The impact of Covid-19 on businesses’ expectations: evidence from the Decision Maker Panel

Quarterly Bulletin 2020 Q3
Published on 25 September 2020

By Nicholas Bloom (Stanford University), Philip Bunn (Bank of England), Paul Mizen (University of Nottingham), Pawel Smietanka (Bank of England) and Gregory Thwaites (University of Nottingham).

  • In partnership with academics from Nottingham and Stanford Universities, the Bank of England set up the online Decision Maker Panel (DMP) Survey in 2016 to poll senior executives in British businesses about their expectations for business conditions and uncertainty. The survey receives around 3,000 responses a month.
  • Businesses reported that the spread of Covid-19 and the measures to contain it led to a fall in sales of around 30% in 2020 Q2. Sales are expected to recover gradually, but further reductions in employment are expected in the second half of 2020.
  • There has been a large increase in uncertainty among businesses about their prospects for the next year.
  • Businesses’ expectations of future sales and employment growth became skewed to the downside. This means that businesses thought that there was an increased risk of bad outcomes occurring.

Overview

The prospects for the economy depend crucially on the spending decisions that households and businesses make. Because the future is uncertain, households and businesses have to make these spending decisions based on their expectations of the future. Those decisions may depend, not just on their views on the most likely future outcome, but also on the degree of uncertainty around their expectations and on the likelihood of things going badly wrong. Measuring and understanding those expectations is therefore very important for economic policymakers.

This article uses data from the DMP to show how the spread of Covid-19 and the measures to contain it have affected the expectations of businesses. It builds on a previous article that assessed how expectations were being affected by the UK’s decision to leave the European Union (EU). The DMP is a large and representative survey of UK businesses. It collects quantitative information on what businesses expect to happen to sales and employment. It also asks about the distribution of those expectations, which reveals the probabilities that businesses attach to negative or positive outcomes. This information is not available from most other business surveys, which tend to report mean values or estimates of whether things are expected to get better or worse rather than distributions of outcomes. Using this information allows us to measure the degree of uncertainty around those expectations, and whether expectations are skewed towards good or bad outcomes.

Introduction

The spread of Covid-19 and the actions to contain it have had a dramatic impact on the UK economy and on many other countries around the world. In the UK, GDP is estimated to have fallen by 20% in 2020 Q2. That was by far the largest quarterly fall on record. The sectors most affected were those where a relatively high proportion of consumer spending involves face-to-face contact and/or social activity (eg accommodation and food or recreational services) and those that were most affected by government restrictions (eg transport).

Output has begun to recover since government restrictions were eased. UK GDP in July was around 18½% above its trough in April and around 11½% below its 2019 Q4 level. High-frequency payments data suggest that household spending has continued to recover during the summer. Employment has fallen since the Covid-19 outbreak, although this has been very significantly mitigated by the extensive take-up of support from temporary government schemes. According to HMRC, some 9.6 million jobs had been furloughed under the Coronavirus Job Retention Scheme (CJRS) by September 2020. Total hours worked have also fallen considerably.

The outlook for the UK economy will depend critically on the evolution of the pandemic, measures taken to protect public health, and how governments, households and businesses respond to these factors. The future is uncertain, therefore households and businesses have to make these spending decisions based on their expectations; for example, households have to decide how much to spend or save, and businesses have to decide how much to invest and how many people to employ. Those decisions may depend not just on their views on the most likely outcome, but also on the degree of uncertainty around those expectations and on the likelihood of an extreme ‘worst-case scenario’ outcome being realised. Households and businesses may be more cautious about making spending commitments when they are less sure about what the future holds or when they think that there is a higher risk of a very bad outcome occurring. Data on the expectations of businesses are therefore very important for policymakers in helping to inform their judgements about the prospects for the UK economy.

In this article, we focus on what we have learnt about the expectations of businesses during the Covid-19 pandemic using data from the DMP. This article builds on a previous Quarterly Bulletin article, which introduced the DMP and summarised some of the early results from the survey including in relation to the UK’s decision to leave the EU.

The DMP was launched in August 2016 by the Bank of England in collaboration with Stanford University and the University of Nottingham.footnote [1] It is closely based on the Survey of Business Uncertainty run in the United States by the Federal Reserve Bank of Atlanta, which is described in Altig et al (2019). The DMP is a large and representative survey of British businesses and is suitable for analysis of the uncertainties facing businesses and how they expect to be affected by that uncertainty. It is a flexible tool for assessing business conditions and, in particular, it has been adapted to collect detailed information on how businesses are being affected by important economic events. For example, it has been used to assess the response to the UK’s decision to leave the EU and most recently the DMP Survey has been refocused to help assess the impact of Covid-19.footnote [2]

The DMP is a monthly online survey of Chief Financial Officers in UK businesses. The panel grew quickly after its launch and has averaged just under 3,000 responses a month since 2019. Businesses with at least 10 employees are randomly selected and invited to participate in the survey. This ensures it provides a representative view across the economy. It covers small, medium and large private sector businesses across all industries. Results are weighted using employment data. The survey methodology is described in more detail in Box A.footnote [3]

The main advantages of the DMP Survey relative to other business surveys are the quantitative data that it provides and the fact that it collects data on the distribution of expectations, not just the single most likely outcome. Many other business surveys tend to focus on questions that ask businesses to indicate whether they expect the conditions they face to get better or worse, rather than by how much they expect them to get better or worse. But the extent to which conditions are better or worse has been particularly important in the context of the Covid-19 pandemic where the size of such changes is much larger than in normal times.

Prior to the launch of the DMP Survey there was relatively little, if any, direct quantitative information about the distribution of expectations of decision makers in individual UK businesses. The DMP Survey was designed to help fill this gap. The DMP is also very timely, with results available to policymakers within days of the survey being completed, and the data collected can be broken down for different sectors of the UK economy. This is important because the impact of Covid-19 has varied considerably across sectors. We publish aggregated data on a monthly basis. Data from the DMP Survey are also often used by the Bank of England’s Monetary Policy Committee (MPC) in publications such as the Monetary Policy Report, minutes of their meetings and in speeches.footnote [4]

The DMP can be used to assess the impact of Covid-19 on businesses’ expectations in a variety of ways. One approach is to ask panel members directly about their best guess for the marginal impact of Covid-19 on their business. A second approach is to ask about their overall expectations. As described in more detail in Box A on the survey methodology, this latter approach also involves asking about the distributions of expected future growth around variables such as sales and employment, whereas the former approach just asks panel members to provide a single point estimate of the expected most likely value.

This article concentrates on better understanding the expectations of businesses rather than how those expectations have affected their behaviour. As well as the most likely outcome, the distribution of future growth, which indicates the range of possible outcomes, and the degree of uncertainty, which we can derive from it, are also likely to affect businesses’ decisions to invest and hire workers. For example, investment or hiring decisions typically involve incurring fixed costs, which cannot be recovered, and so when uncertainty is high there is greater value to taking a ‘wait and see’ approach to make sure that it is the right decision to incur these costs.footnote [5]

Data from the DMP Survey show how Covid-19 led to a very sharp fall in sales in 2020 Q2, although that fall was not quite as large as initially feared. The most recent expectations data show that businesses expect a gradual recovery in sales over the next few quarters, but by 2021 Q2, sales are still expected to remain below what they would have otherwise been in the absence of the Covid-19 shock. Businesses where a large proportion of spending involves face-to-face contact and/or social activity have been particularly heavily affected and that is likely to continue to be the case. Some further reductions in employment are also expected in the second half of 2020. Uncertainty rose sharply in March and April as the extent of the pandemic become clear and it has remained high through to the latest data in August. Businesses’ expectations of future sales and employment growth have become skewed to the downside. This means that businesses thought that there was an increased risk of bad outcomes occurring.

This article describes how businesses have been affected so far by the spread of Covid-19 and considers their expectations for the shape of the recovery, using questions about the marginal impact of Covid-19. The article also discusses the distribution of businesses’ year-ahead expectations for overall sales and employment growth and draws out how those distributions imply a large rise in uncertainty and a skew in risks towards negative outcomes.

Businesses’ expectations and the impact of Covid-19

The spread of Covid-19 has had a large impact on UK businesses (Chart 1). Respondents to the DMP Survey estimated that sales in 2020 Q2 were 30% lower than they otherwise would have been. Employment was reported to have been 5% lower and investment 33% lower. Unit costs were estimated to have increased by around 6%, relative to what would have otherwise happened. In the August survey, a gradual recovery in sales was expected over the next year, although sales were still expected to be 5% lower than they would have been by 2021 Q2 (in the absence of the Covid-19 shock). Investment was expected to recover more slowly than sales, but with a broadly similar profile. Unit costs were expected to begin falling back from a peak in 2020 Q3, although they were still expected to be higher than they would have been out to at least 2021 Q2.

Chart 1 Covid-19 has had a large impact on UK businesses (a)

Lines show change in sales, employment, unit costs and investment. Sales and investment fell in Q2, while unit costs rose.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) The results are based on the questions: ‘Relative to what would otherwise have happened, what is your best estimate for the impact of the spread of Covid-19 on the sales/employment/capital expenditure of your business in each of the following periods?’; and ‘Relative to what would otherwise have happened, what is your best estimate for the impact of measures to contain coronavirus (social distancing, hand washing, masks and other measures) on the average unit costs of your business in each of the following periods?’. Data for 2020 Q2 are from the July DMP Survey. Data for 2020 Q3, 2020 Q4, 2021 Q1 and 2021 Q2 are from the August DMP Survey. Data on the impact of Covid-19 in 2020 Q1 have not been collected in the DMP. Data shown for Q1 are absolute changes in aggregate ONS data for private sector output and business investment between 2019 Q4 and 2020 Q1. The impact on employment is assumed to be zero (as it rose in Q1). The impact on unit costs is also assumed to be zero in Q1.

Covid-19 was expected to lower employment further in the second half of 2020 at the time of the August DMP Survey. The employment effects were expected to increase from 5% in Q2 to a peak of 8% in Q4, implying a further fall in employment of 3%. The winding down of the Government’s CJRS may be a factor affecting these expectations. Under the CJRS, which was launched in March, the Government paid 80% of the income of employees that were placed on furlough (still employed but not required to work any hours), up to £2,500 per month. From August, employers have had to start contributing towards the scheme, which is due to close at the end of October. It is estimated that the effects of Covid-19 on employment will start to ease in the first half of 2021.

Respondents to the DMP Survey estimated that, in April, 36% of employees were on furlough (Chart 2). This percentage has steadily fallen to 12% by August, and is expected to be close to zero by Q4. Businesses planning to reduce employment further in the second half of 2020 were typically those who had previously furloughed a relatively large share of the employees, but these were often also businesses who had seen a large fall in sales in Q2 and expected a relatively slow recovery.footnote [6]

Chart 2 The percentage of employees on furlough has been falling (a)

Chart shows percentage of employees on furlough, working at home, working on business premises and unable to work.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) The results are based on the questions: ‘Approximately what percentage of your employees fall into the following categories in each of the following periods? (i) Still employed but not required to work any hours (eg ‘on furlough’), (ii) Unable to work (eg due to sickness, self-isolation, childcare etc), (iii) Continuing to work on business premises, (iv) Continuing to work from home’.

Although the expected impacts of Covid-19 on businesses have been very large, businesses have become a little less pessimistic over how severely they will be affected over time. For example, in the July survey, sales were reported to have been 30% lower in 2020 Q2 than they would have been, compared to an expectation for the same period of 44% at the time of the April survey when this question was first asked (Chart 3). The extent of the expected reduction in sales has also been revised down further ahead, although to a lesser extent.

Chart 3 Since April, businesses have become less pessimistic about Covid-19’s likely impact on sales (a)

Line chart showing percentage impact of Covid-19 on sales, based on responses to monthly surveys from April to August 2020.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) The results are based on the question: ‘Relative to what would otherwise have happened, what is your best estimate for the impact of the spread of Covid-19 on the sales of your business in each of the following periods?’. The August DMP Survey did not ask about 2020 Q2, data for this period are assumed to be the same as in the July DMP Survey.

One of the advantages of the DMP Survey is that it is very timely and has been able to provide an indication of the impact of Covid-19 on the economy before official statistics are released. But first releases of official statistics for the UK economy are now becoming available for 2020 Q2 and can be compared to results from the DMP Survey. Overall the DMP data appear to suggest a slightly larger impact on sales and employment in 2020 Q2 than the first release of official data, with investment data more closely in line. The ONS estimated that the level of UK GDP in 2020 Q2 was 22% lower than in 2019 Q4. In the DMP, sales in 2020 Q2 are estimated to have been 30% lower than they would have otherwise have been.

However, precise comparisons are complicated by differences in definitions and coverage. For example, official GDP data deduct the costs of intermediate inputs used in the production process from sales, whereas the DMP has just asked about the impact of Covid-19 on sales. And sales would have otherwise increased in the absence of Covid-19 and therefore the absolute fall in sales implied by the DMP would have been smaller than 30%.footnote [7] Aggregate business investment is estimated by the ONS to have fallen by 32% between 2019 Q4 and 2020 Q2, compared to an estimated Covid-19 impact of 33% in 2020 Q2 in the July DMP Survey. But the ONS investment data may also have been affected by the UK leaving the EU. Official data may also be revised before a final comparison can be made and the first releases of official data are also subject to more uncertainty than usual as a consequence of Covid-19.

Expected impacts of Covid-19 by industry

The impact of Covid-19 has been substantial across all industries, but highly consumer-facing businesses where a large proportion of spending involves face-to-face contact and/or social activity have been most severely affected. The analysis in this section divides businesses into four groups:

1. highly consumer-facing non-essential services (including businesses from non-food retail,footnote [8] accommodation and food and recreational services sectors);

2. less consumer-facing non-essential services (for example, businesses in finance and real estate, professional services and information and communication);

3. non-food manufacturing and construction; and

4. businesses involved in the provision of essential services (food manufacturing and distribution, utilities and healthcare).

Chart 4 highlights how highly consumer-facing businesses reported the largest falls in sales in Q2, in the region of 50%, on average, relative to what would have otherwise happened (shown by the red line). Non-food manufacturing and construction was the next most affected sector (sales were around 35% lower) followed by less consumer-facing service businesses (20%). Businesses in essential services reported the smallest reduction in sales in Q2, at around 10%.

Averaging across the July and August DMP Surveys, some recovery in sales was expected for all industry groups over the year to 2021 Q2. But across the next year, the sales impacts were expected to remain larger in the groups that had seen bigger falls in sales in Q2.footnote [9] By 2021 Q2, sales in essential services were expected to be 4% lower than they would have been, compared to 8% in more consumer-facing services.

Chart 4 Sales in consumer-facing industries have been hardest hit by Covid-19 (a)

Line chart showing percentage impact of Covid-19 on sales, split by industry sector.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) The results are based on the question: ‘Relative to what would otherwise have happened, what is your best estimate for the impact of the spread of Covid-19 on the sales of your business in each of the following periods?’. Data are averages from the July and August DMP Surveys.

All industry groups reported a material reduction in employment in 2020 Q2, relative to what would have otherwise happened (Chart 5). Looking ahead, Covid-19 was expected to lower employment further, but these effects were expected to be largest in the groups that had experienced the largest falls in sales. The impact on employment was expected to be largest in consumer-facing businesses up to 2021 Q2. The largest reductions in employment between 2020 Q2 and 2020 Q4 were also expected in consumer-facing businesses.

Chart 5 Consumer-facing businesses expect to see the largest decline in employment (a)

Line chart showing percentage impact of Covid-19 on employment from 2020 Q2 to 2021 Q2, split by industry sector.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) The results are based on the question: ‘Relative to what would otherwise have happened, what is your best estimate for the impact of the spread of Covid-19 on the employment of your business in each of the following periods?’. Data are averages from the July and August DMP Surveys.

The distribution of overall expectations

The previous section focused on questions that asked panel members to provide a point estimate of their best guess for the marginal impact of Covid-19 on various aspects of their business. This section discusses the data on businesses’ overall expectations, from which we can infer Covid-19 impacts.

Rather than asking for a point estimate, this method asks about the distribution of expectations for year-ahead growth for each business. A key advantage of this approach is that it can also be used to show how uncertain businesses are about the future, as well as to assess whether the distribution of expectations are particularly skewed towards negative or positive outcomes. The survey methodology box (Box A) contains more detail on how the DMP Survey asks about the distribution of expectations. Box C evaluates the accuracy of past forecasts based on this approach.

The distribution of expectations for year-ahead sales and employment growth are shown in Charts 6 and 7 respectively. These show how expectations have developed as the pandemic has evolved. Both show a fall in mean and median expected sales and employment growth. Sales growth expectations data collected between May and July referred to expected growth in sales between 2020 Q1 and 2021 Q1.footnote [10] Employment growth expectations refer to annual growth between employment at the time of the survey and the same month a year ahead. The overall expectations data are not directly comparable to the Covid-19 impacts reported above. But if they are combined with data on reported sales and employment growth over the recent past and we assume that things would have otherwise been the same as the average of the two years prior to the Covid-19 outbreak, the overall implied Covid-19 impacts from both approaches are broadly similar.

There have also been some striking changes to the shape of the distributions. First, both distributions have become much wider, consistent with businesses becoming more uncertain about their expectations for the future. Second, that widening of the distribution has been larger in the bottom half of the distribution, particularly in the case of employment growth. That is consistent with the distribution of sales and employment growth becoming more skewed towards negative outcomes. The remainder of this section considers these developments in uncertainty and in the skew of the distribution in more detail.

Chart 6 The distribution of expected sales growth has widened (a)

Distribution of expected sales growth, shown by percentile. Chart shows fall in mean and median expected sales.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) The results are based on the questions: ‘Looking a year ahead from the first/second/third/fourth quarter of this year to the first/second/third/fourth quarter of next year, by what percentage do you expect your sales revenue to have changed in each of the following scenarios: lowest, low, middle, high and highest?’. Respondents were then asked to assign a probability to each scenario. Data up to July 2020. Data from the May to July 2020 DMP Surveys refer to growth in sales between 2020 Q1 and 2021 Q1. February to April Survey data refer to sales growth between 2019 Q4 and 2020 Q4.

Chart 7 The distribution of expected employment growth has widened, particularly towards the bottom of the distribution (a)

Distribution of expected employment growth, shown by percentile. Chart shows fall in mean and median expected employment.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) The results are based on the questions: ‘How many people does your business currently employ?’; and ‘Looking ahead, 12 months from now, how many employees would your business have in each of the following scenarios: lowest, low, middle, high and highest?’. Respondents were then asked to assign a probability to each scenario. Employment growth in each scenario is calculated using Davis, Haltiwanger and Schuh (DHS) growth rates. This is the change between two periods, divided by the average of those two periods. Data up to August 2020.

Uncertainty

Uncertainty exists because we cannot know what the future holds. An increase in uncertainty implies that a wider range of outcomes has become plausible and can be characterised by a widening in the distribution of expected outcomes, as demonstrated in Charts 6 and 7. To demonstrate this point more easily, we can use data on the distribution of expectations to construct an average standard deviation to more easily track developments in uncertainty. These measures of ‘subjective uncertainty’, are shown in Chart 8.footnote [11]

Uncertainty around both future sales and employment growth rose sharply in March and April as the extent of the Covid-19 pandemic became clear and as several measures were introduced to contain it (Chart 8). Subjective uncertainty has since fallen back a little from its April peak, more so for the employment measure, but it has remained high through to August.footnote [12]

These measures of uncertainty relate to the overall level of uncertainty facing businesses, and so, for example, will also include uncertainties associated with other events such as Brexit. We might expect recent increases in uncertainty to be primarily Covid-19 related, and that view is supported by the fact that 86% of DMP members reported that Covid-19 was their largest current source of uncertainty back in April. This proportion had fallen by August to 54%, but 88% still said that Covid-19 was in their top three sources of uncertainty in that August survey. By way of comparison, Brexit was reported to be in the top three sources of uncertainty for 47% of businesses in August, which was slightly higher than in the first half of the year. But only 3% of businesses reported that Brexit was their largest source of uncertainty in August.

Chart 8 Measures of uncertainty have increased sharply in response to Covid-19 (a)

Line chart shows an increase in uncertainty across three categories (sales, employment and overall) in 2020.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) Sales and employment uncertainty data are based on the questions described in the footnotes to Charts 6 and 7. The overall uncertainty data is based on the question ‘How would you rate the overall uncertainty facing your business at the moment?’. Respondents could select one of the following options: (i) Very high – very hard to forecast future sales, (ii) High – hard to forecast future sales, (iii) Medium – future sales can be approximately forecasted, (iv) Low – future sales can be accurately forecasted, (v) Very low – future sales can be very accurately forecasted. The overall uncertainty line shows the percentage of respondents who reported uncertainty to be high or very high.

These measures of subjective uncertainty derived from the distribution of expected sales and employment growth may not perfectly capture all of the aspects of uncertainty that businesses face. For example, they refer to a particular point in time and so might not capture uncertainty about how long Covid-19 will last.footnote [13] Chart 9 shows that there are a range of views among businesses about when Covid-related uncertainty will be resolved. Respondents to the DMP are also trying to quantify a distribution and so might miss aspects of uncertainty that are hard to quantify, so-called Knightian uncertainty.footnote [14] However, measures of uncertainty based on the distribution of expectations offer a very similar picture of developments in uncertainty to an alternative and much simpler concept based on DMP members’ views about whether they rate the overall level of uncertainty facing their business as high or low (shown by the red line on Chart 8).

Chart 9 There are a range of views about when Covid-19 uncertainty will be resolved (a)

Bar chart showing when businesses expect Covid-related uncertainty to be resolved. Most said by June 2021 or later.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) The results are based on the question: ‘When do you think it is most likely that the coronavirus-related uncertainty facing your business will be resolved?’. Data are from the August 2020 DMP Survey.

Skewness

Skewness is a measure of the symmetry of the distribution. A positive skew means that expectations are skewed towards positive outcomes, eg higher or positive growth. A negative skew means that expectations are skewed towards negative outcomes, eg lower or negative growth. Skewness can therefore be seen as an indicator of ‘tail risk’ or ‘disaster risk’. Skewness that is more negative indicates that businesses think there is an increasing risk of lower negative outcomes occurring.

We construct a measure of skewness from DMP data using the distribution of sales and employment expectations (Chart 10).footnote [15] Prior to the Covid-19 pandemic, expectations were generally positively skewed. But they became negative after March 2020 as the scale of the pandemic became clear, particularly for employment. That implies that businesses thought that there was an increased risk of negative outcomes. On a practical level, this could, for example, be associated with the risk of a second wave of Covid-19.

Chart 10 Businesses thought that there was an increased risk of bad outcomes occurring (a)

Chart shows skewness for both sales and employment, with a negative skew for both since March 2020.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) Chart shows Kelley’s measure of skewness based on the distributions shown in Charts 6 and 7. It is calculated as [(p90-p50)-(p50-p10)]/(p90-p10) where p90, p50 and p10 represent the 90th, 50th and 10th percentiles of the distribution.

Conclusion

This article has summarised evidence from the DMP on how Covid-19 has affected the expectations of UK businesses. Understanding these expectations is important for policymakers as they are a key determinant of the decisions that businesses make in the current environment, for example around how much to invest and how many people to employ. The DMP collects quantitative information on what businesses expect to happen to things like their sales and employment, and also asks about the distribution of those expectations. This is not typically available from most other business surveys. It allows the uncertainty around those expectations to be measured and it can be used to show whether expectations are skewed towards good or bad outcomes.

Data from the DMP Survey show how Covid-19 led to a very sharp fall in sales in 2020 Q2, although that fall was not quite as large as initially feared. The most recent data on expectations show that businesses expect a gradual recovery in sales over the next few quarters, but sales are still expected to remain below what they would have been (in the absence of Covid-19) by 2021 Q2. Consumer-facing businesses have been particularly heavily affected and that is likely to continue to be the case. Some further reductions in employment are also expected in the second half of 2020. Uncertainty rose sharply in March and April, as the extent of the pandemic became clear and it has remained high through to August. At the same time, the risks around businesses’ expectations for their sales and employment became skewed to the downside.

The Bank of England’s MPC regularly use data from the DMP Survey, alongside a range of other information, to help them to assess the economic outlook. In their August 2020 Monetary Policy Report, the MPC judged that the current outlook was unusually uncertain and that the risks were skewed to the downside. That was consistent with the expectations of businesses in the DMP Survey.

Box A: Survey methodology

The Decision Maker Panel was launched in August 2016 by the Bank of England, Stanford University and University of Nottingham, supported by funding from the Economic and Social Research Council. The sampling frame for the DMP is the population of all 48,000 active UK businesses with at least 10 employees in the Bureau van Dijk FAME database.footnote [16] Businesses are selected randomly from this sampling frame and are invited by telephone to join the panel by a team of trained analysts based at the University of Nottingham. The invitation to join the survey is typically made to the Chief Financial Officer (CFO) of the business, or the Chief Executive Officer (CEO). Approximately 85% of respondents are in these two positions (70% are CFOs and 15% are CEOs). Once they have agreed to participate, panel members receive a monthly email with a link to the latest survey, which is carried out online. Each survey takes five to 10 minutes to complete.

The DMP grew quickly after its launch to a sample of approximately 8,000 businesses and has received, on average, just under 3,000 responses per month since early 2019 (Chart A). That makes it one of the largest regular UK business surveys. The number of responses fell from 2,900 at the start of 2020 to 2,500 in March, the point at which widespread restrictions to stem the spread of Covid-19 were introduced in the UK. But it has since recovered to around 2,800 a month and there is no evidence of a significantly larger fall in response rates in sectors more heavily exposed to Covid-19. The pandemic therefore appears to have only had a small impact on DMP response rates, perhaps because it is carried out online.

Chart A The DMP Survey receives close to 3,000 responses a month

Bar chart showing number of survey responses per month from 2016 to 2020, with around 2,800 responses per month now.

Footnotes

Source: Decision Maker Panel Survey.

The DMP covers small, medium and large private sector businesses across all industries. Results are weighted using employment data to ensure that it is representative. The survey has a rotating three-panel structure. This means that each business only receives one third of the survey questions in any given month, so that within each quarter all businesses rotate through all of the questions.footnote [17] So, with a sample of 3,000 businesses, around 1,000 would respond to questions in each of the panels in any given month. This approach allows a wider range of questions to be asked, but keeps the survey relatively short in order to encourage panel members to complete it.

Since its launch, the DMP has included questions asking about recent and expected year-ahead growth in sales, prices, employment and investment. These expectations questions ask about the distribution of expectations rather than asking for a point forecast. Taking sales as an example, we ask participants to provide five expected outcomes for sales growth in their business over the next year: a lowest, low, medium, high and highest scenario. Panel members are then asked to assign probabilities to each of these five scenarios, where those probabilities must sum to 100%. From these responses it is possible to compute the average expected growth rate of sales, by calculating a weighted average of sales growth in each of the five scenarios, using the probabilities attached to each scenario as weights. Similarly, it is possible to calculate the standard deviation of expected sales growth for each businesses. The scenarios and associated probabilities for each business can also be pooled together to produce an estimate of the distribution of expected sales growth for the representative business. This approach gives respondents a large amount of flexibility to report different types of distribution without framing the question in any particular direction.

In addition to the regular questions described above, the DMP Survey also contains questions relating to special topics. Some of these questions are frequently included in the survey, for example a question on Brexit uncertainty is one of the longest running special questions. Other special questions may only be asked once. The first special questions relating to Covid-19 were introduced in March 2020, with the majority of the special questions being on this topic since April. Typically these questions have asked panel members to provide an estimate of how Covid-19 has affected different aspects of their business, relative to what would have otherwise have happened. Different questions have been asked about the impact on sales, employment, investment, unit costs, supply capacity, supply disruption, credit constraints, hours worked and the proportion of employees on furlough.

Box B: Working from home

Data from the DMP Survey show how around 40% of employees in UK businesses were estimated to have been working from home at the height of the Covid-19 pandemic in April 2020. At this time, around 20% were still working on businesses’ premises and 40% were not working, either because they were on furlough or unable to work (see Chart 2 in the main text). Although increasing numbers of employees are reported to have returned to businesses premises, by August 2020 around 35% were estimated to still be working from home. This proportion is still expected to be around 30% in Q4. It remains to be seen how persistent this will be, but there may be some parts of the economy where home working will continue to be more widespread than it was prior to Covid-19.

Working from home has been most prevalent in service sector industries that require less face-to-face contact with customers (Chart A). Information and communication, finance and insurance, real estate and professional services are the industries where the highest proportion of employees have worked from home. These industries also had the highest proportions of employees who had previously ever worked from home prior to the Covid-19 pandemic. They are also the sectors where permanently higher levels of home working might be more likely.

Chart A The percentage of employees working from home varies considerably by sector (a)

Percentage of employees working at home, split by sector. Also shows percentage who expect to be working at home in 2020 Q4.

Footnotes

Sources: Decision Maker Panel Survey, Labour Force Survey (LFS) and authors’ calculations.

(a) The results are based on the question reported in the footnote to Chart 2. DMP data are a percentage of all employees, including those on furlough. DMP expectations data for 2020 Q4 are averages from the July and August DMP Surveys.

Box C: How reliable are businesses’ expectations?

To test how informative the data on businesses’ expectations collected in the DMP Survey are, expectations can be compared to what actually happened. This is possible because the DMP Survey is a panel tracking the same businesses through time. This exercise can also provide some validation that businesses are able to properly answer the questions where they are asked to provide the distribution of their expectations.

Chart A shows that there has been a strong positive correlation between actual year-on-year sales growth and expectations of future sales growth for the same period taken one year earlier. This analysis pre-dates the Covid-19 shock as few businesses anticipated it, but in general it suggests that businesses do make relatively accurate predictions about their future sales. However, as shown, in Chart B those predictions are less accurate when uncertainty (standard deviation of expected sales) around the initial prediction was high. This demonstrates the value of also collecting information on the distribution of expectations and not just on the most likely outcome.

The analysis above relates to sales, but similar conclusions apply to predictions for prices, employment and investment. Expectations for each of these variables can help predict future outcomes, but those predictions are less accurate when uncertainty around them is high. In relative terms, businesses are able to more accurately predict growth in employment and prices than growth in sales. But expectations for future investment growth are typically less accurate than those for sales.

Chart A Businesses’ achieved sales growth is strongly correlated with their past predictions (a)

Charts show strong positive correlation between realised sales and expected sales growth a year earlier.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) Results are based on the questions described in footnote to Chart 6 and a question about the past sales growth. Forecasts made between November 2016 and April 2019. Each dot represents 1% of observations grouped by expected sales growth.

Chart B Businesses make less accurate forecasts when uncertainty is high (a)

Chart plots sales growth absolute forecast error against expected sales growth standard deviation a year earlier.

Footnotes

Sources: Decision Maker Panel Survey and authors’ calculations.

(a) Results are based on the questions described in footnote to Chart 6 and a question about the past sales growth. Forecasts made between November 2016 and April 2019. Each dot represents 1% of observations grouped by expected sales growth standard deviation.
  1. The DMP also receives financial support from the Economic and Social Research Council.

  2. See Bloom et al (2019) for analysis of the impact of Brexit on UK businesses using data from the DMP.

  3. More details on the DMP are also available at www.decisionmakerpanel.co.uk.

  4. For example, the August 2020 Monetary Policy Report contained seven charts that used data from the DMP Survey.

  5. See Bloom (2014) for an overview of the literature on uncertainty and its effects.

  6. Chart 2 also highlights how almost 40% of employees were working at home at the peak of the pandemic. Box B discusses this in more detail.

  7. There are other differences too. For example, DMP data relate only to the private sector. The ONS data quoted above are for the whole economy and also include imputed rental income of owner occupiers, which the DMP does not. DMP data are also weighted by employment, which increases the size of the aggregate sales estimate because sales are estimated to have fallen by more in labour intensive sectors. Each of these points would help to narrow the discrepancy between ONS GDP and DMP sales. However, in the other direction, increased costs of intermediate inputs from Covid-19 will lower GDP relative to sales.

  8. This will include spending online as well as in physical shops.

  9. Data from the July and August surveys are pooled together for this sectoral analysis to increase the sample size for this more disaggregated analysis. Differences in aggregate sales and employment expectations were relatively small between the two surveys. Only the July survey asked about 2020 Q2. Only the August survey asked about 2021 Q2.

  10. Sales growth expectations collected in August referred to growth in sales between 2020 Q2 and 2021 Q2. These are not shown on Chart 6 because they refer to expected growth in sales relative to the peak impact of Covid-19 and therefore become strongly positive as they become about expected growth during the recovery.

  11. These measures can be constructed for each business and then averaged, or they can be calculated for the representative business. The measures in Chart 8 use the former approach, although both lead to similar conclusions.

  12. See Altig et al (2020) for a more detailed review of developments in other measures of uncertainty during the Covid-19 pandemic and some analysis of how these subjective uncertainty measures compare to other metrics.

  13. For example, data from the May to July 2020 DMP Surveys refer to growth in sales between 2020 Q1 and 2021 Q1. February to April Survey data refer to sales growth between 2019 Q4 and 2020 Q4. Employment data refer to expected growth in employment between the survey month and the same month a year ahead.

  14. See Knight (1921).

  15. To calculate skewness we use Kelley’s measure. It is calculated as [(p90-p50)-(p50-p10)]/(p90-p10) where p90, p50 and p10 represent the 90th, 50th and 10th percentiles of the distribution.

  16. This database is based on information from Companies House and includes information on the characteristics of businesses and information from their accounts.

  17. New panel members are randomly assigned to one of the three panels when they join the survey.

Share your thoughts with us at QuarterlyBulletin@bankofengland.co.uk.