Recent Updates

  • PAKISTAN: Industrial Production (Sep), BOP (Oct)
  • Taiwan: Consumer Confidence Index (Nov)
  • Korea: Housing Price Indexes (Nov)
  • Australia: Retail Trade Prelim (Oct)
  • New Zealand: Employment Indicators (Oct)
  • more updates...

Economy in Brief

Bank of France Indicator Erodes in July
by Robert Brusca  August 9, 2021

Both orders and expected employment have steadily improved since the French recession that was further intensified by the introduction of the Covid virus. The headline series (see graph) weakened sharply when the Covid virus struck then rebounded smartly and quickly. But since then, the level reading has faltered a bit then risen and stabilized showing a hint of step back as the index had fallen back to 104.7 in July from 106.9 in June.

Notice the difference in the path between the level and the percentage change plots. The levels plot get back to normal very quickly by mid-2020. But the percentage change plot falls sharply then recovers back to normalcy fairly quickly; but it takes it a year for it to show a strong-spike gain followed by a sharp pullback as the base for the calculation of ‘level change’ drops into the trough of the recession pushing up the percentage change one year later then fading as that base climbs in the recovery path to a more normal level. Although we like to use percentage changes on data, that approach as a gauge lags the real effects by a year. In fact, the sharp percentage in the year ahead (in this case in early-2021) has almost no connection to the ongoing reality (the actual level of the survey) which has been steady and level for the ‘better part of a year.’

Setting that observation aside, let’s focus on the table and the monthly values therein plus their connections to past values with past benchmarks constructed in various ways to provide different ‘angles’ for assessment.

For example, we see that the current survey index at 104.7 is lower than its June value of 106.9 and lower than its May value as well. But it is higher than it has been ‘on average’ over the last 12 months (compared to 102.2); it is also above its long-term average which is calculated on data back to August 1990 and stands at 99.6. Put in an array of past data, and ranked back to August 990, the current value of the survey indicator has a 72.3 percentile standing. That tells us that this month, among all historic reading on that timeline, has been this high or higher only 27.7% of the time (100%-72.3%). This statistic puts the current number in better more comprehensible historic context for us. The survey indicator has the business sector well above its median. The median for ranking data lies at the 50th percentile in all cases (minor exception: when there are ties for which number is in the middle).

We can similarly peruse the other data in the table. Right off we see that only two components in improve month-to-month: the output change that rose to 18.2 from 17.6 and the change in finished inventories that logged a slightly smaller negative reading. However, only one component, expected employment, stands above its level of May 2021 as of June… (expected production falls short, too). Standing above their 12-month averages are total output change, order books, capacity utilization, and expected employment. The higher standing of the overall index above its 12-month average is on the back of four indicators: total output, total orders books, capacity utilization, and expected employment.

Compared to their long-term average, the readings that are higher in July are: output change, order books, change in new orders employment vs. last month and employment expectations – that’s five of nine components.

But in some cases, the level of the overall index has been boosted most substantially by a category or two that stand most strongly above their own historic observations. The ranking data reveal that best. Here we find the greatest relative strength right now is for expected employment (95.3%) followed by output change (87.1%) and order books (76.9%). Fourth is employment last month (65.9%). The remaining five observations stand below their respective historic median vales. Remember that the median is the value in the exact middle of the distribution of all observations. It is often similar to the average, but the existence of a cluster of extremely high or low observations can have much more impact on the average making comparisons with the median in some cases a more meaningful one with history.

For now the French business sector is being driven by upbeat employment expectations, strong changes in output, and solid increases in order books. Ironically, holding the index back the most are expected production, changes in finished inventories, and capacity utilization.

Some of this is odd. How can expected production be so weak and yet expected employment be so strong? And while the level of aggregate orders is solid with a 76.9 percentile standing, the change in foreign and total orders both are below their respective medians. That suggests that with output rising briskly order books are being satisfied while new orders are not coming in fast enough setting the stage for a decline the sufficiency of orders overall. Indeed, we do see some slight erosion monthly in order books. While employment monthly has been positive, it is fading; yet, expectations for employment strengthen even as output expectations have fluctuated. Some of these observations are simply survey inconsistencies.

To understand how various portions of the survey relate to other portions, we can look at a correlation matrix. This table (below) is symmetric so we only need to look at one half of it. I have chosen the upper portion above the diagonal. It contains 45 unique observations.

The survey headline has a strong (70% or higher) correlation with 6 of its nine components).
Output change has only one strong correlation with another component that is with order books.
Expected production has no high correlation with a component or even with the survey headline; it has a below 0.5 correlation with expected employment. That low correlation between two related forward-looking variables is unsatisfying and undermines the veracity of the survey- even though some disconnect between these two gauges is quite plausible.
Order books have a strong correlation with output change, capacity use, and the change in employment month-to-month among components.
• The change in new foreign orders has strong correlation to change in new orders (which it should since it is part of that series) and the change in employment vs. last month. The correlation to employment is a very reasonable connection and it is heartening to see the connection is between orders today and the monthly change in employment, so there is some contemporaneousness in it.
• Not surprisingly the change in total new orders correlates with what the change in new foreign orders correlates with since that overall category contains new foreign orders.
• The change in finished inventories is highly correlated with nothing (in fact, all very low correlations).
Capacity utilization is relatively highly correlated with the level of orders in order books- that too is sensible.
• The change in employment last month is correlated with order books, change in new orders, change in foreign orders and change in employment month-to-month.
• The final category is the expected change in employment which we have seen is only highly correlated with the change in employment month-to-month, a sort of accelerator relationship- employment rises and then is expected to rise further. Expected employment has very low correlation with everything else including the headline except that it is also has a high (but not quite ‘strong’) correlation with the order books. In U.S., data correlations between order backlogs and employment are an established statistical relationship.

Survey Interactions

On balance, the correlation matrix does give us a bit of insight into the inner working of the BOF business survey. When analyzing it, we can look at the overall reading and then treat the two expectations components for production and employment as separate indexes because they are so poorly correlated with the BOF index itself. This month the headline weakness and both the expectations series weakened.

large image