Job postings are a leading indicator. Most investors aren't reading them.

Hiring activity encodes information that financial statements cannot. When companies post jobs, they reveal operational priorities, growth expectations, and strategic direction, often weeks or months before those signals appear in reported earnings.

Corporate job postings have long been studied as a leading indicator of firm-level performance. But most prior research has focused narrowly on aggregate posting volumes, leaving the richer information embedded in job descriptions (required skills, qualification levels, geographic distribution, and role composition) largely untapped. Advances in large-scale NLP have changed that calculus.

Hiring dynamics reflect management expectations, operational scaling, and strategic positioning: a forward-looking lens on firm activity not fully captured in traditional financial statements.

RavenPack Jobs Analytics provides a structured, investable layer on top of raw posting data, covering more than 200 million job listings across 100,000+ companies in 150 countries since 2007. This post summarizes key findings from our latest whitepaper, which backtests four distinct signal strategies across the U.S. Top 1000 universe from January 2014 to February 2026.

Four signals, one dataset

The research constructs four independent strategies, each extracting a different dimension of information from the same underlying dataset. All portfolios are sector-neutral and dollar-neutral, with daily rebalancing across the top and bottom 20% of firms within each sector.

01. Hiring growth + geographic consistency: Monthly hiring growth is positively correlated with future stock performance, but not all growth is equal. Firms scaling headcount within their existing geographic footprint outperform those entering new markets, where uncertainty and investment horizons are longer. Conditioning on location cosine similarity improves annualized returns from 1.32% to 2.26%. IR: 0.91.

02. Role-specific demand: Aggregate hiring growth obscures sector-level heterogeneity. A systematic framework, identifying the most in-demand O*NET occupation groups within each sector using only prior-year data, delivers materially better signal quality. Computer Occupations in Technology Services, Engineers in Producer Manufacturing, and Financial Specialists in Finance are selected in over 90% of periods. IR: 1.10, annualized returns: 2.28%.

03. Soft skill stability: Job descriptions provide orthogonal alpha. Firms with stable soft skill distributions — measured via year-over-year cosine similarity of quarterly skill vectors — outperform peers with shifting profiles. The strategy exhibits low daily turnover (~4%), implying effective holding periods exceeding one month. IR: 0.69, annualized returns: 1.43%.

04. Qualification breadth: Firms increasing the diversity of qualifications demanded across their open roles — a proxy for an upgrade in expected future hire quality — outperform peers. The signal spans all six dimensions of the RavenPack Jobs Taxonomy: skills, experience, education, abilities, personality traits, and knowledge. IR: 0.92, annualized returns: 1.84%.

Why this matters for systematic managers

The four signals display relatively low cross-correlation, enabling meaningful diversification in a combined portfolio. After equal-weighting across all four strategies and applying factor risk adjustment, the combined approach retains the majority of its raw performance — factor-adjusted IR of 1.22 at a one-week horizon — indicating the signals capture idiosyncratic alpha rather than known risk premia.

For systematic managers seeking uncorrelated alpha, hiring data represents a meaningful expansion of the alternative data toolkit: one that is forward-looking, high-frequency, and structurally connected to the drivers of firm fundamentals.

Download the full whitepaper to get the complete methodology, signal construction formulas, sector-level performance decomposition, multi-factor risk attribution, and robustness tests across portfolio sizes and percentile thresholds.