KEY DATA: ISM (Manufacturing): +2.6 points; Orders: +4.7 points; Production: +5.7 points; Employment: +4.6 points/ Construction: -2.9%; Residential: -4.5%; Nonresidential: -1.8%
IN A NUTSHELL: “The reopening is slowing the decline in activity.”
WHAT IT MEANS: The economy is reopening and we are starting to see the economic numbers look less ugly. Not, they are not good, but terrible can be an improvement from horrendous, so I will take it. Consider the state of manufacturing. The Institute for Supply Management’s May index rose. On the surface, that looks good and in this case, up is always better than down. But while just about every component increased, the levels still point to massive problems in the sector. Consider new orders, which seemed to be a lot better. Not really. While the percent of respondents saying demand and production improved, the percent saying orders and output were down remained above fifty percent. What is happening is that conditions are worsening at a slower pace. That was also true with employment. The share saying payrolls were down fell to 41% from 46% in April, indicating that lots of firms are still downsizing.
As for construction, it is actually holding in pretty well. Yes, activity declined in April but less than expected. Most of the drop was in residential. Given the collapse of starts, the drop was much lower than forecast. States had started allowing construction to resume earlier than other activities, so I don’t expect construction to be the weakest link this quarter.
Commentary: Don’t take the data at face value.
We are entering a difficult time for the data mills, which includes the government, the Fed, industry groups and those private sector companies that produce numbers. Government actions have created special circumstances that raise questions about the true meaning of the numbers.
The first and maybe biggest issue is that the seasonal adjustment factors may be largely meaningless. Changes in activity are likely occurring more because of states allowing activities to occur rather than normal seasonal changes. Seasonal factors cannot adjust for that. Thus, while all of the data creators are doing the best they can, the seasonally-adjusted data they produce are suspect. The employment numbers on Friday are likely to be really ugly, but they will suffer from an inability to decipher what was going on due to seasonal changes and what was happening as a result of government decisions or economic conditions.
The second issue, which is something only we data wonks worry about, affects certain types of measures, such as diffusion indices. For example, the ISM report referenced above is a typical diffusion index in that it asks if the measure under consideration (orders, production, employment etc.) was better (higher), the same or lower (worse). The difference between better and worse is calculated, seasonally adjusted and turned into an index. Consider the case of a firm that closed down its production line in April. That month, it reported that output declined, which lowered the overall index. But in May, if it was still shuttered, it reported that output remained the same i.e., zero. With a diffusion index, going from down to unchanged actually increases the index level. That would happen despite the fact that the firm still produced nothing. In normal times, the ebb and flow of these changes even out. In times like this, looking at the overall index level rather than the components can be very misleading. Unfortunately, under the current circumstances, even the data nerds who dig into those details are having trouble understanding what is happening.
The final issue is the one I mentioned in my unemployment claims report. To the extent that government policy was designed in a way that intentionally or unintentionally masks the true nature of the data, the numbers become suspect. If the PPP causes firms to hire people that don’t actually produce anything, and that is likely given the spending requirements, the private sector employment numbers are corrupted. A worker gets moved from unemployment, lowering that number, and is counted as employed, raising that number. But that worker effectively remains on the government payroll (paid by a PPP grant) and still doesn’t produce anything. The employment and unemployment numbers are artificially improved, but nothing changed. We have no measure of those workers, so the best we can say over the next few months is that the payroll data overstate the true state of things while the unemployment rate understates what is happening. Keep that in mind when you see the next few months of employment reports.