CSO assessment of Occupations with Potential Exposure to COVID-19

Guest post by Reamonn Lydon (Central Bank).

[Disclaimer: this blog post represents my personal views and not those of the Central Bank of Ireland or the European System of Central Banks]

The CSO has just released an experimental analysis of Occupations with Potential Exposure to COVID-19. This is useful data for anyone who wants to understand how the Covid-19 shock interacts with the structure and composition of employment.  It provides important information on which sorts of occupations and workers have been most directly affected by the restrictions to limit the spread of the virus.

Using O*NET data on the task-related content of four-digit occupations, the CSO construct an Proximity index and an Exposure to diseases index.  Here, I focus on the proximity index, although a similar analysis of the Exposure index would also be of interest.

Quoting from the background notes:

In the O*NET data “Respondents score their job on a scale of one to five where, for proximity, one indicates that the respondent does not work near other people (beyond 100 feet) while five indicates that they are very close to others (near touching) … The data is harmonised on a scale ranging from 0 to 100 by using the following equation: S = ((O-L)/(H-L)) * 100 where S is the standardised score, O the original rating score between one and five, L the lowest possible score (one) and H the highest possible score (five). Under this new classification, the standardised physical proximity measure is defined by:

0 –     I do not work near other people (beyond 100 ft.)

25 –   I work with others but not closely (for example, private office)

50 –   Slightly close (for example, shared office)

75 –   Moderately close (at arm’s length)

100 – Very close (near touching)”

The CSO has constructed a proximity score for 296 four-digit SOC10 occupations. Crucially, it then maps these to total employment, percentage female, over-55 and non-Irish using Census 2016.

Using employment weights, the median proximity score for all workers is 57.6; the mean is 61.8. The four digit occupation at the median is Sales related occupations not elsewhere classified. The lowest scoring occupation (least proximate) is Artists (21.5); the highest scoring is Paramedics (97).

The chart below shows the cumulative share of employment (y-axis) by proximity score (x-axis) for the characteristics provided by the CSO. The variation across charactistics is striking: female workers are more likely to be in ‘lower proximity jobs’, almost 60 per cent are below the median score for all workers. It is hard to pin-point a single occupation that contributes to this result for females, but a relatively higher concentration in occupations like Chartered and certified accountants, Cleaners and domestics and Administrative Occupations do stand out.  By contrast, male, younger and non-Irish workers are all more likely to be in high-proximity jobs, with just 40 per cent below the median.  The relatively higher share of younger workers on the Pandemic Unemployment Payment (PUP) – 41 per cent of under-25s are on PUP currently, compared with 21 per cent of over-25s – tallies with the observation that more of these workers tend to be in higher proximity occupations, and therefore more impacted by the Covid-related restrictions.

Chart 1: Cumulative share of employment by proximity score.
Source: Own calculations using CSO 2020.

For those groups with a greater concentration in high-proximity jobs – that is, male, younger and non-Irish – there is a step-jump around 70. In terms of the occupations arround this jump, for males it includes sports and leisure activities, skilled trades, constrction and protective services. For non-Irish nationals, who make up around 15% of employment (in the 2016 data, it is closer to 20% now) it is a broadly similar set of jobs, but also including a range of food services occupations.  

Finally, the CSO has also published the median annual earnings by occupation. Chart 2 shows the average of median annual earnings by occupation for each quartile of the proximty score distribution (weighted by employment). Higher proximity occupations tend to lower paid.  For example, in the top 25 per cent of jobs by proximity score (which also happens to be a score at 75 or above), the average of earnings per occupation is around €33,500 (in 2016). The average for the bottom 25 per cent occupations (a proximity score of around 49) is €42,300.  When we control for all characteristics such as female, share of over-55s and non-Irish by population, we find that going from the least proximate quartile score (49) to the most proximate quartile score (75) is associated with earnings being around 20 per cent lower on average.

Chart 2: Earnings by proximity score
Source: Own calculations using CSO 2020.


Information on the task-related content of occupation is vital for understanding which sort of jobs are affected by the social distancing restrictions put in place to fight the Covid pandemic.  Similar work in Adrjan and Lydon (2020) shows how countries with a higher concentration of ‘high-proximty’ employment experienced a larger negative shock to labour demand when the crisis first hit. This includes Ireland. Analysing the occupational breakdown from the CSO highlights that younger, male and non-Irish workers are more concentrated in ‘high proximity roles’. These roles are also lower paid on average. This provides crucial insight into who is most affected by the Covid shock, and what sort of policies might be put in place to help certain groups of workers.

Measuring national income in the time of COVID

National accounts are useful.  Yes, they have their limitations, and, particularly in the case of Ireland, can be subject to distortions but they are useful.

One of those uses is measuring changes in living standards.  If the growth of the inflation-adjusted measure of national income exceeds the growth rate of the population then it is likely that living standards are rising, at least on average.  This is usually taken as the real growth rate of per capita GDP (or another variant).

For Ireland, this averaged around one per cent per annum for the first three decades post independence, it rose by an average of three per cent per annum over the next thirty years and has averaged around five per cent per annum in the period since the late 1980s which is where it was before the current crisis hit.  These are useful summaries of our economic performance, though as is well known, such long-term averages do belie some significant volatility that Irish growth rates have exhibited.

2020 seems set to add to that volatility but let’s consider two things that are likely to muddy the link between the change in real per capita national income (as measured by, say,  GNI*), and the impact of the crisis on living standards:

  • Food consumption and the exclusion of domestic household production from national income;
  • Education and the cost-based approach to including public production in national income.

Food Consumption

Restaurants have been closed for dining in since March.  The CSO’s Monthly Services Index shows that the turnover value of services in Restaurants, Event Catering and Other Food Service Activities was over 50 per cent lower in April than in the same month last year.  The contribution to value added and national income from restaurants will be significantly lower this year.

However, this does not mean we are eating less or even consuming fewer food-related services.  The composition has changed.  The CSO’s Retail Sales Index shows that our retail purchases from Food Businesses were 17 per cent higher in April than in the same month last year.  The food we are not consuming in restaurants and other outlets has been replaced by food we are buying in shops.

In the national accounts, both the food and labour inputs are counted when measuring the value added of restaurants.  For food we buy in the shops the domestic labour input used to turn that food into a meal is omitted from national income, but it still contributes to our living standards.

There’s no doubt there’s more to restaurants than the food we eat and the cooking and cleaning services provided to us.  That is why we are willing to pay more for dining in.  But we are still eating the meal we would have had in the restaurant or cafeteria so someone is still doing the cooking and someone is still doing the cleaning.  It still adds to our living standards, it’s just that it has switched from market production to household production.

The drop in living standards implied by the fall in value added from restaurants in the national accounts won’t be as large as the figures suggest.  We have been forced to move to something that does not have its value added included in the national accounts (nor generate as much Value Added Tax for the government which is also counted as value added when measured at market prices.)

And, separately, the employees who would have been paid from that lost value added have had a large part of their income replaced with government transfers.

Public Education 

For market-provided services the value added is essentially the value of the output produced less than cost of intermediate consumption. 

The value of market output is estimated using the prices people for it.  After intermediate consumption has been subtracted from total revenue, value added is divided between labour through compensation of employees, government through taxes on products, and capital through gross operating surplus.  Net operating surplus remains after a deduction for the consumption of fixed capital (depreciation).  The additional value added that goes to users above the price paid (consumer surplus) is not measured. 

Still, value added is a useful concept and represents a large share of the living standards and welfare benefits of the goods and services we produce and consume in market settings.

This does not hold for publicly-produced services such as education.  These are paid for from general taxation.  We do not have prices and revenues to provide an estimate of the value people place on these services (nor how much they would be willing to buy).  Market prices might be absent but they do contribute to living standards.

The value added for public services in national accounts is essentially the sum of compensation of employees and depreciation, that is, it is the cost not the benefit that is included. 

The value added of education is simply put at the pay bill for teachers and the cost of maintained school buildings.  No benefit above that is included in national income aggregates.

Schools have been closed since the middle of March.  Just like restaurants there has been a switch to domestic production.  Yes, some online supports have been provided but the value of this is unquestionably lower (just as we are only willing to pay lower prices for takeaway meals).  The shift to home-schooling has had a huge impact on living standards.

However, the value added of publicly-provided education services will be largely unaffected.  The fact that the school children aren’t in school doesn’t matter for national accounts; all that matters is that teachers get paid.


The provisional Quarterly National Accounts for Q1 2020 show that constant price gross value added in Distribution, Transport, Hotels and Restaurants was down 10 per cent compared to the first quarter of 2020.  This reflects the forced closure of most of these services towards the end of the quarter.

On the other hand the gross value added in Public Admin, Education and Health was up four per cent compared to the same period a year ago.  This is despite the fact that schools were closed from the 12th of March.

This isn’t necessarily an argument to change the way national accounts are compiled.  Should household production be included in national income? Maybe.  Should the added value of public services be more than pay and depreciation costs? Maybe. For the time being we’ll be satisfied with an understanding of what the figures as currently compiled actually mean.

The drop in value added from restaurants doesn’t mean that we are not eating.  The stability in value added from education doesn’t mean that our kids are being taught.  National accounts are useful and the changes in the aggregates can be a useful proxy for changes in living standards. But not always.

Animated Irish Fiscal Data

Jacopo Bedogni and Darren Lawlor from the Parliamentary Budget Office put this series of visualisations together. Slide 19 in particular is a jaw-dropper. Well done both.

Made with Flourish

131st Barrington Medal, 2020/2021

Call for entrants

The Barrington Medal is awarded annually by the Council of the Statistical and Social Inquiry Society of Ireland, under the auspices of the Barrington Trust (founded in 1836 by the bequest of John Barrington). The award, which includes both a silver medal and €1,000, is intended to recognise a promising new researcher in the economic and social sciences in Ireland. This will be the 172nd anniversary of the lecture series and the recipient will be the 131st Barrington Lecturer. Recipients in the past 35 years include:

Deirdre McHugh Don Thornhill George Lee Alan Joyce Daniel McCoy
Brian Lucey Kevin H. O’Rourke Siobhan Lucey Mary Walsh Philip Lane
Aidan Kane Donal O’Neill Peter Clinch Colm Harmon Ronnie O’Toole
Cathal O’Donoghue Paul McNicholas Mary Keeney Liam Delaney Martina Lawless
Cal Muckley Orla Doyle Yvonne McCarthy Ronan Lyons Mark McGovern
Rebecca Stuart Karina Doorley Daragh Clancy Barra Roantree  

The lecture should be based on a paper of not more than 7,500 words addressing a topic of relevance to economic or social policy and of current interest in Ireland. In treating the issue of economic or social policy, the paper may either report the findings of a statistical research study dealing with some aspect of the problem or deal with the underlying theoretical considerations involved, or preferably combine these two approaches. It should be written in a manner that makes it accessible to non-specialists in the area. More technical material may be included in an appendix. The paper is published in the Journal of the Society, so it should not have been published before, nor should it be published subsequently without the prior consent of the Council of the Statistical and Social Inquiry Society of Ireland. Candidates, who at the time of their submission must be not more than 35 years of age, should at least submit a detailed abstract of approximately 1,000 words on the proposed lecture, with preference being given to full papers. A short CV and the name of a proposer who is familiar with their work should also be submitted.

Entries will be accepted from 1st June to 31st August 2020 and should be sent to the Honorary Secretaries of the Society, via email, using the email address secretary@ssisi.ie.