Investors are much like a two-year old, going from one shiny thing to the next and grabbing onto whatever may be close at hand to satisfy some unknown craving. So it is with economic data, with the quarterly reports from companies and GDP reports getting intense scrutiny for maybe a whole day then onto the next report. The monthly reports on the economic health fare a bit better, with discussions around employment, inflation and housing data given a couple of days, with employment getting the bulk of the weekend to mull over investment and economic implications. However it is the weekly reports that get investors excited, as these reports provide the buzz that may wear off soon, but come next week, they’ll get another hit. There are plenty of these that have changed in importance over time. Many years ago, investors would await the monetary supply M1 and M2 reports after Thursday’s market close to determine any changes in Fed policy. Of course this was before they were so transparent (and trampled the various M’s). Focus has swung around to the reports on mortgage activity (refinancing and new purchases) as well as the initial jobless claims report, which has moved to the “granddaddy” of the weekly releases.
The issues surrounding the jobless claims figures are not on the release for the week, but the revisions to past weekly releases, which have been predominately higher over the past year than initially reported. In a bit of mathematical juggling, the headlines tout a decline of 5,000 from the prior week, however the prior week was revised higher by 5,000, meaning the current release is essentially unchanged from the prior week. In fact, the Wall Street Journal did an article highlighting the weekly revisions to the seasonally adjusted figures with 56 of the past 57 adjusted higher. What is missing in that analysis is that the weekly claims are still declining, continuing the trend from early 2009. At the heart of the matter may be the seasonal adjustments, holidays hitting different parts of the year (Easter for example) or over/under estimating the Christmas hiring season. Instead of all the hand wringing over the adjustments and lack of “intelligence” behind the changes, it might be better to look at the raw data without all the seasonal adjustments. While this method does miss the holiday changes from year to year, there are some very distinct patterns that are at the heart of the adjusted data.
There have been some weekly adjustments to even the non-seasonal figures above; however they do not change the characteristics of the chart or the slope of the past four years. The busy arrows in the above chart highlight the seasonal effects of this series, the peaks ALL are a week or two into the New Year, corresponding to the seasonal layoff of Christmas workers. Take a guess at the lows – right around Labor Day (get it?!). This corresponds to the peak summer employment part of the year, from this point until January; there is a rather steep rise into the January peak and a few steps lower into the September annual low. These characteristics were in place prior to the recession as well as following. The data is available back to 1967 and these seasonal factors have been in place throughout. So, what’s the point? The seasonally adjusted figures try to account for these regular changes to this series to capture a “truer” picture of the weekly data. Unfortunately when these adjustments are “out of step”, the revisions lead many to believe the system is broken or more seriously, that the reports are somehow politically motivated.
Is there a better way to compensate for what maybe faulty seasonal adjustments? I think so. Take a look at the chart below, it is the year over year comparison of the non-seasonally adjusted chart above.
Obviously the major highlight is the huge spike in weekly claims corresponding with the recession beginning actually early in 2008. By the end of March ’08, the y/y change in weekly claims surpassed any slowdown since the 2001-2003 recession. This was an early sign that things were different in the economy well ahead of the market collapse later in ’08. Today, the non-seasonally adjusted figures have been running 9-10% below year ago levels, consistent with an improving economy. If we compare today’s level to that of the period between 2003 and 2007, the weekly figures are about 5% above similar dates during that period. At least those figures are in the ballpark of the recent past, meaning employment is getting back to “normal”.
The weekly figure does not square with the monthly jobless report, so using a particular level in the weekly data as corresponding to a certain level of unemployment does not work. What the historical and data above indicate is that the direction of claims does correspond to the direction of the monthly report. The most recent month was a disappointing report; however jobs were added, albeit less than were anticipated. The key in looking at the weekly data is changes from year ago periods that signal changes in the monthly reports – either adding or losing jobs. For example, during those periods where the year over year changes in weekly claims were positive the monthly reports showed a slowing or decline in employment growth. Again, the correlation is not on a one-to-one, but directional. So based upon the still declining weekly figures, I would anticipate the monthly jobs report to continue to show growth in payrolls.