On Thursday 6 March 2023 the National Statistician, Professor Sir Ian Diamond, gave a speech to the Royal Society on The Future of Economic Statistics. The full text of his speech is published below.
The Future of Economic Statistics
Colleagues, thank you for coming and thanks in particular to Sir Adrian Smith and the Royal Society for hosting us this evening.
Earlier this year I found myself on the panel of the BBC’s Any Questions when they visited Newport. Although Newport’s industrial base has declined over the past 40 years, it does hold the powerhouse of the UK’s digital economy in the form of the ONS (Office for National Statistics). And maybe it is an indicator of our changing economy that the position of ‘local employer’ that would have previously been occupied by the manager of a local steel works or manufacturer, was instead taken by someone from the world of data.
One of the questions related to data, specifically GDP: we had just produced the second quarter estimate and naturally people were worried about ‘recession’. However, the Assembly Member Delyth Jewell took the discussion in a really interesting direction.
She asked about the meaning of our economic statistics, what are we actually trying to measure, and invoked Bobbie Kennedy’s speech at the University of Kansas in 1968, where he famously declared that GDP ‘measures everything, except that which makes life worthwhile’.
And that is really my motivation for this evening’s talk. I think we are in the middle of the most exciting period in the development of economic statistics, certainly since Kuznets and Stone began laying the foundations during and after World War 2. Our methods are becoming sharper and more granular, better able to capture the changing nature of our society and economy.
We might not yet be able to measure the quality of our poetry or strength of our marriages, but across whole ranges of our lived experience, from the environment through to human happiness and wellbeing, we are closer than ever to meeting Kennedy’s challenge.
It is worth spending some time setting out how we have come to this position.
Britain has one of the longest histories of economic statistics in the world; a real treasure trove for statisticians and economic historians. Data for the UK national debt extends back to 1691, while government revenue and expenditure statements have been published annually since 1900. But the full suite of economic statistics did not emerge until after World War II.
Over a century of public sector borrowing data
Of course, Kennedy’s critique of GDP and the wider national accounts was slightly unfair. We have known since the very beginning that GDP measured output not wellbeing; Kuznets himself, as early as 1934, recognised that the welfare of a nation could ‘scarcely be inferred from a measure of GDP’.
How to pay for the war
That said, the rise of GDP is both a towering intellectual achievement and also the product of wartime needs. It was the best method to identify the scale and nature of production at the time, whether that was planes, tanks and munitions for the war effort, or food, fuel and basic necessities for the civilian sector, up to 1945 when we really needed to know that. And, when you look back, it is astonishing how we move from Keynes’ calculations of national income in How to Pay for the War in 1940 and here is how it looked.
The 1950s and 1960s...rapid progress in economic statistics.
To move through to Frank Campion’s first blue book in 1952; a period that included the first estimates of national income, first retail prices index, a monthly index of production, a first balance of payments and census of distribution. An amazing display of rigour and creativity; with no calculators, no spreadsheets, no databases and no thumping great manuals from international organisations telling you how to do it.
And it didn’t stop there. Post war economic management and the structural changes in the economy led not just to a deepening of the work on national accounts but to a broader widening of the scope of economic statistics. By 1955 GDP is produced quarterly.
The 1950s boost in housing construction also generated new demands for building statistics.
The ‘balance of payments’ is produced quarterly in 1959 and was to plague successive governments who struggled with maintaining sterling’s position.
By the mid 1960s, we are producing measures of public sector expenditure, and the white heat of technology prompts a greater demand for structural economic statistics through the business statistics office. By the 1970s we are responding to demands for greater granularity, and in 1972 some of you will recall that this marked the first ‘early estimate’ of quarterly GDP.
Economic statistics in 1972
1973 brought regional GDP and more sectoral breakdowns of outputs such as investment and the capital stock In the 1980s, we have big improvements in trade stats, and in 1987 we produce the first computer readable Blue Book - and what a bargain it was!
The first computer readable Blue Book
In the 90s, we have the Interdepartmental Business Register (IDBR), we produce the Consumer Prices Index (CPI) for the first time and we consolidate structural surveys into a single annual business enquiry. And really importantly, in 1996 we created the first environmental accounts; trying to bring a sense of the wider resource framework into the economic statistics system.
But from the 2000s, things start to slow down. There are improvements, but the period begins to be characterised by structural changes in the Office itself – the creation of the Authority, independence and the move to Newport - rather than big improvements in outputs. There are some incredibly important reports – Alsopp and Atkinson were serious pieces of work, looking at the challenges and opportunities facing economic statisticians and we also developed the monthly Index of Services – but rather than the rapid deployment of new and better economic statistics, we were treading water.
That is a problem. A dynamic, evolving economy needs a dynamic evolving statistical office: the economy changes so the statistics capturing that economy also have to change. If they don’t change they cease to be valid. They cannot perform their purpose of showing policymakers what is happening to the economy, and thereby guide wise policy development.
And the 2000s were a really bad time for that dynamism to decay. If one looks at the Blue Book delivered in 2021, which contained three core changes, double deflation, first considered in 1974, the financial services survey, which brought us up to date in that sector with the ‘Big Bang’ reforms of 1986, and telecommunications prices, where the 3G auctions of 2000 had raised £22.5bn for the Exchequer – possibly a sign of dramatic market change - we can see the cumulative delay to addressing these three issues alone was 103 years.
At the precise moment from around 2005 to 2015, when the digital economy takes hold, which throws up fascinating measurement challenges – such as how do you measure ‘free goods’ – through to the explosion of new economic data, the ONS struggled to keep pace. And that is just on ‘classic’ economic statistics. When you add to that the path breaking work from people like Lord Layard on wellbeing and happiness economics- work which basically didn’t even exist in the 1990s - and you can that as a society we had become complacent and largely stopped investing in economic statistics.
This is the world that Professor Sir Charlie Bean revealed in his reports in 2016.
The Bean Review
The opportunities for economic statistics had never been greater, but the ONS had never been so badly positioned to exploit those changes. And it was why he called for us to go ‘back to the future’ to capture that pioneering spirit of Keynes, Stone and Campion and set up a new era for economic statistics in the UK.
The pandemic pivot
The Bean Review set a range of workstreams underway: admin and commercial big data, statistics modernisation, data science and closer collaboration with the academic community among them. Bean and Treasury provided the initial impetus and investment to give staff the opportunity to begin to think on a larger canvas once again, and to lay down fundamental changes to our ways of working so when COVID-19 hit the UK in March 2020 we were better placed than many to respond.
In the same way as the War provided the animating energy for the pioneers of economic statistics, the pandemic providing a huge galvanising force for us to innovate. It gave us purpose and the insatiable demand from the public reminded us all of the vital role that data plays in our policymaking.
Card data during the pandemic
Our financial transactions and Data Science Campus teams supplied daily and weekly data feeds to government on the impact of the pandemic. This included web scraping Google Mobility data, so we could see the impact of lockdowns, and publishing the code on GitHub so that other countries could use it.
Near to real time surveys
We rolled out regular new online business and household surveys at breakneck speed which enabled us to work across government and ask a wide range of questions. Remember in the March 2020 GDP was falling by 1% a day – we could trace the impacts of that through our new data.
Our monthly real time indicators bulletin, populated by a handful of measures, was switched to a weekly release, adding dozens of indicators over time.
Drawing on those new data sources, we beefed up our analysis to provide more timely, relevant and granular on the social and economic impact of the pandemic.
COVID-19 Infection Survey
And of course, we rolled out the COVID Infection Survey, widely considered to be the ‘gold standard’ in providing timely and statistically reliable estimates. While not an economic statistic itself, its impact cannot be under-estimated.
None of this was easy.
Within three weeks, ONS moved from an organisation that was predominately site based, to one that supported remote working for over 5,400 staff.
We talked to dozens of companies who turned up offering their data to try and help out and, where it was useful, negotiating data sharing agreements with them and found ways to securely but quickly ingest their data.
And we had to free our capacity to produce all of these new surveys, indicators and analysis by streamlining our existing work and redeploying our staff.
The ONS played a useful role in helping to provide timely data and insight at a time of national crisis. We weren’t simply ‘the palace of statistics’ anymore.
As the pandemic started to ease and some degree of normality returned, many national statistics organisations started to wind up their new surveys, indicators and analysis. Crisis over?
But the clear message we at the ONS received, from across government and our many wider users, was that we should retain many of those new methods and data because of their continuing value and relevance.
Government departments, for example, have found our business and household surveys invaluable in informing the current cost of living crisis.
Our financial transactions analysis, based on UK card payments, have provided insights as diverse as the impact of Eat Out to Help Out, fuel shortages, inbound tourism and recent industrial disputes.
And we are now working more closely with government departments and the devolved administrations, which helps ensure our work directly contributes to policy making across the country in a way it had not before.
While we were dealing with the pandemic and its aftermath, we were also working to drive up quality.
As I mentioned, the ONS have an ambitious multi-year programme to modernise all our core economic statistics: the labour market, consumer prices, public sector finances, research and development, trade and investment amongst them.
This work has already seen real improvements. Our trade statistics, for example, have increased hugely in both quality and granularity with a 100-fold increase in series published since the start of our transformation project, helping to provide Global Britain with the accurate and detailed statistics it needs as we navigate our post-EU future.
We are only one of two NSIs (national statistics institutes) in the world to produce monthly GDP, and we have made major improvements to our measurement of the digital economy.
But there is a challenge here. As I said earlier, when the economy changes, our statistics need to change too. If they don’t, they risk becoming unmoored from the economic phenomena they’re meant to capture. But when we reform the data, unless we can do this at pace, when they become better attached to the underlying economic reality, we run the risk of major revisions and potentially damaging confidence and public trust.
We had a live example of this last year. We sit here in the Royal Society of London for Improving Natural Knowledge; the oldest continuously existing scientific society in the world. So I imagine you are all familiar with our research and development statistics? And of course last year, after a long period of peer review and consultation we revised those statistics by 40%. And I know that caused some raised eyebrows from some of you in this audience and by looking at this chart you can see why.
No Statistical Institute wants to revise a figure by 40%. So, I want to be transparent with you about why we did it, and also to be clear that as we go forward, as our own natural knowledge of the economy improves, then some of the data will change.
The core issue with our R&D data was that the fundamental survey had not been revised for 30 years. So it was based on a view of the economy where a small number of large firms did R&D. A world of no SMEs, no spin outs. Just the big research labs of BT or GEC. It was an accurate survey of a world that no longer existed.
And that meant we had to update our estimates, and that inevitably revised the data. That wasn’t pain free, but it was the right thing to do. I do not believe that we will have to make another change which is that significant, but as we implement new approaches, as we find new data, and as we implement new methods, the statistics will change. But my commitment to you and all our users is that we will use the best scientific methods and complete transparency to enable you to retain confidence in us as we make those changes.
Transforming economic statistics doesn’t just involve overhauling outdated surveys. It also means automating functions and more innovative use of data.
It also means making greater use of administrative and commercial data and fewer and smaller surveys – resulting in more accurate statistics and a lower burden on business.
Range of tools to support Cost of Living debate
For me, improving quality also means that people can see themselves in our data. That was our motivation in responding to the cost of living problems in producing our inflation calculator, our least cost items analysis and our household cost index.
It is also the aim behind our work on 'subnational' statistics and analysis: neighbourhood, local and regional data to help individuals and communities understand their local area, and the online app we developed so firms could enter their basic turnover and workforce data to identify how productive they are compared to others in their sector.
In December 2021, we launched the cross-government subnational data strategy to drive improvements in local data across the whole statistical system. We also published a data visualisation tool – the subnational indicators explorer – that brings together over 40 data sets aligned to the levelling up missions from across government into a single place and was first published alongside the white paper. This is now being developed into a more fully developed Explore Subnational Statistics service that builds on the tools we have developed for exploring the Census data.
We are now producing very granular statistics, such as estimates of gross value added, down to the level of 400 to 1,200 households, so that people can use these to build their own user-defined areas that mean something to them. We are innovating with newly available methods and data, for example, to understand card spending in hyperlocal areas, or consider how far away can people live and commute to a certain area within a given time on public transport at different times of the day.
We continue to develop analysis of towns and high streets, considering different aspects of local economies, such as how employment compares in towns versus in out-of-town locations, and what the night-time economy is like at levels of geography that are meaningful to people.
And ONS Local is our new analytical advisory service, with staff based across every nation and every region of the UK. It aims to ensure local leaders and the teams that support them, have access to the data, statistics, and analysis they need to take evidence-based decisions. By having eyes and ears in-area, we can build our understanding of how our data, combined with local intelligence can really help local policy-making, as well as influence our priorities accordingly too.
Building on our early successes
What the pandemic showed is the real benefit of taking a more nimble and radical approach to our mission.
We need to retain that drive – recognising that we have only begun to exploit the many opportunities that new data and methods can offer.
We also need to keep that wider engagement and collaboration with our data suppliers, policy makers and researchers. The ONS can't afford to be insular.
That is why I was delighted to see our recent data partnership with Visa, who have agreed to provide us with aggregated and anonymised data to better understand UK consumer spending patterns and improve the quality of our statistics. Here is just one example how these data can provide great real time insight, where we look at the demand for fuel, which was a very hot topic at certain points over the last year or so. Using Visa data to estimate real-time fuel demand
So, what about the future?
With economies and societies changing so quickly and with technological advancements continuing at an astonishing pace, it is fair to say that what the future holds for economic statistics is difficult to predict. However, that doesn't mean I'm not going to try; here are a few thoughts.
It seems clear to me that a more holistic approach to measurement will be at the heart of economic statistics as we move forward and arguably it is already here.
Indeed arguably it has always been with us. The fact that many types of data aren't included in the System of National Accounts does not imply the original drafters did not consider them important; they just didn't view them as best served in a framework denominated in money terms. That's why, after twenty years of working on SNA, culminating with the publication of the 1968 version, Richard Stone moved directly to working on a System of Demographic and Social Statistics, focussed on people, rather than money, delivering this to the UN in 1974.
Even back then users and statisticians could see the need for a sophisticated understanding of economic, environmental and social progress in a far more informed and helpful way than can be captured by the latest percentage point move in the economy alone, but they also came to see that large parts of the world were not ready for the computational complexity and data-hungry systems required. Countries such as the UK developed these methods to meet domestic needs, but today that demand is growing alongside the realisation that the data are now available to be far more ambitious. So, what have our users told us is important to them as we move forward?
What is 'beyond GDP' and will it really matter?
Well, the list is challenging and expectations are high but I wouldn't want it any other way. Here are some of the key priorities that I am hearing loudly and clearly:
estimates of national well-being, recognising the diversity of factors which can affect how people view their lives, both in the here and now, but also considering distributional and sustainability factors, permitting comparison either with different parts of the UK, or with the future
a holistic view of the 'environment', 'society' and the 'economy' and most importantly the trade-offs between these three domains.
supporting this, a more robust, inclusive, analysis of where value comes from in our society. This means data on 'The Missing Capitals', an umbrella description for:
- intangible capital: these are assets with no physical form so includes intellectual property, brand recognition, designs, and organisational capital.
- human capital: the value of the skills, knowledge and experience of people created by education and on-the-job training.
- natural capital: the economic value resources such as plants, animals, air, water, soil and minerals providing a service to humankind, and
- social capital: the value derived from positive connections between people.
These are seen as increasingly vital for understanding our modern and rapidly changing economy, society and environment.
But do these chime with you? If not do let us know your thoughts. I want to generate a debate on where we are heading to make sure our destination is right.
Are we waiting for the future?
Absolutely not. We are already working hard to meet the needs and expectations of our users. The ONS has long been committed to measuring the progress of the UK, and we already provide a wide range of data, statistics and insights that do exactly this and have radical and ambitious plans to do so much more.
Measuring national well-being
Under the banner of the 'Quality of life in the UK' we have already begun publishing fresh estimates of national well-being on the ten areas that matter most to people's well-being as set out in our Measures of National Well-being framework. These cover personal well-being, our relationships, health, what we do, where we live, personal finance, economy, education and skills, governance, and the environment.
We are not resting on our laurels, though, as we will in spring 2023 publish the outcome of our review of 'Measuring National Well-being', which will help shape the way forward for us on this agenda.
In parallel, we will publish estimates during the second quarter of 2023 of Gross Inclusive Income and Net Inclusive Income; these will integrate existing economic, social and environmental data to provide exactly the holistic insight requested by users. Methodologically equivalent to GDP but covering the flows of benefits received by citizens from human and natural capital, this will allow us for the first time to demonstrate in really simple terms the trade-offs between economic growth and environmental damage. Imagine being able to say to a Minister something like 'GDP went up 0.3%, but because of offsetting environmental damage Gross Inclusive Income only went up 0.1%'. Being able to put data into a wider context is the key benefit of having a wide-ranging national statistics institute.
The missing capitals
To be able to produce this analysis we rely on work from leading thinkers and academics in this field via the Economic Statistics of Excellence and other government departments to develop estimates right across the missing capitals. In 2023, our plans already include new estimates for natural capital, human capital and a full satellite account on UK Households, which will provide detailed estimates of the value of the work done by households and how it impacts the economy. And I think when you look at the figures involved on this slide and the great quote, it just reinforces the importance. Also in 2023 we will be consulting on the transformation of our population and migration statistics, where our strategic aim is to produce more frequent, timely and inclusive statistics about the population and its characteristics. Administrative data will be at the heart of this system.
Climate Change Portal
However, I want to return to the importance of climate change. We have moved quickly to step up to the challenge of better measurement in this field. We have launched a brilliant cross-government UK Climate Change Statistics Portal that brings together climate change related data and statistics to improve coherence, accessibility and comparability, and we already have plans in 2023 to produce new and better estimates of greenhouse gas emissions and air pollution.
We have also recently started to publish a quarterly Climate Change Insights publication that brings together statistics to provide timely climate change insights. Where possible, we set these in the context of the UK's environmental and natural capital accounts, enabling direct comparison with GDP.
And that brings us to the international update of the System of National Accounts, which is aiming for completion in 2025. Implementing the new guidance will be a long and arduous process requiring significant research and feasibility testing, but it is clear to me that the most important revision is the aim to include depletion of natural resources within the national accounts as what is called a 'cost of production'.
Whilst ambitious, it is important to note two key issues with this approach. Firstly, unlike the rest of the accounts where we see the benefits we receive from the varied assets as well as their deterioration, this will not be a symmetric treatment. We won't include the value of the local park as you receive it, only the loss of it if it is destroyed.
Secondly, the current proposal will conform with current national accounts identities and practices. As such the natural resources lost will only be visible in Net Domestic Product, not Gross Domestic Product. This means for this change to have any influence on policy-makers institutionally ONS will need to work to support any number of partners transition from using GDP to using NDP in their models.
At ONS we have been carefully considering this proposal, and the risk that it may appear we are kicking this issue into the long grass. There is a real danger that the international statistical community ends up saying that we prefer to prioritise established national accounts methods more than meeting the policy need, even when there is no more pressing issue than climate change.
There is an alternative. We all know that GDP is a gross concept by design, but we also know what our predecessors like Keynes would say. When faced with the reality of the Second World War, Keynes saw the importance of unpaid household work but did not include it in what became GDP, because what mattered was industrial production. He did not accept that the perfect should be the enemy of the good enough.
This may be another moment where we are forced to set aside the perfection of our methods and permit an exception to our normal practice, by accepting the depletion of natural resources within our definition of GDP. This would be a large step, and it is one we would appreciate user views on. Is including the environment in NDP but not GDP enough, or should one break with tradition and accept the depletion of a very tightly specified set of resources into GDP?
Communication – what does the future hold?
Someone once said to me that you can have the best data sources and most sophisticated methods available but if you can't explain and interpret the results in a clear way or make them easily accessible, you have failed.
This may be a touch harsh but there is a lot of truth in this thinking, and this is why alongside our intellectual thinking and delivery on measurement, we must equally invest in how we engage on it.
I've been accused of perhaps oversimplifying this, but I see a threefold approach as we move forward:
Insight: providing intelligent interpretation of the results.
Clarity and Transparency: describing our methods in a way so they are clearly understood and their relative strengths and limitations are known. To me a vital responsibility we have to our users.
Accessibility: the more discussions I have, the more I am completely convinced that high quality, easy to use Application Programming Interfaces (APIs) are what we need to provide to our users to access the data that they need.
So, there we have it. Again, I think we have made progress in recent times on all three but we, again, have more to do. The institution that produced a computer readable Blue Book 35 years ago, should be able to have all of its data easily available through APIs in the next five years.
What does a future successful National Statistical Institute look like?
So, bringing it all together what does this means for National Statistical Institutes for the future? For me, a successful one will have these following characteristics:
a centre of excellence and destination of choice: for data, statistics, and analysis.
independent of government. Something I cherish in the UK and is critical to user trust.
convenor and enabler: an institution that is looked to, to bring organisations, people and data together for the public good.
educator: supporting future generations in understanding all things in the data and statistics world. ONS truly is a great place to be; it would be brilliant if we continue to encourage more and more people to be part of it.
I began this talk with a quote from Bobbie Kennedy, and I hope I have illustrated that today's wider economic statistics are much closer to the aspirations and challenges that Kennedy set out in 1968. But when I think about a guiding principle for the work of economic statisticians now and in the future, we need to go even further back to one of the founders of economics itself, Alfred Marshall. For him economics was really about the 'study of mankind in the ordinary business of life'.
Economics is not a dry discipline concerned with market transactions, but about the welfare of our people. And as a group of economic statisticians we should always be striving to understand that 'ordinary business of life' and use the best tools we have to shine a spotlight upon it.
Sometimes that might need us to be willing to break the mould. We have to be willing to have that debate if we are to remain true to our mission of measuring the right things well enough, rather than the wrong things perfectly. Thank you.