Statistical Office of the Republic of Slovenia (SURS).

Business Statistics Division.

Short-term Business Statistics Section.

Litostrojska cesta 54, 1000 Ljubljana, Slovenia

Industrial production index (IPI). The source is the Monthly Questionnaire on Turnover and Value of Stocks in Industry. The IPI is compiled from data taken from this source.

Statistical classification of economic activities NACE Rev. 2. The national version of NACE used is The Standard Classification of Activities (SKD).

The monthly survey covers enterprises and their establishments classified in Sections B, C and D of NACE Rev. 2, excluding Group 35.3 (Steam and air conditioning supply).

Observation units are selected into the survey with a threshold coverage based on the number of persons employed and yearly turnover. We included all companies and establishments that had at least 18 or 22 employees (18 for the units already included in the survey in the previous year and 22 for new units; source: the Statistical Register of Employment - SRDAP). The limit for the yearly turnover is minimum 420,000 EUR. The stratum is determined on a 2- level NACE Rev. 2. To achieve good representativeness of data, we ensured at least 75% coverage of the number of employees at the level of division (2-level NACE Rev. 2), which means that we also included units with less than 18 employees. Coverage can vary depending on a division.

The industrial turnover index is an important indicator showing the current demand for industrial products and services on the domestic and foreign markets and also on the euro and non-euro markets.

The stocks value index shows the trend of products and services stocks. It indirectly shows the production and sale dynamics.

The data on stocks and turnover enable the calculation of the industrial production index. The purpose of the industrial production index (IPI) is to monthly monitor the trends in industrial production, assessed on the basis of production value. In theory, the IPI shows the movement in the amount of value added in factor costs in the month observed. The aim of the monthly monitoring of the movement of IPI is early discovery of changes in the economic development, therefore, it is appropriate for short-term observation and analyses of the economic development. It is also used for calculation of the quarterly gross domestic product. The basis for the calculation of the IPI is data on the value of production, which is calculated according to the following formula:

PRODUCTION VALUE_{month x} = TURNOVER_{month x} + STOCKS VALUE_{month x} - STOCKS VALUE_{month x-1}

The reporting unit is an enterprise and its establishments.

The observation unit is the KAU.

Industrial enterprises performing one or more activities from the fields of Mining and quarrying (B), Manufacturing (C) and Electricity, gas, steam and air conditioning supply (D) of NACE Rev.2. Group 35.3 Steam and air conditioning supply is excluded. In 2021 approximately 25,050 units were included in the statistical frame (the size of the population).

Variables cover activities on the whole national territory. Turnover from the non-domestic market also covers national enterprises creating turnover performing industrial services outside the national territory.

The present survey dates from 2004. Data from this source for July 2005 were first published in September 2005. Before the present survey IPI was calculated from quantity data (the previous questionnaire). Data series for IPI starts from January 2000.

The base and the reference year is 2015.

All data are in the form of indices.

The reference period is the beginning and the end of a calendar month.

European legal basis: European Business Statistics (EBS) Regulation (EU) 2019/2152 of the European Parliament and Council of 27 November 2019 and the Commission Implementing Regulation 2020/1197 laying down technical specifications and arrangements pursuant to the mentioned EBS Regulation (General Implementing Act).

The former legal basis for the STS indicators is the Council Regulation No 1165/98 of 19 May 1998 concerning short-term statistics and subsequent amending regulations.

National legal basis: National Statistics Act (OJ RS, No. 45/1995 and No. 9/2001) and the current Annual Programme of Statistical Surveys.

Certain data are also transmitted to UN, OECD and IMF.

All data collected and published by the Statistical Office of the Republic of Slovenia are governed by the National Statistical Act (OJ RS No. 45/95 and No. 9/01), Article 50. At the European level confidentiality is governed by the Council Regulation No 223/2009, Chapter V (OJ No L 87/09, p.164).

The lowest level of published data is a 2-digit NACE Rev. 2. Data are confidential dependent on the number of KAUs in one activity or in aggregate and also dependent on the whole population in that activity or aggregate. Data cells for less than 3 enterprises are suppressed.

Preliminary dates are fixed for a period of one year ahead.
An advance notice of release dates for First Release is displayed on SURS' website.

http://www.stat.si/StatWeb/en/releasecal

The data are released on the same date to all interested parties. No user has prior access to the data. Users can be notified of the data release via e-mail if interested.

No internal government access to data before release.

Monthly data on IPI are transmitted to Eurostat on the same day they are published in Slovenia (First Release). The transmission is done by eDAMIS using the SDMX format.

Monthly.

First Release is a monthly publication in Slovenian and in English and accessible on SURS' website http://www.stat.si/StatWeb/en/Field/Index/7/89. Data on industrial production, industrial turnover and the value of stocks are presented in the form of a news release. Seasonally adjusted data for Total industry, Section levels and the Main industrial groupings are presented. They are presented in the form of indices with regard to the different base periods: index for current month compared to the previous month, current month compared to the same month of the previous year, current month compared to the base year 2015, period from January to the current month compared to the same period of the previous year. A time series chart for the last 6 years is published showing the monthly industrial production index in both seasonal and trend-cycle format.

Stat`o`book, electronic publication, annuall (Slovenian, English): Stat’o’book is a publication that presents a statistical overview of Slovenia in the past year, replacing two of our former publications, the Statistical Yearbook of the Republic of Slovenia and Slovenia in Figures. Accessible on SURS' website https://www.stat.si/StatWeb/en/Catalogue/Index. 

Online database SiStat is updated monthly, available in Slovenian and in English and accessible on SURS' website https://pxweb.stat.si/SiStat/en under Industry / Production and turnover in industry. There are unadjusted, seasonally adjusted and working day adjusted data series. We publish unadjusted production indices for Total Industry, NACE Rev. 2 Section and Division levels, the Main industrial groups from 2000 on and by technological intensity from 2010 on. Seasonally adjusted indices and working day adjusted indices are available for Total industry, the NACE Rev. 2 Sections and the Main industrial groups from 2005 on, and for production by technological intensity from 2010 on.

Access to statistically protected microdata is possible only for registered researchers and institutes. The statistically protected microdata are available for researchers in a "secure room" and by "distance access". In the statistical protection of data against direct identification we remove name and surname, address, single ID number of the citizen and the business entity, tax number and similar identifiers. In statistical protection against indirect identification we review some rare features, or the combination of features, and we statistically protect the appropriate data.

Data are also used by other institutes and disseminated in their publications, for instance United Nations, OECD, IMF. Data are also transmitted to Eurostat and published in their news release and database.

The main methodological reference is Eurostat's manual – European business statistics methodological manual for short-term business statistics, 2021 edition. Methodological explanations on the industrial production index are available on SURS' website http://www.stat.si/StatWeb/en/mainnavigation/methods-and-classifications/methodological-explanations, theme Industry, sub-theme Production and turnover in industry, Indices of industrial production, turnover and stocks in industry. In methodological explanations there is a description of the purpose of the statistical survey, legal basis, observation units, coverage, sources and methods of data collection, definitions, explanations, publishing, revisions and other methodological materials.

The quality reports for statistical surveys are available on SURS' website for monthly survey on Turnover and value of stocks in industry (IND-PN/M) http://www.stat.si/StatWeb/en/mainnavigation/methods-and-classifications/quality-reports, theme Industry, sub-theme Production and turnover in industry.

All information about quality assurance is available on http://www.stat.si/StatWeb/en/FundamentalPrinciples/FundPrinc. For IPI we follow this general policy and each year we estimate quality indicators which are published in a quality report.

In the last decade the statistical profession has made great progress towards broader understanding of the quality of statistical data. SURS adopted the Eurostat’s common quality definition. According to this definition the quality of statistical data is composed of the following quality dimensions: relevance, accuracy of estimates, timeliness and punctuality of publication, accessibility and clarity of information, comparability of statistics and coherence of results. More information can be found on http://www.stat.si/StatWeb/en/FundamentalPrinciples/QualityInStat.

All user needs which we detected are met, data are compliant with European legislation, we try to minimize statistical errors, data are published on time, there are no level shifts that would be a result of methodological changes. Data series are comparable between countries. We are also attentive to similar data in other domains and to consistency between different levels of aggregates. Consistency is relatively high. For the IPI indicator we have implemented various actions in order to improve the response rate for the survey. The response rate is relatively high, around 96%. We have also implemented the use of Statistical business register to avoid incorrect registration of an activity in Business register of Slovenia.

National users are Chamber of Commerce and Industry of Slovenia, Bank of Slovenia, Institute of macroeconomic analysis and development, Fiscal Council, various researchers and media. Foreign users are UN, IMF, OECD, Eurostat, credit rating agencies and similar private institutions. According to Council Regulation about STS indicators and national users all the requirements are met.

Even though the formal ways of cooperation with our users are well developed (the Statistical Council, the Statistical Advisory Committees, working groups and direct bilateral meetings with some larger users), the recognition of users’ demands and satisfaction is becoming increasingly difficult. The needs, demands and even the manner of practice vary a lot among the user groups. In order to improve our user registers we constantly improve and supplement our segmentation of users.

The monitoring of the users of SURS is implemented on different stages:
• Analysis of written and telephone requests for statistical data in SURS’s Information center.
• Analysis of the website statistics.
• Conducting user satisfaction surveys.

We are constantly monitoring the satisfaction and needs of the users of our statistical data and services, and improving our products and services based on their opinions and proposals. The level of general user satisfaction with SURS's products and services was high in 2021. Respondents assessed general satisfaction with SURS with the average score of 8.3 (on a scale from 1 – disagree completely to 10 – agree completely). Summary of findings from annual User Satisfaction Survey for 2021 is available on https://www.stat.si/StatWeb/en/News/Index/10342.

All indicators required by European legislation are calculated and published.

Errors can happen due to misinterpretation of definitions written in a questionnaire, incorrect registration of an activity in business register or missing questionnaires (non-response). Statistical errors are relatively small due to several activities such as the establishment of the Statistical business register and verification of data.

Since the survey is based on threshold coverage and the estimates do not have a sampling error, we estimated the bias of one of the statistics - turnover which is a variable for further calculation of IPI. Bias which is the consequence of the threshold coverage was estimated on the basis of data that enterprises send to the Tax Administration for calculation of the value added tax. From these data we can get a quite good approximation for the variable monthly turnover. Bias was estimated for the turnover index for the current month compared to the previous month.

When the data are first published (provisional data), the unweighted response rate is around 96%. At the end (final data), the unweighted response rate is around 98%.

Response rates can only be calculated for the final data. The variable used for the calculation of the weighted response rate is the number of employees which is available when defining the observation units.

All respondents who have not returned the questionnaire receive a friendly reminder on the day of the deadline. Two days after they receive another reminder. Another two days after they are contacted by phone (key-respondents are contacted preferentially). Approximately 60% of units return the questionnaires before the deadline. We also try to reduce non-response by informing reporting units on the purpose of the statistics and by providing the statistical data to reporting units free of charge. All non-responses of a unit are imputed.

In the survey, there is no case of item non-response as it is already checked in the stage of organizing the data whether the questionnaires are filled out completely. If this is not the case, the unit concerned is contacted by telephone and the missing data are obtained from it. The stage of statistical processing is thus only entered by the data for units where all the required data are reported.

Over-coverage rate is computed for each month. The average over-coverage rate in 2021 was 2.2%. We always try to detect the real activity of an enterprise since sometimes enterprises are incorrectly registered.

Editing rate for a certain variable expresses the ratio between the number of reporting units where we, in the phase of arranging data, corrected the original data and the number of all reporting units which completed the questionnaire. In the beginning of the year when new reporting units are included in the sample, there are more checks of data and more explanations for the reporting units. In 2021, the average editing rate for domestic turnover was 4.5%, for non-domestic turnover 5.5% and for stock value it was 6.9%.

The industrial production index (IPI) is published and sent to Eurostat not later than 1 month and 10 days after the end of the reference month.

Web questionnaries for a reference month (i.e. January) are available in the first working days of the following month (i.e. February). The deadline for completing the questionnaire is the 20th of the following month (i.e. February), but later submissions are acceptable (the web questionnaire remains open for the past three months).

Final data are released once a year for the entire previous year.

All releases are delivered as scheduled.

The statistics are in compliance with the STS requirements in the EBS Regulation and the EBS methodological guidelines. This ensures a good comparability between national data and good-quality European aggregates.

Industrial production indices - time series of indices calculated according to the methodology described - are available from January 2000 until the current month. In mid-2005 the methodology for the industrial production index was changed. Before this period industrial production was calculated with quantitative data and from 2005 on with value data. All time series have been recalculated and there was no break in series due to the change of methodology. From 2009 on data are published according to the 2008 Standard Classification of Activities and all time series have been recalculated for the period before 2009. When the base year changes, we recalculate all data series. In cases when methodological changes could lead to breaks in series, we recalculate the series in a way that comparability over time is assured.

A relevant comparison for the IND-PN/M survey is the survey on Business Tendency in Manufacturing (PT). PT survey is a monthly survey and is carried out on a sample of enterprises which are classified in Manufacturing (C). Questions are qualitative and are answered by managers. We compared the industrial production index (IPI) from the IND-PN/M survey with results/answers to the question: Our production in the last month compared to the previous month was: a) lively, b) equal, c) weak. Because results are shares of answers for a certain option, we calculate the balance which is the difference between the share of negative and positive answers. Therefore, for comparison of the data only graphical comparison is shown.

Surveys differ in methodology since IND-PN/M is a quantitative survey for monitoring variables in values and PT is a qualitative survey for monitoring managers` opinions about individual economic variables. Deviation is partially due to the difference in samples since PT monitors opinions only in enterprises which are classified in manufacturing. For this reason the comparison between the two surveys is only for manufacturing. Results from PT have a one month time lag because managers` opinions in month x actually refer to production in month x-1. We compared monthly chained IPI with balance for the question on how production was in the last month compared to the previous month. Correlation between the two series of results for the period January 2017-December 2021 was 0.65.

A direct comparison between IND-PN/M and PT is not possible but from the chart and correlation we can conclude there is good accordance of observed variables. Despite different methodology, both selected surveys give similar and acceptable results.

Chart 1: Industrial production indices (IND-PN/M), previous month=100, and production balances (PT), original data, Slovenia, January 2017–December 2021

Statistical outputs within the data sets are overall consistent. Sub-aggregates are mostly consistent with main aggregates. There can be small inconsistency between aggregates when seasonal adjustment is performed. During the whole process of calculation we try to provide for the highest level of internal coherence - during aggregation and seasonal adjustment.

In 2021 SURS spent 5,688 hours working on the survey Turnover and Value of Stocks in Industry. Approximately half of these working hours were spent on industrial production indices. There were 1,993 response units in the sample. The survey is monthly.

There were 1,877 response units on average that completed the questionnaire each month. Considering one response unit has to fulfil 12 questionnaires a year and spends 23 minutes on average per questionnaire, the total time spent was 8,635 hours. With the questionnaire we collect data on turnover and stocks in industry and IPI is derived from these data. Burden for response units is calculated for the entire survey.

Types of data revisions in relation to planning:

a) Planned data revision

Planned data revision is subject to the following reasons:
I.  Due to the needs of users for timely information, data are published that meet the criteria of the quality of official statistical data, but do not meet the quality that can be met with additional statistical procedures. Final data are based on more complete answers about the phenomenon and/or analyses and are published later on;
II. Seasonal adjustment and/or elimination of calendar effects;
III. Change in methodology and classifications.

b) Unplanned data revision: Unplanned data revision is not part of the regular statistical process. It appears due to unpredictable changes in the methodology, unpredictable emergence of new and better data, unpredictable changes regarding reporting units that transmit their data to the Office, unpredictable obstacles in data processing and publishing, and errors in data processing and publishing (e.g. a key unit corrects its data for the past few months, an unpredictable change in the administrative data source).

Types of data revisions in relation to time of implementation:

a) Regular revisions: inclusion of a more complete/additional data source or a change in the data source, seasonal adjustment and/or elimination of calendar effects;

b) Occasional revisions are a consequence of including a new/more complete/additional data source that becomes the standard in later data releases or a consequence of an unpredictable obstacle in data processing and publishing, and change in methodology.

Types of data revisions in relation to the purpose:
a) Inclusion of a more complete/additional data source or a change in the data source;
b) Seasonal adjustment and/or elimination of calendar effects;
c) Transition to a new base period;
d) Improvement of methodology due to a change in the statistical method or a change in classifications, concepts and definitions;
e) Elimination of errors.

Vintage data are available in our database in archives and documents about revisions are also kept. General information about revision is published in methodological explanations.

There is no revision calendar since except for January data revision is due to inclusion of additional or revised data that reporting units send us.

The same revision policy is applied to data released nationally and transmitted to our users (including Eurostat).

More information on SURS'website: 

https://www.stat.si/StatWeb/File/DocSysFile/5299/General_ME_Revision_of_statistical_data.pdf

Changes and revisions of methodology are announced in advance in our publications. Before that the Advisory Committee for Industry discusses the proposed methodological changes.
If a revision is necessary due to an error, a new edition with explanatory notes is published as soon as possible.

Before implementing any changes we always use benchmarking to decide whether backcasting is necessary or not.

Regular revisions of the IPI are due to inclusion of better and more complete data (because of unit non-response). Regular revisions of provisional data are on a monthly basis and are quite small and on average have no impact on the highest aggregates. When data for June are published (in August), data for the previous year become final. When January data are published, there is also recalculation of indices for the previous year (data for the previous year are still provisional) due to the new available data on weights (value added from SBS) and the review of data series in JDemetra + (the tool for seasonal adjustment).

Occasional revisions are due to methodological changes. If possible and if the revision has a great impact on the data series, we recalculate the whole series (for instance when NACE Rev. 2 was implemented). In 2018 there was a change in the base year. In 2013 in the process of deflation of non-domestic turnover, euro-zone turnover and non-euro-zone turnover we replaced deflation using producer price indices of manufactured goods on domestic market with deflation using producer price indices of manufactured goods on non-domestic market. We took the new deflator into account for indices from January 2010 on. Consequently, time series for non-domestic turnover, euro-zone turnover, non-euro-zone turnover and industrial production indices for some activities and aggregates could be slightly changed from January 2010 on. JDemetra + was implemented at the beginning of 2016.

The same revision policy is applied to data released nationally and transmitted to our users (including Eurostat). Whether data is final or provisional, we inform our users.

Coherence between Provisional and Final Data

In the first half of the year the data for the current year and for the previous year are provisional, other data are final. Data for the previous year become final with the release of data for June while data for the current year remain provisional. With each release provisional data can be replaced with new, revised data. Revision of data is planned. Regular revisions include better data sources, including weights and changes of data due to elimination of impacts of the season and calendar. Data are revised because new data can significantly contribute to the quality of decision making. Since data releases are in accordance with European legislation and therefore very short, data can be completed.

We show coherence for the industrial production index (IPI) for 3 various aggregates, from highest to lowest. Data refer to 2021 since for 2022 data are still provisional.

Mean Absolute Revision (MAR) and Mean Revision (MR)

The values for the IPI index are:

• Growth rates for original data series (YoY):

MAR = 1.4732, MR = 0.3567

• Growth rates for calendar/working day adjusted (YoY):

MAR = 1.6964, MR = 0.3746

The source is a statistical survey. Frame on which the source is based, is the Statistical Business Register of Slovenia. Census above a cut-off is used and a stratum is determined according to the 2-digit NACE Rev. 2. We use threshold values of employees (18 for the units already included in the survey in the previous year and 22 for new units) and annual turnover (minimum 420,000 EUR). Updating of the sample is once a year, at the beginning of the year. In 2021 1,993 observation units were included in the sample.

Monthly.

One standard questionnaire - Monthly Questionnaire on Turnover and Value of Stocks in Industry (IND-PN/M) is used. Before May 2013 data were collected only with postal questionnaires, from May 2013 by postal and electronic questionnaires and from 2020 only electronic questionnaires remain available.

We monitor unit non-response each month and have various methods for decreasing it starting with important reporting units. All respondents who have not returned the questionnaire receive a friendly reminder on the day of the deadline. Two days after they receive a reminder. Another two days after they are contacted by phone (key-respondents are contacted preferentially). We also try to reduce non-response by informing reporting units on the purpose of the statistics and by providing the statistical data to reporting units free of charge.

Data are checked at the micro level with various error detection (intra-dataset checks). All potential errors, light and heavy, need to be checked in telephone contact with a response unit. Data are validated on a macro level as well. Once a year we also compare data with the data received via Annual report on industry. Inconsistencies need to be solved. We try to follow similar data in other statistical domains as well. The files which are transmitted to Eurostat are validated using SDMX Converter.

The non-response is treated on a micro level. In case of non-response, units are contacted by phone and asked to provide missing figures. Data for units with non-response are imputed. The historical value and the mean value methods are used. Since 2020, part of the data in cases of unit non-response is imputed with the administrative data from value added tax (VAT) declarations. Because there is no random sampling and non-response is treated by imputation procedure, there is no need for grossing-up. The production index is calculated according to the Laspeyres formula. Weights are used (which are shares of value added) when aggregating the data and at calculation of IPI for the higher levels of classification. Shares are calculated twice a year and are based on the data on value added as soon as the latest data is available (usually T-1). We use chain-linking, the one-month overlap chained Laspeyres indices. Weight source is structural business statistics.

Calendar adjustment:

Country-specific (national holidays) regressor used, adjustment for moving holiday effect - Easter effect, adjustment for Leap year, calendar adjustment is performed for all data series, regARIMA calendar adjustment is used.

Other pre-adjustment:

We detect and treat additive outliers (AO), temporary changes (TC) and level shifts (LS).

Outliers are a consequence of economic changes, collective holiday (greater effect than anticipated), extraordinary events in enterprises, etc. 

Most outliers occur during holiday periods when enterprises have collective leave days (usually August and December). All deviations are checked.

Seasonal adjustment:

We use software JDemetra+ 2.2.0, TRAMO/SEATS method.

Model/filter selection is manual but some automatic tests are used for help (test for transformation, automatic detection of outliers, automatic selection of ARIMA model).

The models and the respective parameters are set annually (mainly in the beginning of the year) but re-estimated monthly for each new release of unadjusted data.

Seasonally adjusted time series are revised dependent on unadjusted data. For each new release of unadjusted data, the whole seasonally adjusted time series are revised. 

10 time series have multiplicative decomposition and 2 time series have additive decomposition. Data series for total industrial production, capital goods, durable and non-durable consumer goods have multiplicative seasonal adjustment decomposition while data series for intermediate goods and energy have additive seasonal adjustment decomposition.

The model used is ARIMA; critical value for outlier detection depends on the time series, usually it is between 2.5 and 3.5. Most of the time series do not have the possibility of outlier detection. Filter length is automatically chosen and there are no seasonal breaks in the time series.

All the time series are seasonally adjusted directly, so seasonally adjusted data of an aggregate are not composed of seasonally adjusted data of its components. When models are selected or changed, connection between an aggregate and its components is taken into account (similar time sereies have similar models...).

Residual seasonality is checked when the model is selected. Afterwards, residual seasonality diagnostics are taken into account.

Aggregated time series are seasonally adjusted and no further aggregation is needed. Since direct adjustment is used, there is different ratio between seasonally adjusted time series when breaking series to the different levels than between unadjusted time series (consistency differs).

No comment.