.

Upon receipt of Danish supply and use tables as well as input-output tables, Eurostat performs a further series of validity and consistency checks. The error messages and warnings that come back from Eurostat can be divided into three categories:

• Explainable deviations in Danish data from Eurostat's normal. These can be, for example, negative values ​​in certain investment categories. This occurs in Danish data when, for example, investment goods (transport equipment) or large consumer goods (e.g. yachts) are sold (disinvested). These deviations are explained to Eurostat and entered on a list of repeated deviations in the Danish data so that Eurostat does not have to notice them again the following year.

• Errors that occurred in connection with the transformation of Danish data into the format in which data must be reported, including the attachment of all necessary metadata. These errors are corrected as soon as information about them is received and the data is resent to Eurostat.

• Actual errors in supply and use tables and/or input-output tables. These errors are extremely rare, as they would normally have been discovered already in Statistics Denmark's work on compiling and publishing these tables. The few errors that have occurred have been quite small and have already been known to us. In the Danish national accounts, errors in already published data are only corrected in connection with the 5-yearly benchmark revisions, and it is therefore normally difficult to meet Eurostat's desire to correct such errors before the next benchmark revision.

Data compilation

The number of industries used in the compilation of the supply and use tables is 117 and the number of products are approximately 2 350 for the final year (t-3). The number of groups of household consumption is 83. There is 10 different COICOP based categories of NPISH consumption, and 32 COICOP based groups of individual government consumption. The 10 groups of government collective consumption is based on COFOG. (See also S.3.2). Finally, investment is split in 13 different categories, like e.g. Buildings, Machinery, R&D etc. For the first preliminary year (t-2) a more aggregated SUT table with 81 industries, 72 consumption groups and around 350 products is compiled.

Current prices For the final year initial and independent estimates for GDP by the production approach (GDP(P)) and expenditure approach (GDP(E)) at the level of target totals are compiled as described below:

• GDP(P): After initial processing of the various primary statistics (accounting statistics) data are transferred to a common accounting plan and national accounts industries in the so called Intermediate System. Then relevant conceptual adjustments to arrive at national accounts concepts for production and intermediate consumption are made.

• GDP(E): The largest share of expenditure-based GDP is calculated using a direct estimate for the various expenditure components. The most important exceptions are household final consumption expenditure, for example hotel and restaurant services and dwelling services.

Then the SUT is compiled in the following steps:

1. Target totals and values at the product level that can be entered directly into the system as predetermined are collected. Predetermined values are e.g. everything related to energy. The energy part of the supply and use tables is being compiled in an independent system of Energy Accounts.

2. An initial version of the SUT is compiled in three steps:

3. Firstly, the supply matrix, with supply of products by supplying industry is established. This matrix is normally fixed from this point and not touched any further in the supply and use table compilation process.

4. Secondly, an initial version of the use table is compiled as the use table from the previous year inflated to the current years price level. Then the new target totals for intermediate consumption by industry, as well as target totals for final use by COICOP/COFOG groups and investment categories, are used to update the initial use matrix using a "vertical" proportional distribution.

5. Thirdly, a “horizontal distribution”. The intermediate and final use of products is updated through a proportional horizontal distribution of the total supply by product. The supply and use of products balance. After this step the sum of columns will usually differ from the targets on the use side. Total trade and transport margins and total VAT may also differ from their respective targets.

6. Then follows a step, where the initial version of the SUT is adjusted manually. The unsolved problems are examined closely. In many cases the problems reveal errors in the calculations and primary data. Adjustments and corrections to the initial balances are entered into the system to create a new - but not yet final - version. This step is referred to as "Manual balancing"

7. In the last step the differences between totals and targets are removed except where such differences are deemed acceptable. In this step trade and transport margins and VAT are finally adjusted to their targets. This step is referred to as "Final balancing" and is more or less an automated process.

As a final step, exports by product is split between what is domestically produced and what is imported (re-exports). The split is based on how the split was made the previous year. The original splitting that goes a number of years back in time was based research done by the staff of the national accounts. Recently, the size and distribution of re-exports has been confronted with data for re-exports compiled by the department of foreign trade statistics at Statistics Denmark. The finding was that the numbers were surprisingly close.

Previous years prices

All products in the current price SUT enter into the deflation process and every product is assigned a price index at basic prices (e.g. production, imports, exports, consumption).

Price indices from the supply side, i.e. from production and imports is the primary source. A few exceptions exists, including energy where volume indicators is used directly.

Domestic use at basic prices is deflated with the implicit price index of the residual between total supply and exports which makes sure that each product is balanced in basic prices. Government individual consumption expenditure is deflated from the output side using volume indicators.

Output and input are deflated independently so there is no need for special indices for GVA – it is calculated in this double deflation procedure.

Input-output-tables

The Danish input-output tables are based almost entirely on the supply and use tables. The compilation grossly follows the guidelines in the manuals Eurostat and UN. The transformation of supply and use tables into input-output tables requires a couple of rather strong assumptions about how the products flow in the economy. So one of the most important aspects for the compilation is the choice of method for this transformation. There are generally four different methods laid out in the manuals mentioned covering various assumptions. Statistics Denmark produces input-output tables based on method D. This method build on the assumption that "each product has its own fixed sales structure, regardless of which industry produces it". This means that if industry 1 produces 80 per cent of a product and industry 2 produces the remaining 20 per cent, then all users of this product will get 80 per cent from industry 1 and 20 per cent from industry 2. Obviously, this does not hold for all products all the time in reality, but we consider it the method that tweaks the supply and use data the least. Method D leads to an industry-by-industry input-output table.

1. First step is to split the basic price supply table in a domestically produced part and an imported part. Statistics Denmark does not compile a real matrix of imports by intermediate or final use categories. What we have is a vector of imports by products (approximately 1,950 products) and a similar vector of custom duties on the imported products. Therefore, we need to create a vector of import coefficients that tells us how large a share of the supply of each product is imported. Because exports is already split in domestically produced and imported in the supply and use tables we only need to consider imports for domestic intermediate and final use. Thus, the formula looks like this

import_coefficients = (imports – re-exports) / ((domestic output-exports) + (imports – re-exports))

In a matrix formulation, the use matrix excluding exports should be premultiplied with this vector put on the diagonal of a matrix of zeros in order to get the import matrix. The domestically produced part is off course the remaining part of the initial use matrix.

1. Second step is the transformation of the two new use tables from a productuse category to an industryuse category table using method D. In order to do that we must calculate a market-share matrix based on the supply matrix. Such a product*industry market share matrix gives for every product the share that was produced in each of the 117 industries. For all the products that we do not produce in Denmark the matrix initially just have zeros in the rows. In order to cope with that we keep and update a list of all the products in the system and their “characteristic industry”. This means the industry where the product most likely would have been produced. On the list is bananas which we do not produce in Denmark, and its characteristic industry is Agriculture. With the market-share matrix, we transform the domestic use table as well as the imported use table from product to industries. Thus, for the imported products that we also produce in Denmark, we assume that they are produced by foreign industries in the same relative way as they are in Denmark. This means that if industry 1 in Denmark produces 80 per cent of a product and industry 2 in Denmark produces the remaining 20 per cent, then we assume that 80 per cent our imports of this product comes from the foreign industry 1 and 20 per cent from the foreign industry 2. In case we do not also produce an imported product in Denmark, we assume it is imported entirely from its characteristic industry.

2. Thirdly, we must implement the other price layers between basic and purchasers prices in the table. The column sums in an input-output table are in purchasers prices so in addition to all the cells in basic prices we must add information about margins and taxes used as well. In the Danish supply and use table system, there are information for each product on basic prices, wholesale and retail margins, product taxes, VAT and purchasers prices. In the Danish supply and use tables there are no transport margins. We considered them a part of wholesale trade margins. Therefore, this text only refers to trade margins. We then compile matrices of margins, product taxes and VAT in more or less the same way as the basic price matrices.

The matrix of trade margins is summed and placed into these three industries:
- G45 (450000) produces wholesale as well as retail trade margins selling cars and related items - G46 (460000) produces wholesale trade margins - G47 (470000) produces retail trade margins

Wholesale trade margins obviously goes into G46 and retail trade goes into G47. We determine the part of margins that goes into G45 in a calculation that takes some margins from G46 and some from G47 based on some investigations that was made years ago into what products that are actually fully or partly sold by enterprises that resides in industry G45.

The matrices of product taxes, net and VAT are also summed and placed at the bottom of the input-output table.

In order to produce input-output tables in previous years prices the exact same procedure is carried out based on the supply and use tables in previous years prices.

As an example of the use of input-output tables, Statistics Denmark publishes calculations of import-adjusted growth contributions using the input-output method. In this method, input-output tables and the input-output model are used to calculate how much import the individual components of final demand have given rise to, both directly and indirectly. These imports are then deducted from the individual components of final use, so that the Danish-produced part of final demand remains. Based on these components, which together constitute Danish GDP, it can finally be calculated how large a contribution to GDP growth comes from each individual component of final demand.

Adjustment

All relevant adjustments are described in the sections above.