Reality: Unemployment stats continue to improve
Reported: Unemployment stats look terrible
So, what happened? A flawed misinterpretation of the data – once again.
The good facts, the bad analysis and the ugly headlines
Let’s start with the ugly headlines…
- The Wall Street Journal – “Rise in Weekly Unemployment Claims Points to Faltering Jobs Recovery”
- CNN Business – “American unemployment claims are on the rise again for the first time in 4 months”
- Yahoo Finance – “Jobless claims: Another 1.416 million Americans filed new unemployment claims last week, as labor market recovery sputters”
- USA Today – “‘The jobs just aren’t there’: Number of Americans filing for unemployment rises for first time since March, even as aid is set to shrink”
Now to the bad analysis…
The ugly headlines are all based on a single, flawed comparison: The seasonally-adjusted weekly initial unemployment claims as of July 18 compared to those of July 11.
That’s it. Just two numbers. No look back at the multi-week trend.
Worse, there is no consideration of what’s happened to all those who filed initial claims in the preceding weeks. Those still-unemployed are tracked and reported by the “continuing claims” they must file weekly to continue receiving benefits. The importance of these numbers is that they reveal the dynamics of the unemployed, especially those who have moved on, primarily to new employment.
Finally, the bad data
No thought is given to the use of seasonally-adjusted data. It is improper because the seasonal adjustments are based on normal, seasonal employment shifts, and 2020 is wildly abnormal. Therefore, the huge coronavirus-shutdown unemployment numbers should not be inflated or deflated by former seasonal trends adjustments.
Now to the misinterpretations of the latest weekly unemployment report
First, the misapplied seasonal adjustments produced an increase from a decrease
Start with normality…
In normal times, seasonal adjustments allow better short-term comparisons, where seasonal factors produce regular cycles. The following graph shows the weekly adjustment factors for initial unemployment claims, covering the past three years, overlaid with 2020 through July 18.
Now to 2020…
It’s the 2020 unemployment adjustments that are wrong. What we are witnessing are the huge, abnormal effects caused by coronavirus-caused business conditions – not the usual seasonal hiring and layoff patterns associated with holidays and vacation times.
Therefore, the true measure of what’s happening now is the not-seasonally-adjusted claims figures. There is no rationale to support inflating or deflating those abnormal numbers by those normal seasonal factors. (Just look at the huge increases ahead, as high as 30%, coming in the normally low unemployment summer weeks. Now imagine the scary headlines coming when those seasonally-adjusted amounts are reported.)
The following two graphs show the projected seasonal adjustments if the NSA initial claims were 1 million per week (down from the average 1.4 million over the past five weeks). Note the increases coming over the next 14 weeks to adjust for the normally low number of claims. Clearly, increasing the NSA numbers is wrong, yet those recalculated SA numbers will be used by the media as proof of tougher times.
Now to the data behind the Thursday (July 23) report for the week ended July 18…
(The source of the data is the Department of Labor “Unemployment Insurance Weekly Claims” report)
The previous (July 11) weekly not-seasonally-adjusted (NSA) initial claims were 1.51 million. However, that reading was seasonally-adjusted (SA) downwards by -13.6% to 1.31 million.
In the recent (July 18) weekly report, the one that produced the negative media reports, NSA initial claims were 1.37 million, well below the previous week’s NSA 1.51 million. However, it was adjusted upwards by 3.3% to 1.42 and then compared to the 1.31 figure.
So, the actual weekly -0.14 million decrease was morphed into a misleading 0.11 million increase, fodder for the many negative reports (even labeled an unemployment “spike” by some).
Moreover, because the predominate seasonal adjustment has been upwards throughout the coronavirus period, the cumulative initial claims number has been an ever-increasingly overstated amount. Here is the graph showing the two cumulative amounts – NSA and SA.
Second, the focus on initial claims (newly unemployed) ignores the strong employment dynamics at work
Initial claims do not mean forever unemployed. Once a person is unemployed, stuff happens – most often a new job. And that has been happening regularly even through the coronavirus period. But to get to that dynamic, we need to go beyond the initial claims report and, especially, the cumulative initial claims amount.
This means watching the other figure oft-ignored “continued [weekly unemployment] claims.” Although they appear in the same report as the initial claims, they lag by a week, so tend to be overlooked. However, they are key to understanding the unemployment picture.
Start with the initial vs continuing claims figures
This first graph shows the cumulative NSA initial claims (orange line) since mid-March, when the coronavirus-shutdown effect took hold. The NSA continuing claims (green bars) show the current number still receiving unemployment benefits.
Note the huge difference now between all those who initially filed and those who are still filing. As of July 11, the cumulative initial claims were a whopping 47.7 million, but the continuing claims were “only” 16.4 million, a decline of almost 2/3. Clearly, many people have found work and income elsewhere.
But don’t those high numbers (1-1/2 million initial claims and 16 million continuing claims) means unemployment is still a problem?
Yes, but neither number is stagnant. There clearly dynamics at work when 1-1/2 million initial inflow does not push continuing claims up. Necessarily it means that there is a similar 1-1/2 million (about 10%) who are leaving the continuing claims pool each week. This outflow is the great unmentioned good news that is rarely (never?) mentioned.
So, here is a graph makes this issue clear. The green bars are the continuing claims, and the orange line shows the cumulative departures of people from the unemployed pool. Note that the ratio is ~2:1 – that is, the number of the formerly unemployed is double the number of the still unemployed.
So, what is the conclusion from all the above?
Ignore the weekly message of “Yet again, __ million of newly unemployed!” It is both overstated and overly simplistic. To see the true unemployment trend, focus on the numbers ignored by the media:
- The not-seasonally-adjusted (NSA) initial claims
- The NSA continuing claims
- The outflows of the formerly unemployed
And here is one more thing, not discussed above…
Study the monthly “Employment Situation” report from the U.S. Bureau of Labor Statistics. A fulsome data-gathering process is conducted each month, then analyzed and reported in two steps: Preliminary findings in the following month and final results in the next month. (August 7 is when preliminary July 2020 and final June 2020 statistics are reported.)
The BLS provides employment and unemployment breakdowns by age, education, gender, ethnic group, etc. Unemployment numbers are total, regardless of whether unemployed are receiving benefit payments or not. (Graph below shows difference in numbers.)
Examining the report reveals longstanding facts, current trends and a reality that is lost by simplified, one-number reporting (e.g., the unemployment rate of 11.1%).
The bottom line: We need to conduct our own unemployment analysis
For an accurate view of the unemployment picture, we need to ignore the media’s myopic view of the one-week change in the seasonally-adjusted initial unemployment claims.
Why must we do it ourselves? Because today’s news operates under the 15-minute rule. Publish a news report over 15 minutes after the event (e.g., the release of the weekly claims data) and the number of viewers drops off significantly. Obviously, that leaves no time for thoughtful analysis. It’s also why headlines look so similar nowadays.