Inferring recession probabilities from one indicator is a dodgy way to monitor business-cycle risk, but it’s forever popular. Last week I criticized an effort to estimate recession risk by using the ISM Manufacturing Index in isolation; today we focus on a similar attempt to use the Labor Market Conditions Index (LMCI) as the basis for deciding if the US economy has slipped over to the dark side.

Earlier this week, Bloomberg reported that “recession risk is rising,” based on analysis of LMCI.

After falling just three times from 2012 to 2015, the index has fallen every month of 2016 except for one, July. And in July the annual change in the LMCI, from July 2015, turned negative.

That’s only the eighth time in nearly 40 years the index was down on a year-over-year basis, Deutsche Bank Chief U.S. Economist Joseph LaVorgna wrote in a note to clients today. Of the seven previous occasions, LaVorgna wrote, “four were soon followed by recession.”

LMCI is published monthly by the Federal Reserve. This multi-factor benchmark is a valuable tool for analyzing the labor market and the economy, but like any data set it suffers limits for capturing the breadth and depth of the US macro trend. Granted, no serious business-cycle model can ignore job creation and related metrics, and by that logic LMCI is a worthy time series. But it’s not flawless, and so the possibility of false signals is forever lurking in real-time analysis. That’s hardly a surprise—every indicator falls short of dispensing reliable and timely warnings of recession warnings.

The solution is to build a business cycle benchmark based on a range of indicators and monitor the results regularly. There’s always a degree of uncertainty, but combining indicators is a solid foundation for robust analytics. Modeling the data from different perspectives further enhances the dependability of the signals. As an example, the US Business Cycle Risk Report combines analysis of two broadly defined indexes published by Federal Reserve banks with a pair of proprietary benchmarks designed by The Capital Spectator. But I digress.

Let’s take LMCI at face value and reflect on the implied message. Looking at the raw values, it’s clear that labor conditions are relatively soft. LMCI has been printing at modestly negative values for most of the year to date, slipping to -2.2 in September (the October release is scheduled for Nov. 7). The red ink marks the eighth month in the last nine that LMCI has posted a mildly sub-zero reading.

Clearly, employment conditions have weakened in recent months. But is the deceleration weak enough to trigger a recession? At the moment, no, based on running the raw LMCI numbers through a probit model in context with NBER’s historical data on recession dates. As the next chart below shows, the probability that the US entered a recession in September on this basis is virtually nil.

Let’s re-run the numbers using year-over-year changes in LMCI. The results for this methodology show a slightly higher recession risk for last month—roughly 7%–but that’s still too low to sound the alarms.

To be fair, LMCI’s trend of late doesn’t look encouraging. But as the first chart above reminds, this indicator has dipped below zero several times through the years without a subsequent recession. In fact, LMCI has been considerably lower at various points with no downturn in the immediate aftermath. Does that mean that the current weakness in LMCI is another false signal? Not necessarily, but it’s premature to rule out the possibility that the macro trend may reaccelerate.

Meantime, based on what we know now, LMCI isn’t signaling a recession. Even if that changes in the next update, there’d still be room for debate about the economy’s near-term prospects. As I noted last week, current projections for the US business cycle—based on a broad set of figures—suggest that a modest growth bias will prevail through November.

The obvious caveat: nothing’s written in stone. But at the moment, it’s too soon to assume that an NBER-defined recession has started or is about to start. The real challenge, as always, is deciding when a broad set of indicators tells us that the current expansion has expired. That’s not easy if we’re looking for high-confidence signals that are also timely, but success on those dual mandates is nearly impossible if we rely on one or two indicators.