Central Statistics Office (CSO)

(i)  Business Statistsics Division Data Collection Uniit

(ii) Business Statistics Division Results, Analysis and Publication

Central Statistics Office, Skehard Road, Cork T12 X00E, Republic of Ireland.

The Retail Sales Index (RSI) is the official short-term indicator of changes in the level of consumer spending on retail goods. It measures the trend in the level of average weekly sales for each month, after allowances are made for calendar composition.

The primary purpose of the RSI is to provide a short-term indication of changes in the value and volume (or quantity) of retail sales in Ireland.

The Retail Sales Inquiry collects the data on sectors G45, G47 and I563.

The Monthly Services Inquiry (MSI) collects data on the Services Sector including G46 Wholesale

Statistical Classification of Economic Activities in the European Community, Rev. 2 (2008) (i.e. NACE Rev. 2). 

Nace Ref                                 NACE REV 2 Description 

Retails Sales Inquiry (collection)

Retail Sales Index (dissemintation)

45                                           Sale Maintenance and repair of motor vehicles and motorcycles

47.11                                       Retail sale in non-specialised stores with food, beverages or tobacco predominating

47.19                                       Department Stores

47.2                                         Retail sale of food, beverages and tobacco in specialised stores

47.3                                         Retail sale of automotive fuel

47.73-5                                    Retail sale of pharmaceutical and medical goods, cosmetics and toilet articles

47.51, 47.71-2                          Retail Sales of Clothing Footwear & Textiles

47.59                                       Retail sale of furniture, lighting equipment and household articles

47.52                                       Retail Sale of hardware, paints and glass

47.41-3, 47.54                          Retail sale of electrical household appliances and radio and television goods

47.61-2                                    Retail of books newspapers and stationery

47.76-8                                    Other Retail Sales

47.91                                       online and mail order

Monthly Services Inquiry (collection)

Monthly Services Index (dissemination)

G46                                          Wholesale

Retail Sales Index

 Size Coverage: All enterprises with one or more persons employed.

Size classes are determined by turnover.  There are 3

Size Class Annual   Turnover Thresholds

1                          € 999,000 > Turnover

2                          € 1,000,000 ≤ Turnover < € 4,999,999

3                          € 5,000,000 ≤ Turnover

Monthly Services Index

Size coverage enterprises with turnover of €500,000 and over.

Variables: Turnover.

Planned changes: None.

Accounting conventions.

Respondents are offered the option of responding in:

1. a 4-4-5 week pattern
           or
2. a calendar month basis

Reporting unit: The reporting unit is the Kind of Activity Unit (KAU) if identified otherwise it is the enterprise.

Observation unit: The observation unit is the Kind of Activity Unit (KAU) if identified otherwise it is the enterprise.

Note Each year a thorough inverstigation is done involving statisticains from Business Demography, Short Term Statistics and Profiling unit to identify KAUs or potential KAUs in each short term survey.

Enterprises that were 20% of their NACE division were identified and investigated, Our LCU companies (mainly in ICT in Services) were investigate and either no KAU was found or the Enterprise  was not able to provide data at KAU level as it is a statistical concept not an accounting concept.  Therefore Enterprise = KAU.

Note no enterprise in either G45, G46 or G47 is 20% of its division by turnover, number of employees or Gross Value Added.

The Retail Sales Index which covers G45, G47 and I563.

The survey population is made up of all enterprises in NACE 45, 47 and 56.3.  This target population is approx. 25,400 enterprises from the Annual Services Inquiry.

The RSI sample is approx. 1,700 enterprises.  This gives an average sampling fraction of 7%.

If coverage is viewed from a turnover perspective, the coverage is considerably higher, with an average sampling fraction of approximately 50%.

Within each NACE Classification there are 3 Size Classes. The Size Classes are categorised into 3 groups (or size classes) according to their turnover.

 The 3 size classes are:

Size Class

Annual Turnover Thresholds

1 € 999,000 > Turnover

2 € 1,000,000 ≤ Turnover < € 4,999,999

3 € 5,000,000 ≤ Turnover

Generally a census of all enterprises within Cell Size 3 is conducted. A sample of enterprises with a turnover of less than €5.0m is then conducted. This sample is a stratified sample.  Enterprises are stratified initially by their NACE classification and then within each NACE sector by turnover.

The retail sales registry is updated annually with relevant enterprises in the retail sector.

Commencements of businesses, changes of business activity and cessations of businesses are updated at this time. Ad-hoc changes are also made to the register when relevant information becomes available from individual enterprises or from other sources. The retail sales registry is a subset of the Central Business Register (CBR). The CBR is updated on an on-going basis using administrative and survey data.

The monthly Services Inquiry is how sector G46 is collected

The smaple is of enterprises with an annual turnover value of more than €500,000 or enterprises with more than 100 persons engaged

Note the RSI collects and disseminates I56.30 for national purposes.  

The geographical area covered is the Republic of Ireland.

The first monthly RSI index figures were published in March 1963 and the RSI has been published continuously since then. The first figures published were value figures only and date back to January 1961. In March 1977 volume figures were first produced and at this time the first seasonally adjusted figures were also produced. Seasonally adjusted and unadjusted RSI value and volume indices for “Total Retail Sales” were produced with figures going back to January 1968.

The monthly services Index has been published since 2009.

Base year 2015=100. 

Collection : Euro.

Dissemination: Indices and percentage changes.

G45 and G47 (Retail Sales Inquiry)

Is either a 4 week or 5 week period depending on the month.

Respondents are offered the option of responding in:

1. a 4-4-5 week pattern
           or
2. a calendar month basis 

G46 (Monthly Services Inquiry) is calendar Month

National: The RSI is a statutory inquiry, collected under the Statistics (Retail Sales Inquiry) Order 2018 No. 432/2018. 

http://www.irishstatutebook.ie/eli/2016/si/118/made/en/print

The Monthly Statistics is a statutory inquiry, collected under the Statistics (Monthly Services Inquiry) Order 2018 (S.I. No. 431/2018)

https://www.irishstatutebook.ie/eli/2018/si/431/made/en/print

Legislative Basis European:

European: Council Regulation (EC) No. 1165/98 and Commission Regulation No. 472/2008 concerning short-term statistics.

https://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CONSLEG:1998R1165:20090101:EN:PDF

REGULATION (EU) 2019/2152 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 20 June 2019

https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A32019R2152

Section 34 of the 1993 Statistics Act states:

“The Office may provide, for statistical purposes only, information obtained in any way under this Act or the repealed enactments, in such form that it cannot be directly or indirectly related to an identifiable person or undertaking, to such persons and subject to such charges, conditions and restrictions as the Director General may determine.”

The provision on statistical confidentiality is regulated by Sections 32 and 33 of the 1993 Statistics Act. It states:

Section 32 states:

“All information furnished by a person, undertaking or public authority under this Act shall be used only for statistical compilation and analysis purposes.”

 Section 33 states:

“(1) No information obtained in any way under this Act or the repealed enactments which can be related to an identifiable person or undertaking shall, except with the written consent of that person or undertaking or the personal representative or next-of-kin of a deceased person, be disseminated, shown or communicated to any person or body except as follows:

 (a)   for the purposes of a prosecution for an offence under this Act;

(b)   to officers of statistics in the course of their duties under this Act;

(c)   for the purposes of recording such information solely for the use of the Office in such form and manner as is provided for by a contract in writing made by the Director General which protects its confidentiality to his satisfaction.      

   (2) The Office may, for statistical purposes only, assign codes derived from information collected under this Act classifying undertakings listed in the administrative systems of other public authorities by economic activity and size (persons engaged) categories.

   (3) The Taoiseach may by order prescribe such further prohibitions on the disclosure of identifiable records or information obtained under this Act or the repealed enactments for such periods as may be prescribed.

   (4) Nothing in this Act shall be construed to require any person or undertaking to provide information in relation to a matter on which information was sought in circumstances that would entitle the person or undertaking to decline to give the information in a civil proceeding in any court or on grounds of privilege.”

The retail turnover data provided by the respondent enterprises are treated as strictly confidential in accordance with Part V of the Statistics Act, 1993 and cannot be accessed under the terms of the Freedom of Information Act, 1997. Data are not disclosed by the CSO to any other Government Department or outside body.

An annual calendar showing release dates is publically available and updated on the CSO website. In addition, a weekly release calendar which contains a schedule of precise release dates for those statistics that are to be published by the CSO during the following week is issued each Thursday by e-mail to the media and all other interested parties.

The annual release calendar can be accessed on the CSO website.

http://www.cso.ie/en/csolatestnews/releasecalendar/

The CSO disseminates the RSI publication on the CSO website at 11am (local time) on the day of publication (see point 8.1).

RSI data is transmitted to Eurostat after 11am on the day of publication or under embargo, if the national RSI publication is scheduled to be published after the deadline for transmission of data to Eurostat.

MSI is published on the CSO website at 11am GMT on the day of publication (see point 8.1)

MSI data is transmitted to Eurostat after publication or under embargo, if the national MSI publication is scheduled to be published after the deadline for transmission of data to Eurostat.

Monthly.

None.

Data on retail sales are released in the following locations on the CSO website:

On the banner of the CSO home page / landing page.

 “Retail Sales Index” - “Latest Releases”/”Latest Headline Figures” on the CSO homepage http://www.cso.ie/en/index.html and also on http://www.cso.ie/en/statistics/services/

The quality description of the statistics, as well as concepts and definitions, are available on:

http://www.cso.ie/en/methods/services/

The data that is published in the RSI monthly publication is also made available on PxStat with each aggregated index also published as an unadjusted value index on PxStat only.

 PxStat is the CSO Data Dissemination Management System, developed by the CSO with the support of the PC-Axis Reference Group and the collaboration of the Open Source community.

 https://data.cso.ie/table/RSM05

Microdata are not publicly available.

Data are sent to Eurostat to be used in European aggregates and/or to be released as national data. There are no planned future changes in national dissemination.

These are available at the following website, under Surveys & Methodologies’:

https://www.cso.ie/en/methods/services/retailsalesindex/

http://www.cso.ie/en/methods/services/retailsalesindex/

This is available at the following internet site:

https://www.cso.ie/en/methods/qualityreports/retailsalesindex/

Guidelines focusing on quality in general and dealing with quality of statistical programmes, including measures for ensuring the efficient use of resources.Guidelines:

The Retail Sales is conducted under the guidelines of the CSO quality assurance framework https://www.cso.ie/en/methods/quality/qualitymanagementframework/ and the European Statistics Code of Practice

Self Assestment Quality questionnaires are carried out annually.  Compliance is monitored in April and November.  International best practice is researched and implemented where possible. 

Response rates are compiled and analysed each month.  Quality checks and validation of data are done throughout the whole compilation process. Details are available on http://www.cso.ie/en/methods/services/retailsalesindex/ in the “Standard report on Methods and Quality for Retail Sales Index” - Chapter 5 “Quality” – a summary of main points are given here):

1. Accuracy and Reliability

 1.1 Sampling Effect & Representivity

The population of retail enterprises in Ireland is approximately 25,400. The RSI sample comprises about 3,000 enterprises. This gives an average sampling fraction of 7%, however this fraction can differ significantly from sector to sector and within size class.

If coverage is viewed from a turnover perspective, the coverage is considerably higher, with an average sampling fraction of approximately 50%.

1.2 Non-Sampling Effects

Bias may be introduced as a result of the matched sample approach. It is difficult to completely represent the population due to the ongoing births and deaths of enterprises within the sector. This may result in the consistent under-estimation of growth during sustained periods of growth.

 1.3 Quality of the Data Sources used (other than survey register)

Monthly product price data from the CSO’s consumer price section is supplied to the retail sales section in order to calculate RSI deflators. The data supplied is thoroughly cleaned and edited.

 1.4 Register Coverage

The retail sales registry is updated annually with relevant enterprises in the retail sector. Commencements of businesses, changes of business activity and cessations of businesses are updated at this time. Ad-hoc changes are also made to the register when relevant information becomes available from individual enterprises or from other sources. The retail sales registry is a subset of the Central Business Register (CBR). The CBR is updated on an on-going basis using administrative and survey data.

 1.5 Non-response (Unit and Item)

High quality of data is ensured due to a high response rate (60%). The response rates for the provisional survey are typically in the order of 50% covering about 85% of the value of the turnover surveyed. The corresponding figures for the final month are approx.60% covering about 90% of the value of the turnover.

The retail sales index adopts a matched sample approach and therefore there is generally no need to conduct imputation. Occasionally, however, some ad-hoc imputation is carried out for some significant firms. In these cases each firm is looked at individually and an extension of nearest neighbour and last observation carried forward techniques to impute the missing values. Non response bias is not measured in the survey.

 1.6  Measurement Errors

Measurement Error is not regarded as a major concern for this survey. The following should be noted:

1. Questionnaire – the questionnaire is clear with the period requested clearly defined. The design of the questionnaire is monitored and changes made if necessary. Occasionally enterprises may provide turnover inconsistent with the period covered. These errors are usually highlighted at the data-cleaning stages.
2. Data collection method – the collection method is by eform and by post. These methods are unlikely to be a source of measurement error.
3. Respondent – survey data can contain sensitive and confidential information due to the fact that it is a short-term survey and may be requested prior to annual accounts being completed. The survey publishes aggregated indices at 4 digit NACE (may publish at 3 or 4 digit NACE, or may aggregate more that one-two digit NACE group – due to confidentiality).

 1.7 Processing Errors

1. Data capture errors are likely to be low risk as survey postal forms are keyed and edited immediately.

2. No manual coding is done.

4. Data editing- edits are run which test the returns against a range of values, e.g.;

• Turnover is unusually high.
• Large increase/decrease in turnover compared to the corresponding month last year.
• If turnover is greater than 0 then number of outlets must also be greater than 0.

 5. Contact is made with company to confirm returned figures if required. Data are manually changed if required.

 1.8 Coherence

Consistency checks are done on an ad-hoc basis with the Annual Service Inquiry.

Differences are verified and this leads to improved data quality.

In certain instances comparison with other data sources such as Revenue Commissioners data and Industry data are conducted. These are investigated where necessary leading to better quality data. In general the same trends are evident in the various data sources.  Data coherence checks are done with other areas of the office eg Vehicle licensing compared with Motor Trade data.

The primary purpose of the RSI is to provide a short-term indication of changes in the value and volume (or quantity) of retail sales in Ireland. In doing so the RSI provides a leading monthly indicator on economic activity. It provides an accurate and objective measure of retail trading and supplies a valuable guide to consumer spending behaviour in the Irish economy. More generally, in conjunction with several other monthly and quarterly economic indicators published by the CSO, the RSI offers a valuable tool for better understanding the general economic climate and performance in Ireland.

 It is mainly of relevance to the following user groups:

1. Eurostat,

2. Government,

3. Economists,

4. Other CSO divisions eg National Accounts,

5. The general public,

6. Professional Bodies,

7. University students and other interest groups for research purposes,

8. Media.

Due to Globalisation of the Irish Economy the Retail Sales is also seen as a main indicator of the domestic economy without the Global influence.

No information available.

The index is fully compliant with requirements of the Council Regulation (EC) No 1165/98 concerning short-term statistics and the subsequent amendments and Council Regulation (EU) 2019/2152 concerning European Business Statistics 

For interpretation of results, users have to be aware that every index draws a picture of reality in a simplified way.

There is no statistical model which is able to represent the complexity of reality in total.

The CSO tries to reduce sampling errors by using a sample of enterprises that is as large as possible while taking burden on enterprises and time and resource constraints into account.  The sample is designed to focus on the larger enterprises which have the largest turnover and therefore the greatest impact on the statistics.

Non-sampling errors are not quantified. The CSO tries to reduce non-sampling errors through continuous methodological improvements and survey processing improvements. Bias may be introduced as a result of the matched sample approach. It is difficult to completely represent the population due to ongoing births and deaths of enterprises within the retail sector. This may result in the consistent under-estimation of growth during sustained periods of growth.

Provisional Results: The preliminary data are published by the CSO within T + 28 days after the end of the reference month.

Final Results: The final data are published by the CSO within T + 58 days after the end of the reference month (i.e. the final results are published at the same time as the provisional results for the subsequent month).

The results are transmitted to Eurostat within T + 28 days after the end of the reference.

RSIs have always been published nationally on the pre-announced release dates and RSIs have always been transmitted to Eurostat within the agreed time frame even during COVID. 

In general, the reporting of retail sales is similar throughout Europe but the underlying methodologies may differ. Many countries will not include Motor Trade and Bar sales in their retail figures. Therefore to compare Ireland’s retail sales figures with those of other countries in Europe the retail sales figure as reported in the ‘All Businesses excluding Motor Trades and Bars’ is widely used.

The RSI is a monthly turnover index. This presents a comparability problem as months differ in length i.e. the number of days in each month. A further difficulty arises from what those days are i.e. how many Fridays, Saturdays etc. are in a particular month. This is critical for retail trade as a trade varies on certain days.

To overcome this difficulty, the RSI indices are compiled using standardised reporting periods (SRPs) of 4, 4 and 5 weeks, i.e. the first two months of every quarter comprises of 4 weeks while the third month has 5 weeks. With this SRP approach the number of days in every month is equalised. So not only does each month have a standardised number of weeks, turnover is “trading day” adjusted so that effectively, each of those weeks are identical – every week begins with a Sunday and finishes on a Saturday.

For each period, enterprises have the option of reporting their turnover using either the standardised month or the actual calendar month. About 20% of respondents, particularly large enterprises, supply data corresponding to the 4-4-5 pattern. The remaining enterprises provide calendar month data. This calendar month is then adjusted to a standardised month using calendar correction factors.

These calendar correction factors are based directly on trading day micro data provided by enterprises on the RSI sample. Every 5 years, as part of the rebasing process, enterprises are asked to distribute the average weekly sales over the 7 days of the week. This data is then compiled to construct the trading day weights. These fixed trading weights are then used to calculate calendar correction factor for each month.

The 4-4-5 pattern adds up to 364 day year and consequently requires a re-calibration every 5th or 6th year (depending on when leap years fall) to account for the missing week. Here the exact 52 week year is replaced by an exact 53 week year. This additional week is added to February, replacing the 4-4-5 pattern with a 4-5-5 pattern for the 1st quarter of the re-calibrated year.

In certain instances comparisons with other data sources such as Revenue Commissioners data and Industry data are conducted. These are investigated where necessary leading to better quality data. In general the same trends are evident in the various data sources.

Data is tested for coherence at individual enterprise level each month comparing data with the corresponding data from the same month of the previous year and any inconsistencies are followed up with the enterprise.

Consistency checks are done on an ad-hoc basis with the Annual Service Inquiry. Differences are verified, leading to improved data quality.  Coherence checks are done with Vehicle Licensing.

No information available.

The current months data is always provisional and only becomes final when the next month data is finalised.  The seasonally adjusted turnover and volume indices are revised each month as another month’s data is included.  Any national revisions are transmitted to Eurostat at the earliest opportunity.  This usually is on the day of publication or the next transmission date.

Each month the provisional data from the previous month is revised on the receipt of late returns or amendments of existing returns.  

For 2022 the following were calculated for the volume of retail sales

• the mean average revision was 0.1%
• the mean absolute revision was 0.9%
• range of revisions were -2.1% to 2.4%

For 2022 the following were calculated for the value of retail sales

• the mean average revision was 0.1%
• the mean absolute revision was 0.9%
• range of revisions were -2.0% to 2.2%

The edit checks performed are checks on consistency of data between months.

The following edit checks are carried out:

1. Large increase/decrease in turnover compared to the corresponding month last year.
2. If turnover is greater than 0 then number of outlets must also be greater than 0.

Each member of staff is responsible for a block of enterprises based on Business Groups. Key enterprises have been identified on each block based on turnover and are prioritised regarding edits. Each edit is checked for possible scanning or verification errors and any comment that may explain an inconsistent figure on the return. Returns for the previous months are also checked for indications of a trend. The enterprise is contacted to resolve any discrepancy where necessary. The return is then manually edited on the basis of the explanation from the enterprise. Once all errors are amended a copy is made of the clean dataset.

Reports are run throughout the survey period outlining the number of edits outstanding and also response rates are tracked.

The Edit rules had to be changed in 2020 and depending on the sector due to the massive impact the COVID 19 pandemic had on turnover.

Seasonal adjustment is conducted using the direct seasonal adjustment approach.  Under this approach each individual series is independently adjusted, e.g. aggregate series are adjusted without reference to the component series. Each individual seasonally adjusted series is calculated based on unadjusted data spanning from January 2010 to the current period.

Seasonal adjustment for the Retail Sales series is complicated by the complex calendar effects induced by the Retail Sales Inquiry’s use of the 4-4-5 standardised reporting period. Each year each standardised month changes its position relative to the calendar month.  The twelve standard periods total to 52 weeks or 364 days compared to 365 days in a year (or 366 in a leap year).  Consequently each year the standardised months used by the Retail Sales slip back one (or two days) every year.  Every five or six years an extra week is added to February to cater for this slippage.

The slippage of the standardised months relative to the corresponding calendar months operates on a cyclical basis.  The cycle should have a cycle of seven years but because of leap years the actual cycle is 28 years.

Each individual Retail Sales series or sub-index is seasonally adjusted separately using the model that best fits the characteristics of that series.  Individual series models will be reviewed once every 12 months and series models and parameters are adjusted if required. 

The revised series are published on the CSO web – www.cso.ie.  Distinct seasonal factors are calculated for each Business Group and each Combined Group.  This reflects the different seasonal patterns exhibited by different Business Groups and Combined Groups. For example, sales in the Motor Trades are low in December, contrasting with sales of Food Beverages & Tobacco, which usually increase in the lead up to Christmas.  Separate factors are also calculated for the value and volume series. 

The adjustments are completed by applying the X-13-ARIMA model, developed by the U.S. Census Bureau to the unadjusted data. This methodology allows seasonal factors to be estimated whilst also taking into consideration factors that impact on the quality of the seasonal adjustment such as:

• Calendar effects, e.g. the timing of Easter,
• The phase shift effect, i.e. the fact the reporting period of the RSI does not coincide with the calendar month and
• Outliers, temporary changes and level shifts in the series.

 Please look at https://www.census.gov/srd/www/x13as/ for additional information on the X-13-ARIMA software.

Identifying and Treating Outliers, Temporary Changes and Level Shifts

Outliers, temporary changes and level shifts are abrupt changes in the underlying series that can affect the quality of the seasonal adjustment if not treated correctly. The X-13-Arima seasonal adjustment program identifies any outliers, temporary changes or level changes and removes them from the original series before the seasonal adjustment factors are calculated.  Once the seasonal adjustment factors are calculated these outliers, temporary changes and level shifts are then re-introduced so they are present in the final seasonally adjusted series.

Estimating Calendar Effect

The calendar effects induced in the series by the 4, 4, 5, recording period can be estimated and removed using the X-13-Arima seasonal adjustment program.  The model of a retail sales series y_t may be represented as follows:

 Where  is the regression effect, and z_t is the regression error term fitted by an ARIMA model.

 In retail sales the following calendar (regression) effects are adjusted for:

1. The phase shift effect resulting from year to year movement of the standard recording periods.
2. Easter effects, resulting from the Easter holiday moving between the Retail Sales’ standardised ‘March’ and ‘April’ periods.
3. An October Bank Holiday effect resulting from the holiday moving between the Retail Sales’ standardised ‘October’ and ‘November’ periods.

In order to estimate these regression effects 16 separate regressors were included initially in the regression model for each series. There are 12 phase shift regressors, 3 Easter regressors and 1 October Bank Holiday regressors. For each series 10 of the phase shift regressors (‘January’ – ‘September’ and ‘December’) are formally tested for statistical significance using the standard t-test.  The ‘November ‘, ‘October’ and October Bank Holiday regressors are jointly tested using the log likelihood ratio test. The 3 Easter regressors are similarly jointly tested.  Only regressors that are proven to be statistically significant are included in the final regression model.

Once the calendar effects are adjusted for the seasonal factors are then calculated the X-13-Arima seasonal adjustment program.

not applicable