Whiplash: Trading on the MTS Rollercoaster

Originally published January 12, 2007

There ain’t no seat-belt hefty enough to keep you in your seat if you decide to take positions based on evidence in
the Monthly Treasury Statements. 

Every couple of months, an analyst seizes on a fluctuation, often a wild fluctuation,
in the Monthly Treasury Statements to make the case that the job market is either far stronger or way weaker
than the Bureau of Labor Statistics’ estimates suggest. Oh, OK, these remarks do come disproportionately
form those on the hunt for evidence that the Bureau of Labor Statistics is underestimating payrolls, witness recent attention to January’s surge in withholding at the federal level, recent stories focused on another strong showing in March receipts, and the fact that once the “hidden strength” story is out in the markets for a given month, the almost inevitable reversal in the following month never makes it to traders’ screens.

We have spent a significant amount of time trying to tease
meaningful trends out of the withholding data in the Monthly Treasury
Statements, and it is generally a fruitless effort. The very issues
that our state contacts work hard to unravel—the percentage of the
gains coming from bonuses, from the high-income groups, and even from
severance payments—a painful trend in the Midwest—run unfettered in
the monthly treasury statements. On many occasions our state contacts
anticipate significant increases or declines in a given month’s
receipts (based on the timing of bonus payments, extra pay-in days and
other calendar issues), and we watch, somewhat bemused, as the
corresponding movement in the MTS flows garner a good bit of attention
in the markets. Our state contacts argue forcefully that such swings
tell you very little about the state of the labor market. Indeed, the
correlation coefficient between yearly changes in employment and
Federal withholding since 1975 is a minuscule 0.196.

A much better fit with employment can be found using FICA receipts,
which reflect employment changes at all but the highest level of the
wage distribution (correlation between the monthly measures of yearly
changes is a much more respectable 0.537, though we have the data only
going back to 1999). In the nearby graph we adjusted FICA receipts for
the number of workdays per month and take a six-month moving average,
which raises the correlation with employment to 0.650. There is also a
decent, but hardly seamless, correlation between the yearly change in
social insurance contributions series in the national income accounts
and payroll employment. (In neither case is the fit perfect; note
misses on the employment decline and recovery in the early 1990s.)


Social insurance contributions and FICA receipts are considerably
below their levels of a year earlier—just like employment. And that’s
just what our tax contacts had forecast.

So if you don’t mind a little whiplash and are happy to lose money,
pay your money and take your chances, but remember, there are far safer

– Philippa Dunne & Doug Henwood