This Forecast in-depth page has been updated with information available at the time of the March 2023 Economic and fiscal outlook.

GDP growth is an important driver of trends in the overall budget deficit and the path of public sector debt. Over a 5-year forecast horizon, GDP growth is largely determined by the growth rate of potential output, so it is necessary – implicitly or explicitly – to make judgements about the economy’s underlying growth potential. We also need to make a judgement about the margin by which economic activity currently exceeds or falls short of its potential or sustainable level (the ‘output gap’). This is because we generally assume that the Monetary Policy Committee at the Bank of England sets monetary policy so that activity will return to its underlying potential level and therefore the size of the output gap will determine whether GDP is expected to grow faster or slower than potential output over the forecast period. We set out those judgements explicitly, in order to be transparent and because they play a role in our fiscal forecasts. Neither potential output nor the gap between it and actual output are directly observable, so estimating them is necessarily challenging. Moreover, neither can be verified, even with the benefit of hindsight.

In economic terms, potential output is the starting point for thinking about how fast an economy can grow over the medium term. To that, one would add or subtract an estimate of the output gap to find the level of actual GDP at present and over the forecast period. In terms of our forecast process, the ordering is reversed. We start the forecast knowing the latest ONS data on the level of GDP and with survey and other indicators (e.g. about companies’ capacity utilisation or recruitment difficulties) that allow us to estimate the size of the output gap. From those two pieces of information, we infer the current level of potential output and its components, from where we apply our judgements about the rate of potential output growth.

Estimating the output gap allows us to judge the size of the structural budget deficit – in other words, the deficit that would be observed if the economy were operating at its sustainable level. When the economy is running below potential, part of the headline deficit will be cyclical (and would therefore be expected to diminish as above-trend growth boosts revenues and reduces spending). If the economy is running above potential, the structural deficit will be larger than the headline deficit as a period of below-trend growth would be expected to depress receipts and push up spending as the output gap returns to zero.

Our estimates of potential output and the output gap are based on whole-economy output. Prior to the November 2022 forecast, we excluded the small and volatile oil and gas sector and then added on a forecast for oil and gas production to complete our GDP forecast. We have stopped stripping out North Sea output from our potential output figures, given the waning influence of North Sea production on output levels in recent decades.

Further information on the concept of potential output and our approach to forecasting it is set out in Briefing paper No. 8: Forecasting potential output – the supply side of the economy.

  The output gap

We normally begin the economy forecast process by estimating the size of the ‘output gap’ – the difference between the economy’s current level of activity (as estimated by the ONS) and the ‘potential’ level consistent with stable inflation in the long term (inferred from the output gap and the level of actual activity). A negative output gap is associated with lower rates of capital and labour utilisation, implying some spare capacity in the economy; a positive output gap is associated with higher rates of resource utilisation and, if sufficiently positive, evidence of ‘overheating’ which would put upward pressure on wage growth and inflation.

We cannot measure the supply potential of the economy directly, so we use various techniques to estimate it indirectly, including statistical filters, cyclical indicators and production functions. The techniques used to construct these estimates are refined from forecast to forecast, so the precise variables and parameters may vary over time. Further information on our approach is set out in Briefing paper No. 2: Estimating the output gap  and Working paper No. 5: Output gap measurement: judgement and uncertainty. Since our December 2014 forecast, we have generally used estimates of the output gap implied by the nine different techniques described in Working Paper No.5 to inform our judgement.

Statistical filters attempt to decompose time-series variables into a cyclical component and an underlying trend, which is typically assumed to evolve relatively smoothly. There are two main types of statistical filters that we use:

  • Univariate filters produce a series for trend output based on the observed path of actual output alone. The filter in effect provides a rule by which fluctuations in observed output are related to changes in potential output, where the output gap is given by the difference between the two. We place least weight on these measures as the estimate of potential output for the most recent data can be overly influenced by recent movements in actual output (the so-called ‘end-point problem’) and can be revised substantially as new output data become available.

 

  • Multivariate filters augment the output data with other information reflecting the economy’s cyclical position. They derive estimates of the output gap from a set of conditioning relationships such as the relationship between inflation and the output gap (the Phillips curve) and the relationship between the output gap and disequilibrium unemployment (Okun’s law). We typically place more weight on these because of the wider pool of information they bring to bear.

Cyclical indicator approaches use a range of indicators of the position of the economy relative to trend, such as survey measures of spare capacity and recruitment difficulties, and combine them to provide an aggregate indicator of the output gap. There are two main types of cyclical indicators that we use:

  • Aggregate composite estimates use the responses given by firms to certain survey questions about capacity utilisation and recruitment difficulties. These are combined by taking a weighted average of the survey indicators, with the weights on each indicator based on factor incomes (i.e. weighting recruitment indicators by labour income and capacity indicators by profits) and sector shares (i.e. manufacturing and services shares of output or income) from the National Accounts.
  • Principal components analysis estimates use survey indicators as well as ONS data, such as average weekly earnings. These are combined using a statistical technique that seeks to identify the common determinant of a number of variables. By choosing only indicators that we expect to relate to the output gap, we assume that the common determinant extracted by the principal components analysis is a proxy for the output gap.

We also estimate the output gap via a production function – an equation that relates inputs to the production process (e.g. workers and the equipment that they use) to output. The level of potential output is assumed to be a function of labour supply, the capital stock and the maximum efficiency with which they can be combined (total factor productivity or ‘TFP’). Univariate statistical filters are used to estimate trend levels for the inputs, so they are subject to similar issues to the simple filtering techniques set out above. They may also imply different judgements to the ones we take in the rest of the forecast.

In practice, every method has its limitations and no approach avoids the application of judgement. The Budget Responsibility Committee (BRC) therefore considers the swathe of estimates produced by the nine techniques alongside a broad range of evidence when reaching a judgement on the initial output gap and the associated level of potential output. We sense-check our judgement by considering the profile it implies in the latest quarter against the profile of output growth and the unemployment rate in the same period.

  Back to top

  Potential output

The latest ONS data for actual output, combined with the estimate of the output gap, allow us to infer the current level of potential output. The growth of potential output from this point is then established by splitting it up into several components that are then analysed and projected separately:

  • population;
  • employment (which is made up of labour market participation less unemployment);
  • average hours worked; and
  • productivity (on an output-per-hour basis).

To project these components, we use a variety of approaches, all involving a degree of judgement. These approaches can and do vary between forecasts:

  • We normally base our population growth assumptions on official ONS population projections, choosing the variant that we consider the best fit with recent data and our judgements on prospects. We typically stick to using ONS variants because the granular details underlying them are important inputs to our fiscal forecasts – for example, state pensions spending is sensitive to the number of people reaching state pension age in each year, which varies depending on the size of different age cohorts. Our March 2023 population growth assumption was based on the latest ONS 2020-based-interim population projection published in January 2023. Considerable uncertainty remains around historical estimates and projections of the UK population, which we will be monitoring in upcoming EFOs.
  • Our projection for the potential participation rate is informed by a ‘cohort model’ that combines the detailed age- and gender-specific population projections with historical age- and gender-specific participation rates derived from the Labour Force Survey. This approach captures the effect of the changing age composition on the overall participation rate – in particular, participation rates tend to be lower at older ages, so as the population ages the participation rate would be expected to fall. It also allows us to factor in the effect of a rising state pension age increasing labour market participation among the affected cohorts. Our March 2023 forecast for potential participation also reflects the recent rise in inactivity driven by long-term sick, much of which is concentrated amongst older age groups and is therefore assumed to prove persistent. It also reflects the estimated effect of the Government’s policy measures to boost participation.
  • Prospects for the equilibrium unemployment rate – sometimes referred to as the sustainable or natural rate of unemployment (although these concepts are not necessarily the same) – are informed by an assessment of past trends in the observed unemployment rate, as well as other recent labour market developments (such as wage growth, changes in the level of long-term unemployment or evidence of labour market mismatch). We revised down our assessment of the equilibrium unemployment rate in the October 2018 forecast when we believed it was around 4 per cent. In addition, certain Government policies can alter our view of the equilibrium rate. For example, the National Living Wage is set to rise faster than productivity growth, which we expected to raise equilibrium unemployment slightly to 4.1 per cent by 2024 in our March 2020 forecast. In our March 2023 forecast, we assumed the that the structural unemployment rate rose during 2020 as a result of the pandemic and associated restrictions, and that it would then gradually fall back as the economy re-opens and reallocates resources, reaching 4.1 per cent at the forecast horizon.
  • Potential average hours  fluctuated significantly over the pandemic period as a result of restrictions and the furlough scheme. They are projected to remain broadly flat in the medium term, in line with the experience seen between the financial crisis and pandemic but in contrast to the downward trend seen before the financial crisis when productivity and income growth was significantly stronger.
  • Our potential productivity growth forecast is informed by considering historical averages of growth rates, as well as judgements about factors that may prevent trend productivity from growing in line with previous trends, including the functioning of the financial system (which is important for allocating resources to their most productive use), the outlook for business investment (which influences the amount of capital available to each worker) and persistently higher energy prices (which lowers the level of output firms are willing to produce for a given price). Productivity growth is the single biggest source of potential output growth – and therefore GDP growth too – so this is the key judgement in our economy forecast and the most important source of uncertainty around medium-term growth prospects. Productivity growth is one of the areas where we adjusted our forecast in light of the UK’s exit from the EU and we expect the level of productivity to be 4 per cent lower in the long term than if the UK had remained a member of the EU. The coronavirus pandemic led us to make a further downward revision to the level of potential productivity largely due to the effects of weaker business investment lessening capital deepening. More recently, we have begun to split productivity into capital deepening and TFP effects. Further information is available in Briefing paper No. 8: Forecasting potential output – the supply side of the economy.

The BRC considers both the individual projections for each component described above and the overall picture for potential output growth before reaching a final judgement.

 

  Back to top