Jobless Claims Plunge To 334K, But July Seasonal Adjustment Is Problematic

by: Doug Short

The Unemployment Insurance Weekly Claims Report was released yesterday morning for last week. The 334,000 new claims number was a 24,000 decrease from the previous week's 358,000 (a downward revision from 360,000). The less volatile and closely watched four-week moving average, which is usually a better indicator of the recent trend, fell by 5,250 to 346,000. Many analysts are skeptical of the July numbers because of seasonal adjustments based on unpredictable schedules of factory shutdowns for annual retooling, especially in the auto industry.

Here is the official statement from the Department of Labor:

In the week ending July 13, the advance figure for seasonally adjusted initial claims was 334,000, a decrease of 24,000 from the previous week's revised figure of 358,000. The 4-week moving average was 346,000, a decrease of 5,250 from the previous week's revised average of 351,250.

The advance seasonally adjusted insured unemployment rate was 2.4 percent for the week ending July 6, an increase of 0.1 percentage point from the prior week's unrevised rate. The advance number for seasonally adjusted insured unemployment during the week ending July 6 was 3,114,000, an increase of 91,000 from the preceding week's revised level of 3,023,000. The 4-week moving average was 3,019,250, an increase of 37,000 from the preceding week's revised average of 2,982,250.

Thursday's seasonally adjusted number was below the forecast of 345K and's 348K.

Here is a close look at the data over the past few years (with a callout for the several months), which gives a clearer sense of the overall trend in relation to the last recession and the trend in recent weeks.

As we can see, there's a good bit of volatility in this indicator, which is why the 4-week moving average (the highlighted number) is a more useful number than the weekly data. Here is the complete data series.

Occasionally I see articles critical of seasonal adjustment, especially when the non-adjusted number better suits the author's bias. But a comparison of these two charts clearly shows extreme volatility of the non-adjusted data, and the 4-week MA gives an indication of the recurring pattern of seasonal change in the second chart (note, for example, those regular January spikes).

Because of the extreme volatility of the non-adjusted weekly data, a 52-week moving average gives a better sense of the long-term trends. I've now added a linear regression through the data. We can see that this metric continues to fall below the long-term trend stretching back to 1968.

A Four Year Comparison

Here is an overlay of the past three calendar years and the beginning of 2013 using the 4-week moving average. The purpose is to show the relative slope of improvement since the peak in the spring of 2009. The latest year was off to an excellent start. It then oscillated a bit, but the latest eight weeks have stalled in a fairly narrow range.

For an analysis of unemployment claims as a percent of the labor force, see my recent commentary What Do Weekly Unemployment Claims Tell us About Recession Risk? Here is a snapshot from that analysis.

For a broader view of unemployment, see the latest update in my monthly series Unemployment and the Market Since 1948.