Supply and demand during and after the pandemic - speech by Michael Saunders

Given during an online webinar
Published on 26 March 2021

Michael talks about why it is difficult to measure the UK economy’s potential output at present.

He looks at what this might imply for the level of spare capacity now and in the year ahead. And he discusses what that might mean for monetary policy.

Speech

As well as reducing economic activity sharply over the last year, the pandemic has also had large effects on the economy’s potential output. footnote [1] In this speech, I want to discuss some issues in measuring potential output and the amount of spare capacity in the economy at present, and some risks around the outlook for supply and demand.footnote [2] I will then turn to some considerations for monetary policy. I will make several main points.

  • The pandemic and associated restrictions have significantly reduced potential output over recent quarters, but the majority of these effects are likely to be fairly temporary.
  • I put more weight on the risk that the central forecast in the February MPR overstated the temporary drop in potential output over the last year (and hence understated the extent of spare capacity in the economy at present) and was overly pessimistic on the path for potential output in the year ahead.
  • In coming months, swings in energy prices are likely to lift CPI inflation, while the easing of restrictions is likely to boost economic activity. With potential output also likely to be recovering, increases in activity and headline inflation by themselves may not signal much either way as to whether spare capacity is rising or falling.
  • The recent level of activity is well below the likely post-pandemic path of potential output, in other words its efficient level once temporary adverse effects on potential output of the pandemic fade.footnote [3]
  • In order for the output gap to close sustainably and return inflation to target on a sustained basis, activity needs to close the shortfall with post-pandemic potential output.footnote [4] In my view, the MPR forecast for unemployment three years ahead (4½%) is likely to be a useful benchmark in judging the extent to which this has been achieved.

Let’s start with the economy’s current position. The pandemic and associated restrictions have caused economic activity to fall sharply, and the level of real GDP in Q4-2020 was 8% down from Q4-2019. GDP in Q1 this year probably has fallen further below the Q4-19 level. These declines are much bigger than in any recession of recent decades. But the pandemic (and associated restrictions) have also reduced potential output. Some parts of the economy have been shut or restricted and, in many others, workers and firms have been unable to operate as normal. In the February MPR, the MPC judged that the recent drop in GDP has been largely matched by a drop in potential GDP, and hence would not put commensurate downward pressure on inflation.

Figure 1. UK – Change in Real GDP and Potential GDP since Q4-2019

Footnotes

  • Sources: ONS and Bank of England.

Estimates of potential GDP are inherently uncertain. In normal times (eg pre-pandemic), potential output and its growth rate are often estimated from a production function approach, using trends in the growth of the workforce, capital stock and total factor productivity. Or they can be derived as the residual from output gap estimates and the level of GDP.footnote [5] The resultant estimates for potential GDP growth usually do not change much from year to year, and are usually not affected much by short-term swings in economic activity. However, the pandemic has affected potential output through various channels that may be hard to model using those traditional approaches. The February MPR includes a range of estimated effects.footnote [6]

  • Furlough. The approach in the MPR is to treat furloughed workers as largely removed from the workforce, except to the extent that they are actively looking for another job (and hence pushing down on pay growth).footnote [7] Hence, with this approach, a rise in furloughed workers largely represents a temporary drop in potential output via reduced labour supply, rather than spare capacity.
  • Public sector measurement. The output of government services that are consumed by individuals is often measured using volume indicators, such as the number of hospital operations and the number of school pupils taught. During the last year, and especially during the first lockdown, a lot of routine healthcare was postponed. Schools have been temporarily closed during some of the lockdowns and have recently reopened. All this affects measured real government output. Provided schools stay open as planned, education output is likely to recover significantly in March and in Q2 this year (although it may continue to be affected by quarantine and self-isolation). In the MPR, these swings are treated as potential GDP and do not affect the output gap.
  • Temporary effects on productivity. It is likely that the pandemic (and associated restrictions) has affected productivity through various channels, some positive and some negative. The MPR assumes that people newly working from home because of government restrictions and social distancing are materially less productive than usual. Moreover, there has been some Covid-related disruption to trade (especially in early 2020), and additional Covid-related costs and workplace inefficiencies which may also have had some temporary adverse effects on productivity. Against this, there is a large positive composition effect on aggregate productivity – akin to a “batting average” effect – because the drop in hours worked has been relatively concentrated in sectors with low average productivity levels. In the MPR, these effects largely offset each other in Q4 last year.
  • Temporary effects on structural unemployment. Increases in medium- and (especially) long-term unemployment may put upward pressure on the NAIRU (the short-term equilibrium jobless rate) and hence reduce potential output.footnote [8] Thus far, long-term unemployment has risen only slightly, but it is likely to rise further as the overall jobless rate increases in coming quarters, and then decline later than the overall jobless rate (so that the long-term unemployed will account for a rising share of total unemployment). In the MPR, this significantly reduces potential GDP for a period late this year and in 2022. But this effect is likely to be temporary. Unless the jobless rate subsequently rises again, medium- and long-term unemployment will gradually fall during next year, such that the NAIRU declines again. In the MPR forecast, this effect is largely unwound at a two-year horizon.
  • Longer term effects on productivity. The drop in business investment during the pandemic, and some future drag on investment from wider credit spreads, are likely to reduce innovation and the capital stock (relative to pre-pandemic expectations). In addition, those who have been unemployed or furloughed during the pandemic are assumed to suffer some loss of human capital compared to those in continuous work.
  • Longer term effects on structural unemployment. Changes in the pattern of demand in the economy are likely to lead to some reallocation of resources that may increase labour market mismatch, albeit much less than in the 1980s. In the MPR, this adds slightly to the structural unemployment rate over time.

Based on this, the February MPR estimated that potential GDP in Q4-2020 was 7-8% below the Q4-2019 level and (given the further rise in restrictions and furlough) that potential GDP in Q1 would be about 11% down from Q4-19.footnote [9] The MPR forecast implies that these adverse effects on potential GDP are likely to be largely temporary. Some of the adverse effects should diminish quite quickly (eg furlough, disruptions to trade, effects on health and education output). Others were expected to fade over the next year or two as the economy recovers (eg WFH effects on productivity, upward pressure on NAIRU from medium- and long-term unemployment).footnote [10] Others were judged likely to increase over the coming year and persist for longer (eg labour market mismatch, lower capital stock). In the MPR, these persistent effects reduce potential output by 1¾% two or three years ahead compared to the pre-pandemic path.

Given this estimated path for potential output, the February MPR estimated the output gap (ie gap between GDP and potential output) was ½%-¾% of GDP in Q4-2020. With GDP in Q1-2021 forecast in the February MPR to be 12% down from Q4-2019, the MPR projected the output gap in Q1 this year would be about 1% of potential GDP. That output gap, while material, is actually not much more spare capacity than the estimated pre-pandemic level of Q4-19 (¼%-½% of GDP).

Table 1. UK – February 2021 MPR Projections

Projections

Q1-2021

Q1-2022

Q1-2023

Q1-2024

GDP growth

-9.2

14.2

1.3

1.3

CPI inflation

0.8

2.1

2.1

2.0

LFS unemployment rate

5.5

5.7

5.0

4.5

Excess supply/Excess demand

-1

0

Bank Rate (market path)

0.1

-0.1

-0.1

0.0

Footnotes

  • Note: modal projections for GDP, CPI inflation, LFS unemployment and excess supply/excess demand. GDP projection is four quarter growth. CPI inflation projection is the four-quarter inflation rate. The path for Bank Rate is the market path at the time, which is the usual conditioning assumption for the MPC’s forecasts. Excess supply/Excess demand is measured as a per cent of potential GDP - a negative figure implies output is below potential and a positive figure that it is above. Source: Bank of England.

The February MPR assumed that restrictions would gradually ease between end-March and end-September, which would lead to a recovery in activity, alongside a boost from the substantial monetary and fiscal policy stimulus in place at the time. In that forecast, GDP and potential output started to recover in Q2, and the recovery would be strong enough to close the output gap – and indeed to lift the economy slightly into excess demand – late this year. In the central forecast, the jobless rate was projected to rise over the summer as the furlough scheme ends, but fall back to about 5¾% in Q1 2022 and to about 4½% three years ahead.footnote [11] In that central forecast, swings in energy prices cause headline CPI inflation to rise close to the 2% target in Q2 this year. Thereafter, the closing of the output gap would lift core inflation such that headline inflation stayed around the 2% target next year and subsequently.

Risks Around the Path and Outlook for Potential Output

The concept that the pandemic and associated restrictions have affected potential GDP (as well as actual GDP) is reasonable in my view. However, the potential GDP estimates in the MPR have large margins of error. This is in part because each component is measured with some uncertainty, and the channels included in the MPR may not cover all the supply-side effects of the pandemic. By varying slightly the estimates, it is not hard to produce potential output estimates that differ by several percentage points of GDP. With actual GDP currently also measured with considerable uncertainty, the resultant output gap estimates are likely to be very uncertain indeed.

For myself, I find it hard to reconcile the top-down estimates that the output gap is modest and little changed over the last year with a range of other indicators which suggest that spare capacity in the economy has risen markedly over the last year. For example, since Q4-2019, the jobless rate has risen by slightly more than 1pp (from 3.8% to 5.0%), while under-employment has also risen on a range of measures.footnote [12] At the same time, business surveys suggest that capacity use in firms (weighted across manufacturing and services) is clearly below average. All this is consistent with a much greater rise (and a higher level) of spare capacity than implied by the estimate in the February MPR.footnote [13]

Figure 2. UK – Guides to Spare Capacity in the Economy

Footnotes

  • Note: Capacity use in firms is weighted across manufacturing, services and (where available) the retail sector. The series is an average of surveys from the CBI, British Chambers of Commerce, BoE Agents and Markit PMI. An increase in the pink and green lines implies a rise in spare capacity, as does a drop in the blue line. Sources: IHS Markit/CIPS, CBI, BCC, ONS and Bank of England.

It is hard to disentangle clear signals of spare capacity from recent price and cost data. This is partly because of lags.footnote [14] It is also because some of the data are more noisy and uncertain than usual at present.footnote [15] In addition, price data may currently be affected by various temporary factors (eg changes in energy prices, taxes, Covid-related health costs), while both prices and pay may be being affected by shifts in the composition of activity.footnote [16] Nevertheless, there are hints of a material rise in spare capacity over the last year. For example, various measures suggest that services inflation, which tends to respond relatively closely to changes in the output gap, has slowed compared to Q4-2019 and compared to the 2018-19 average (a period in which the MPC judge the economy had a modest output gap of about ½% of GDP on average).

Figure 3. UK – Measures of Services Sector Inflation

Footnotes

  • Note: Core CPI services excludes airfares, package holidays, education and the effects of VAT changes. The second median CPI services measure excludes the effects of the Eat Out to Help Out Scheme as well as changes in VAT. The Agents scores are not percentages but ordinal scores based on a range from -5 to +5; a score of 1 is stronger than 0 (no change) but weaker than 2 (average growth rate for that variable). The latest Agents scores are for the three months up to mid-February compared with the same period a year ago. The latest figures are Q4-2020 for services producer prices and the three months ended February 2021 for the CPI series. Sources: ONS and Bank of England.

So what is going on?

There are some measurement issues around the LFS at present. But, in my view, it is unlikely that the signals from these capacity use guides are all wrong.footnote [17] In theory, it is possible that GDP really has fallen more than shown by the ONS data. But I think the most likely outcome is that potential output over the last year has not fallen as much as the estimate in the February MPR. There are several possible factors that could explain this.

One issue is that, while the pandemic and its side effects have probably hit potential output through some channels, they have also prompted technological gains and accelerated the use of existing technologies in ways that have been (and will probably continue to be) positive for potential output.footnote [18] The Banks’ Agents report that the shutdown or forced adjustment of many processes in response to Covid has enabled firms to reassess their efficiency, and to a greater extent than during the financial crisis and recession of 2008-09.

An example is the shift to digital. The DMP survey suggests that, across a wide range of businesses, the proportion of total sales that were delivered online rose from 46% in 2019 to 62% in Q1-2021. Over the long term (ie 2022 and beyond), firms expect more half of total sales (53%) to be delivered online, an increase of 7pp on 2019.footnote [19]

The retail sector – which had already been seeing relatively high productivity growth in the couple of years before the pandemic – has been at the forefront of this shift to digital.footnote [20] With the switch to online sales, productivity (measured by GVA per hour) in the retail sector rose by 20% in the first three quarters of last year, the biggest rise since data began more than 20 years ago and indeed a bigger gain than over the prior ten years combined (the rise from Q4-2009 to Q4-2019 was 12%). Some of this may reflect a composition effect from the forced closure of non-essential stores, and may partly reverse as non-essential stores reopen. But, provided the share of online sales remains well above pre-pandemic levels, some of these productivity gains will probably persist. The shift to online shopping is also likely to promote greater price transparency and reduce barriers to entry (and expansion) among small retailers, hence increasing competitive pressures.footnote [21] Another example of the impact of this shift to digital may be that productivity in office and administrative services in Q3 last year rose by more than 20% YoY. Likewise, the wider use of video calls and reduced business travel is likely to generate substantial time savings.

The effects of working from home (WFH) on potential output during the pandemic also may turn out to be less one-sided than assumed in the February MPR. Moreover, a persistent increase in WFH (from pre-pandemic levels) seems likely, and may well actually support potential output over time.

The DMP survey suggests that remote working only accounted for around 8% of total hours worked in 2019 (and 13.5% of employees were fully or partly remote working). With the Covid pandemic and mandatory “stay-at-home” policies, the share of hours accounted for by remote working rose to 43% in Q1-2021, with 49% of employees fully or partly remote working. It is likely that compulsory WFH on this scale has hurt productivity for some firms, for example because the lack of face to face contact makes close collaboration and knowledge sharing with colleagues more difficult, which may hinder innovation.footnote [22]

Figure 4. UK – DMP Survey on Percentage of Sales that are Online

Footnotes

  • Note: Data are based on responses to the question: “In each of the following periods, approximately what percentage of your sales/services was delivered/do you expected to be delivered in the following ways? Survey period 5-19 Feb 2021. Source: Bank of England.

Figure 5. UK – DMP Survey on Percentage of Total Hours Worked Remotely Among UK Firms

Footnotes

  • Note: DMP survey period was 5-19 February 2021. Source: Bank of England.

However, while a shift to widespread compulsory full WFH probably is not optimal, WFH offers a range of possible advantages for some firms. These include more efficient use of the business capital stock and cost reduction, including savings on city centre office space (or, equivalently, you could regard WFH as an expansion of the business capital stock, because more of the stock of housing, domestic ICT and home furniture is now used for work). It is likely to allow some firms to access a wider pool of staff (for example, people that cannot easily get to a specific work location), and to better match jobs to skills, while leading to reduced absence from sickness (which may include caring responsibilities for family members).footnote [23] For some people, the option to WFH also seems to be good for employee satisfaction and productivity, because of reduced distractions and a quieter work environment, as well as advantages from a better work-life balance.footnote [24] This may lead to better staff retention, and hence improved work skills, as well as cost savings from lower staff turnover. By cutting commuting time and costs, increased WFH is likely to expand labour supply, by allowing more people to enter the workforce (or stay in the workforce for longer).footnote [25]

The ONS BICS results suggest that more firms report adverse effects on their productivity from increased remote working over the last year than report positive effects. However, there is a clear effect of firm size: in general, larger firms are much more positive in their attitude to remote working than smaller firms. Weighted by firm size (number of employees), the share of firms reporting positive effects on productivity from remote working over the last year has been similar to the share reporting adverse effects (averaged over recent months). There is some variation across sectors: adverse effects dominate in, for example, accommodation and food services, arts and entertainment, construction, administration and support services. Positive effects dominate in real estate, professional and scientific activities, ICT and health services. Similarly, a recent survey of people working from home found a modest balance reporting positive effects on their productivity.footnote [26]

The relative scale of these advantages and disadvantages is likely to vary among different jobs and people in the same firm. For the economy as a whole, the relationship between these effects is likely to be hump-shaped. Now that we have overcome initial barriers of learning how to operate home working in many jobs, some degree of WFH and flexibility around its scale probably boosts potential output. But when WFH hours are too high and compulsory WFH predominates, adverse effects take over and potential output might fall.footnote [27]

The DMP survey suggests that on average, firms expect that around 20% of total hours will be worked remotely in in the next few years (2022 onwards) – compared to 8% pre-pandemic. The results suggest the persistent shift to WFH will be especially marked in professional services, real estate services and ICT. The BICS survey gives a similar message, with a sizeable share of firms in all sectors expecting higher remote working in future years than the pre-pandemic level.

Figure 6. UK – BICS Survey Responses on Effects of Remote Working During and After the Pandemic

Footnotes

  • Note: WFH Working from Home. Results weighted by the number of employees and averaged across Waves 16-22, covering the period from mid-Sep 2020 to mid-Jan 2021. The fourth series shows the share of firms that “intend to use increased homeworking as a permanent business model going forward”. This work contains statistical data from ONS which is Crown Copyright. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets which may not exactly reproduce National Statistics aggregates. Sources: ONS and Bank of England.

Figure 7. UK – DMP Survey – Firms’ Expectations for Change in Percentage of Employees that will be Working from Home, Comparing 2022 Onwards with 2019 (as a share of total employees)

Footnotes

  • Note: The results are based on the question ‘How often did your full-time employees work from home/how often do you expect them to work from home in the following periods (2019, 2021 Q1 and 2022+)?’ Respondents were able to provide percentages for the following categories: (i) one day per week; (ii) two days per week; (iii) three days per week; (iv) four days per week; (v) five or more days per week; (vi) work from home rarely. A figure of, for example, 10pp indicate that on average firms expect a 10pp rise in the share of total employees that will be working from home for this specified number of days. Source: Bank of England.

Many firms are looking to make more use of a hybrid model whereby people are partly remote working, to combine the various benefits of WFH with the benefits of collaboration if people work in the same place as their colleagues at least some of the time.footnote [28] The BICS results suggest that, among firms that intend to use more WFH, the main factors cited are reduced overheads, higher employee satisfaction, higher productivity, the ability to access a wider pool of staff, to better match jobs to skills, and reduced absence from sickness. While this shift to persistently higher remote working may create challenges in how firms enable collaborative working, in my view it also may have positive effects on the future path of potential output in terms of labour supply, labour productivity and a more efficient use of the capital stock compared to the old pre-pandemic model (and the February MPR assumptions).

Another issue that may be supporting potential output is that the contribution of furloughed staff to spare capacity (and potential output) may be higher than assumed in the February MPR. This is in part because, with the pandemic (and furlough) having persisted for so long, a rising share of those on furlough are seeking work elsewhere. Since the MPR, new LFS data suggest that the share of furloughed workers searching for another job rose to 13% in Q4-2020, from 10% in Q3 and 5% in Q2. Separate data from a survey for the Resolution Foundation suggest that, among furloughed workers, the share searching for another job is relatively high for those who have experienced a longer cumulative period in furlough over the last year, possibly because this group also perceive the risk of losing their job to be relatively high.footnote [29] As the pandemic has rolled on, a rising share of furloughed workers fall into this persistent furlough category, perhaps accumulating repeated spells of furlough.footnote [30] Job search rates are also relatively high among those who are currently working but have been furloughed previously, especially those furloughed for a long period.

In addition, the job search measure among furloughed workers used in the MPR estimate is quite narrow. It relies on a question in the LFS that asks people whether they looked for another job in the last week, a relatively brief period. By contrast, in judging whether people who are out of work count as actively looking for work for the unemployment figures, the LFS asks whether they looked for work in the last four weeks, a much wider window. The RF survey, which simply asks people whether they are looking for another job, found a job search rate of over 20% among furloughed workers.

Figure 8. UK – Percentage of People Who Are Searching For Another Job

Footnotes

  • Note: Data surveyed 22-26 January 2021. The period for which people have been furloughed is measured on a cumulative basis, and may include more than one spell of furlough. This analysis was undertaken independently by the Bank of England. Base by categories: All people working now = 3419; Working now, never furloughed = 2952; Working now, have been furloughed 6 months = 46; All furloughed workers = 481; Furloughed now, total furlough 6 months = 270. Sources: Resolution Foundation and Bank of England.

Moreover, in my view, some furloughed workers also may contribute to spare capacity even if not searching for another job. Some firms have furloughed staff because of weak demand, even though people are not prevented from buying their goods and services. For example, in January this year, around 10% of jobs were furloughed in construction, real estate services, manufacturing and professional services (sectors that have stayed open). Similarly, roughly 30% of furloughed jobs are flexible furlough (January data), whereby people work part-time and the employer can vary their paid hours. In my view, such workers may in part contribute to spare capacity in the economy regardless of whether they are looking for another job, in the same way that people who would like to work more hours are included in estimates of spare capacity.

Looking ahead, I am also unsure – indeed, a bit doubtful – of the MPR estimate for the extent to which higher unemployment will lift structural unemployment later this year. There is evidence that the job finding rate among longer term unemployed is relatively low.footnote [31] Moreover, the past experience of our standard filter is that the estimated NAIRU (the UK equilibrium jobless rate that best explains the observed behaviour of wages, based on a Phillips curve relationship) tends to drift up when the jobless rate has been elevated for a while.footnote [32] However, when tested directly, the relation between unemployment and pay growth does not seem to be affected by changes in the share of long-term unemployment (see Appendix).footnote [33] In other words, high unemployment tends to reduce pay growth irrespective of whether it is short- or long-term unemployment that is high. If rising long-term unemployment does not lift structural unemployment as much as implied by the MPR (if at all), then this would imply upside risks to potential output late this year and during next year relative to the MPR forecast. This issue that higher long-term unemployment may temporarily reduce potential output may in any case matter less now, given that the extension of the furlough scheme is likely to lower the near term path for unemployment – and hence the path for long-term unemployment – compared to the February MPR.

This is not an exhaustive list of uncertainties over potential output.footnote [34] But, overall, my hunch is that the temporary adverse effects of the pandemic on potential output have been smaller, and will fade more quickly, than the MPR assumption.

The first two factors (the shift to digital, more remote working) – as well as a lower near-term unemployment path – may also imply that the MPR estimate for longer term scarring on potential output (1¾% of GDP) will turn out to be too pessimistic. But for that longer term outlook, there are also some risks on the downside. For example, there are signs of a substantial outflow of foreign workers over the last yearfootnote [35] which, if it has occurred and persists, could imply a smaller workforce. Nor does the MPR allow for the possibility that the pandemic will lower participation, for example because of health effects on some people or decisions by some older workers to retire earlier.footnote [36] There is (and probably will continue to be) plenty of uncertainty around the long-term effects on potential output of the pandemic – but the effects may not all be adverse.

Risks Around the Outlook for Demand

Data so far suggest that Q1 GDP will be less weak than expected in the February MPR, and the outlook for activity in Q2 also is probably better than in the MPR central forecast. This partly reflects the effects of higher government output. But, in addition, since the MPR, Covid infections, hospitalisations and deaths have continued to fall rapidly, while the vaccination program has continued at pace. The current plans envisage that restrictions across the UK will be eased somewhat more rapidly than assumed in the February MPR.footnote [37] Moreover, the recent Budget provided significant further near-term support for the economy.

There are some upside possibilities for demand further ahead. For example, the household saving rate might fall faster over the next year or two than assumed in the MPR, reflecting pent-up demand as households run down a higher share of the stock of savings built up over the last year. If there is less damage to the supply side from the pandemic, then this should also support demand over time, because households and firms will anticipate stronger future real gains in incomes and activity.

But there are also downside risks, on the side of a slower recovery over time. There probably will continue to be lingering risks that viral mutations could trigger a renewed rise in infections and restrictions over the next year or two (which could also affect potential output). Firms may be cautious over hiring and investment until it is clear that the pandemic is definitely finished. The slower pace of vaccinations in many other countries may prolong these risks, and require continued restrictions in other countries which could affect UK exports.

Business spending in the UK may also be restrained by the overhang from the rise in corporate debts during the pandemic.footnote [38] Household spending may be restrained by pressure among some households to rebuild savings that have been run down during the crisis, and fears of job losses after furlough ends. The rise in aggregate household savings has been concentrated in higher income households and the elderly, who typically have a relatively low marginal propensity to consume out of both income and wealth.footnote [39] In total, roughly as many households report their savings have fallen as risen. The share of households reporting their savings have fallen is especially marked in the lower half of the income distribution, as well as among those who have been furloughed and self-employed workers. Moreover, the recent Budget increased the medium-term tightening in fiscal policy.

Figure 9. UK – Percentage of People Intending to Spend More/Less In The Future Compared to Pre-Pandemic Level

Footnotes