Responding to the unprecedented situation resulting from the COVID-19 virus, the Banque de France has made changes to the way it uses and presents the results of its monthly business survey (MBS) in manufacturing industry, market services and construction. Our primary goal is to provide the most detailed picture of activity in each sector of the economy during the lockdown in the second half of March. The MBS was conducted over the phone between 27 March and 3 April, that is, after the lockdown measures came in, and the sample consisted of 8,500 companies. Response rates were highly satisfactory despite the circumstances.
Based on the picture at end-March provided by the MBS, we estimated how much activity was lost by each sector and on an aggregate basis (in GDP terms) during the lockdown period. This estimate was corroborated using other high-frequency data sources, including daily bankcard transactions and “partial activity” applications filed owing to the coronavirus.
We also used this information to estimate the GDP growth rate in the first quarter of 2020. This estimate is different from the monthly index of business activity (MIBA) that we usually publish. It does not use the same approach, as the MIBA methodology cannot accommodate the current break with historical patterns. Nor does it purport to offer the usual level of accuracy. Rather this estimate gives an order of magnitude that is backed by evidence and meaningful when it comes to assessing the most recent economic situation.
As announced in a press release on 18 March, the Banque de France did not publish its customary macroeconomic forecasts at the end of March for the current year and the next two years. With considerable uncertainty still surrounding the lockdown's length and exit procedures, we continue at this stage to prioritise close monthly monitoring of business conditions.
The monthly business survey was conducted between 27 March and 3 April and involved a sample of 8,500 companies and establishments.
In response to the coronavirus outbreak and the lockdown measures that took effect at 12 noon on 17 March, companies shut down sites for multiple days in March and recorded a steep drop in activity.
Manufacturing industry saw an average of 5 exceptional closure days, but the number varied from 1 day in pharmaceuticals to around 8 days in the transport equipment sector.
The number of exceptional closure days (five) should be considered against the total number of usual working days in March (10.5, not including weekends) during the lockdown, meaning that, on average, in manufacturing industry, production sites were open almost 50% of the time.
Sectors with the highest numbers of exceptional closure days were typically those reporting the most negative balances of opinion regarding the change in activity relative to the previous month and the lowest capacity utilisation rates.
Based on the responses provided by business leaders on a seven-step scale, we calculate a balance of opinion that expresses the difference between the proportion of companies that believe that activity increased and those that believe it decreased. Aggregate balances of opinion may range between +200 and -200.
While there were substantial difference between sectors, there was an across-the-board downward movement, with activity shrinking in all sectors, in most cases drastically. This shift was not due solely to the situation in France. Orders from outside the country have also been seriously affected as the pandemic has hit the main trading partners of French businesses.
In manufacturing industry, the sectors hardest hit by the downturn were the automotive industry, metal manufacturing, and machinery and equipment manufacturing.
The overall capacity utilisation rate for manufacturing industry fell from 78 % in February to 56 % in March, an all-time low for the survey.
Market services saw an average of 6 exceptional closure days, and the dispersal was even greater, ranging from 1 day for computer services to 14 days in food services.
In services as well, the sectors with the highest numbers of exceptional closure days were also those reporting the most negative balances of opinion regarding the change in activity relative to the previous month.
Accommodation and food services recorded the sharpest declines. Temporary employment and automotive repair were also severely affected.
Business services were among the least affected sectors.
Construction was also heavily impacted, with many projects shutting down as the lockdown got started.
Companies reported that cash positions deteriorated significantly in March as a direct result of the sharp downturn in activity. This deterioration affected all sectors to varying degrees, but SMEs were particularly impacted.
To cope, companies increased their demand for cash loans over the period: 17 % of small and medium enterprises (+10 points at end-March) and 22 % of mid-sized firms (+12 points at end-March) said they had applied for a cash loan recently. The survey period coincided exactly with the launch of France's government-backed loan scheme: companies have filed applications but not received the funds yet, which can cover up to 25 % of their annual turnover.
Our standard short-term GDP forecasting models are not suited to the current situation, which has prompted us to develop an alternative sector-based approach.
The lockdown is a thoroughly unprecedented and exceptional situation. Our standard short-term GDP forecasting model, the MIBA model, is built to adjust optimally to typical historical patterns, making it inappropriate for the current break in trends.
For this reason, we have developed an alternative method to assess the effects of the lockdown by sectors, drawing chiefly on the information provided by the March MBS. The loss of activity in each sector was estimated at the finest disaggregation level possible based on the number of exceptional closure days reported by companies over the month of March. This information was corroborated by other survey variables and supplemented where applicable by other available sources. Table 1 presents the results.
According to this sector-based approach, the loss of activity for the economy as a whole is estimated at 32 % in a typical lockdown week.
The greatest losses are in construction (loss equivalent to three-quarters of normal activity) and trade. Losses in transportation, accommodation and food services are equivalent to two-thirds of normal activity.
Manufacturing industry is also heavily impacted (loss equivalent to around one-half of normal activity), as are other market services (equivalent to approximately one-third), including artistic activities and entertainment, as well as business services affected by the global economic slowdown.
Combining all of these sectors, namely manufacturing industry, construction and non-financial market services, which together represent 55 % of GDP, the loss of activity represents about one-half of the normal level.
Less severely affected sectors, which represent just under one-half of GDP, include agriculture, manufacturing of food, coking, refining and energy production, non-market services, and financial and real estate services. Financial and real estate services, for example, include actual and imputed rents, which are inherently unaffected by the lockdown.
Over the month of March, the loss of activity is estimated at 17 %, based on two weeks of lockdown and the downturn in activity recorded in certain sectors from the start of the month.
Our previous forecast for quarterly GDP growth in the first quarter, which we published a month ago with the February survey, was +0.1%. The impact on activity from the lockdown measures estimated above has caused us to sharply revise our estimated quarterly GDP growth rate to -6 %.
This estimate does not offer the same accuracy as our usual estimates, but nevertheless provides a useful order of magnitude. It is an approximation extrapolated from the 32 % loss of activity during the lockdown, which is divided by six to reflect a half-month of lockdown over the entire quarter. Besides the mechanical impact of the lockdown, the estimate also factors in the effects of the COVID-19 crisis in the first fortnight of March, even if these effects are considerably smaller; the slowdown in tourism, for example, fed through into slacker activity in several services sectors from the start of the month onwards. Conversely, GDP growth in the first two months of the year was marginal when set against the March collapse and does not affect our calculation.
It is necessary to go back to the second quarter of 1968, which saw popular unrest and widespread protests in the month of May, to find a quarterly decrease in activity of similar size. At that time, GDP contracted by 5.3 %, before rebounding by 8.0 % in the third quarter of 1968.
All other things being equal, each fortnight of lockdown takes around 1.5 % off annual GDP.
This result is obtained by dividing the estimated loss of GDP for one fortnight (32 %) by the number of fortnights in the year (24). However, extrapolations can be taken only so far, and it is possible that the fortnightly loss of GDP will change as the total lockdown period lengthens.
This estimate is backed up by other data sources. Other information taken from the business conditions survey, including the capacity utilisation rate in manufacturing industry, corroborates the order of magnitude obtained.
Today's situation may also be compared to the downturn observed during the 2008 financial crisis. Balances of opinion on activity in the business survey fell by about 1.5 times more in the first quarter of 2020 than they did in the fourth quarter of 2008 in manufacturing industry and by about four times more in market services. A relationship of similar proportions exists between our estimate of the downturn in activity in these two sectors for the first quarter of 2020 and the decrease that they recorded in the fourth quarter of 2008.
Other high-frequency data sources back up the estimate. For example, bankcard payment volumes fell by 50 % on average during the week of 23 to 29 March. Let us assume that these payments are representative of the variable portion of consumer spending, which makes up between 60 and 65 % of total household consumption (the remainder comprising actual and imputed rents and other committed expenses). Under this assumption, total consumption declined by approximately 30 % during the week beginning on 23 March, i.e. on a par with our estimated economic downturn.
Another external data source is information on the “partial activity” applications filed by companies, which concerned 3.9 million employees as at 1 April, according to figures compiled by the Research and Statistics Directorate of the French Labour Ministry (DARES), or around 13 % of the labour force. This is a useful statistic to monitor but tricky to employ in tracking business conditions, particularly since it is increasing very quickly and the average number of partial unemployment hours per beneficiary is not available. When compared against a contraction of approximately 30 % in economic activity, however, this indicator looks consistent with the size of decrease that we estimated.
Updated on: 04/14/2020 08:53