IMPROVING LIVES THROUGH DATA ECOSYSTEMS

1. Background

The personal, social and economic impact of COVID-19 is unlike anything experienced by the world in the past 75 years. Stats SA is aware of the need for timely data to understand these impacts, but is equally constrained by the restrictions put in place to manage the spread of the coronavirus. The task of scheduling collection, analysis and publication of data is made all the more difficult by uncertainty regarding when enterprises will resume operations and when our staff will be able to return to work.

This statement shares our plans for all the economic releases, based on a lifting of the lockdown by 4 May. It also outlines innovative activities undertaken by the organisation during lockdown.

Most publications due in April (but covering earlier reference periods) have been suspended. The extension of the lockdown to the end of April required a rethink about resuming the publication of statistics on the economy as well as the potential impact of these events on the quality of statistics. The impact on each series differs depending on the source of the information.

2. New activities during COVID-19 lockdown

Two new initiatives were launched in recent weeks.

First, an online survey, based on the sample of the annual financial statistics, was administered with a due date of Friday 17 April . It includes qualitative questions on the impact of COVID-19 on the business. We have no means to engage with the respondents in order to ensure a high response rate as many of these businesses are closed, or working with limited staff who may not be those that the questionnaire was emailed to, or open but non-responsive. Stats SA will compile a note on the results of the survey in the following week, with the option of conducting the survey again.

Second, a selection of food and pharmaceutical items that form part of the CPI basket are being priced on a weekly basis. We aim to release a note on price changes for these products while the lockdown is still in operation.

3. COVID-19: current situation
   • Current lockdown period

During the lockdown period of 27 March – 30 April, staff from Economic Statistics cannot continue with data collection and analysis activities as usual as businesses and large parts of government are closed and only limited numbers of Stats SA staff are able to work remotely.

On the assumption of Stats SA becoming fully operational on Monday 4 May, the conditions we expect to face are as follows.

• Series based on administrative data

In the case of administrative surveys where government agencies provide us with raw data, Stats SA anticipates that these institutions will have to return to full functionality before the required data can be sent for processing and analysis.

• Civil cases for debt (P0041). Collection rate for February currently 72%. Planning to publish February on 21 May; and March together with April on 18 June.
• Liquidations and insolvencies (P0043). Planning to publish March on 11 May; and April on 25 May (note: the reference month for insolvencies lags liquidations by one month).
• Mining (P2041). Planning to publish February on 19 May; and March together with April on 11 June, subject to confirming availability of March and April data.
• Export and import unit value indices (P0142.7). Planning to publish February on 30 April; and March on 28 May.
  • Sample surveys

Sample surveys are drawn from the statistical business register and the enterprises in the sample receive the questionnaire by email. The questionnaire is then either completed and returned to Stats SA, or staff make telephone calls to collect the information. The information is then captured and analysed before the releases are developed and published. It therefore requires that staff are able to access both the respondents to collect the information as well as our internal systems for processing. Currently most respondents are not open for business or are working with limited staff and therefore cannot provide the necessary information. It is therefore not possible to produce these releases as usual.

• Electricity (P4141). Planning to publish March on 20 April and April on 4 June.
• The following monthly releases cannot be collected, analysed and published until such time as all staff return to the office and the respondents are available to supply information.
  • Manufacturing (P3041.2). February collection rate currently 72%. Planning to publish February on 19 May; and March together with April on 11 June.
  • Retail trade (6242.1). February collection rate currently 53%. Planning to publish February on 20 May; and March together with April on 17 June.
  • Building (P5041.1). February collection rate currently 82%. Planning to publish February on 21 May; and March together with April on 18 June.
  • Wholesale trade (P6141.2). February collection rate currently 71%. Planning to publish February on 21 May; and March together with April on 18 June.
  • Motor trade (P6343.2). February collection rate currently 72%. Planning to publish February on 21 May; and March together with April on 18 June.
  • Tourist accommodation (P6410). February collection rate currently 82%. Planning to publish February on 25 May; and March together with April on 22 June.
  • Food and beverages (P6420). February collection rate currently 74%. Planning to publish February on 25 May; and March together with April on 22 June.
  • Land transport (P7162). February collection rate currently 76%. Planning to publish February on 25 May; and March together with April on 22 June.
  • Producer price index (P0142.1). Planning to publish March and April on 28 May.
  • Construction materials price indices (P0151.1). Planning to publish March and April on 28 May.
• Manufacturing capacity utilisation (P3043). Published quarterly. February collection rate currently 65%. Planning to publish February on 12 May.
• Quarterly financial statistics (P0044). Planning to publish the March 2020 quarter in the final week of June.
• Quarterly financial statistics of municipalities (P9110). Planning to publish the March 2020 quarter in the final week of June.
• We plan to publish the annual public-sector statistical releases as previously scheduled, namely Financial census of municipalities (P9114), Non-financial census of municipalities (P9115), National government (P9119.3), Extra-budgetary accounts and funds (P9102), Provincial government (P9121), Higher education institutions (P9103.1), Capital expenditure by the public sector (P9101), Consolidated general government (P9119.4).
• Annual financial statistics (P0021). Planning to publish AFS 2019 in November as scheduled.
• Agricultural survey (P1101). Planning to publish the 2019 survey by the end of March 2021.
• Building statistics – annual (P5041.3). Planning to publish 2019 in June as scheduled.
• Building statistics – annual (50-11-01). Planning to publish 2018 in August as scheduled.
• Large sample surveys. Planning to publish in September as scheduled (retail and other trades; personal services).
  • Fieldwork based surveys – consumer price index (CPI)

Produciton of the CPI relies on data collected by fieldworkers who visit retail outlets across the country and data that is sourced online.

All data for March were collected and all except for a small fraction were fully captured and edited. The March CPI should be published on 22 April.

Data are being collected from online sources during March and April for those products that are available for sale during the lockdown. The CPI for April will be released in a limited publication before the end of May. The CPI for May will also depend on online prices to ensure comparability with the April observations.

• Derived statistics

These statistics are based on the different collections described above, as well as information sourced from other agencies. It is therefore not possible to publish these without having access to sufficient basic statistics.

• Quarterly National Accounts

The GDP estimates for the March quarter were due on 2 June 2020. This will no longer be possible and we now plan to publish on 30 June. Depending on the quality of source data, revisions for the March quarter may be larger than usual when the June quarter is published in September.

• Periodic Environmental-Economic Accounts

Natural Capital Accounts have been developed in conjunction with the UNSD and EU, and will probably be published through a series of accounts between July and November 2020.

• Annual Tourism Satellite Account

The publication of the TSA for 2018 remains scheduled for March 2021.

A revised publication schedule for economic statistics is provided in the final section of this statement.

4. Risks

If we resume normal duties on Monday 4 May, and the business community is fully operational, there are still risks associated with meeting the revised publication schedule. These include:

• The availability of administrative data is not guaranteed (liquidations and insolvencies; mining; civil cases for debt).
• Low response rates. Many enterprises in the sample will not prioritise responding to Stats SA questionnaires. We expect that as enterprises open for business, they will have many other priorities that will occupy them. Some companies may not have the information readily at hand as their accounting departments need time to generate their financial statements after their lockdown.
• Most monthly surveys are subjected to seasonal adjustment, a technique that adjusts estimates due to the seasonal nature of the phenomenon that is measured. Typically a seasonal adjustment model is updated annually, and then the adjustment factors are kept in place for 12 months. With the data points for March and especially April expected to be outside the norm, the relevance of the current seasonal adjustment models needs to be re-examined. If not revisited, by treating the data points as outliers as an example, the published statistics could give an incorrect estimate of the actual economic activity that took place.

Stats SA is working on a multi-year project with the International Monetary Fund to harmonise seasonal adjustment techniques in the different monthly indicators with the estimates of National Accounts. This forms part of the benchmarking and long-term revision of National Accounts, due for September 2020 (see further below). The issues around outliers due to the lockdown or the long-terms effects of COVID-19 are the subject of methodological discussions in most statistical organisations. Stats SA is actively following these developments.

5. Longer-term impact

If the lockdown is extended further, the revised publication schedule will no longer be relevant.

Two other issues that we will engage users and the international community on are:

• The scope and coverage of the surveys. All sample surveys are based on the statistical business register, which in turn relies on (amongst others) taxation information received from SARS. After various processes, such as taking a snapshot of all the businesses, the frames and annual samples are drawn based on the specific specifications of each survey.

An enterprise is deemed to still be alive for 18 months after its last VAT payment. However, should there be significant closures of smaller businesses especially in industries that are not dominated by large and medium enterprises, the sample surveys will include these dormant businesses in the estimates. It is not possible to quantify this risk of future over-estimation of the level of economic activity at present. Stats SA is following developments within other national statistical offices on how this phenomenon is being dealt with.

• Benchmarked estimates of National Accounts were scheduled to be published on 8 September. This involves a comprehensive review of the levels of economic activity and the associated patterns of growth. The current estimates will be revised from 2005 with a new base year of 2015. A new series of supply and use tables will be developed for 2011 onwards. The primary data sources include periodic economic large sample surveys, household expenditure surveys as well as the annual financial survey. There are two potential complications.

First, the rescheduling of the forthcoming quarterly GDP will lead to overlapping activities related the regular GDP and activities related to the rebasing and benchmarking. Consequently, the timing for the publication of the benchmarked estimates will need to be rescheduled.

Second, the timing of introducing the benchmarked estimates should always be carefully considered. It is not good practice to make major data revisions during a period of abnormality as we have now. Stats SA depends on a range of monthly indicators to estimate the quarterly GDP. In the event that the information for the months of March through to June/July is volatile and does not represent a normal state of affairs, this may lead to questionable historic estimates. In addition, the latest quarterly GDP estimates based on short-term indicators may be revised substantially when comprehensive information from annual surveys is available later in the year. Annual surveys are more accurate than the high-frequency series given the larger sample size, the comprehensive nature of the questionnaire, and their use of financial statements rather than monthly key variables.

Therefore the option of postponing this comprehensive review of the GDP time-series will be considered for the reasons outlined above.

Users of economic statistics will be kept informed as the situation changes and as we are able to confirm dates for the publication of official statistics and other notes that are being developed.

6. Revised publication schedule for economic statistics

Please note: in some cases data for two months are scheduled for publication in one release, for example manufacturing for March and April in one release on 11 June.