Jobless claims

Two series arrive in the same weekly release from the United States Department of Labor. Initial claims count first-time filings made during the week that ended the previous Saturday. Continuing claims count people still receiving payments and refer to a week further back again, because a person has to have filed before they can be counted as continuing. Both are published seasonally adjusted, and the unadjusted counts are published beside them.

Frequency is the whole point of the release. A monthly reading such as non-farm payrolls dates a change in the labour market to a month, while a weekly filing count dates it to a week, which is why the series is treated as a leading indicator of the monthly employment figures. The reading convention is the four week moving average rather than any single week, because one week is moved by public holidays, seasonal plant shutdowns, strikes, severe weather and processing backlogs in individual states, and because states that do not report in time are estimated and then corrected in the following week.

The point most often missed is that the series counts an administrative act, not unemployment. Only workers who are eligible and who actually file appear in it, which leaves out the self employed, people who never qualified, and anyone whose entitlement has run out, and eligibility rules differ between states and change with legislation. A low count is therefore consistent with a weak labour market in which comparatively few of the people losing work qualify to claim, and comparisons across decades measure a shifting population as much as a shifting economy.

Where practitioners disagree is on what any given level means. One approach reads the series against its own recent range, on the grounds that the level that once signalled contraction has drifted with the size of the workforce and with the rules; another reads a sustained rise over several weeks as the only reliable signal in the data, treating the level itself as uninformative. Both readings are current, and they routinely produce opposite descriptions of the same release.