Updated: Oct 21, 2019
Our custom labor cost indicator, which can forecast hourly earnings growth up to two quarters in advance, suggests moderating wage inflation
An Economic Conundrum
One of the biggest mysteries facing economists since the financial crisis has been the decoupling of wages and employment. For ages, economists had relied on the trusted Phillips Curve, which describes an inverse relationship between wage inflation and the unemployment rate. However, this relationship has broken down over the past several years: despite the U.S. experiencing the tightest labor market in decades, wages have failed to meaningfully and sustainably increase.
While analysts have posited several competing explanations, such as weaker worker bargaining power or changes in labor force participation, nobody has developed a solid working model of worker pay in this economy, making it hard to forecast wage inflation reliably. This is why we thought extracting labor cost guidance would pose a valuable inaugural use-case of our technology, which allows us to infer macro insights from individual pieces of commentary.
In other words, we wanted to answer the question: does management rhetoric matter when it comes to forecasting wage inflation?
It turns out that it does.
Leveraging Natural Language Processing
We started with a simple premise: public companies whose management teams forecast higher labor costs will on average experience higher labor costs down the road. Aggregate measures of this guidance should therefore correlate positively with wage inflation indicators (of course, company payroll numbers also factor into labor costs, but the correlation should still hold).
What do these forecasts look like? Anything akin to "we anticipate higher labor costs into 2019" or "We have paid out higher wage rates and expect this trend to continue." Companies can talk about labor costs in all sorts of ways, and we didn't want to simply produce an advanced Control+F.
We didn't want to simply produce an advanced Control+F
Using our advanced natural language processing engine, we then derived a custom indicator by scanning for higher labor cost guidance in over 600,000 corporate documents in the Russell 3000 over the past 12 years. The indicator represents a simple aggregate: the percent of all Russell 3000 forecasting a lift in wages.
A Robust Leading Indicator
We then compared year-over-year change of this indicator to that of the official average hourly earnings time series published by the Bureau of Labor Statistics, with fascinating results:
Our indicator forecasts the official BLS wage inflation by a remarkable two quarters. The forecast period nevertheless represents what we feel is an intuitive and plausible management planning period.
A simple correlation coefficient between our indicator and the official BLS statistic (2 quarters ahead) comes out to 0.75, all the more impressive given the independent nature of these data sources.
Preliminary data suggests that wages are set to moderate over the next six months.
While the decline is consistent with the recent global growth slowdown (particularly in manufacturing), we think the tapering partly reflects the base effects of minimum wage hikes unveiled a few years ago by several local and state jurisdictions. In other words, the minimum wage is becoming less of a hot topic on earnings calls and financial disclosures than it was a year ago.
Equity Price Signal
Our wage indicator also seems to provide useful signal on share price movements.
A priori, we would expect investors, on average, to downgrade shares of companies announcing higher anticipated labor costs due to resulting pressure on margins. Indeed, a simple backtest shows that underweighting such stocks for 90 days after these firms release their guidance would have outperformed the broader Russell 300 index. Specifically, an equal-weighted long/short strategy (i.e. the Russell basket versus shares of a company that released higher labor cost guidance) would have yielded an annualized 2.0% return on average.
Of course, these high-level results still require further exploration. For example, we can further refine our asset price study by sector and guidance rationale to formulate a more nuanced trading strategy. However, it’s clear to us that even simple, high-level aggregates provide substantial forecasting power and a potentially useful macro overlay investors can leverage for sophisticated trading strategies.