Reference metadata describe statistical concepts and methodologies used for the collection and generation of data. They provide information on data quality and, since they are strongly content-oriented, assist users in interpreting the data. Reference metadata, unlike structural metadata, can be decoupled from the data.
Industrial production index (Indice della produzione industriale).
The survey describes the development of productive activity in Italy through the collection of production volumes.
The 2024 indices are calculated using the chaining method introduced by the publication of the indices relating to the month of January 2022.
Starting from the publication of the January 2024 data, the industrial production indices are released in the new reference base 2021=100. The calculation base has also been updated and now relates to 2023.
The innovations introduced concern not only the change in the reference base but also the sources used, the updating of the sample of enterprises used in the survey, the revision of the weighting system, the updating of the basket of goods.
Starting from this edition, for some groups of economic activity, the survey on industrial production uses information deriving from the survey on industrial turnover, appropriately deflated.
Hence the 2024 indices are calculated using the chaining method - the weighting reference is year 2023 - according to the Laspeyres formula using a structure of weights updated. annually.
3.2. Classification system
NACE Rev.2.
3.3. Coverage - sector
Product coverage: Industrial goods classified in Sections B, C and D.
Link period 2023: indices for Year 2024
Activities covered: sections B, C and D except for Divisions 05, 07 and 09
Out of a total of 229 classes in Section C, IPI detects 194: they correspond to 97.9% of the 2021 added value (of which 16 classes/17 groups are detected as seen through the turnover volume indices: they correspond to 4.2%).
Nine classes under observation for a weight of 0.8%.
Starting from the calculation base 23, 4 classes were excluded from the detection (2 index, 2 under observation) because of low weight.
The classes not detected are 26 in total: they are considered residual and weigh 1.3% of the added value
Distribution of classes of section C in the IPI survey based on calculation 2023
Typology
Number
% turnover volume 21
In Index
194
97,9
of which detected as seen through the turnover volume
16 classi/ 17 gruppi
4,2
Observation
9
0,8
Not detected
26
1,3
Total
229
100,0
With the transition to the 2023 calculation base, the previous basket of 612 macro items has changed, also following innovations on the sources used. Against 58 eliminated macro itemes, 40 were inserted: the number of elementary macro items of the new base is therefore equal to 594. Almost the whole of 598 elementary macro-products for which index numbers are calculated have a coverage of about 65% in terms of value of production in each Class (4-digit level) (more than 80%). Size classes covered: Units employing 20 employees or more. In some activities, lower thresholds apply.
3.4. Statistical concepts and definitions
List and definition of variables: The index shows the evolution of gross production in volume terms.
For the Link period 2023, indices for Year 2024
Starting from the 2023 base, the survey of industrial production is carried out by a panel of companies that provide information on just under 1,196 products (of which 105 car models) on a monthly basis. These products are grouped into homogeneous groups to form a basket of goods, which represents the elementary level of aggregation (macroprodotti/macro items).
Information is collected on more than 9,500 basic observations for about 5,800 Respondent units product sand .5,400 Enterprises
Accounting conventions: The target variable is measured as a flow during the calendar month. Products, which need more than a month for the production, are recorded in value or in worked hours only for the part accrued to the reference period. Production in value is deflated by output prices. Hours worked are adjusted with productivity coefficients based on short-term indicators.
With the update of the weighting structure, the incidence of the proxies used to measure the trend of industry output have changed compared to the previous base of 2022. As can be imagined, increases the share –expressed in terms of the relative weight on the general index – of products recorded in value to the detriment of the other two measures, due to the use for some groups of economic activity of industrial turnover, appropriately deflated. The incidence of products recorded in quantity remains predominant, however decreasing compared to the previous base (72.7% compared to 76.3%), followed by products recorded in value of production (15.1%), whose share instead, it grows by almost 5 percentage points compared to the 2022 base. Finally, the incidence of products in hours worked is 12.2%.
TABLE. Proxies used to measure products. Weighting structure of fixed base 2015 and link period 2021, 2022 and 2023
TYPE OF PROXY
Fixed base2015
Link period 2021
Link period 2022
Link period 2023
Physical quantities (a)
76.4
76.6
76.3
72.7
Hours worked
13.1
13.6
13.5
12.2
Value of production
10.5
9.8
10.1
15.1
TOTALE
100.0
100.0
100.0
100.0
Planned changes in information collected: None.
3.5. Statistical unit
Reporting unit: KAU or local unit KAU.
Observation unit(s): Unit of homogeneous production (UHP) or Local unit of homogeneous production (LUHP).
3.6. Statistical population
Target population corresponds to enterprises with 20 or more employees in 2022 for sections covered by this survey. Total number is about 31,200 enterprises.
3.7. Reference area
Whole national territory.
3.8. Coverage - Time
The Industrial production survey was established in 1928. Series of indices with base 2021=100 start from January 1990.
3.9. Base period
The base year is 2021
Index
Month
6.1. Institutional Mandate - legal acts and other agreements
Legal basis
National level: This is one of the compulsory surveys fixed every year by a Decree of the President of the Republic (D.P.R.).
European level:
The legal basis for the STS indicators are the Regulation (EU) 2019/2152 of the European Parliament and of the Council on European business statistics, repealing 10 legal acts in the field of business statistics (EBS-Regulation) and the Commission Implementing Regulation 2020/1197 laying down technical specifications and arrangements pursuant to Regulation (EU) 2019/2152 (General Implementing Act).
Obligation on units to provide data: Surveys are mandatory. Penalties are foreseen for non-response.
Planned changes in legal basis, obligation to respond and frame used: None.
6.2. Institutional Mandate - data sharing
Data are sent only to Eurostat (not to other international organizations).
7.1. Confidentiality - policy
Several national legal acts guarantee the confidentiality of data requested for statistical purposes. According to art. 9, paragraph 1 of the Legislative Decree n. 322 of 1989, personal data can only be disseminated in aggregate form, in order to make it impossible to identify individuals. The Personal Data Protection Code (introduced by the Legislative Decree no. 196 of 30 June 2003) provides special rules concerning the processing of personal data for statistical purposes within the National Statistical System (Sistan). Further, in order to make statistical secrecy and protection of personal data effective, Istat is currently taking appropriate organizational, logistical, methodological and statistical measures in accordance with internationally established standards.
Links to relevant legal acts on statistics are also listed on the website of Sistan.
7.2. Confidentiality - data treatment
Survey elementary data are treated as confidential. Aggregated data are disseminated on Istat's website.
Some variables can be transmitted to Eurostat as confidential when they are not disseminated in Italy (8 of 341series are confidential as well as the 594 elementary macro-product index. All non confidential data, from Nace Rev.2 four digit up to General Index, are available on IstatData.
8.1. Release calendar
Every year, in December Istat publishes a release calendar providing the precise release dates for the coming year. During the first part of the year the required one-quarter-ahead approximate release dates are disseminated on the Internet on the "IMF's Dissemination Standards Bulletin Board".
Data are released simultaneously to all interested parties by issuing the press release "Produzione industriale" which is provided, on an embargo basis, to the media 30 minutes prior to general release. The press release (Italian version and English version) is subsequently posted on the Istat's website. Data are transmitted to Eurostat by tele transmission no later than 40 days from the end of the reference month. If the National release is scheduled later than the Eurostat transmission, time data are sent to Eurostat under embargo. Data are sent in xml file according SDMX standard and disseminated as national data in the Eurostat's database (Eurobase, theme Industry, trade and services, sub-theme Short-term business statistics).
Monthly
10.1. Dissemination format - News release
A monthly press release "Produzione industriale" where figures and tables are disseminated with the data to facilitate the analysis. Analysis of current-period developments is included with dissemination. The Italian version of the Industrial production press release includes also a data file, the weighting structure and a methodological note.
The monthly industrial production indices are available at IstatData, the Istat data warehouse. The on-line database is open and free, statistics are organised by theme in a two-level hierarchical tree. Data are presented in aggregate form in multidimensional tables; acting on variables, reference periods and the arrangement of heads and sides users can create custom tables and graphs. The system can be searched by keyword. A wide range of standard metadata facilitates the retrieval and understanding of statistics by users.
Before data dissemination, validated microdata are stored in the repository of the Institute, ARMIDA. The repository ARMIDA (Archive of validated microdata) was established with the main objective of preserving and documenting the data produced by Istat surveys, and has subsequently supported the objective of disseminating data. The data archived in ARMIDA supplies, in fact, the different channels for the dissemination of microdata (for internal use at the Institute through the "Memorandum of access to microdata of ARMIDA, internal users ", for authorities in Sistan, for research files, for standard files, etc.).
A short description of the methodology and characteristics of the survey is contained in the technical explanatory note attached to the press release updated regularly. Generally, major changes in methodology are announced a short period in advance in the information note publication.
A description of methodology and sources used in preparing statistics is also contained in the Information System for Survey Documentation and Quality Control (Siqual); it is available on Istat's website. This system manages qualitative information related to each single phase of a survey's production process (SIQual). More information about procedures applied for quality management and quality assessment in the document "Quality Guidelines For Statistical Processes" - December 2012
11.1. Quality assurance
Essential points of Istat quality policy are: process quality, product quality, documentation, respect for respondents, strengthening of statistical literacy, users' orientation.
11.2. Quality management - assessment
As regards main issues of quality, index is compliant with a good standard. Eurostat's requirements in terms of time series are fulfilled, timeliness is respected, cross-domain checks are used for monitoring a good level of coherence. Some information on data quality are available in SIQual.
12.1. Relevance - User Needs
Main users of industrial production index are:
Institutions: government departments, ECB, national central bank, international organisations, etc.
Social actors: employers' organisations, trade unions, lobby groups, etc.
Media
Researchers & students
Enterprises/businesses
Internal users
Eurostat
12.2. Relevance - User Satisfaction
Not available.
12.3. Completeness
Eurostat's requirements in terms of time series are fulfilled.
13.1. Accuracy - overall
The samples are designed with the cut-off method and therefore sampling errors cannot be measured. Non-sampling errors are reduced in the collection phase both using an electronic questionnaire that helps debugging and aiming at an increase of the response rate of enterprises.
13.2. Sampling error
It is not possible to compute sampling errors because a non-probability sampling method (cut-off) is used.
13.3. Non-sampling error
When the first provisional data are released, the weighted overall response rate for the period 2021-2024 was 95,8%; while for only year 2024 was 95.1%. For revised data the weighted overall response rate for the period 2021-24 - was 98.0%; while for only year 2024 was 98.1%. To speed up and increase the rate of response reminders and intensive follow ups by phone are addressed to the MIB (Main Industrial Business) units. These units are the most significant (in terms of production realized in the year before that of reference) for the specific products concerned.
14.1. Timeliness
The official release is published on average 40 days after the end of the reference period; data are transmitted to EUROSTAT within 40 days, using, when needed, an embargo clause; data are provisional at the time of the Press Release in which they are reported.Until 2021 these revisions typically occurred when releasing the indices concerning the month of February and incorporated both the new estimates of the short-term statistics for the three previous years and the corrections based on the late responses received and on the corrections to already received information. Starting from year 2022 retrospective revisions concern the previous year only.
14.2. Punctuality
All releases delivered according to the timetable.
15.1. Comparability - geographical
European definitions are used and they apply in the whole national territory
15.2. Comparability - over time
Consistent time series data are available for a long period (since January 1990). When changes in source data, methodology and statistical techniques are introduced, historical series are reconstructed as far back as possible.
15.3. Coherence - cross domain
Comparisons are made with the evolution of turnover. Results for these comparisons show good level of coherence for main aggregates.
15.4. Coherence - internal
Raw data aggregates are always consistent with their sub-aggregates but for seasonally adjusted and calendar adjusted data internal coherence is not ensured as seasonally adjustment direct method is used.
The cost estimate is compiled in terms of total number of hours worked (THW). It is equal to the product of three terms: (total) Annual Working Days (AWD), (total) Hours Worked per Day (HWD), Number of Persons involved in the Survey (NPS).
AWD
HWD
NPS
Cost estimate (THW)
220
7.12
10.5
16,447
The burden estimate is compiled in terms of total number ofminutes (TNM) taken to fill in the (web) questionnaire. The burden is applied to respondent, i.e. the person who fills in the questionnaire on behalf of the enterprise. To estimate it, two elements have been taken into account. The former distinguishes between beginner and expert respondent, to point out respectively the person who fills in the web questionnaire for the first time and the one who is familiar with filling in the form. The latter, closely concerns the information that has to be monthly provided. On average, by combining these two elements, the time spent to fill in the questionnaire (just the burden) is estimated in terms of total number of minutes between a minimum and a maximum threshold.
Type of Respondent
Burden estimate (TNM)
Beginner
20 min. ≤ burden≤ 30 min.
Expert
5 min. ≤burden≤ 10 min.
17.1. Data revision - policy
The general revision policy is predetermined and reasonably stable from year to year and is adopted in order to incorporate updated information. It consists of:
Routine revisions
The industrial production indices of the most recent month are provisional and subject to a revision occurring with the dissemination of the following month and considering additional information received from enterprises (revised indices are disseminated in the press release).
Second type of revision
A second type of revision occurs annually and concerns the time series of the indices. This revision aims at incorporating four types of information, which become available after the publication of the first revision. Specifically, the elements considered in the revision process are the following:
The release of the most recent data to define the weighting structure. As previously highlighted, at the moment for the indices of the year t there is a misalignment between the year of the link period (t-1) and the data available for calculating the structure of their weights (t-3 for the SBS Frame and t-2 for ProdCom). When the indices for the year t+1 are released (where link period is t) also data t-2 for the SBS Frame and t-1 for the ProdCom become available, therefore data for year t are revised to update the weighting structure.
The updating and periodic revision of the short-term statistics (turnover index and hours worked) on which the used annual productivity coefficients are based, as mentioned above, for the products surveyed through the monthly flows of hours worked. These products that, as previously reported, weights for 12.2% belong to few sectors (in particular, machinery and mechanical equipment, electrical equipment, transport vehicles, repairs and systems installation). Consequently, the effect of the coefficient revision may be significant for those specific sectors.
Responses received from the enterprises after the dissemination of the revised indices (which usually occurs around 60 days past the reference period); it involves a very limited number of responses, which accounts, on average, for approximately 2.3% of the sample (measured in production volume terms) but which may determine corrections on the disaggregated indices.
Subsequent corrections of information previously received from enterprises that have been reported as inaccurate by consistency checks. Usually these modifications have a minor effect on the aggregated indices, however they occasionally may cause significant revisions for specific sectors.
Coherence of the revision policy is respected for data released nationally and transmitted to Eurostat.
17.2. Data revision - practice
A paragraph called 'Revisioni' is attached to every press release in Italian and English version. It contains a methodological note on revisions (revision policy, source, periodicity, etc.) and the results of the revision analysis for the current period presented. Major revision dues to mistakes are made known to the public after the event. The annual revision is disseminated through the bulletin released on January 2025 and concerns monthly indices of 2024. The comparison between the 2024 average annual rate of change of the indices released on 12 February and the revised ones – data corrected for calendar effects – shows a worse annual trend for the general index (from -3.5% to -4.0%).Here – as mentioned – the effect of three factors comes into play: the late responses which also have a very marginal impact, productivity, which has its effects above all on capital goods (whose variation goes from -4.7% to -5.7%) and the role of the updated weights, visible for example in the decrease in the variation of energy (from +0.6% to +0.2%).
TABLE. INDUSTRIAL PRODUCTION index BY MAIN INDUSTRY GROUPING. Data corrected for calendar effects. Revisions for the year 2024. Average annual changes year 2024/2023
MAIN INDUSTRIAL GROUPINGS
AVERAGE ANNUAL CHANGES
2024
(published on February
12, 2025)
2024
(revised data)
Consumer goods
-3.3
-3.8
Durable
-4.8
-3.8
Non-durable
-2.9
-3.6
Capital goods
-4.7
-5.7
Intermediate goods
-3.5
-3.6
Energy
0.6
0.2
Total
-3.5
-4.0
The Mean Absolute Revision (MAR), Mean Revision (MR) and RMAR (Relative Mean Absolute Revision) are compiled for totals.
Unadjusted
Period
MR
MAR
RMAR
MAXAR
2021-2024
0.1
0.267
7.4%
1.2
Working-day adjusted
Period
MR
MAR
RMAR
MAXAR
2021-2024
0.122
0.289
10.3%
1.2
Seasonally adjusted
Period
MR
MAR
RMAR
MAXAR
2021-2024
-0.042
0.297
25.9%
1.3
18.1. Source data
Data are collected through a statistical survey. In addition data are obtained: - for electric energy, from TERNA (Rete Elettrica Nazionale); - for mining and quarrying, from Uffici nazionali minerari, idrocarburi e geotermia (Ministero delle attività produttive); - for meat production, from the Istat survey on slaughtered animals for meat production.
The PRODCOM annual survey is used as frame.
Sample or census: The industrial production is a cut-off survey, with a non-probabilistic sample. For each product, enterprises are ranked according to the percentage of the value of production they represent for this product. Then the highest ranked units are drawn until 70% of the value of production is reached.
Criteria for stratification: not applicable
Threshold values and percentages: In total around 5,700 KAU are surveyed each month.
Frequency of updating the sample: As a consequence of the adoption of the concatenated method was updated the sample of companies used in the survey. The following table shows between the two bases for enterprises, respondent units and production.
18.2. Frequency of data collection
Monthly
18.3. Data collection
Self-administered data collection via electronic transmission, data are collected through a web portal called Portale Imprese. Respondents need to login for insert data in the web questionnaire.
Accurate checks on time series are used for plausibility of data to validate. Integrity of files structure and used codes is ensured by extractions made from a database respecting integrity constraints.
18.5. Data compilation
Estimates for non-response: For estimating non responses concerning a given product, usually an automatic estimation criterion is applied on the basis of the average percentage variation m/(m-1) calculated on responses available for the same product. Estimates for grossing-up to population levels: None. Type of index: Chaining method introduced by the publication of the indices relating to the month of January 2022 Laspeyres index. Method of weighting and chaining: The elementary products are aggregated up to the Class level using as weights the output value (from PRODCOM survey); the aggregation from the Classes up to the total is based on value added at factor costs weights drawn from SBS data.
18.6. Adjustment
Data are disseminated in unadjusted form, adjusted for working days (total index, indices by MIGs, branches at 4-digit level of NACE Rev. 2 classification), and in a seasonally-adjusted form. The raw indices are trading days adjusted and corrected for seasonal effects by TRAMO-SEATS procedure, version 942 for Linux. Working days adjustment series and seasonal adjustement series are estimated down to 4 digit level. Adjustment models are completely reviewed at the beginning of each year (with the release of February data) and parameters are re-estimated every month. Each month, seasonally adjusted series are revised from January 2001.
A Reg-Arima approach is used for working days and seasonal adjustment. Calendar adjustment is performed by using our country specific calendar which also take into account national holidays. No calendar adjustment is performed where no significant effect is found. Leap year and moving holidays like Easter are adjusted. Outliers (additive outliers, temporary changes, level shift) are manually fixed at the beginning of the year when seasonal adjustment models are revised. The critical value for the outlier detection and the filter length are automatically chosen, whereas model/filter selection is manual. Both additive and multiplicative seasonal adjustment decomposition are used. All series are directly adjusted. Whenever indirect adjustment is performed, a residual seasonality test is performed and checked.
Industrial production index (Indice della produzione industriale).
The survey describes the development of productive activity in Italy through the collection of production volumes.
The 2024 indices are calculated using the chaining method introduced by the publication of the indices relating to the month of January 2022.
Starting from the publication of the January 2024 data, the industrial production indices are released in the new reference base 2021=100. The calculation base has also been updated and now relates to 2023.
The innovations introduced concern not only the change in the reference base but also the sources used, the updating of the sample of enterprises used in the survey, the revision of the weighting system, the updating of the basket of goods.
Starting from this edition, for some groups of economic activity, the survey on industrial production uses information deriving from the survey on industrial turnover, appropriately deflated.
Hence the 2024 indices are calculated using the chaining method - the weighting reference is year 2023 - according to the Laspeyres formula using a structure of weights updated. annually.
13 June 2025
List and definition of variables: The index shows the evolution of gross production in volume terms.
For the Link period 2023, indices for Year 2024
Starting from the 2023 base, the survey of industrial production is carried out by a panel of companies that provide information on just under 1,196 products (of which 105 car models) on a monthly basis. These products are grouped into homogeneous groups to form a basket of goods, which represents the elementary level of aggregation (macroprodotti/macro items).
Information is collected on more than 9,500 basic observations for about 5,800 Respondent units product sand .5,400 Enterprises
Accounting conventions: The target variable is measured as a flow during the calendar month. Products, which need more than a month for the production, are recorded in value or in worked hours only for the part accrued to the reference period. Production in value is deflated by output prices. Hours worked are adjusted with productivity coefficients based on short-term indicators.
With the update of the weighting structure, the incidence of the proxies used to measure the trend of industry output have changed compared to the previous base of 2022. As can be imagined, increases the share –expressed in terms of the relative weight on the general index – of products recorded in value to the detriment of the other two measures, due to the use for some groups of economic activity of industrial turnover, appropriately deflated. The incidence of products recorded in quantity remains predominant, however decreasing compared to the previous base (72.7% compared to 76.3%), followed by products recorded in value of production (15.1%), whose share instead, it grows by almost 5 percentage points compared to the 2022 base. Finally, the incidence of products in hours worked is 12.2%.
TABLE. Proxies used to measure products. Weighting structure of fixed base 2015 and link period 2021, 2022 and 2023
TYPE OF PROXY
Fixed base2015
Link period 2021
Link period 2022
Link period 2023
Physical quantities (a)
76.4
76.6
76.3
72.7
Hours worked
13.1
13.6
13.5
12.2
Value of production
10.5
9.8
10.1
15.1
TOTALE
100.0
100.0
100.0
100.0
Planned changes in information collected: None.
Reporting unit: KAU or local unit KAU.
Observation unit(s): Unit of homogeneous production (UHP) or Local unit of homogeneous production (LUHP).
Target population corresponds to enterprises with 20 or more employees in 2022 for sections covered by this survey. Total number is about 31,200 enterprises.
Whole national territory.
Month
The samples are designed with the cut-off method and therefore sampling errors cannot be measured. Non-sampling errors are reduced in the collection phase both using an electronic questionnaire that helps debugging and aiming at an increase of the response rate of enterprises.
Index
Estimates for non-response: For estimating non responses concerning a given product, usually an automatic estimation criterion is applied on the basis of the average percentage variation m/(m-1) calculated on responses available for the same product. Estimates for grossing-up to population levels: None. Type of index: Chaining method introduced by the publication of the indices relating to the month of January 2022 Laspeyres index. Method of weighting and chaining: The elementary products are aggregated up to the Class level using as weights the output value (from PRODCOM survey); the aggregation from the Classes up to the total is based on value added at factor costs weights drawn from SBS data.
Data are collected through a statistical survey. In addition data are obtained: - for electric energy, from TERNA (Rete Elettrica Nazionale); - for mining and quarrying, from Uffici nazionali minerari, idrocarburi e geotermia (Ministero delle attività produttive); - for meat production, from the Istat survey on slaughtered animals for meat production.
The PRODCOM annual survey is used as frame.
Sample or census: The industrial production is a cut-off survey, with a non-probabilistic sample. For each product, enterprises are ranked according to the percentage of the value of production they represent for this product. Then the highest ranked units are drawn until 70% of the value of production is reached.
Criteria for stratification: not applicable
Threshold values and percentages: In total around 5,700 KAU are surveyed each month.
Frequency of updating the sample: As a consequence of the adoption of the concatenated method was updated the sample of companies used in the survey. The following table shows between the two bases for enterprises, respondent units and production.
Monthly
The official release is published on average 40 days after the end of the reference period; data are transmitted to EUROSTAT within 40 days, using, when needed, an embargo clause; data are provisional at the time of the Press Release in which they are reported.Until 2021 these revisions typically occurred when releasing the indices concerning the month of February and incorporated both the new estimates of the short-term statistics for the three previous years and the corrections based on the late responses received and on the corrections to already received information. Starting from year 2022 retrospective revisions concern the previous year only.
European definitions are used and they apply in the whole national territory
Consistent time series data are available for a long period (since January 1990). When changes in source data, methodology and statistical techniques are introduced, historical series are reconstructed as far back as possible.
